Mutation patterns in small cell and non-small cell lung cancer patients suggest a different level of heterogeneity between primary and metastatic tumors
Mutation patterns in small cell and non-small cell lung cancer patients suggest a different level of heterogeneity between primary and metastatic tumors
Ali Saber 2
T.Jeroen N.Hilterman 1 2
Klaas Kok 0 2
M.Martijn Terpstra 0 2
Kim de Lange 0 2
Wim Timens 2
Harry J.M.Groen 1 2
Anke van den Berg Emaila:. 2
0 Department of Genetics, University of Groningen, University Medical Centre Groningen , 9700 RB Groningen , The Netherlands
1 yD, epartment of Pulmonary Diseases
2 Department of Pathology and Medical Biolog
Several studies have shown heterogeneity in lung cancer, with parallel existence of multiple subclones characterized by their own specific mutational landscape. The extent to which minor clones become dominant in distinct metastasis is not clear. The aim of our study was to gain insight in the evolution pattern of lung cancer by investigating genomic heterogeneity between primary tumor and its distant metastases. Whole exome sequencing (WES) was performed on 24 tumor and five normal samples of two small cell lung carcinoma (SCLC) and three non-SCLC (NSCLC) patients. Validation of somatic variants in these 24 and screening of 33 additional samples was done by single primer enrichment technology. For each of the three NSCLC patients, about half of the mutations were shared between all tumor samples, whereas for SCLC patients, this percentage was around 95. Independent validation of the non-ubiquitous mutations confirmed the WES data for the vast majority of the variants. Phylogenetic trees indicated more distance between the tumor samples of the NSCLC patients as compared to the SCLC patients. Analysis of 30 independent DNA samples of 16 biopsies used for WES revealed a low degree of intra-tumor heterogeneity of the selected sets of mutations. In the primary tumors of all five patients, variable percentages (19-67%) of the seemingly metastases-specific mutations were present albeit at low read frequencies. Patients with advanced NSCLC have a high percentage of non-ubiquitous mutations indicative of branched evolution. In contrast, the low degree of heterogeneity in SCLC suggests a parallel and linear model of evolution.
Lung cancer is the main cause of cancer-related deaths in the cell evolution; i.e. linear versus non-linear or branched ev-olu
world with approximately 1.4 million patients per year1)(. Two tion (
). In the linear model, tumor cells acquire sequential
main histological subtypes are recognized with approximately genomic changes (mutations or copy number variations (CNVs))
15% of the patients presenting with small cell lung carcinoma over time and the clone that contains most favorable genomic
(SCLC) and 85% with non-small cell lung carcinoma (NSCLC). changes will become dominant 6(). The branching evolution
The latter group is further subdivided into three subgroups, i.e. model assumes development of multiple parallel tumor cell
large cell carcinoma (LCC), squamous cell carcinoma and aden-o clones that all acquire specific genomic aberrations over time.
carcinoma (AC) (
). These tumor clones exist independently within the tumor mass
Recent developments in next generation sequencing te-ch (
). A branched evolution has been shown in pancreatic8,(
niques in combination with improved bioinformatics tools clear cell renal cance1r0(
), breast cancer1(
) and lung cancer
have greatly advanced the field of tumor genetics. These de-vel (
). A branched evolution results in extensive subclonal di-ver
opments have resulted in a better understanding of tumor sity and intra-tumor heterogeneity. In contrast, a linear evolution
Abbreviations diagnostic procedure using mucin stains and immunohistochemistry for
p40/p63, TTF1 and CK5-6 for NSCLC and with CD56, synaptophysin and
AC adenocarcinoma chromogranin for SCLC.
LCC large cell carcinoma To exclude the presence of a minor population of tumor cells in frozen
MRF mutant read frequency normal tissue samples of the SCLC patients, immunostaining was pe-r
NSCLC non-small cell lung carcinoma formed using anti-cytokeratin (clones AE1/AE3, Dako, Glostrup, Denmark)
SCLC small cell lung carcinoma and anti-CD56 (Becton Dickinson, NJ) antibodies using standard protocols.
SPET single primer enrichment technology Anti-Cytokeratin was added as an extra precaution to identify atypical
WES whole exome sequencing cells in particular of use in the brain tissue as normal control. Briefµlmy, 3
frozen sections were fixed in acetone for 10 min. Next, slides were in-cu
bated with primary antibody at room temperature (RT) for 1 h (dilutions
results in a more homogeneous tumor mass. Linear evolution were 1:100 and 1:50 for Cytokeratin and CD56, respectively). After blocking
in combination with an incomplete clonal sweep, i.e. failure of of endogenous peroxidase activity, secondary and tertiary antibodies were
a subclone to completely outcompete its ancestral clones, may incubated at RT for 30 min and slides were stained with AEC
(3-aminoresult in a lower degree of intra-tumor heterogenei1t4y,1(
). 9-ethylcarbazole) and hematoxylin.
