UCLA
UCLA Previously Published Works
Title
Spatial intratumoral heterogeneity and temporal clonal evolution in esophageal squamous
cell carcinoma.
Permalink
https://escholarship.org/uc/item/5g283207
Journal
Nature genetics, 48(12)
ISSN
1061-4036
Authors
Hao, Jia-Jie
Lin, De-Chen
Dinh, Huy Q
et al.
Publication Date
2016-12-01
DOI
10.1038/ng.3683
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
Spatial intratumor heterogeneity of genetic, epigenetic
alterations and temporal clonal evolution in esophageal
squamous cell carcinoma
Jia-Jie Hao
1,10
, De-Chen Lin
2,3,10,11
, Huy Q. Dinh
4,10
, Anand Mayakonda
5,10
, Yan-Yi
Jiang
5,10
, Chen Chang
1
, Ye Jiang
1
, Chen-Chen Lu
1
, Zhi-Zhou Shi
6
, Xin Xu
1
, Yu Zhang
1
, Yan
Cai
1
, Jin-Wu Wang
7
, Qi-Min Zhan
1
, Wen-Qiang Wei
8,11
, Benjamin P. Berman
4,11
, Ming-Rong
Wang
1,11
, and H. Phillip Koeffler
2,5,9
1
State Key Laboratory of Molecular Oncology, National Cancer Center/Cancer Hospital, Chinese
Academy of Medical Sciences and Peking Union Medical College, Beijing, 100021, China
2
Cedars-Sinai Medical Center, Division of Hematology/Oncology, UCLA School of Medicine, Los
Angeles, USA
3
Guangdong Provincial Key Laboratory of Malignant Tumor Epigenetics and Gene Regulation,
Medical Research Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Guangzhou,
China
4
Center for Bioinformatics and Functional Genomics, Biomedical Sciences, Cedars-Sinai Medical
Center, UCLA School of Medicine, Los Angeles, USA
5
Cancer Science Institute of Singapore, National University of Singapore, Singapore
6
Faculty of Medicine, Kunming University of Science and Technology, Kunming, Yunnan, China
7
Department of Pathology, Linzhou Cancer Hospital, Henan, China
8
Department of Cancer Epidemiology, National Cancer Center/Cancer Hospital, Chinese
Academy of Medical Sciences and Peking Union Medical College, Beijing, 100021, China
Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research,
subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms
11
Correspondence should be addressed to M-R.W. (wangmr2015@126.com), B.P.B (Benjamin.Berman@csmc.edu), D-C.L.
(dchlin11@gmail.com) or W-Q W. (weiwq2006@126.com).
10
These authors contributed equally to this work.
URLs
BWA-MEM,
http://arxiv.org/abs/1303.3997v2.
fpFilter perl script, https://github.com/ckandoth/variant-filter.
Bam-readcount, https://github.com/genome/bam-readcount.
PHYLIP, http://evolution.genetics.washington.edu/phylip.html.
Accession codes. Digital sequencing and HM450 Bead array files have been deposited into Sequence Read Archive (SRP072112) and
Gene Expression Omnibus (GSE79366), respectively.
AUTHOR CONTRIBUTIONS
M.-R.W., D.-C.L., B.P.B. and H.P.K. conceived and designed the experiments. J.-J.H., D.-C.L., H.Q.D., W.-Q.W. B.P.B., M.-R.W., and
H.P.K. wrote the manuscript. J.-J.H., D.-C.L., Y.J., C.C., C.-C.L., X.X., Y.C. performed the experiments. J.-J.H., H.Q.D., A.M., B.P.B.,
and Z.-Z.S. performed statistical analysis. J.-J.H., D.-C.L., H.Q.D., Y.-Y.J., B.P.B. and H.P.K. analyzed the data. X.X. contributed
reagents. W.-Q.W. contributed materials. J.-W.W. and J.-J.H. read the H&E slides. D.-C.L., Y.Z., Q.-M.Z. and H.P.K. jointly
supervised research.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
HHS Public Access
Author manuscript
Nat Genet
. Author manuscript; available in PMC 2017 April 17.
Published in final edited form as:
Nat Genet
. 2016 December ; 48(12): 1500–1507. doi:10.1038/ng.3683.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
9
National University Cancer Institute, National University Hospital Singapore, Singapore
Abstract
Esophageal squamous cell carcinoma (ESCC) is among the most common malignancies, but little
is known about its spatial intratumor heterogeneity (ITH) and temporal clonal evolutionary
processes. To address this, we performed multiregion whole-exome sequencing on 51 tumor
regions from 13 ESCCs, and multiregion global methylation profiling on three of these 13 cases.
