multiregional sequencing reveals genomic alterations and ... · primary malignant melanoma of the...
TRANSCRIPT
1
Multiregional sequencing reveals genomic alterations and clonal dynamics in primary malignant 1
melanoma of the esophagus 2
Running title: Genetic evolution of PMME 3
Jingjing Li1*, Shi Yan2*, Zhen Liu1*, Yong Zhou2, Yaqi Pan1, WenQin Yuan1, Mengfei Liu1, Qin Tan1, Geng 4
Tian3, Bin Dong4, Hong Cai1, Nan Wu2, Yang Ke1. 5
1 Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education/Beijing), 6
Laboratory of Genetics, Peking University Cancer Hospital & Institute, Beijing, China. 7
2 Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education/Beijing), 8
Department of Thoracic Surgery II, Peking University Cancer Hospital & Institute, Beijing, China. 9
3 Geneis Co. Ltd., Beijing, China. 10
4 Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education/Beijing), Central 11
Lab, Peking University Cancer Hospital & Institute, Beijing, China. 12
Correspondence and requests for materials should be addressed to Nan Wu, MD, Department of 13
Thoracic Surgery II, Peking University Cancer Hospital & Institute, Fucheng road 52, Haidian District, 14
Beijing, 100146, China. E-mail: [email protected]; Yang Ke, Professor, Laboratory of Genetics, 15
Peking University Cancer Hospital & Institute, Fucheng road 52, Haidian District, Beijing, 100146, 16
China. E-mail: [email protected]. 17
* These authors contribute equally to this work. 18
Key words: primary malignant melanoma of the esophagus, multiregional sequencing, tumor 19
evolution 20
Abbreviations: 21
BAF, B allele frequency 22
CCF, cancer cell fraction 23
ESCC, esophageal squamous cell carcinoma 24
FFPE , paraffin-embedded 25
GI, gastrointestinal 26
IHC, immunohistochemistry 27
ITH, intra-tumor heterogeneity 28
LRR, Log R ratio 29
OG, oncogene 30
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
2
PCM , primary cutaneous melanoma 31
PMM, primary mucosal melanoma 32
PMME, primary malignant melanoma of the esophagus 33
SCNAs, somatic copy number alterations 34
SNVs, somatic nucleotide variants 35
TSG, tumor suppressor gene 36
VAF, variant allele frequency 37
38
Grant support: 39
This work was supported by the UMHS-PUHSC Joint Institute for Translational and Clinical Research 40
(grant number BMU20140483) (Y.Ke), Beijing Municipal Science and Technology Commission [grant 41
number Z141100002114046] (Y.Ke), and the Charity Project of National Ministry of Health [grant 42
number 201202014] (Y.Ke). 43
44
45
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
3
Abstract 46
Primary malignant melanoma of the esophagus (PMME) is a rare and aggressive disease with high 47
tendency of metastasis. To characterize the genetic basis and intra-tumor heterogeneity of PMME, we 48
performed multiregion exome sequencing and whole genome SNP array genotyping of 12 samples 49
obtained from a PMME patient. High intra-tumor heterogeneity was observed in both somatic 50
mutation and copy number alteration levels. Nine geographically separate samples including two 51
normal samples were clonally related and followed a branched evolution model. Most putative 52
oncogenic drivers such as BRAF and KRAS mutations as well as CDKN2A biallelic inactivation were 53
observed in trunk clones, whereas clinically actionable mutations such as PIK3CA and JAK1 mutations 54
were detected in branch clones. Ancestor tumor clones evolved into three subclonal clades: clade1 55
fostered metastatic subclones that carried metastatic features of PIK3CA and ARHGAP26 point 56
mutations as well as chr13 arm-level deletion, clade2 owned branch-specific JAK1 mutations and PTEN 57
deletion, and clade3 was found in two vertical distribution samples below the primary tumor area, 58
highlighting the fact that it is possible for PMME to disseminate by the submucosal longitudinal 59
lymphatic route at an early stage of metastasis. These findings facilitate interpretation of the genetic 60
essence of this rare melanoma subtype as well as the pattern of cancer evolution, thus reinforcing the 61
therapeutic challenges associated with PMME. 62
Introduction 63
Mucosal melanoma is most often the result of metastases from primary cutaneous melanoma (PCM), 64
whereas primary mucosal melanoma (PMM) is an exceedingly rare disease that behaves more 65
aggressively and has a poorer prognosis(1). The molecular and pathological characteristics of 66
cutaneous melanoma have been well studied, and more than one thousand PCM exomes have been 67
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
4
sequenced(2,3). Key cellular pathways, such as CDKN2A/CDK4/CCND1/RB1, MAPK and PI3k/AKT, and 68
significant genes, such as BRAF, NRAS, TERT, TP53, NF1, and RB1, are considered to play critical roles in 69
PCM tumorigenesis(4-7). Recently, two studies with very limited sample sizes have claimed that 70
distinct mechanisms drive PMM and PCM(8,9). Moreover, melanoma subtypes affecting different 71
body regions exhibit different genetic features, even those of the same histological type(2,10,11). 72