Two models have been suggested to contribute to
intratumor heterogeneity; genetic and non-genetic16(). In the DNA isolation and WES
genetic model, a monoclonal tumor population can become For the WES procedure, we aimed to isolate DNA from a tissue sample
heterogeneous due to genomic instability. In the non-genetic with at least 80% tumor cells. For blocks with more than 80% tumor
model, factors like availability of nutrition, alteration of oxygencontent, we used the complete tissue section, otherwise we performed
level, interaction of tumor cells with neighboring cells, presence macro-dissection or laser microdissection (LMD6000, Leica, Wetzlar,
of cancer stem cells and location within the primary tumor can Germany). For validation of non-ubiquitous mutations, we used both total
lead to preferential outgrowth of some cell populations res-ult tissue sections and laser microdissected tissue for DNA isolation using
standard laboratory procedures. WES was carried out using standardized
ing in intra-tumor heterogeneity16(). protocols of the UMCG genome facility as described previousl1y7(). Briefly,
The observed evolution patterns might be affected by expe-ri 0.3–3 μg genomic DNA was randomly fragmented by ultrasonic nebu-li
mental factors such as sampling bias, and by biological factors zation (K7025-05, Life Technologies, Paisly, UK). Library preparation was
such as exposure to carcinogens, drug treatment and tumor done using Agilent library prep kits (Agilent technologies, Santa Clara,
). So far, most studies have focused on CA) for samples with high input DNA (>500 ng) and with Mondrian library
analysis of single biopsies of the tumor mass, which could result prep kits (NuGEN Technologies, San Carlos, CA) for samples with low DNA
in underestimation of the actual degree of tumor heterogeneity. input (<500 ng). For both procedures, fragments of ~300 bp were isolated
The aim of this study was to gain insight in the evolution using the PerkinElmer labchip XT gel system. Enrichment of exons was
pattern of lung cancer by studying genomic heterogeneity in the for both approaches performed using the Agilent SureSelect All exon
primary tumor and its metastases by whole exome sequencing V5 kit (Agilent technologies, Santa Clara, CA) on equimolar pools of PCR
products of four independent samples. Purified samples were subjected
(WES) of samples derived from multiple tumor locations in five to paired-end sequencing on the HiSeq2500. Image Files were processed
lung cancer patients. WES data were validated using targeted using standard Illumina® base calling software and subsequently de-m-ul
re-sequencing of non-ubiquitous somatic mutations. In add-i tiplexed (Illumina Inc., San Diego, CA).
tion, we used the validation data to identify presence of minor Reads were aligned to the human 1000 genomes reference based on
metastasis-resembling clones in the primary tumor samples. the GRCh37 build using BWA 5.9rc1(
). Picard tools were used for format
conversion and marking duplicate reads. Genome Analysis Toolkit (GATK1)
was used for indel realignment and base score quality recalibration (BSQR)
Materials and methods (
) by Molgenis Compute 4 (
). After using custom scripts in the VCF
tools library2(1) for the VCF files, variant calling was done using the GATK
Samples unified genotyper. Annotation of the variants was performed using snpeff/
Multiple tumor samples of five deceased lung cancer patients undergoing snpsift 3.5 (22) with the ensembl release 74 gene annotationhst(tp://www.
autopsy were included in this study. One patient presented with LCC, one ensembl.org/index.html), dbNSFP2.3 (23), and GATK with annotations
with squamous cell carcinoma, one with AC and two with SCLC. These five from the Database of Single Nucleotide Polymorphisms (dbSNP) Bethesda
patients were selected from a larger group of patients based on having stage (MD): National Center for Biotechnology Information, National Library
IV lung cancer with metastases in mediastinal lymph nodes and in different of Medicine (dbSNP Build ID: 137) and CosmicCodingMuts_v62 2(
visceral organs, in combination with having frozen tissue samples of normal addition to the snpeff and snpsift, we also annotated the variants using
tissue, primary tumor location and at least two metastatic locations. Tissue ANNOVAR (25). A combination of three different filtering steps has been
samples of normal, primary lung tumor and multiple metastatic locations applied on the list of mutations. We removed variants tha1t) (were
prewere all obtained at autopsy with exception of patient #1 for whom we also sent in the Caucasian based 1000-Genome 2(
) with allele frequency larger
included a primary tumor sample obtained at diagnosis. Autopsy was done than 0.2% (
), map in non-coding regions and (
) were synonymous.
