Content uploaded by Paulo De Sousa
Author content
All content in this area was uploaded by Paulo De Sousa on Jan 15, 2018
Content may be subject to copyright.
Content uploaded by Paulo De Sousa
Author content
All content in this area was uploaded by Paulo De Sousa on Jan 15, 2018
Content may be subject to copyright.
The new england
journal of medicine
n engl j med 376;22 nejm.org June 1, 2017
2109
established in 1812
June 1, 2017
vol. 376 no. 22
The authors’ full names, academic de-
grees, and af filiations are listed in the
Appendix. Address reprint requests to
Dr. Swanton at the Translational Cancer
Therapeutics Laboratory, Francis Crick
Institute, 3rd Fl. SW, 1 Midland Rd., Lon-
don NW1 1AT, United Kingdom, or at
charles . swanton@ crick . ac . uk.
* A complete list of investigators in the
Tracking Non–Small-Cell Lung Cancer
Evolution through Therapy (TRACERx)
Consortium is provided in Supplemen-
tary Appendix 1, available at NEJM.org.
Drs. Jamal-Hanjani, Wilson, McGranahan,
Birkbak, and Veeriah and Mr. Watkins con-
tributed equally to this ar ticle.
This arti cle was published on Apr il 26, 2017,
at NEJM.org.
N Engl J Me d 2017;376:2109-21.
DOI: 10.1056/NEJMoa1616288
Copyright © 2017 Massachusetts Medical Society.
BACKGROUND
Among patients with non–small-cell lung cancer (NSCLC), data on intratumor heterogeneity
and cancer genome evolution have been limited to small retrospective cohorts. We wanted
to prospectively investigate intratumor heterogeneity in relation to clinical outcome and to
determine the clonal nature of driver events and evolutionary processes in early-stage NSCLC.
METHODS
In this prospective cohort study, we performed multiregion whole-exome sequencing on
100 early-stage NSCLC tumors that had been resected before systemic therapy. We sequenced
and analyzed 327 tumor regions to define evolutionary histories, obtain a census of clonal
and subclonal events, and assess the relationship between intratumor heterogeneity and
recurrence-free survival.
RESULTS
We observed widespread intratumor heterogeneity for both somatic copy-number altera-
tions and mutations. Driver mutations in EGFR, MET, BRAF, and TP53 were almost always
clonal. However, heterogeneous driver alterations that occurred later in evolution were
found in more than 75% of the tumors and were common in PIK3CA and NF1 and in genes
that are involved in chromatin modification and DNA damage response and repair.
Genome doubling and ongoing dynamic chromosomal instability were associated with
intratumor heterogeneity and resulted in parallel evolution of driver somatic copy-number
alterations, including amplifications in CDK4, FOXA1, and BCL11A. Elevated copy-number
heterogeneity was associated with an increased risk of recurrence or death (hazard ratio,
4.9; P = 4.4×10
−4
), which remained significant in multivariate analysis.
CONCLUSIONS
Intratumor heterogeneity mediated through chromosome instability was associated with
an increased risk of recurrence or death, a f inding that supports the potential value of
chromosome instability as a prognostic predictor. (Funded by Cancer Research UK and
others; TRACERx ClinicalTrials.gov number, NCT01888601.)
abs tr act
Tracking the Evolution of Non–Small-Cell Lung Cancer
M. Jamal-Hanjani, G.A. Wilson, N. McGranahan, N.J. Birkbak, T.B.K. Watkins, S. Veeriah, S. Shafi, D.H. Johnson,
R. Mitter, R. Rosenthal, M. Salm, S. Horswell, M. Escudero, N. Matthews, A. Rowan, T. Chambers, D.A. Moore,
S. Turajlic, H. Xu, S.-M. Lee, M.D. Forster, T. Ahmad, C.T. Hiley, C. Abbosh, M. Falzon, E. Borg, T. Marafioti,
D. Lawrence, M. Hayward, S. Kolvekar, N. Panagiotopoulos, S.M. Janes, R. Thakrar, A. Ahmed, F. Blackhall,
Y. Summers, R. Shah, L. Joseph, A.M. Quinn, P.A. Crosbie, B. Naidu, G. Middleton, G. Langman, S. Trotter,
M. Nicolson, H. Remmen, K. Kerr, M. Chetty, L. Gomersall, D.A. Fennell, A. Nakas, S. Rathinam, G. Anand,
S. Khan, P. Russell, V. Ezhil, B. Ismail, M. Irvin-Sellers, V. Prakash, J.F. Lester, M. Kornaszewska, R. Attanoos,
H. Adams, H. Davies, S. Dentro, P. Taniere, B. O’Sullivan, H.L. Lowe, J.A. Hartley, N. Iles, H. Bell, Y. Ngai,
J.A. Shaw, J. Herrero, Z. Szallasi, R.F. Schwarz, A. Stewart, S.A. Quezada, J. Le Quesne, P. Van Loo, C. Dive,
A. Hackshaw, and C. Swanton, for the TRACERx Consortium*
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2110
The
new england journal
of
medicine
L
ung cancer is the leading cause of
cancer-related death worldwide,
1,2
with non–
small-cell lung cancer (NSCLC) being the
most common type. Large-scale sequencing stud-
ies have revealed the complex genomic landscape
of NSCLC
3-6
and genomic differences between lung
adenocarcinomas and lung squamous-cell carcino-
mas.
7
However, in-depth exploration of NSCLC
intratumor heterogeneity (which provides the fuel
for tumor evolution and drug resistance) and can-
cer genome evolution has been limited to small
retrospective cohorts.
8,9
Therefore, the clinical
signif icance of intratumor heterogeneity and the
potential for clonality of driver events to guide
therapeutic strategies have not yet been def ined.
Tracking Non–Small-Cell Lung Cancer Evolu-
tion through Therapy (TRACERx)
10
is a multi-
center, prospective cohort study, which began
recruitment in April 2014 with funding from
Cancer Research UK. The target enrollment is
842 patients from whom samples will be obtained
for high-depth, multiregion whole-exome se-
quencing of surgically resected NSCLC tumors in
stages IA through IIIA. One primary objective of
TRACERx is to investigate the hypothesis that in-
tratumor heterogeneit y — in terms of mutations
(single or dinucleotide base substitutions or small
insertions and deletions) or somatic copy-number
alterations (reflecting gains or losses of chromo-
some segments) — is associated with clinical
outcome. Here, we report on the f irst 100 patients
who were prospectively recruited in the study.
Methods
Patients and Tumor Sample s
We collected tumor samples from 100 patients
with NSCLC who had not received previous sys-
temic therapy (Fig. 1A; and Fig. S1 in Supplemen-
tar y Appendix 1, available with the full text of
this article at NEJM.org). Identifiers of patients
were reassigned to protect anonymity and were
ordered according to intratumor heterogeneity
and histologic subtype. Eligible patients were at
least 18 years of age and had received a diagno-
sis of NSCLC in stages IA through IIIA (except
Patient CRUK0035, whose tumor was classified
as stage IIIB on the basis of postoperative histo-
logic analysis). The cohort was representative of
a population of patients with NSCLC who were
eligible for curative resection. Histologic data
were confirmed on central review by a lung pa-
thologist. (Details regarding the study design are
provided in the protocol, available at NEJM.org.)
To assess intratumor heterogeneity, samples
of at least two tumor regions that were separated
by a margin of 0.3 cm to 1 cm (depending on the
size of the tumor) had to be available for study.
None of the tumors carried a translocation in
ALK, ROS1, or RET on the basis of sequencing.