We found an average of 35.8% heterogeneous somatic mutations with strong evidence of ITH.
Half of driver mutations located on the branches targeted oncogenes, including
PIK3CA, NFE2L2,
MTOR, etc
. By contrast, the majority of truncal and clonal driver mutations occurred in tumor
suppressor genes, including
TP53, KMT2D, ZNF750, etc
. Interestingly, the phyloepigenetic trees
robustly recapitulated the topologic structures of the phylogenetic ones, indicating the possible
relationship between genetic and epigenetic alterations. Our integrated investigations of the spatial
ITH and clonal evolution provide an important molecular foundation for enhanced understanding
of the tumorigenesis and progression of ESCC.
Esophageal carcinoma is among the most common human cancers, causing over 400,000
deaths worldwide annually
1,2
. The highest-risk areas are located in Eastern Asia, as well as
Eastern and Southern Africa; and the most prevalent type is esophageal squamous cell
carcinoma (ESCC)
1,2
. The five-year survival rates of ESCC patients undergoing surgery is
below 30%, because a large proportion of the tumors are unresectable or have already
metastasized before diagnosis
3
.
Recently, several large-scale genomic studies including ours have characterized ESCC
genomes with hundreds of somatic mutations and copy number alterations, and have
identified significantly mutated genes, including
TP53, PIK3CA, ZNF750
, etc.
4–9
. The
APOBEC signature is a predominant mutational spectrum, and contributes to the mutagenic
processes of ESCCs
6,8
. However, the genomic alterations identified in all of these studies
were obtained using only single samples representing individual cases, and little is known
about the spatial intratumor heterogeneity (ITH) and the temporal clonal evolutionary
processes of mutational spectrum in ESCC. Moreover, although alterations in DNA
methylation have been observed in ESCC, the ITH of these epigenetic changes is still
unknown, and whether it correlates with the genetic architecture remains unexplored.
Precise understanding of both the genomic and epigenomic architecture of primary ESCC
tumors is crucial for developing personalized patient treatment and molecular-based
biomarkers
10
. Furthermore, an integrated investigation of the genomic and epigenomic
evolutionary trajectory of ESCC may also reveal new insights into the relationship between
the genome and epigenome. In the present study, we address these critical issues through
integrative molecular approaches, including multiregional whole-exome sequencing (M-
WES), global methylation profiling, as well as phylogenetic and phyloepigenetic tree
construction.
Hao et al. Page 2
Nat Genet
. Author manuscript; available in PMC 2017 April 17.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
RESULTS
Spatial ITH of ESCC
M-WES was performed on the genomic DNA from 13 primary ESCC patients, and the
clinical-pathological parameters of these patients were listed in Supplementary Table 1. In
total, 51 tumor regions and 13 matched morphologically normal esophageal tissues (four
tumor regions and one matched normal tissue per case, with the exception of ESCC04,
which only had three tumor regions) were sequenced, with the mean coverage depth of
150×. A total of 1,610 non-silent somatic mutations (affecting 1,427 genes) and 568 silent
mutations were identified, with a validation rate of 90% (Supplementary Tables 2–3).
To explore ITH and genomic evolution of ESCC, phylogenetic trees were constructed based
on somatic mutations (both silent and non-silent) identified from each tumor region. The
trunk, shared branch and private branch of the tree represented mutations in all tumor
regions, in some but not all regions, and only in one region, respectively. As shown in Fig.
1a and Supplementary Fig. 1, phylogenetic trees varied extensively among different cases,
and all of the 13 ESCCs showed evidence of spatial ITH, with an average of 35.8%
(780/2,178; range, 8.0%–60.9%) of somatic variants having spatial heterogeneity.
Characterization of the relative timing of mutations affecting driver genes with possible
biological relevance is essential for revealing the evolutionary processes of the cancer
genome, as well as further improving precision medicine strategies. To address this, we
identified potential driver mutations according to recent large-scale ESCC sequencing
data
4–8
, COSMIC gene census
11
and Pan-cancer analysis
12
; this was followed by tracing
them within the phylogenetic trees (See Methods). Overall, driver mutations were
significantly more enriched in the trunks than were passenger mutations (77.8% vs. 63.8%;
P
= 0.023; Fig. 1b). This indicates that drivers are mutated as relatively early events during
the evolutionary process of the tumors, which is in accordance with previous findings in
other tumor types
13
. We next separated putative driver mutations into those occurring either
in oncogenes or tumor suppressor genes (TSGs). Importantly, half of the driver mutations
(50.0%) that mapped to the branches were within oncogenes, including
PIK3CA, KIT,
NFE2L2, MTOR
and
FAM135B
. In comparison, only 22.4% of driver mutations located on
the trunks affected oncogenes, and the rest were in TSGs. For example,
TP53
mutations
were present in twelve of the thirteen cases, and were truncal in all of the mutated cases, in
agreement with recent reports
14,15
. It is worthwhile to note that potentially actionable
mutations such as those targeting
PIK3CA
and
MTOR
tended to be oncogenic branch
events. These findings highlight the extra caution needed when considering inhibiting these
mutants in ESCC, given previous studies showing that suppressing subclonal drivers led to
growth acceleration of non-mutated subpopulations
16
.