within 6 h after death. All tissue samples were snap frozen. In total, 45-fro Personal variants in the tumor were removed based on a z-score ≥
zen tissue blocks were obtained, of which 29 (5 normal and 24 tumor blocks) −3 in normal samples using the formula:Z = (allele frequency − 0.5)/SD×
were subjected to WES. The additional tumor blockns =( 16) were used in (SQRT of total reads). The remaining somatic variants were included in the
the validation and heterogeneity analyses (Supplementary Tables 1 and 2, downstream analysis.
available atCarcinogenesis Online). Each patient gave informed consent for
this study. All procedures and protocols were performed according to the Correction for normal cell admixture
guidelines of the Medical Ethics Committee.
To correct for admixture of normal cells, we deduced the approximate p-er
centage of admixture of normal cells based on the mutant read frequency
Immunohistochemistry (MRF) of the true mutations. We based these calculations on the mean
Hematoxylin and eosin (HE) sections were made for each sample to co-n MRF of the somatic mutations in the 25–75% interquartile range assuming
firm diagnosis and to determine the histological subtype based on the that these represent heterozygous mutations for which the MRF should
WHO classification. In addition, HE sections were made for each normal be 0.5 in a sample with 100% tumor cells. The lower and higher quartiles
sample, which was derived from normal lung for patients #1 and #3, no-r were excluded as they might be derived from gene loci that have copy
mal liver for patient #2, and normal brain for patients #4 and #5. This number gain or loss or represent homozygous mutations or minor clones.
classification was confirmed in diagnostic biopsies as part of the routine The mean value multiplied by 2 was considered as an estimation of the
actual percentage of tumor cells. For example, a mean MRF of the 25–75% Heterogeneity analyses
interquartile range of 0.40 would correlate with a tumor cell percentageIntra-tumor heterogeneity was assessed for tumors for which additional
of 80%. For each sample, we recalculated the mutant allele frequency by blocks or subsequent tissue sections were available for DNA isolation and
reducing the total number of reads with the normal cell admixture-per targeted sequencing. We included all variant positions with a total read
centage while keeping the original number of mutant reads. count of ≥30 in all samples of the patient. We calculated the percentage
of SPET validated mutations in the additional samples of the same tumor
High- and low-confidence mutations using the original WES sample as a reference.
Criteria to call mutations as high- or low-confidence were based on
the mutant allele count and frequency after correction for normal cellIdentification of metastases-specific mutations in the primary
admixture. Mutations with a mutant read count ≥5 and a MRF ≥20% were tumor samples
denoted as high-confidence mutations. Mutations with mutant read Metastases-specific mutations were called in the primary tumor samples
count of 1–4 or a MRF <20% were referred to as low-confidence. In the based on presence of at least two mutant reads in the primary tumor sa-m
absence of mutant reads and a total number of reads <30, the presumed ples. Presence of such mutations in the primary tumor indicated presence
mutation was called inconclusive (high chance of being absent). If the of minor metastasis resembling clones.
mutant read count was zero and the total read was ≥30, the presumed
mutation was called absent. Phylogenetic tree
Overlap with lung cancer mutated genes from
COSMIC and other sources
Hierarchical cluster analysis was done using R 3.3.1 and the APE 3.5 pa-ck
age using the binary distance matrix containing presence or absence of
somatic mutations based on the combined results of the WES and SPET
analysis to create phylogenetic tree3s2(,33).
The list of genes mutated in the primary tumorsn( = 1074) was compared
with the top 20 most commonly mutated genes in lung cancer according
to the COSMIC database v.73 and to the list of most commonly mutated Intra- and inter-tumor heterogeneity definition
genes (n = 64) of five studies on lung cancer including 110–230 tumor sa-m
ples (27–31). Overlap of genes mutated in these lists and in one of the five
primary lung tumors was identified by using an online Venn diagram tool
Heterogeneity within one specific location has been referred to as
‘intra-tumor heterogeneity’ and heterogeneity between different tumor
locations within a single patient has been referred to as ‘inter-tumor
Validation of somatic mutations and heterogeneity analysis using single primer enrichment technology
To confirm inter-tumor heterogeneity, all non-ubiquitous single nu-cle Patient characteristics
otide variants were validated using the Ovation© target enrichment
procedure that is based on the single primer enrichment technology All patients presented as stage IV lung cancer. Four of the five
(SPET) (NuGEN Technologies, San Carlos, CA). We included five ubiq-ui lung cancer patients (#1, #2, #4 and #5) were smokers and had
tous mutations for each patient as positive controls. This resulted in a chronic obstructive pulmonary disease (COPD). Patient #3 was
selection of 467 SNVs. In total, we re-analyzed 228 mutations of tumor the only non-smoker and had advanced adenocarcinoma of the
samples of patient #1, 152 of patient #2, 47 of patient #3, 23 of patient #4 lung (Supplementary Table 1, available aCtarcinogenesis Online).