This finding was confirmed for ALK and ROS1
with the use of immunohistochemical testing.
All the patients provided written informed con-
sent. The clinical characteristics of the patients
and the study criteria are provided in Tables S1
and S2 and in the Experimental Procedures sec-
tion in Supplementar y Appendix 1.
Multiregion Whole-Ex ome Sequencing
We used the Illumina HiSeq to perform whole-
exome sequencing on multiple regions collected
from each tumor. We sequenced 327 tumor re-
gions (323 primary tumor regions and 4 lymph-
node metastases) and 100 matched germline sam-
ples derived from whole blood (median number,
3 regions per tumor; range, 2 to 8), to a median
depth of 426× (Table S3 in Supplementary Ap-
pendix 1). Orthogonal validation was performed
(Table S4 and Fig. S2 in Supplementary Appen-
dix 1). All sequencing data have been deposited
in the European Genome–Phenome Archive under
accession number EGAS00001002247.
Results
Intratumor Heterogeneity in NSCLC
Genetic diversity within tumors can act as a
substrate for natural selection and tumor evolu-
Figure 1 (facing page). Overview of the Demographic
and Clinical Characteristics of the Patients
in the TRACERx Study.
Panel A shows the demographic and clinical character-
istics of the 100 patients in the study, including diag-
nosis, tumor stage, and smoking status. Panel B shows
how multiregion sequencing was performed on surgi-
cally resec ted tumors to analyze somatic mutations
and copy-number alterations, which facilitated the as-
sessment of intratumor heterogeneity and phylogenetic
reconstruction. Stars on the schematic chromosomes
indicate mutations, where yellow represents clonal pre-
genome doubling mutations, pink represents clonal
postgenome doubling mutations, and red represents
subclonal mutations. Panel C shows the key clinical
questions that were addressed in the study.
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2111
Tracking the Evolution of Non–Small-Cell Lung Cancer
ATRACERx 100 Cohort
BMultiregion Intratumor Heterogeneity Analysis
CClinical Questions
R2
R2
R2
R2
R1
R2
R3
R4
Multiregion Sampling
Multiregion Mutation and
Copy-Number Analysis Clonal Hierarchy and Phylogeny
Surgery with
Curative Intent
Intratumor Heterogeneity and Survival Causes of Intratumor Heterogeneity Census of Clonal and Subclonal Drivers
R1 R2 R3 R4
Genome
doubling
Subclonal
mutations
Late clonal
mutations
Early clonal
mutations
Genome doublingMutational heterogeneity and survival Clonal status of targetable alterations
Chromosomal instability
Mutational processes
GCGATCACGAC
CGCTAGTGCTG
GCGATTACGAC
CGCTAATGCTG
R1 R2
R3 R4
Copy-number heterogeneity and survival
Time
% Alive
Time
% Alive
Never smoked (N=12) Former smoker (N=48) Current or recent smoker (N=40) 62 Men, 38 Women
1B
3A
2B
Lung Adenocarcinoma (N=61)
Stage 1A (N=26) Stage 1B (N=36) Stage 2A (N=13) Stage 2B (N=11) Stage 3A (N=13) Stage 3B (N=1)
Other (N=7)
Lung Squamous-Cell Carcinoma (N=32)
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2112
The
new england journal
of
medicine
tion. We performed multiregion whole-exome
sequencing on 100 TRACERx tumors and classi-
fied somatic mutations, which were def ined as
coding and noncoding single-nucleotide variants,
and copy-number alterations, which were measured
as a percentage of the genome affected by such
alterations, as clonal (present in all cancer cells)
or subclonal (present in a subset of cancer
ce lls) (Fig. 1).
We observed extensive intratumor heteroge-
neity, with a median of 30% (range, 0.5 to 93) of
somatic mutations identified as subclonal and a
median of 48% (range, 0.3 to 88) of copy-number
alterations as subclonal (Fig. 2A, and Fig. S3 in
Supplementary Appendix 1). This finding sug-
gests that genomic-instability processes at the
mutational and chromosomal level are ongoing
during tumor development. Considerable varia-
tion in intratumor heterogeneity among tumors
was also observed, with the number of subclonal
mutations ranging from 2 to 2310 and the per-
centage of the genome affected by subclonal copy-
number alterations ranging from 0.06 to 81%
(Fig. 2A). Without multiregion whole-exome se-
quencing, 76% of subclonal mutations could have
appeared to be clonal, which suggests the selec-
tion of subclones within individual tumor regions
(Fig. S4 in Supplementary Appendix 1). Signif i-
cantly more mutations were identified with the
use of multiregion whole-exome sequencing than
with single-sample analysis (median number, 517
vs. 398; P = 0.009) or with the use of single NSCLC
samples obtained from the Cancer Genome Atlas
(median number, 207; P<0.001) (Fig. S5 in Supple-
mentary Appendix 1). The Cancer Genome Atlas
research network (http://cancergenome . nih . gov)
was retrieved through dbGaP authorization acces-
sion number phs000178.v9.p8.
Squamous-cell carcinomas carried significant-
ly more clonal mutations than did adenocarcino-
mas (P = 0.003) (Fig. S6 in Supplementary Ap-
pendix 1). This finding potentially ref lects
differences in smoking history, with a median of
32 pack-years for adenocarcinomas and 41 pack-
years for squamous-cell carcinomas (P = 0.047)
(Fig. S7 in Supplementary Appendix 1). There were
no significant differences between squamous-cell
carcinomas and adenocarcinomas in the number
or proportion of subclonal mutations (P = 0.72)
(Fig. S6 in Supplementar y Appendix 1) or within
specif ic adenocarcinoma histopathological sub-
types (Fig. S8 in Supplementary Appendix 1). In
squamous-cell carcinomas, no significant rela-
tionship was observed between intratumor hetero-
geneity and clinical variables (Table S5 in Sup-
plementary Appendix 1).
In adenocarcinomas, tumor stage positively
correlated with the proportion of subclonal copy-
number alterations, and Ki67 staining positively
correlated with the burden of both clonal and
subclonal mutations, as well as with the propor-
tion of subclonal copy-number alterations (Table
S5 in Supplementary Appendix 1). Furthermore,
in adenocarcinomas, a signif icantly higher clonal
and subclonal mutational burden was observed
in smokers than in patients who had never smoked
(Fig. S9 in Supplementar y Appendix 1).
There was no signif icant association between
the proportion of subclonal mutations (median in
the cohort, 30%) and relapse-free survival (Fig.
2B). However, in this preliminary analysis, patients
who had tumors with a high proportion of sub-
clonal copy-number alterations (≥48%, the co-
hort median) were at higher risk for recurrence
or death than those with a low proportion (haz-
ard ratio, 4.9; 95% confidence interval [CI], 1.8 to
13.1; P = 4.4×10
−4
) (Fig. 2C). The median time
until recurrence or death was 24.4 months in the
higher risk group of patients compared with a
median that was not reached in the lower risk
group. This finding remained significant in a
multivariate analysis after adjustment for age,
pack-years of smoking, histologic subtype, adju-
Figure 2 (facing page). Genomic Heterogeneity
of Tumors Obtained from Patients with Non–Small-
Cell Lung Cancer (NSCLC).
Panel A shows the number of coding and noncoding
mutations that were detec ted in each tumor region in
the study, according to tumor stage, smoking history,
outcome of recurrence or death, and number of regions
affected. The percentages of somatic mutations and
copy-number alterations that were found to be clonal
or subclonal in each tumor are shown below the num-
ber of mutations. The percentages of study patients
who were disease -free over a 30-month period are
shown according to whether the patients had a high
proportion (above the median) or a low proportion
(below the median) of subclonal mutations (Panel B)
or of subclonal copy-number alterations (Panel C).
There was no signif icant association between the pro-
portion of subclonal mutations and relapse-free sur-
vival (P = 0.70), but patients who had tumors with a
high proportion of subclonal copy-number alterations
were at significantly higher risk for recurrence or death
than those with a low proportion (P = 4.4×10
−4
).
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2113
Tracking the Evolution of Non–Small-Cell Lung Cancer
vant therapy, and tumor stage (hazard ratio,
3.70; 95% CI, 1.29 to 10.65; P = 0.01) (Table S6
in Supplementary Appendix 1). A static measure
of chromosome disruption (describing the mean
proportion of the genome that was aberrant
across tumor regions) was not associated with
No. of Coding and Noncoding
Mutations per Tumor
4000
2000
3000
1000
0
Copy Number,
Percentage
Subclonal
Tumor Stage
Pack-Years
Recurrence or Death
No. of Regions
100
60
80
40
20
0
Mutation,
Percentage
Subclonal
100
60
80
40
20
0
AIntratumor Heterogeneity
Subclonal
Clonal
No Yes
Disease-free Survival (%)
100
60
80
40
20
0
05 10 15 20 25 30
Months to Death or Recurrence
BDisease-free Survival According to Percentage of Subclonal Mutations
Hazard ratio, 0.86 (95% CI, 0.40 –1.85)
P=0.70
No. at Risk
Low
High
49
51
40
49
36
43
31
35
21
21
7
4
0
0
Low
High
Low
High
Disease-free Survival (%)
100
60
80
40
20
0
05 10 15 20 25 30
Months to Death or Recurrence
CDisease-free Survival According to Percentage of Subclonal
Copy-Number Alterations
Hazard ratio, 4.9 (95% CI, 1.8 –13.1)
P=4.4×10 − 4
No. at Risk
Low
High
47
45
42
40
40
32
36
24
19
18
5
4
0
0
13001a 1b 2 3 4 5 6 7 8
Adenocarcinoma Squamous-Cell Carcinoma Other
Patients 1– 61 Patients 62– 93 Patients
94– 100
Tumor Stage Pack-Years Recurrence or Death No. of Regions
2a 2b 3a 3b
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2114
The
new england journal
of
medicine
survival, which suggests that the rate of ongoing
dynamic chromosomal instability, rather than the
state of the genome, is prognostic (Fig. S10 in
Supplementary Appendix 1).
Evolutionary Histories and Tumor Clonal
Architecture in NSCLC
The number or proportion of subclonal muta-
tions does not fully capture the extent of intra-
tumor heterogeneit y, since these measures do not
ref lect the number or prevalence of genetically
distinct subclones that evolve in space and time.
To elucidate subclones within regions and map
the evolutionary history of each tumor, we clus-
tered mutations according to their cellular prev-
alence. Each cluster represents a node on the
phylogenetic tree of the tumor and a subclone
that is present in the tumor population or has
existed during its evolutionary history (Table S7,
Figs. S11 and S12, and the Experimental Proce-
dures section in Supplementary Appendix 1).
We identified 525 mutation clusters, with a
median of 5 per tumor (range, 2 to 15). Most
tumor regions (86%) were found to carry sub-
clones from only a single branch of the phyloge-
netic tree, which emphasizes the limitations of
a single diagnostic biopsy sample in accurately
capturing the true extent of intratumor hetero-
geneity. Without the use of multiregion whole-
exome sequencing, 65% of branched subclone
clusters could have erroneously appeared to be
clonal.
Causes of Intratumor Heterogeneity in NSCLC
Mutational Processes
Understanding how mutational processes shape
tumor evolution may inform strategies to limit
tumor adaptation in the clinical setting.
11
Using
published mutational signatures,
12
we analyzed
clonal and subclonal mutations to determine
which mutational processes contributed to intra-
tumor heterogeneit y.
The number of early mutations (accumulated
before genome doubling or copy-number change)
signif icantly correlated with the burden of muta-
tions associated with smoking (mutational sig-
nature 4), with Spearman’s rank correlations of
0.90 (P<1.1×10
−16
) for adenocarcinomas and of
0.84 (P = 3.9×10
−9
) for squamous-cell carcinomas.
This finding was consistent with the identif ica-
tion of mutations induced by tobacco carcino-
gens as being a key influence on trunk length
(i.e., the number of mutations found in the most
recent common ancestor of all cancer cells) and
was ref lected in the significant correlation be-
tween pack-years and truncal signature 4 muta-
tions in adenocarcinomas (Spearman’s rank cor-
relation, 0.63; P = 5.3×10
−8
). In samples obtained
from 7 of 12 patients with adenocarcinomas who
were long-term former smokers (with >20 years
since last tobacco exposure), a smoking signa-
ture could be detected in late clonal mutations
(>30% with signature 4). This finding was sug-
gestive of a long period of tumor latency in the
evolution of lung adenocarcinomas before clini-
cal presentat ion.
In squamous-cell carcinomas, no signif icant
correlation was observed between pack-years and
smoking-related signature 4 (Spearman’s rank
correlation, 0.10; P = 0.57), and the timing of
genome doubling (ratio of the number of early
mutations to the number of late mutations) was
signif icantly later than in adenocarcinomas (Fig.
S13 in Supplementary Appendix 1). Intriguingly,
Patient CRUK0093, who had squamous-cell car-
cinoma, had a large burden of clonal signature
4 mutations (>1000) despite having been identi-
fied as a lifelong nonsmoker. This patient’s oc-
cupational history indicated exposure to chemicals
that included arsenic, benzene, bisphenol, and
polybrominated diphenyl ethers and coal tar,
which may mimic the mutagenic effects of to-
bacco exposure.
There were signif icant correlations between
the subclonal mutation burden and the number
of subclonal mutations that were classified as
clocklike signatures 1A (spontaneous deamina-
tion of methylated cytosines) and 5 (of unknown
cause).
13
The number of subclonal mutations was
also significantly correlated with signatures 2 and
13 (induced by APOBEC, a family of cytidine
deaminase enzymes involved in messenger RNA
editing) but not with signature 4 (smoking)
12
(Fig.
S14 in Supplementary Appendix 1). (APOBEC cy-
tidine deaminases, which are usually involved in
innate immunity and RNA editing, have been
found to be enriched in several tumor t ypes and
act as an important source of mutagenesis.
14
)
Tumors with the largest subclonal mutation bur-
den had extensive APOBEC-mediated mutagene-
sis (e.g., those obtained from Patients CRUK0001,
CRUK0006, CRUK0020, and CRUK0063), and
spatial heterogeneity in APOBEC mutations was
observed in 15 tumors (Figs. S11 and S14 in
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2115
Tracking the Evolution of Non–Small-Cell Lung Cancer
Supplementary Appendix 1). Tumors obtained
from 19 patients had subclonal driver mutations
that could be attributed to APOBEC activity,
which illustrates how APOBEC mutagenesis may
frequently induce a subclonal driver event that
may contribute to subclonal expansions.
Chromosomal Instability and Genome Doubling
Given the association between intratumor hetero-
geneity characterized by copy-number alterations
and shorter relapse-free survival, we further ex-
plored the dynamics of chromosomal alterations
in different tumor regions and the extent to
which chromosomal instability may drive intra-
tumor heterogeneity. By leveraging germline
heterozygous single-nucleotide polymorphisms
in tumors by means of multiregion whole-exome
sequencing, it is possible to determine whether
the same or distinct parental alleles are gained or
lost in distinct subclones on different branches
of the phylogenetic tree of a tumor. Specifically,
if the maternal allele is gained or lost in a sub-
clone in one region, yet the paternal allele is
gained or lost in a different subclone in another
region, it will result in a mirrored subclonal al-
lelic imbalance profile (Fig. 3A and 3B). Such an
imbalance, which indicates additional ongoing
chromosomal instability, may also reflect parallel
Figure 3. Drivers of Intratumor Heterogeneity.
Panel A shows an example of mirrored subclonal allelic imbalance. This occurs when the maternal allele is gained or lost in a subclone in
one region and the paternal allele is gained or lost in a different subclone in another region. Such imbalance indicates additional ongoing
chromosomal instability and can be inferred through multiregion whole -exome sequencing by using the frequencies at which heterozy-
gous germline single-nucleotide polymorphisms (SNPs) (termed B-allele frequency [BAF]) are detected. The BAF of heterozygous SNPs
is plotted in the same color as their parental chromosome of origin. Panel B shows the BAF profile across the genome of a tumor sample
obtained from Patient CRUK0062. Areas of BAF in regions (including tumor regions R1 through R7 and a germline [GL] reference region)
that have mirrored subclonal allelic imbalance are highlighted in blue or orange. Events that showed mirrored subclonal allelic imbalance
were identified in more than 40% of the genome. Panel C shows phylogenetic trees that indicate parallel evolution of driver amplifica-
tions detec ted through the observation of mirrored subclonal allelic imbalance (arrows). Subclones that are colored blue carr y a cancer
driver event, and those that are colored gray carry no driver event; black outlining of the circles indicates that the subclone appears to be
clonal in at least one tumor region.
AMirrored Subclonal Allelic Imbalance BBAF Profile in a Single Tumor Sample
CPhylogenetic Trees Indicating Parallel Evolution of Driver Amplifications
12345678910 11 12 13 14 15161718
19 22
21
20
Chromosomes
R2
R3
R1
R4
R5
R6
R7 0.7
0.3
0.7
0.3
0.7
0.3
0.7
0.3
0.7
0.3
0.7
0.3
0.7
0.3
0.7
0.3
GL
0.7
0.3
R2
1 maternal
2 paternal
CRUK0062 Region BAF
Germline
BAF
Maternal chromosome
Paternal chromosome
1
1
0.7
0.3
0.7
R1
2 maternal
1 paternal
0.3
CRUK0012
MUC1
amp
MUC1
amp
CRUK0083
CCNB1IP1
CHD8
NKX2-1
FOXA1
amp
CCNB1IP1
CHD8
NKX2-1
FOXA1
amp
CRUK0072
BCL11A
REL
XPO1
amp
BCL11A
REL
XPO1
amp
CRUK0009
RHOH
PHOX2B
amp
RHOH
PHOX2B
amp
CRUK0001
CDK4
LRIG3
amp
CDK4
LRIG3
amp
CDK4
LRIG3
amp
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2116
The
new england journal
of
medicine
evolution involving multiple distinct events con-
verging on the same genes in different subclones
(Fig. S15 in Supplementary Appendix 1). This
phenomenon was obser ved in 62% of 92 tumors
with copy-number data on multiregion whole-
exome sequencing (found in 30 adenocarcino-
mas, 23 squamous-cell carcinomas, and 4 other
samples). In total, we detected 375 mirrored
subclonal allelic imbalance events that varied in
size from focal to whole chromosome and in-
volved 1 to 43% of affected tumor genomes (Fig.
S16 in Supplementary Appendix 1).
Chromosomal instabilit y may also directly con-
tribute to mutational heterogeneity through loss
of genomic segments carr ying clonal mutations.
Overall, a median of 13% of subclonal mutations
(range, 0 to 56) per sample are probably sub-
clonal as a result of loss events associated with
copy-number alterations, which suggests that
chromosomal instability may be an initiator of
both copy-number and mutational heterogeneity
(Fig. S17 in Supplementary Appendix 1).
Accumulating evidence suggests that genome-
doubling events are associated with the propaga-
tion of chromosomal instability by cancer cells
and may predict a poor prognosis.
15 -17
Genome-
doubling events were identified in 76% of tumors
and appeared to be clonal in all but three of
these t umors (from Patients CRUK0011, CRUK0062,
and CRUK0063), which suggests that whole-
genome duplication is an early event in NSCLC
evolution. In adenocarcinomas, we observed a
significant association bet ween genome doubling
and the frequency of both subclonal mutations
(P = 0.02) and subclonal copy-number alterations
(P = 0.003) (Fig. S18 in Supplementary Appendix 1).
Moreover, mirrored subclonal allelic imbalance
was signif icantly enriched in genome-doubled
tumors (P = 0.004 by Fisher’s exact test) (Fig. S16
in Supplementary Appendix 1).
Selection and Parallel Evolution
Deciphering evidence of ongoing selection in tu-
mors may shed light on evolutionary constraints,
which may identify therapeutic targets. Con-
straints and selection are exemplified by the oc-
currence of parallel evolution, in which somatic
events in distinct branches within a single tumor
converge on the same gene, protein complex, or
pathway.
No evidence of parallel evolution was found
at the mutational level. However, focal amplifi-
cations of different parental alleles in distinct
subclones occurred in 5 tumors and affected
known cancer genes, including MUC1, CDK4,
CHD8, and N K X2-1 (Fig. 3C, and Fig. S19 in Sup-
plementary Appendix 1). At the chromosome-arm
level, potential parallel evolution was observed in
13 tumors (5 adenocarcinomas, 6 squamous-cell
carcinomas, and 2 other tumors). Most parallel
evolution of chromosome-arm gains (in 10 of 11
samples) and losses (in 6 of 8 samples) have
been previously classified as significantly gained
or lost in NSCLC,
3,7
a finding that is consistent
with positive selection operating later in tumor
evolution (Fig. S20, S21, and S22 in Supplemen-
tar y Appendix 1).
To empirically estimate positive selection at
the mutational level, we used a ratio of substitu-
tion rates at nonsynonymous sites to those at
synonymous sites (dN/dS) that accounts for the
trinucleotide context of each mutation and de-
termines whether there is an enrichment of
protein-altering mutations as compared with the
background mutation rate.
18
Evidence for positive
selection (dN/dS, >1) was observed when all exonic
missense mutations were considered (Table S8
in Supplementary Appendix 1). This finding sug-
gests that mutations may be shaped by selection
in NSCLC. However, when mutations were tem-
porally dissected, signif icant positive selection
was observed for late, but not early, mutations.
Consistent with this f inding, nonsense mutations
were found to be depleted (dN/dS, <1) early but
not late in tumor development. These data fur-
ther suggest that selection is persistent in NSCLC
evolution and that constraints shape evolution-
ary trajectories. Depletion of early nonsense mu-
tations (dN/dS, <1) was greater in squamous-cell
carcinomas than in adenocarcinomas, and t he
rate of acquisition of clonal driver mutations (as
determined by the ratio of driver mutations to
passenger mutations) was significantly greater
in adenocarcinomas than in squamous-cell car-
cinomas (P = 0.001 by the Wilcoxon test).
Clonal and Subclonal Driver Alterations
and Timing of Genomic Events
Determining whether a cancer driver event oc-
curs early or late can indicate whether it is in-
volved in tumor initiation or maintenance, and
its clonality may inform potential therapeutic
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2117
Tracking the Evolution of Non–Small-Cell Lung Cancer
strategies, since subclonal alterations will be
present in only a proportion of cells and when
targeted may result in reduced treatment effi-
cac y.
19
We identified 795 driver events (range in
adenocarcinomas, 1 to 19; range in squamous-
cell carcinomas, 2 to 21). Of these events, 219 in
77 tumors were found to be subclonal (range in
adenocarcinomas, 0 to 10; range in squamous-
cell carcinomas, 0 to 12) and 576 to be clonal
(range in adenocarcinomas, 1 to 18; range in
squamous-cell carcinomas, 1 to 14) (Fig. S23 in
Supplementary Appendix 1 and Table S9 in Sup-
plementary Appendix 2). Significantly more driver
alterations were identif ied with the use of multi-
region whole-exome sequencing than with single-
sample analysis (P = 0.004 by the Wilcoxon test)
(Fig. S24 in Supplementary Appendix 1).
Alterations in cert ain cancer genes were not
only primarily clonal but almost always occurred
before genome duplication, which suggests in-
volvement in tumor initiation (Fig. 4). In adeno-
carcinomas, these alterations included targeta-
ble mutations or amplif icat ions in EGFR, MET, and
BRAF, as well as amplifications in TE R T, 8p loss,
and 5p gain. In squamous-cell carcinomas, muta-
tions in NOTCH1, amplifications in FGFR1 and in
the 3q region (which includes SOX2 and PIK3CA),
and loss of 3p, 4p, 5q, and 17p were early clonal
events. Mutations in T P53 were predominantly
clonal and early for both subtypes. Conversely,
other driver events, including mutations in K MT2C
and COL5A2 in adenocarcinomas and in PIK3CA
in squamous-cell carcinomas, while predomi-
nant ly clonal, often occurred after genome du-
plication, which suggests their involvement in
tumor maintenance or progression. Except for
alterations in T P53, ATM, CHEK2, and MDM2,
51% of 72 driver alterations affecting chromatin
remodeling, histone methylation, or DNA dam-
age response and repair were subclonal or late in
both histologic subtypes (23 of 41 events in ad-
enocarcinomas and 14 of 31 events in squamous-
cell carcinomas) (Fig. S25 in Supplementary
Appendix 1). UBR5, with a known role in dif-
ferentiation and DNA damage response, was one
of the most frequently altered genes later in evolu-
tion in both adenocarcinomas and squamous-cell
carcinomas. Other genes that were subject to
frequent subclonal or late alterations in adeno-
carcinomas included NF1 and NOTCH1, along
with 3p, 13q, and 21p loss and 7q and 8q gain,
whereas in squamous-cell carcinomas, alterations
in MLH1 and KRAS, along with 10q loss and 7p,
8q, and 20q gain, were late events.
Driver mutations that occurred early showed
a signif icantly greater tendency to occur in estab-
lished histologic-subtype–specific cancer genes
than did late or subclonal driver mutations,
which affected a broader selection of pan-cancer
genes
20
(Fig. S26 in Supplementary Appendix 1).
These data are consistent with the dN/dS muta-
tion-selection analysis and suggest that constraints
inherent in cancer evolution var y as tumors de-
velop, which potentially renders more evolution-
ary paths permissive for progression.
Overall, 86 of the 100 tumors in our study
had alterations that are being investigated in
NSCLC in genomically profiled drug studies, in-
cluding the National Lung Matrix Trial (NLMT)
21
and the Molecular Analysis for Therapy Choice
(MATCH) trial.
22
Of these 86 tumors, 17 (20%)
had subclonal targetable mutations and copy-
number alterations. In 12 of these 17 tumors
(71%), both a clonal and a subclonal targetable
alteration were present, which indicates how tar-
gets might be prioritized for therapeutic inter-
vention (Fig. S27 in Supplementary Appendix 1).
Discussion
Intratumor heterogeneit y provides the fuel for
tumor evolution and drug resistance.
23
Here, we
have provided an analysis of NSCLC evolution,
which has shown that intratumor heterogeneity
and branched evolution are almost universal
across the cohort. We also observed a common
pattern of early clonal genome doubling, followed
by extensive subclonal diversification.
These data may have important implications
for our understanding of tumor biolog y and
therapeutic control in NSCLC. Certain targetable
driver mutations, including those in EGFR, MET,
and BR AF, were almost exclusively clonal and
early, which explains the robust and uniform
responses that are often seen across multiple
sites of disease when these alterations are tar-
geted.
24 -26
However, more than 75% of the tumors
in our study carried a subclonal driver alteration,
including in genes such as PIK3C A, NF1, KRAS,
TP53, and NOTCH family members. Moreover, a
large fraction of subclonal driver mutations ap-
peared to be clonal in a single region but were
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2118
The
new england journal
of
medicine
Genome
doubling
Late
drivers
Mutations Copy-Number Events
Copy-Number Events
Copy-Number Events
Mutations
Mutations
Mutational Processes
Mutational Processes
Mutational Processes
7q
8q
3p
13q
21p
9p
15q
7p
17p
9q
1q
5p
8p
14/18
12/16
12/19
12/20
12/21
11/20
11/20
17/31
6/15
7/20
5/16
4/22
2/16
PPFIBP1 (12p)
EIF3E (8q)
KRAS (12p)
COX6C (8q)
RSPO2 (8q)
NKX2−1 (14q)
HEY1 (8q)
TERT (5p)
EGFR (7p)
FOXA1 (14q)
2/6
2/6
2/7
2/7
1/6
1/6
1/7
1/10
1/10
0/7
CS ES NS
CS ES NS
CS ES NS
CIC
EP300
FLT4
NOTCH1
PTPRC
SMAD4
DNM2
PASK
UBR5
KMT2C
BAP1
PLXNB2
ARID2
CTNNB1
NCOR1
NCOA6
COL2A1
COL5A2
U2AF1
KMT2D
PIK3CA
NF1
MGA
DOT1L
FUBP1
CREBBP
PRF1
RASA1
WRN
NOTCH2
NRAS
RNF43
SMARCA4
ARHGAP35
RAD21
SERPINB13
RB1
ATRX
KDM5C
WAS
FANCM
ARID1B
STK11
ATM
2/2
2/2
2/2
2/2
2/2
2/2
2/2
2/2
2/2
3/3
3/4
3/4
3/4
2/3
2/3
2/3
2/3
2/3
2/3
2/3
3/5
4/7
5/9
1/2
1/2
1/2
1/2
1/2
1/2
2/4
1/2
1/2
1/2
2/4
1/2
1/2
2/5
1/3
1/3
1/3
1/3
1/3
2/7
1/4
FAT1
EGFR
TP53
KRAS
KEAP1
BCOR
RBM10
APC
CHEK2
FBXW7
PHOX2B
TSC2
BRAF
CMTR2
MET
PRDM1
1/5
2/13
4/29
3/26
1/9
0/2
0/5
0/4
0/2
0/2
0/2
0/2
0/4
0/3
0/2
0/2
Pre-GD
initiating
drivers
Post-GD
clonal/
subclonal
Genome
doubling
Mutations Copy-Number Events
Mutational Processes
Mutational Processes
Mutational Processes
2p
7p
8q
15q
12p
20q
16q
7q
22q
1p
10q
9/9
10/10
12/13
7/8
6/7
9/11
8/10
5/7
7/10
6/10
8/14
5p
20p
4q
9p
10p
18p
13q
9q
11p
18q
21p
21q
5q
17p
3q
3p
10/20
6/12
5/10
7/14
8/16
5/10
7/15
5/11
3/7
3/7
6/15
4/10
4/17
3/13
2/9
4/18
4p 2/13
CDKN2A (9p)
BCL11A (2p)
FIP1L1 (4q)
IL7R (5p)
AKT2 (19q)
CD79A (19q)
FOXL2 (3q)
CCND1 (11q)
MYC (8q)
FGFR1 (8p)
PIK3CA (3q)
SOX2 (3q)
2/3
2/3
2/3
LSM14A (19q)
TERT (5p)
REL (2p)
IKBKB (8p)
CEBPA (19q)
LIFR (5p)
HOOK3 (8p)
2/5
2/5
1/3
1/4
1/4
1/5
1/5
1/6
1/6
0/3
0/3
0/4
0/4
0/8
0/13
0/14
CS ES
CS ES
CS ES
CYLD
KRAS
MLH1
UBR5
CBLB
PIK3CA
MGA
NCOA6
PLXNB2
ERCC5
CUL3
COL2A1
NOTCH2
COL5A2
FAT1
CDKN2A
BRIP1
DNM2
FANCM
WRN
CUX1
DICER1
NFE2L2
RASA1
NOTCH1
TP53
CREBBP
KEAP1
LATS1
SMAD4
KMT2D
FBXW7
PTEN
WT1
2/2
2/2
2/2
4/4
2/2
5/7
2/3
2/3
2/3
1/2
1/2
1/2
1/2
2/4
4/8
4/8
1/2
1/2
1/2
1/2
1/2
1/2
3/7
2/5
1/5
2/27
0/2
0/2
0/2
0/2
0/3
0/2
0/3
0/2
Late
drivers
Pre-GD
initiating
drivers
Post-GD
clonal/
subclonal
Pre-GD clonal
somatic event
Untimed clonal
somatic event
Post-GD clonal
somatic event
Subclonal
somatic event
Signature
Unclassified
Signature 1A
(mitotic clock)
Signature 2/13
(APOBEC)
Signature 4
(smoking)
Signature 5
(unknown)
AAdenocarcinoma BSquamous-Cell Carcinoma
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2119
Tracking the Evolution of Non–Small-Cell Lung Cancer
absent or subclonal in other regions, which con-
firmed the limitations of sampling single tumor
regions and emphasized the ability of multi-
region whole-exome sequencing to def ine the
clonality of driver events for prioritization of
drug t argets.
Late mutations in tumor-suppressor genes that
occur after genome doubling often affected only
one allele, which potentially left the wild-type
alleles intact. Although this finding could indi-
cate that late tumor-suppressor mutations are
often passenger events that do not contribute to
tumor progression, it is also plausible that germ-
line defects, subclonal copy-number loss, haplo-
insuff iciency, or transcriptional regulation may
act to limit wild-type expression. In contrast to
early mutations, late driver mutations were not
specific to the NSCLC subtype and often occurred
in cancer genes that have been identif ied in
other tumor types; a high proportion occurred
in genes that are involved in the maintenance of
genome integrit y through DNA damage response
and repair, chromatin remodeling, and histone
methylation. Such mutations may remove tissue-
specif ic constraints on the cancer genome and
provide advantages to emerging subclones later
in evolution. However, the observation of paral-
lel evolution of driver copy-number alterations
that were identified through mirrored subclonal
allelic imbalance, including in CDK4, FOXA1, and
BCL11A, suggests that despite extensive diversity,
specific constraints, which could be therapeuti-
cally exploited, may operate later in tumor evo-
lution.
Tumors with the highest subclonal mutational
burden had extensive APOBEC-mediated muta-
genesis, and 19 tumors carried subclonal driver
mutations within an APOBEC context. This find-
ing suggests that targeting the enzymatic activ-
ity of APOBEC may provide a means of limiting
subclone diversif icat ion. The clonal mutation
burden was significantly enriched in patients
with a smoking history. Conceivably, this f ind-
ing could be exploited for therapeutic benef it
through the use of peptide vaccines or adoptive
cell therapy against clonal neoantigens that are
present in every tumor cell. However, the obser-
vation that clonal mutations can be lost owing to
later copy-number events could limit the eff icacy
of such strategies, especially in tumors with high
chromosome instability.
Finally, although a single sample can provide
a static measure of chromosomal complexity,
27
the use of multiregion whole-exome sequencing
enables the assessment of dynamic chromosome
instability, which may lead to differences in chro-
mosomal karyotypes between NSCLC subclones.
The onset of chromosome instability appears to
have a considerable effect on the evolution of
NSCLC; such instability appears to be the pre-
dominant driver of parallel evolution and can
lead to both mutational and copy-number diver-
sity among subclones. Elevated copy-number het-
erogeneity was associated with shorter relapse-
free survival, which suggests that patients who
have early-stage tumors with high levels of copy-
number heterogeneity may represent a high-risk
group who may benefit from close monitoring
and early therapeutic intervention during follow-
up. We are continuing to assess this association
in the next 742 patients enrolled in TRACERx.
Whether noninvasive prognostic approaches, such
as liquid biopsy, can be used to prospectively
assess the levels of chromosomal instability in
the clinical setting warrants further attention.
28
In addition to ongoing efforts to target single
genetic alterations, there is a need to develop a
greater understanding of chromosomal instabil-
ity, which can alter the copy number of a multi-
tude of genes simultaneously. Indeed, therapeu-
tic efforts that can attenuate this process may
limit the ensuing heterogeneit y and tumor evolu-
tion that drive poor rates of relapse-free survival.
In the analysis presented here, we provide a
Figure 4 (facing page). Timing of Somatic Events
in NSCLC Evolution.
A diagram of tumor evolution in adenocarcinoma
(Panel A) and squamous-cell carcinoma (Panel B)
shows the approximate timing of genomic aberrations
with respect to the cancer life history. The timing of
mutations and copy-number events is shown as bars
indicating whether the events are clonal or subclonal.
Clonal mutations and chromosome-arm events are fur-
ther timed as early or late with respect to genome dou-
bling (GD). The frequency of mutations and copy-num-
ber alterations (subclonal and total) is indicated on the
right side of the bars. Pie charts show the fraction of
estimated mutations for each signature, averaged
across current smokers or recent ex-smokers (CS),
long-term (>20 years) former smokers (ES), and life-
long never smokers (NS) at three different time points.
Only genes that were mutated in at least two patients
or that had copy-number alterations in at least 20% of
the patients in the cohort are shown.
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2120
The
new england journal
of
medicine
census of driver events in early-stage NSCLC in
relation to clonality and show that chromo-
somal instability is not only a significant driver of
parallel evolution but also a predictor of poor
outcome.
Supported by Cancer Research UK (CRUK), the CRUK Lung
Cancer Centre of Excellence, t he University Col lege London Hos-
pitals Biomedical Rese arch Centre, the CRUK University College
London Experimental Ca ncer Medicine Centre, the Rosetrees
Trust, t he Francis Crick Institute (which receives it s core fund-
ing from CRUK [FC001169 and FC001202]), the U.K. Medical
Research Counci l (FC001169 and FC001202), and the Wellcome
Trust (FC001169 and FC001202). Dr. Swanton is a Royal Society
Napier Chair in Oncolog y and is funded by CRUK (TRACERx
and CRUK Cancer Immunotherapy Catalyst Net work), the Na-
tiona l Inst itute for Healt h Research, Novo Nord isk Foundation
(ID16584), the European Research Council, and PloidyNet (a
Marie Cu rie Initia l Train ing Net work). Dr. Van Loo is a Winton
Group Leader in recog nition of the suppor t of the Wi nton
Charit able Foundation i n the est ablishment of the Francis Crick
Institute.
Disclosure forms provided by the authors are available with
the fu ll text of this article at NEJM.org.
We thank a ll the patients who participated in this study and
represent atives of Illumin a and Agi lent who prov ided sequenc-
ing inf rast ruct ure suppor t.
Appendix
The authors’ full names and academic degrees are as follows: Mariam Jamal-Hanjani, M.D., Ph.D., Gareth A. Wilson, Ph.D., Nicholas
McGranahan, Ph.D., Nicolai J. Birkbak, Ph.D., Thomas B.K. Watkins, M.C.I.T., Selvaraju Veeriah, Ph.D., Seema Shafi, Ph.D., Diana H.
Johnson, B.Sc., Richard Mitter, M.Sc., Rachel Rosenthal, M.Sc., Max Salm, Ph.D., Stuart Horswell, M.Math., Mickael Escudero, M.Sc.,
Nik Matthews, B.Sc., Andrew Rowan, B.Sc., Tim Chambers, M.Sc., David A. Moore, M.D., Samra Turajlic, M.D., Ph.D., Hang Xu, Ph.D.,
Siow-Ming Lee, M.D., Ph.D., Martin D. Forster, M.D., Ph.D., Tanya Ahmad, M.D., Crispin T. Hiley, M.D., Ph.D., Christopher Abbosh,
M.D., Mary Falzon, M.D., Elaine Borg, M.D., Teresa Marafioti, M.D., David Lawrence, M.D., Martin Hayward, M.D., Shyam Kolvekar,
M.D., Nikolaos Panagiotopoulos, M.D., Sam M. Janes, M.D., Ph.D., Ricky Thakrar, M.D., Asia Ahmed, M.D., Fiona Blackhall, M.D.,
Ph.D., Yvonne Summers, M.D., Ph.D., Rajesh Shah, M.D., Leena Joseph, M.D., Anne M. Quinn, M.D., Ph.D., Phil A. Crosbie, M.D.,
Ph.D., Babu Naidu, M.D., Gary Middleton, M.D., Gerald Langman, M.D., Simon Trotter, M.D., Marianne Nicolson, M.D., Hardy Rem-
men, M.D., Keith Kerr, M.D., Mahendran Chetty, M.D., Lesley Gomersall, M.D., Dean A. Fennell, M.D., Ph.D., Apostolos Nakas, M.D.,
Sridhar Rathinam, M.D., Girija Anand, M.D., Sajid Khan, M.D., Peter Russell, M.D., Ph.D., Veni Ezhil, M.D., Babikir Ismail, M.D.,
Melanie Irvin-Sellers, M.D., Vineet Prakash, M.D., Jason F. Lester, M.D., Malgorzata Kornaszewska, M.D., Ph.D., Richard Attanoos,
M.D., Haydn Adams, M.D., Helen Davies, M.D., Stefan Dentro, M.Sc., Philippe Taniere, M.D., Ph.D., Brendan O’Sullivan, B.Sc., Helen L.
Lowe, Ph.D., John A. Hartley, Ph.D., Natasha Iles, Ph.D., Harriet Bell, M.Sc., Yenting Ngai, B.Sc., Jacqui A. Shaw, Ph.D., Javier Herrero,
Ph.D., Zoltan Szallasi, M.D., Roland F. Schwarz, Ph.D., Aengus Stewart, M.Sc., Sergio A. Quezada, Ph.D., John Le Quesne, M.D., Ph.D.,
Peter Van Loo, Ph.D., Caroline Dive, Ph.D., Allan Hackshaw, M.Sc., and Charles Swanton, M.D., Ph.D.
The authors’ affiliations are as follows: the Cancer Research UK Lung Cancer Centre of Excellence (M.J.-H., G.A.W., N. McGranahan,
N.J.B., S.V., S.S., D.H.J., R.R., S.-M.L., M.D.F., C.A., S.M.J., C.D., C.S.), London and Manchester, Good Clinical Laboratory Practice
Facility, University College London (UCL) Experimental Cancer Medicine Centre (H.L.L., J.A.H.), Bill Lyons Informatics Centre (J.H.),
and Cancer Immunology Unit (S.A.Q.), UCL Cancer Institute, the Translational Cancer Therapeutics Laboratory (G.A.W., N. McGranahan,
N.J.B., T.B.K.W., A.R., T.C., S. Turajlic, H.X., C.T.H., C.S.), Department of Bioinformatics and Biostatistics (R.M., M.S., S.H., M.E.,
A.S.), Advanced Sequencing Facility (N. Matthews), and Cancer Genomics Laboratory (S.D., P.V.L.), Francis Crick Institute, the Renal
and Skin Units, Royal Marsden Hospital (S. Turajlic), the Departments of Medical Oncology (M.J.-H., S.-M.L., M.D.F., T.A., C.A., C.S.),
Pathology (M.F., E.B., T.M.), Cardiothoracic Surgery (D.L., M.H., S. Kolvekar, N.P.), Respiratory Medicine (S.M.J., R.T.), and Radiol-
ogy (A.A.), UCL Hospitals, Lungs for Living, UCL Respiratory, UCL (S.M.J.), the Department of Radiotherapy, North Middlesex Univer-
sity Hospital (G.A.), the Department of Respiratory Medicine, Royal Free Hospital (S. Khan), and UCL Cancer Research UK and Cancer
Trials Centre (N.I., H.B., Y.N., A.H.), London, Cancer Studies, University of Leicester (D.A.M., D.A.F., J.A.S., J.L.Q.), the Department
of Thoracic Surgery, Glenfield Hospital (A.N., S.R.), and the Medical Research Center Toxicology Unit (J.L.Q.), Leicester, the Institute
of Cancer Studies, University of Manchester (F.B.), the Christie Hospital (F.B., Y.S.), the Departments of Cardiothoracic Surgery (R.S.)
and Pathology (L.J., A.M.Q.) and the North West Lung Centre (P.A.C.), University Hospital of South Manchester, and Cancer Research
UK Manchester Institute (C.D.), Manchester, the Departments of Thoracic Surgery (B.N.) and Cellular Pathology (G.L., S. Trotter),
Birmingham Heartlands Hospital, Molecular Pathology Diagnostic Services, Queen Elizabeth Hospital (P.T., B.O.), and Institute of Im-
munology and Immunotherapy, University of Birmingham (G.M.), Birmingham, the Departments of Medical Oncology (M.N.), Cardio-
thoracic Surgery (H.R.), Pathology (K.K.), Respiratory Medicine (M.C.), and Radiology (L.G.), Aberdeen University Medical School and
Aberdeen Royal Infirmary, Aberdeen, the Department of Respiratory Medicine, Barnet and Chase Farm Hospitals, Barnet (S. Khan), the
Department of Respiratory Medicine, Princess Alexandra Hospital, Harlow (P.R.), the Department of Clinical Oncology, St. Luke’s
Cancer Centre, Guildford (V.E.), the Departments of Pathology (B.I.), Respiratory Medicine (M.I.-S.), and Radiology (V.P.), Ashford and
St. Peters’ Hospitals, Surrey, the Department of Clinical Oncology, Velindre Hospital (J.F.L.), the Departments of Radiology (H.A.) and
Respiratory Medicine (H.D.), University Hospital Llandough, the Departments of Pathology (R.A.) and Cardiothoracic Surgery (M.K.),
University Hospital of Wales, and Cardiff University (R.A.), Cardiff, and Wellcome Trust Sanger Institute, Hinxton, and Big Data Insti-
tute, University of Oxford, Oxford (S.D.) — all in the United Kingdom; the Center for Biological Sequence Analysis, Department of
Systems Biology, Technical University of Denmark, Lyngby (Z.S.); the Computational Health Informatics Program, Boston Children’s
Hospital and Harvard Medical School, Boston (Z.S.); MTA-SE-NAP, Brain Metastasis Research Group, 2nd Department of Pathology,
Semmelweis University, Budapest, Hungary (Z.S.); Berlin Institute for Medical Systems Biology, Max Delbrueck Center for Molecular
Medicine, Berlin (R.F.S.); and the Department of Human Genetics, University of Leuven, Leuven, Belgium (P.V.L.).