Clonal status of putative driver mutations
We next investigated the clonal status of somatic mutations within individual regions.
Cancer cell fraction (CCF) in each tumor region was calculated as described previously
through integrative analysis of local copy number, variant allele frequency (VAF) and tumor
cell purity
16,17
. Several driver mutations were subclonal and possibly occurred as late events
Hao et al. Page 3
Nat Genet
. Author manuscript; available in PMC 2017 April 17.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
in ESCC, including
MTOR, KEAP1, PTPRB
and
FAM135B
. In contrast, cancer genes on
the trunks, such as
TP53, NOTCH1, CREBBP, KMT2D
and
ZNF750
, were predominantly
mutated in a fully clonal manner (Fig. 2), further verifying our earlier phylogenetic tree
analysis showing that these mutations were possibly early lesions during ESCC
development. Of particularly noteworthy distinction, a number of driver variants detected as
clonal within some individual tumor regions, were absent in others from the same individual,
producing an “illusion” of clonal dominance. For example, a
PIK3CA
hotspot mutation
(M1043I) was undetectable in tumor region T2 and T3 in case ESCC13 but was clonally
dominant in the other two regions. Likewise, a hotspot mutation in
KIT
gene (E601K) was
present in 100% tumor cells in regions T1 and T3 in case ESCC08, yet was absent in the rest
of the tumor regions. Such clonal dominance was also observed in
NFE2L2
in case
ESCC12. Our results suggest that driver mutations can have mixed and complex
intratumoral clonal status in ESCC, and that current single-sampling approach may
misinterpret these critical genomic lesions because of the “illusion” of clonal dominance.
We further investigated all the non-silent variants within genes and related pathways that
have potential targeting approaches. As shown in Supplementary Fig. 2, mutations affecting
members in PI3K/MTOR pathway,
KIT, AURKA
and
CCND2
were always late events
(branched/subclonal). By contrast, variants in
ERBB4, FGFR2, BRCA2, ATM
and
TP53
,
were mutated as early events (truncal/clonal), suggesting their potentials as candidate
actionable targets for ESCC.
ITH of copy number alterations (CNA)
We next analyzed the ITH at the copy number level (Supplementary Table 4). First, recurrent
copy number alterations which involve important cancer genes in ESCC were identified
based on our previous results
6
, and we confirmed that the present cohort harbored these
recurrent CNAs with similar frequencies (Supplementary Fig. 3). Although CNAs were
generally more similar within cases than between different cases, we found extensive CNA
ITH, with 90% (9/10) of all recurrent CNAs being spatially heterogeneous. For example in
ESCC08, chr7p11.2 amplification (encompassing
EGFR
) was observed in regions T1 and
T4, but not in regions T2 and T3. Similarly, deletions of chr9p21.3 (harboring
CDKN2A/B
)
were ubiquitous in some cases but also occurred as heterogeneous aberrations in other
samples. The only driver CNA found as consistently ubiquitous was the copy number gain
of 11q13, which encompassed a number of oncogenes including
CCND1, ANO1
18–20
and
CTTN
21,22
, highlighting the importance of this aberration as a founder genomic lesion in the
development of ESCC. These results suggest that similar as somatic mutations, CNAs also
show significant spatial ITH, concordant with the observations in several other types of
cancers
23–25
.
The within-patient mutational rate (mean = 168) was higher than the within-region rate
(mean = 139, Supplementary Table 5), highlighting the improved resolution of our multi-
biopsy approach for genomic interrogation. Particularly, in the case of branched cancer
genes, current M-WES approach markedly increased the sensitivity of the detection rate
(Table 1). For example,
ATR
and
TSC1
mutations, which were detected in only 2% of tumor
regions (in agreement with previous results), occurred in 7.7% of cases. In addition, the
proportion of subclonal mutations detected in each tumor region was much lower than that
Hao et al. Page 4
Nat Genet
. Author manuscript; available in PMC 2017 April 17.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript