Background
Lung cancer is a leading cause of cancer death worldwide. In 2012, 1,824,701 new lung cancer cases were diagnosed and 1,590,000 patients died of this devastating disease worldwide [
1]. The discovery of oncogenic driver mutations has contributed to the determination of the pathogenesis of lung cancer, development of therapeutic target agents, and prediction of treatment responses [
2‐
4]. The identification of actionable oncogenic driver mutations that guide selection of target agents has improved clinical outcomes of lung cancer patients, making tumor genotyping routine clinical practice [
5].
However, variable responses and eventual development of resistant clones to target agents are critical challenges to the drug development and therapeutic application. Recent advances from clinical studies have led investigators to posit that targeted therapies may fail to cure on account of tumor heterogeneity. Tumor heterogeneity implies that there are distinct morphological and phenotypic profiles of cellular morphology, gene expression, metabolism, motility, proliferation, and metastatic potential between tumors of the same type in different patients (intertumor), and between cancer cells within a single tumor of a patient (intratumor) [
6]. Tumor heterogeneity challenges not only the relevance of driver mutations to the clinical outcome and to response to target therapies, but also the development of diagnostic and therapeutic biomarkers [
7‐
11].
Development of heterogeneous clones inside tumors has been modeled by cancer stem cell (CSC) models and clonal evolution models. In the CSC model, heterogeneity among cancer cells is the result of differences in the stem cells from which they originate, and the variation of stem cells is explained by (1) epigenetic changes and (2) clonal evolution of the CSC population. On the other hand, in the clonal evolution model, tumors arise from a single mutated cell that accumulates diverse mutations with its progression of time in heterogeneous microenvironment. In the clonal evolution model, the spatial differences in exogenous mutational pressure such as cigarette smoking [
12‐
15] and endogenous processes of up-regulation of APOBEC (apolipoprotein B mRNA-editing enzyme, catalytic polypeptide-like) cytidine deaminase activity contribute to the regional variation in mutational burden over time [
16‐
18]. The spatial and temporal diversity in lung cancer evolution is attributed to a decrease in smoking-related mutation accompanied by an increase in APOBEC-associated mutation [
15].
The mutagenic effects of cigarette smoking are associated with C > A transversion [
14] and the proportion of C > A transversion is decreased in branched (late) mutations, which is indicative of a regional heterogeneous mutational burden attributed to smoking [
15]. APOBEC cytidine deaminase is a class of cytidine deaminases that convert cytosine to uracil during RNA editing in the context of a T
CW motif (where W corresponds to either A or T; the mutated base underlined). APOBEC-associated mutagenesis contributes to tumor progression and a large portion of subclonal diver mutations occur in the APOBEC context, which eventually contributes to a different mutational spectrum across regions and time [
19].
We doubt that a biopsy at a single site represents the genetic variation of the entire tumor. To clarify this, genetic profiles of primary tumor and metastatic lymph node (L/N) in the surgically resected pStage IIA-IIIA non-small cell lung cancer (NSCLC) were compared using next-generation sequencing (NGS)-based repeated deep sequencing of 16 genes containing actionable oncogenic mutations. The mutation profile of primary tumors and corresponding metastatic L/N was examined in respect of smoking related and APOBEC mediated signature. The difference in the genetic profiles between lesions was frequently observed in the genes that have important roles in cancer biology and it might influence responses to therapeutic agents.
Methods
Patient characteristics and tumor DNA samples
A total of 59 patients who met the following criteria were randomly selected from tissue archives of affiliated hospitals of Yonsei Medical Center: (1) a pathologically confirmed diagnosis of pStage IIA ~ IIIA lung adenocarcinoma, (2) history of curative surgical resection, (3) availability of tissues of both primary tumor and metastatic L/N, (4) confirmation of known oncogenic driver mutation either at ALK, EGFR, or KRAS, (5) history of platinum-based adjuvant chemotherapy and (6) submission of informed consent for tissue collection. After analysis of NGS data, 18 cases were excluded from further analysis because a driver mutation had not been confirmed at both the primary tumor and metastatic L/N. For the extraction of cancer enriched DNA from tissue block, paraffin-embedded tissue samples were loaded onto silanated slides as 4-μm-thickness sections. One slide of each block was lightly stained with H&E, re-examined for the presence of cancer cells and then the cancer-enriched area was marked by an independent lung pathologist. The cancer cell-enriched areas were scrapped with blades, and DNA was extracted using a QIAamp DNA FFPE Tissue Kit (Qiagen, Valencia, CA, USA). This study was approved by the ethical committee of Gangnam Severance Hospital (IRB #3-2013-0298).
Library preparation, NGS with Ion Torrent, and variant calling
Ten ng of gDNA was amplified by the Ion AmpliSeq™ Custom Panel (Life Technologies, Carlsbad, CA). This panel contains 16 genes that contain actionable mutations;
AKT1, ALK, BCL2L11, BRAF, DDR2, EGFR, ERBB2, FGFR1, KRAS, MAP2K1, MET, NRAS, PIK3CA, PTEN, ROS1, and RET [
20]
. The process for generating sequencing data using the Ion Torrent PGM sequencer is described elsewhere
. Briefly, multiplex pools were purified with Agencourt AMPure XP beads (Beckman Coulter Incorporated) and ligated with Ion Xpress barcode adapters (Life Technologies). The fragment size and quantity of each library were analyzed with a BioAnalyzer using a High Sensitivity Chip (Agilent, Santa Clara, CA). The library was diluted, and emulsion PCR was performed with the OneTouch™ reagent kit (Life Technologies). The emulsion PCR product was enriched using DynabeadsR MyOneTM Streptavidin C1 beads (Life Technologies). The final enriched ion spheres were mixed with a sequencing primer and polymerase and loaded onto five 318v2 chips. The libraries were sequenced with the Ion Torrent PGM sequencer at deep coverage (aiming for 1000×) using the Ion OneTouch 200 Template Kit v2 DL and Ion PGM Sequencing 200 Kit v2 with the 318 v2 chip kits (all from Life Technologies). The sequencing reads were aligned to the human reference GRCh37 genome, and base calling was performed using the Ion Torrent Suite V3.4.2 using tmap-f3 on the Ion Torrent server. The Ion Torrent Variant Caller (ITVC) v3.4 was used for the detection of mutations, requiring a frequency greater than 5% for a variant to be called. BAM (Binary sequence Alignment/Map format) and FASTQ files (alignment) were generated based on the base calling results and were used to report variant calling, including single nucleotide polymorphisms (SNPs) and insertions/deletions (INDELs).
Statistical analysis
Categorical and continuous variables were compared using χ2-tests and t-test, respectively. Differences in distribution of continuous variables between two independent samples were assessed by Mann–Whitney U test, and the Kaplan–Meier estimator was used for survival analysis. All analyses were performed with IBM SPSS Statistics version 20 (IBM Corp.). All statistical tests were two-sided, and a P-value <0.05 was considered to be statistically significant.
Discussion
Despite the continuous development of new therapeutic modalities for lung cancer, the outcomes still fall short of our expectations and a cure is seldom observed in patients treated with targeted therapy. One possible explanation for the heterogeneous responses and eventual development of resistance clones to the target agents is the existence of various types of tumor heterogeneity. Initial phenotypic heterogeneity and changes in cellular phenotypes resulting from adaptation to response and selection for resistant phenotypes need to be accounted for in order to achieve substantial improvements in therapeutic outcomes [
11]. At the same time, given the link between genetic heterogeneity and poor prognosis, a measure of heterogeneity by itself may be useful as a prognostic marker [
11].