References
1. Siegel R, Naishadha m D, Jemal A.
Cancer st atistics, 2013. CA Cancer J Clin
2013; 63: 11-30.
2. Jemal A, Br ay F, Center MM, Ferlay J,
Ward E, Forman D. Global ca ncer statis-
tics. CA Cancer J Clin 2011; 61: 69-90.
3. Cancer Genome Atlas Research Net-
work. Comprehensive molecular profiling
of lung adenoc arcinoma. Nature 2014;
511: 543-50.
4. Cancer Genome Atlas Research Net-
work. Comprehensive genomic cha racter-
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.
n engl j med 376;22 nejm.org June 1, 2017
2121
Tracking the Evolution of Non–Small-Cell Lung Cancer
izat ion of squamous cell lung canc ers.
Natu re 2012; 489: 519-25.
5. Imielinski M, Berger A H, Hammer-
man PS, et a l. Mapping t he hal lmarks of
lung adenocarcinoma with massively par-
allel sequencing. Cell 2012; 150: 1107-20.
6. Govind an R, Ding L, Grif fith M, et al.
Genomic landscape of non-small cell lung
cancer in smokers a nd never-smokers. Cell
2012; 150: 1121-34.
7. Campbel l JD, Alexandro v A, Kim J, et al.
Distinct patterns of somatic genome
alt erations in lung adenocarci nomas and
squamous cell carcinomas. Nat Genet
2016; 48: 607-16.
8. de Bruin EC, McGranahan N, Mitter
R, et al. Spat ial and temporal di versit y in
genomic instabi lit y processes def ines lu ng
cancer evolut ion. Science 2014; 346: 251-6.
9. Zhang J, Fujimot o J, Zhang J, et a l.
Intratumor heterogeneity in localized
lung adenocarcinomas delineated by multi-
region sequencing. Science 2014; 346: 256-9.
10. Ja mal-Hanjani M, Hackshaw A, Ngai Y,
et al. Track ing genom ic cancer evolution
for precision medicine: the lung T RACERx
stud y. PLoS Biol 2014; 12(7): e1001906.
11. A lexa ndrov LB, Nik-Zaina l S, Siu HC,
Leung SY, Stratton MR. A mutationa l sig-
nature in gastric cancer suggests t hera-
peutic strategies. Nat Commun 2015; 6:
8683.
12. Alexand rov LB, Nik-Zainal S, Wedge
DC, et al. Sig natu res of mutational pro-
cesses in human c ancer. Nature 2013; 500:
415-2 1.
13. Alexandrov LB, Jones PH, Wedge DC,
et al. Clock-like mutat ional processe s in
human somatic cells. Nat Genet 2015; 47:
14 0 2- 7.
14. Robert s SA, Law rence MS, K limczak
LJ, et al. An A POBEC cytidi ne deam inase
mutagene sis pattern is w idespread in hu-
man ca ncers. Nat Genet 2013; 45: 970-6.
15. Carter SL, Cibulskis K, Helman E, et al.
Absolute qu anti fication of somatic DNA
alterations i n human cancer. Nat Biotech-
nol 2012; 30: 413-21.
16. Dewhurst SM, McGranahan N, Burrel l
RA, et a l. Tolerance of whole-genome
doubling propagates chromosomal insta-
bilit y and accelerates cancer genome evo -
lution. Cancer Discov 2014; 4: 175-85.
17. Fujiwara T, Bandi M, Nit ta M, Ivanova
EV, Bronson RT, Pellman D. Cy toki nesis
failu re generating tetraploids promotes
tumor igenesis i n p53-null cel ls. Nature
2005; 437: 1043-7.
18. Ma rtincoren a I, Roshan A, Gerst ung
M, et al. Tumor evolution: hig h burden
and pervasive positive selection of somat-
ic mutat ions in nor mal human skin. Sci-
ence 2015; 348: 880-6.
19. Lohr JG, Stojanov P, Carter SL, et al.
Widespread genetic heterogeneity i n mul-
tiple myeloma: implications for ta rgeted
therapy. Cancer Cell 2014; 25: 91-101.
20. Lawrence MS, St ojanov P, Mermel CH,
et al. Discovery and saturat ion ana lysis of
cancer genes across 21 tumou r ty pes. Na-
ture 2014; 505: 495-501.
21. Middleton G, Crack LR, Popat S, et al.
The National Lung Matrix Trial: t ranslat-
ing the biology of st ratification in ad-
vanced non-sma ll-cel l lung ca ncer. Ann
Oncol 2015; 26: 2464-9.
22. Abrams J, Conley B, Mooney M, et a l.
Nationa l Cancer Institute’s Precision Medi-
cine Initiat ives for the new Nat ional Clin-
ical Tri als Net work. Am Soc Clin Oncol
Educ Book 2014; : 71-6.
23. Greaves M. Evolut ionar y determi-
nants of cancer. Cancer Discov 2015; 5:
80 6 -20.
24. Chapma n PB, Hauschild A, Robert C,
et al. Improved sur viva l with vemurafenib
in melanoma wit h BRAF V600E mutat ion.
N Engl J Med 2011; 364: 2507-16.
25. Mok TS, Wu Y-L, Thongprasert S, et al.
Gefitin ib or carboplatin–paclita xel in pul-
monar y adenocarcinoma. N Engl J Med
2009; 361: 947-57.
26. Cao Y, Xiao G, Qiu X, Ye S, Lin T. Ef-
fic acy and s afet y of crizotinib among Ch i-
nese EML4-ALK-positive, advanced-stage
non-small cell lung cancer patient s. PLoS
One 2 014; 9(1 2): e114 00 8.
27. McGranahan N, Burrell RA, Endes-
felder D, Novelli MR, Swa nton C. Cancer
chromosomal instability: therapeutic a nd
diagnostic ch allenges. EMBO Rep 2012;
13: 528-38.
28. Ni X, Zhuo M, Su Z, et al. Reproducible
copy number variation pat terns among
single circulating t umor cells of lung can-
cer patient s. Proc Nat l Acad Sci U S A
2013; 1 10: 210 83-8.
Copyright © 2017 Massachusetts Medical Society.
ARTICLE METRICS NOW AVAILABLE
Visit the article page at NEJM.org and click on the Metrics tab to view
comprehensive and cumulative article metrics compiled from multiple sources,
including Altmetrics. Learn more at www.nejm.org/page/article-metrics-faq.
The New England Journal of Medicine
Downloaded from nejm.org at UNIVERSITY COLLEGE LONDON on July 13, 2017. For personal use only. No other uses without permission.
Copyright © 2017 Massachusetts Medical Society. All rights reserved.