and 17 of patient #5. In total, 57 DNA tumor samples originating from Patient #1, diagnosed with LCC, had the longest interval between
40 tumor blocks were analyzed for validation and heterogeneity ass-ess diagnosis and death, i.e. more than 6 years, whereas patient #5,
ment. For some of the tumor blocks, we did multiple independent DNA diagnosed with SCLC had the shortest interval from diagnosis
isolations. It should be noted that all DNA samples were analyzed for all
selected variants. to death, i.e. only 6 days F(igure 1). Treatments were cisplatin
Ovation© target enrichment system landing-probes were designed by and gemcitabine in the patient with squamous cell carcinoma
NuGEN at a distance of 10–20 nucleotides downstream and upstream of (#2), cisplatin and pemetrexed for AC patient (#3), cisplatin and
each single nucleotide variant to be validated. In a few cases, two variants etoposide for both patients with SCLC (#4 and #5) and surgery for
were too close together to be able to design separate landing probes. For the poorly differentiated LCC (#1). Multiple samples of the p-ri
459 of the variants, two landing probes could be designed. Five variants mary tumor, different metastatic sites and normal tissue were
were targeted by only one landing probe. For three variants, no landing obtained from each patient at autopsy (Supplementary Tables 1
probes could be designed. Validation could be performed for 464 somatic and 2, available atCarcinogenesis Online). In addition, we obtained
mutations. Target enrichment and library preparation was done acco-rd a sample from the primary tumor at diagnosis from patient #1.
ing to the manufacturer’s protocol (NuGEN Technologies, San Carlos, CA).
Briefly, genomic DNA was fragmented, end-repaired and ligated with WES and Mutation detection
forward barcoded adaptors, which was followed by a beads purification
step. The barcoded adaptor contained both a 8-nt sample-specific barcode WES resulted in a mean coverage ranging from 4×3 to 89× and
and a 6-nt molecular barcode. The latter was used to identify and remove a >20× coverage ranging from 83 to 96% of the target region
duplicate reads. Samples were combined for multiplex target enrichment. (Supplementary Table 3, available aCtarcinogenesis Online). Based
The reverse adaptor was annealed to the target regions and extended. on a z-score of >−3, 31–34% of the variants were considered as
Library amplification step was done using 22–25 PCR cycles depending on personal variants in patients #1, #2, #4 and #5. In patient #3, 74%
the quality of DNA pools as determined by qPCR; followed by beads pu-ri of the identified variants were denoted as personal variants
fication and sequencing. (Supplementary Table 4, available atCarcinogenesis Online). This
much higher percentage of personal variants is probably due to
Validation the non-Caucasian background of this patient. The number of
We did not discriminate between low and high-confidence mutations for somatic mutations ranged from 92 (patient #3) to 462 (patient
the independent validation, as we did not perform laser micro-dissection #1) (Supplementary Tables 3 and 5, available atCarcinogenesis
for all samples and tumor percentage might be relatively low at least in
part of the samples. We considered a mutation as validated when the MRF Online).
was ≥0.05 and the total read count ≥30. We calculated the percentage of The mean of interquartile mutant read frequencies ranged
validated mutations for each of the four categories, i.e. high-confidence, from 49 (SD = 0.05) to 18 (SD = 0.02) indicating an admixture
low-confidence, inconclusive and absent, separately based on the total of normal cells ranging from 2 to 64% (Supplementary Table 3
number of mutations with enough coverage. and Supplementary Figure 1, available atCarcinogenesis Online).
shared between primary tumor and at least one of the metas-ta
sis. The tumor samples of patient #3 with AC had 92 mutations
in 91 genes, of which 49 (53.3%) were present in all samples. No
primary tumor-specific mutations were observed in this patient.
Thirty-one (33.7%) mutations were metastases-specific, while
12 (13%) mutations were shared between primary tumor and at
least one of the metastasis. To explore potential differences in
the impact of ubiquitous versus non-ubiquitous mutations in the
three NSCLC patients, we calculated the percentage of damaging
and non-damaging mutations for all seven annotation tools.This
revealed no differences between the impact of ubiquitous and
non-ubiquitous mutations (Supplementary Figure 2, available at
In SCLC patient #4, we identified 359 mutations in 345
genes (Figure 2A). Ninety-five percent n( = 341) of the somatic
mutations were present in all tumor samples. In addition, we
found four mutations to be specific for the primary tumor. In
the metastasis, we found four mutations to be specific for the
mediastinal lymph node, one mutation specific for liver and one
for the adrenal metastasis. In SCLC patient #5, we identified 271
mutations in 249 genes. Two hundred and fifty nine (95.6%) of
the somatic mutations were present in all tumor samples. No
primary specific mutations were identified in this patient. Four
mutations were metastases-specific and the other eight mu-ta
SPET analysis of all non-ubiquitous and five ubiquitous
mutations per patient resulted in a mean coverage of 97–288
and a 50× coverage ranging from 77 to 95% for the five patients
To allow a reliable comparison of shared and sample-specific (Supplementary Table 4, available aCtarcinogenesis Online). Using
somatic mutations, we corrected the MRF for the estimated this approach, we confirmed presence of ‘high-confidence’
admixture of normal cells. In total, we found 1511 unique mutations in more than 99%. The presumed mutations called
somatic mutations in the five patients. SNPEFF, SIFT, MetaSVM, as ‘absent’ in the WES analysis were indeed absent in more than
Polyphen2, CADD, GERP++ and PROVEAN scores of all mutations 99% of them (Table 1). Almost all positions (99.7%) referred to as
are given in Supplementary Table 5, available aCtarcinogenesis ‘inconclusive’ in the WES study due to lower coverage, turned
Online. out to be absent. Mutations referred to as ‘low-confidence’ in the
Comparison of the genes mutated in at least one of the five WES study were confirmed in only about 30% of the cases, pro-b
primary tumors to the top 20 most commonly mutated genes ably due to the mild criteria (low MRF) used for calling variants
in lung cancer present in the COSMIC database revealed that in this category.
11 out of 20, includingTP53, KRAS, RB1 and ALK, were mutated
in one or more of the five primary tumor samples in this study Mutation pattern after integration with validation
(Supplementary Table 6, available aCtarcinogenesis Online). Most results
of these mutations were ubiquitous. Sixteen out of 64 genes
related to lung cancer according to five large stud2ie7s–3(1)
were mutated in at least one of the five primary tumor samples
(Supplementary Table 7, available aCtarcinogenesis Online).
Inter-tumor heterogeneity based on WES
A complete list of the ubiquitous and non-ubiquitous somatic
mutations detected by WES is given in Supplementary Table 5,
available atCarcinogenesis Online. We found 462 mutations in
430 genes in patient #1 with LCC. Approximately, 50% n( = 237)
of these mutations were ubiquitous, 67 (14.5%) were specific for
the primary tumor and 129 (27.9%) were present in one or more
metastases (metastases-specific mutations). The remaining 29
(6.3%) mutations were shared between the primary and one of Intra-tumor heterogeneity based on validated
the two metastasesF(igure 2A). The tumor samples of patient #2 mutations
with squamous cell carcinoma contained a total of 331 somatic We used targeted re-sequencing to determine intra-tumor h- et
mutations in 300 genes. More than half of these mutations (55%) erogeneity in 30 additional DNA samples obtained from 16 of
were ubiquitous and 23 (7%) of the mutations were specific for the the samples used for WES. More than 75% of the selected
WESprimary tumor. One hundred and three (31%) of the mutations based mutations were present in the additional samples, with
were metastases-specific. The remaining 23 (7%) mutations were no obvious differences between NSCLC and SCLC patients.
The percentage of metastases-specific mutations based on the
combined WES and SPET results, was significantly higher in the
metastasis of the three NSCLC patients as compared to the p-er
centages in the metastases of the two SCLC patientPs <( 0.01)
(Figure 2B). This indicates much more inter-tumor heterogen-e
ity in NSCLC. Consistent with these findings, we showed a s-ig
nificantly higher percentage of ubiquitous mutations (>95%) in
SCLC as compared to NSCLC (around 50%) (P < 0.01) (Figure 2C).
Phylogenetic trees confirmed the larger genetic distance
between primary tumor and different metastases in NSCLC as
compared to SCLC patients (Figure 2D).
These analyses revealed an overall low degree of intra-tumor respectively F(igure 3 and Supplementary Table 9, available at
heterogeneity for the selected sets of mutationFsig(ure 3 and Carcinogenesis Online).
Supplementary Table 8, available aCtarcinogenesis Online).
Identification of metastases-specific mutations in Discussion
the primary tumor samples The aim of this study was to explore tumor heterogeneity by
Twenty-one of the 110 WES-based seemingly metastases- comparing the mutation patterns of tumors obtained from
specific mutations (19.1%) could be detected at low read f-re different locations within a patient. We observed significantly
quencies in one or more of the primary samples obtained at more heterogeneity between the primary tumor and the d-if
time of diagnosis of patient #1 using the targeted re-seque-nc ferent metastatic sites (inter-tumor heterogeneity) in NSCLC
ing approach. In the sample of the primary tumor obtained compared to SCLC patients. Moreover, we detected a proportion
at autopsy of this patient, the mutation pattern was more of the seemingly metastases-specific mutations in the primary
similar to the lymph node metastasis F(igure 3). In NSCLC tumors of all five patients by targeted re-sequencing.
patients #2 and #3 these percentages were 65.7% (65/99) and About 50% of the somatic mutations in the three NSCLC
27.6% (8/29), respectively. In the SCLC patients, 2 out of 8 and patients are non-ubiquitous, which indicates a marked degree
2 out of 3 metastases-specific mutations were present at low of inter-tumor heterogeneity and an ongoing evolution of the
read frequencies in the primary tumors of patients #4 and #5, tumor cells over time. Presence of tumor location-specific
aNumber of unique mutations in each category.
bSubsequent sections for independent DNA isolations with the same conditions as in the WES experiment. LDM was performed, if isolated DNA for WES had been
originated from a LDM sample.
cNo mutation in this category selected for validation. The first number in each sample name indicates a tissue block and the second number after ‘dash’ symbol
shows the subsequent sections for independent DNA isolation.
mutations has been shown previously in pancreatic cancer9)(. analyses revealed mutation patterns highly similar to the o-rigi
The high percentage of non-ubiquitous mutations in NSCLC nal WES sample. This indicates an overall low degree of
intrapatients suggests a branched evolution model and implies tumor heterogeneity for the selected non-ubiquitous mutations.
that the ubiquitous mutations may not be sufficient to induce However, the number of samples was limited, and the samples
development of metastasis. Thus, NSCLC tumor cells need were mostly obtained from the same area as the samples used
time to acquire additional genomic aberrations enabling them for WES. In two previous studies of 73 multiregional primary
to migrate and survive in a new environment. The ubiquitous lung tumor samples of 18 NSCLC patients, variable degrees of
mutations most likely originate from founder clones and might intra-tumor heterogeneity and coexistence of several subclones
be associated with tumor formation. However, overall no -dif within the primary tumors have been reported1(
differences were observed between the predicted impact of the ferences of our intra-tumor heterogeneity results with these
mutations on the protein for the ubiquitous as compared to the two studies might be related to our specific focus on WES-based
non-ubiquitous mutations. non-ubiquitous mutations and on the relative close vicinity of
In SCLC, almost all mutations were ubiquitous indicating a the additional tumor samples compared to the initial WES sa-m
homogenous nature of the tumor mass at different anatomical ple. In our study, the degree of intra-tumor heterogeneity may
locations within a patient resembling a linear evolution model. have been underestimated.
The homogeneous nature may be indicative of the short-time A proportion of the WES-based seemingly
metastases-speinterval between tumor formation and the parallel deve-lop cific mutations could be detected in the primary tumors of the
ment of multiple distant metastases. This finding is consistent NSCLC and SCLC patients albeit as minor clones in the targeted
with the aggressive nature of SCLC. The highly similar mu-ta re-sequencing analysis. This indicates that the metastasis se-ed
tion profile of the primary tumor as compared to the metast-a ing clones were already present in the primary tumor sample.
sis suggests that the dominant clone of the primary tumor has This finding is consistent with the results of Yachideta al. (
already all the characteristics needed for development of the who showed presence of metastasis seeding subclones in pr-i
metastases. Overall, our data suggests a branched model of e-vo mary pancreatic tumors of all seven patients. Even for patient
lution of NSCLC and linear in SCLC, which is also reflected in the #1, with the primary tumor sampled at diagnosis 6-year before
phylogenetic trees. autopsy, we were still able to identify some of the
Analysis of additional samples of both the NSCLC and SCLC specific mutations in the primary tumor. The mutation pattern
patients, i.e. subsequent tissue sections or additional tissue of the primary tumor sample obtained at autopsy more closely
samples, for 16 of the 24 tumor samples included in the WES resembled the metastases than the primary tumor at diagnosis
for the subset of mutations included in the targeted re-seque-nc also be indicative of a relative high mutation rate within the
ing approach. This finding together with the high percentage of tumor cells and thus of an increased chance of developing drug
ubiquitous mutations in the primary tumor at diagnosis iden-ti resistance-inducing mutations. Thus, inclusion of multiple sa-m
fied by WES suggests that a clone within the primary tumor at ples from different geographical regions of the primary tumor
diagnosis caused the relapse and the metastases in this patient. and applying targeted next generation sequencing tests at high
Despite the relative low number of metastases-specific mut-a read depth should be considered to increase the probability to
tions in the two SCLC patients, we were able to confirm presence detect minor therapy resistant genomic aberrations. For sel-ect
of part of them in the primary tumor samples of both patients. ing the most optimal primary treatment one needs to know the
Tumor heterogeneity, as observed in our study, can have presence of trunk mutations that are drivers and druggable.
therapeutic implications for treatment of patients. Driver mu-ta In conclusion, the high number of non-ubiquitous mutations
tions found as major clones in one part of the tumor might be in NSCLC indicates an ongoing and branched evolution with a
present as minor clones or even be completely absent in another propagation of many different subclones. In contrast, the low
part of the tumor 1(
). This indicates that analysis of a single number of non-ubiquitous mutations in SCLC suggests a pa-r
biopsy might not be sufficient for diagnostic purposes. In co-n allel and linear model of evolution. These data suggest that
trast, Zhanget al. (34) concluded that a single biopsy is sufficient characterization of multiple regions of NSCLC tumors might be
for clinical evaluation based on results of a study of 11 localized important to improve treatment planning in a clinical setting.
AC lung patients. These opposing conclusions might be related to
differences in read depth that might or might not allow detection Supplementary material
of minor clones. Overall, there is ample evidence for the presence
of metastatic ancestor or resistant subclones within the primary Supplementary data are available aCtarcinogenesis online.
). These subclones may become dominant dur- Conflict of Interest Statement: None declared.
ing further tumor development and selective pressure due to
A high degree of tumor heterogeneity will increase the 1. Ju, Y.S. et al. (2012) A transforming KIF5B and RET gene fusion in lung
chance of resistant clones being already present at start of t-reat adenocarcinoma revealed from whole-genome and transcriptome
ment. On the other hand, a high degree of heterogeneity may sequencing. Genome Res., 22, 436–445.
2. Travis , W.D. et al.; WHO Panel . ( 2015 ) The 2015 World Health Org-ani 22 . Cingolani , P. et al. ( 2012 ) A program for annotating and predicting zation classification of lung tumors: Impact of genetic, clinical and the effects of single nucleotide polymorphisms, SnpEff: SNPs in the radiologic advances since the 2004 classification . J. Thorac . Oncol., 10 , genome of Drosophila melanogaster strain w1118; iso -2 ; iso-3. Fly 1243-1260 . (Austin)., 6 , 80 - 92 .
3. Saber , A. et al. ( 2015 ) Genomic aberrations guiding treatment of non- 23. Liu , X. et al. ( 2013 ) dbNSFP v2 . 0: a database of human non-synonymous small cell lung cancer patients . Cancer Treatment Commun., 4 , 23 - 33 . SNVs and their functional predictions and annotations . Hum. Mutat.,
4. Yates , L.R. et al. ( 2012 ) Evolution of the cancer genome . Nat. Rev. Genet ., 34 , E2393 - E2402 . 13 , 795 - 806 . 24 . Forbes , S.A. et al. ( 2015 ) COSMIC: exploring the world's knowledge of
5. Swanton , C. et al. ( 2016 ) Clinical implications of genomic discoveries in somatic mutations in human cancer . Nucleic Acids Res ., 43 ( Database lung cancer . N. Engl. J. Med ., 374 , 1864 - 1873 . issue), D805 - D811 .
6. de Bruin , E.C. et al. ( 2013 ) Intra-tumor heterogeneity: lessons from 25 . Wang , K. et al. ( 2010 ) ANNOVAR: functional annotation of genetic varimicrobial evolution and clinical implications . Genome Med ., 5 , 101. ants from high-throughput sequencing data . Nucleic acids research ,
7. Polyak , K. ( 2008 ) Is breast tumor progression really linear? Clin. Cancer 38 , e164 . Res., 14 , 339 - 341 . 26 . Abecasis , G.R. et al. ( 2012 ) An integrated map of genetic variation from
8. Campbell , P.J. et al. ( 2010 ) The patterns and dynamics of genomic inst-a 1,092 human genomes . Nature , 491 , 56 - 65 . bility in metastatic pancreatic cancer . Nature , 467 , 1109 - 1113 . 27 . Ding , L. et al. ( 2008 ) Somatic mutations affect key pathways in lung
9. Yachida , S. et al. ( 2010 ) Distant metastasis occurs late during the adenocarcinoma . Nature , 455 , 1069 - 1075 . genetic evolution of pancreatic cancer . Nature , 467 , 1114 - 1117 . 28 . Imielinski , M. et al. ( 2012 ) Mapping the hallmarks of lung adenocar-ci
10. Gerlinger , M. et al. ( 2012 ) Intratumor heterogeneity and branched noma with massively parallel sequencing . Cell , 150 , 1107 - 1120 . evolution revealed by multiregion sequencing . N. Engl. J. Med ., 366 , 29 . TCGA. ( 2012 ) Comprehensive genomic characterization of squamous 883-892. cell lung cancers . Nature , 489 , 519 - 25 .