In this study, spatial genetic heterogeneity of lung adenocarcinoma was compared by analyzing differences in 16 genes with known actionable mutations between primary and metastatic L/N. NSCLC tissues with confirmed
EGFR, KRAS, and
ALK driver mutations were randomly selected from our institutional tissue archives. The studies from the East Asian NSCLC patients showed that the mutation profile of East Asian NSCLC is different from that of Western population. Excluding the cases without identifiable driver mutations, the mutation profile of this study cases was
EGFR 85.3%,
KRAS 12.2%, and
ALK break-apart 2.4 %, which is comparable to the previous reports from East Asia. One study that recruited 1420 NSCLC patients revealed that 82 (5.8%) cases harbored
KRAS mutations either in codon 12 or 13. The low frequency of
KRAS mutation and high
EGFR mutation rate is consistent to the other reports from East Asia [
21‐
25]. When a confirmed driver mutation was not validated in both the primary lesion and metastatic lesion by the study panel, it was excluded from further analysis, based on the concept that these known driver mutations occur early in cancer development and further propagates through branched clonal evolution [
15,
26‐
28]. Using this criterion, 18 of 59 (30.5%) pairs were excluded from analysis. This simple but conservative study design may harbor risks that it might miss new clinically relevant mutations in the subclones. However, it might be a useful indicator of the presence of cancer DNA in the tissue samples mixed with a small number of cancer cells. Indeed, a recent meta-analysis on the differences in
EGFR and
KRAS mutation between primary and corresponding metastatic NSCLC tumors indicated that there is no difference between lesions [
26]. Compared to genome-wide approaches, this study adapted a focused approach with a specific set of genes that contain actionable mutations [
6]. We also adapted sequencing of pooled DNA from regional sampling instead of single cell sequencing, or longitudinal sampling approaches [
11,
29‐
31]. This approach has limitations to the analysis and classification of heterogeneity that may be attributed to diverse causes, but it may be the most relevant method applicable to the current medical fields.
In the primary tumor of the current smoker, where the mutagenic pressure of cigarette smoking is stronger than that of metastatic site, the proportion of C > A transversion is higher than that of the metastatic sites [
14,
15]. Interestingly, this study revealed that C > A transversion was more frequently detected in L/N metastasis rather than primary tumor, suggesting that other mechanisms that cause heterogeneity exist even considering the majority of study population are composed of non-smokers. The APOBEC deaminase activity which result in increase of C > T:G > A and C > G:G > C mutation in the context of T
CW motif may be attributed to the different regional accumulation of missense mutation with cancer progression [
15]. This study applied less stringent condition of APOBEC mediated mutation and showed that APOBEC signature is more prominent in the primary tumor rather than metastatic L/N. This can be assumed that the primary tumor has larger tumor burden with more complex microenvironment than the metastatic site.
Genetic heterogeneity may be implicated in poor clinical outcome. This study population did not show differences in the DFS and OS between those with homogenous mutation profile and those with heterogeneous mutation profile. It may originate from that the study population composed of pStage IIA-IIIA NSCLC patients treated with curative surgical resection and adjuvant chemotherapy based on platinum. Further studies are required on the patients with stage IV NSCLC who are treated with target agents.
Spatial heterogeneity, indicating that a biopsy of single site does not represent the whole biology of cancer, may pose significant challenges in the personalized medicine. The cases with large tumor burden and those with treatment failure require profound consideration for the number and location of target biopsy lesion, which guide treatment planning and improve the clinical outcome of the patients. It needs to be determined which approaches to genetic heterogeneity lead substantial improvement in clinical outcome in an economical way.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
EYK participated in the design of the study, reviewed clinical data, carried out the IRB process, and the manuscript draft. ENC participated collection of clinical specimens and clinical data. HSP reviewed clinical specimens and marked cancer enriched area. AK, JYH, and SL extracted DNA from clinical samples and generated library, and sequenced with the Ion Torrent PGM sequencer. JPY carried out quality control, interpretation, and analysis of data. SYH participated in the study design, statistical assistance, and interpretation of the data. YSC conceived the study, and participated in its design and coordination and helped to draft the manuscripts. All authors read and approved the final manuscript.