11. Gerlinger , M. et al. ( 2014 ) Genomic architecture and evolution of clear 30 . TCGA. ( 2014 ) Comprehensive molecular profiling of lung adenocar-ci cell renal cell carcinomas defined by multiregion sequencing . Nat. noma. Nature , 511 , 543 - 50 . Genet., 46 , 225 - 233 . 31 . George , J. et al. ( 2015 ) Comprehensive genomic profiles of small cell
12. Nik-Zainal , S. et al. ( 2012 ) Mutational processes molding the genomes lung cancer . Nature , 524 , 47 - 53 . of 21 breast cancers . Cell , 149 , 979 - 93 . 32 . Paradis , E. et al. ( 2004 ) APE: analyses of phylogenetics and evolution in
13. de Bruin , E.C. et al. ( 2014 ) Spatial and temporal diversity in genomic R language . Bioinformatics , 20 , 289 - 290 . instability processes defines lung cancer evolution . Science , 346 , 251 - 256 . 33. R. Core .Team. R: A Language and Environment for Statistical Compu-t
14. Swanton , C. ( 2014 ) Cancer evolution: the final frontier of precision ing . Vienna, Austria: R Foundation for Statistical Computing, 2016 . medicine? Ann. Oncol., 25 , 549 - 551 . 34 . Zhang , J. et al. ( 2014 ) Intratumor heterogeneity in localized lung-ade
15. Burrell , R.A. et al. ( 2013 ) The causes and consequences of genetic h- et nocarcinomas delineated by multiregion sequencing . Science , 346 , erogeneity in cancer evolution. Nature , 501 , 338 - 345 . 256 - 259 .
16. Neelakantan , D. et al. ( 2015 ) Intratumoral heterogeneity: clonal -coop 35 . Ding , L. et al. ( 2010 ) Genome remodelling in a basal-like breast cancer eration in epithelial-to-mesenchymal transition and metastasis. Cell metastasis and xenograft . Nature , 464 , 999 - 1005 . Adh. Migr., 9 , 265 - 276 . 36 . Shah , S.P. et al. ( 2012 ) The clonal and mutational evolution spectrum of
17. Liu , Y. et al. ( 2014 ) The mutational landscape of Hodgkin lymphoma primary triple-negative breast cancers . Nature , 486 , 395 - 399 . cell lines determined by whole-exome sequencing . Leukemia , 28 , 37 . Diaz , L.A. Jr et al. ( 2012 ) The molecular evolution of acquired resistance 2248-2251. to targeted EGFR blockade in colorectal cancers . Nature , 486 , 537 - 540 .
18. Li , H. et al. ( 2010 ) Fast and accurate long-read alignment with Burrows- 38. Misale , S. et al. ( 2012 ) Emergence of KRAS mutations and acquired Wheeler transform . Bioinformatics , 26 , 589 - 595 . resistance to anti-EGFR therapy in colorectal cancer . Nature , 486 ,
19. McKenna , A. et al. ( 2010 ) The Genome Analysis Toolkit: a MapReduce 532-536. framework for analyzing next-generation DNA sequencing data . 39 . Inukai , M. et al. ( 2006 ) Presence of epidermal growth factor receptor Genome Res ., 20 , 1297 - 1303 . gene T790M mutation as a minor clone in non-small cell lung cancer .
20. Boomsma , D.I. et al. ( 2013 ) The genome of the Netherlands: design, and Cancer Res ., 66 , 7854 - 7858 . project goals. Eur. J. Hum. Genet ., 22 , 221 - 227 . 40 . Hashida , S. et al. ( 2014 ) Presence of the minor EGFR T790M mutation
21. Danecek , P. et al. ( 2011 ) The variant call format and VCFtools. Bioin-for is associated with drug-sensitive EGFR mutations in lung adenoca-rci matics , 27 , 2156 - 2158 . noma patients. Oncol. Rep ., 32 , 145 - 152 .