Despite these findings, there is a paucity of data elucidating the etiopathogenesis and predictive 73
factors, as well as the effectiveness of therapeutic options, for these diseases. 74
Primary malignant melanoma of the esophagus (PMME) is a rare gastrointestinal (GI) mucosal 75
melanoma, representing 0.1-0.2% of all esophageal malignancies, with only 339 cases reported by 76
2016 (12-14). It is reported that PMME has a tendency for vertical growth within the esophageal 77
wall(15). Melanosis is another important feature favoring PMME over a metastatic melanoma to the 78
esophagus(15). However, the molecular background of PMME is largely unknown because of its rarity. 79
Recent reports have included only small series or case studies with analyses of a single or few 80
molecular aberrations. The BRAF, NRAS, KRAS, and c-Kit genes have been reported to be the most 81
commonly altered genes in PMME(16,17). Although evidence indicates that GI (gastrointestinal) 82
mucosal melanoma arises primarily at certain areas within the GI system, it is still difficult to 83
distinguish PMM from metastatic melanoma in the GI tract that may be derived from an unknown or 84
regressed cutaneous primary tumor. Tumorigenesis and metastatic mechanisms, as well as other 85
clinical performance of PMME are lack of scientific interpretation due to the vacancy of studies. 86
Here, we applied exome sequencing and chromosomal aberration analysis to study multiple spatially 87
separated resected samples from a PMME patient, including samples of the primary tumor, metastatic 88
lymph nodes, normal esophageal mucosal tissues, and non-neoplastic melanosis lesion. The presence 89
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
5
of a precursor lesion or melanosis could be a marker for GI mucosal melanoma(18); however, this 90
notion is still under debate, and more evidence is needed. In general, this PMME exhibited distinct 91
genetic features, with both similarities to and differences from those of cutaneous melanoma. 92
Significant driver genes in PCM, such as BRAF and CDKN2A, were determined to be potential driver 93
genes in this PMME, whereas infrequently altered genes in PCM, such as KRAS, also played a role in 94
PMME tumorigenesis. By characterizing the genetic basis and intra-tumor heterogeneity (ITH), we 95
found that the genome of this PMME fit a branching evolution model, as ancestor clones evolved 96
along different trajectories and diversified into primary tumor subclones and early and late metastatic 97
subclones. Interestingly, the mutational signature switched from a C>T to a C>A transition with tumor 98
progression, indicating that different inducers promoted each trunk and branch mutation during the 99
mutational processes. 100
Comprehensively resolving the genetic architecture and temporal dynamics of the mutagenic 101
processes in PMME could help us better understand the genetic basis of this rare malignancy. With 102
more PMME cancer genomes being deciphered, advanced diagnostics and therapeutics could be 103
developed in the near future. 104
Material and Methods 105
Patient and samples 106
A 59-year-old Chinese female was admitted to the Thorax Surgical II Department of Beijing Cancer 107
Hospital in October 2013 with a 3-month history of dysphagia. Physical examination and CT scan 108
showed no evidence of skin melanoma lesions. Esophagogastroduodenoscopy revealed a 3.0 × 2.5-cm 109
polypoidal tumor with black pigmentation; 3 cm below the bottom edge of the main tumor, a flat, 0.5 110
× 1.0-cm mucosal melanosis lesion was observed. The patient did not receive any treatment before 111
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
6
surgery. The patient was informed and signed an informed consent form before surgery. This study 112
was approved by the Medical Ethics Committee of Peking University Cancer Hospital (#2013KT31) and 113
conducted in accordance with the Declaration of Helsinki. 114
The resected esophageal specimen was photographed and marked with a vertical ruler. Then, spatially 115
separated samples, including four primary tumor samples (T1-T4), five normal mucosal samples 116
(N1-N5), and one mucosal melanosis sample (ME1), were collected discretely. Lymph node samples 117
(LN1-LN4) were isolated from surgical specimen tissue blocks, and a peripheral blood sample was 118
acquired after surgery (Figure 1A&Figure S1). Each tumor and normal mucosal sample (10-20 mg) was 119
cut into thirds; one piece was used for pathological analysis, the second was used for genomic DNA 120
extraction, and the third was kept in a storage solution (RNAlater, Thermo Fisher Scientific) as an 121
extra piece. Experienced local pathologists performed a histopathological review of all primary tumors, 122
including immunohistochemistry (IHC) for detection of HMB-45, Melan-A, and S-100 protein 123
expressions (Supplementary Figure S2). The ME1 sample had a jelly-like texture, which may have led 124
to its failure in the formalin-fixed, paraffin-embedded (FFPE) process; thus, no further pathological 125
examination was performed on this sample. Genomic DNA was isolated using a DNeasy Blood and 126
Tissue Kit (QIAGEN) or a QIAamp DNA FFPE Tissue Kit (QIAGEN). 127
Multiregion exome sequencing 128
Exome capture was performed using 3-6 μg of genomic DNA per sample with an Agilent SureSelect 129
Human All Exon V4 Kit according to the manufacturer’s instructions. Briefly, genomic DNA was 130
fragmented with an ultrasonicator (Covaris) to obtain fragments that were primarily between 250 and 131
300 bp in length. These fragments were amplified using an NEBNext Ultra DNA Library Prep Kit (NEB) 132
and then hybridized with an Illumina TruSeq Capture Kit for enrichment. Each captured library was 133
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
7
loaded onto an Illumina HiSeq2000 platform, and paired-end sequencing was performed to the 134
desired median sequencing depth (>200×). Raw image files were processed using HCS1.4.8 for base 135
calling with the default parameters. 136
The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive 137
in BIG Data Center, Beijing Institute of Genomics (BIG), Chinese Academy of Sciences, under accession 138
number CRA000346, that are publicly accessible at http://bigd.big.ac.cn/gsa. 139
Somatic mutation calling 140
The sequencing reads were aligned to a human reference genome (hg19) using BWA(19). Mutect 141
(v1.16) was employed for somatic mutation calling. The following filtering criteria were applied to the 142
mutation profiles produced with Mutect to obtain highly probable somatic nucleotide variants (SNVs): 143
(i) only SNVs marked with a ‘keep’ status were considered; (ii) minimum depths of 10× and 14× were 144
required for the normal and tumor samples, respectively; and (iii) a minimum of four reads was 145
required to support the mutant allele. Small indels were identified using Pindel version 0.2.4(20) in 146
paired tumor-normal mode. Highly probable indels were identified using an in-house pipeline as 147
follows: i) all identified indels were required to have a minimum coverage of 20 reads in both the 148
tumor and normal samples; and ii) a minimum of 5 independent reads was required to support the 149
indel event in the tumor sample, with no evidence of such an event in the paired normal sample. The 150
qualified variants were annotated with oncotator(21). 151
Variant classification 152
To identify possible driver events, we classified the altered genes in this PMME according to reports in 153
the literature, as well as databases. Mutated genes were classified into one of six tiers: 1) Tier A: 154
cutaneous melanoma-related gene(2); 2) Tier B: probability score for oncogene (OG) or tumor 155
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
8
suppressor gene (TSG) > 0.75, as indicated by TUSON Explore (22); 3) Tier C: suggested TSG or OG by 156
Vogelstein et al(23); 4) Tier D: suggested significant cancer gene by Lawrence et al(24); 5) Tier E: 157
cancer-related gene in the Cancer Gene census catalog (http://cancer.sanger.ac.uk/census); and 6) Tier 158
F: recurrent gene mutation in mucosal melanoma(8). SIFT(25), PolyPhen-2(26), and 159
MutationAssessor(27) were used to predict the potential biological function changes of each missense 160
mutation. If any mutation received a rating of "Damaging" in SIFT, "Damage"/"Possible Damage" in 161
Polyphen-2, or "Medium"/"High" in MutationAssessor, it was considered to have a high functional 162
impact (HiFI). Nonsense and frameshift mutations, as well as splicing mutations, were also regarded as 163
HiFI alterations. Non-HiFI mutations were not considered to contribute to PMME tumorigenesis or 164
invasion. 165
Copy number alteration 166
Whole-genome SNP array for each samples was performed on Illumina Human OmniZhongHua 167
BeadChips (Illumina, San Diego, CA, USA). Raw intensity values were processed to obtain a normalized 168
B allele frequency (BAF) and a Log R ratio (LRR) for each probe using the Genome Studio Software 169
V2011.1. LRR values were segmented by the ASCAT algorithm(28). Tumor and matched normal 170
samples were used in pairs to obtain somatic alterations. 171
Inferring cancer cell fraction and clonal status 172
Somatic copy number alterations (SCNAs) were predicted with ADTEx(29) using both depth of 173
coverage ratios and B allele frequencies (BAFs) calculated from the WES data. Briefly, ADTEx consists 174
of two hidden Markov models (HMMs) to predict SCNAs and genotypes in WES data. One HMM is 175
used to predict CNAs, which, in combination with BAFs, can be used to estimate tumor ploidy and 176
predict absolute copy numbers. The second HMM is used to predict the zygosity or genotype of each 177
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
9
CNA segment. In this study, ADTEx v2.0 was run with the BAM and BAF inputs using the default 178
parameters. 179
ABSOLUTE (v1.0) was used to predict the tumor content, ploidy, and absolute copy number based on 180
the copy number profile and somatic SNVs for each tumor. The cancer cell fraction (CCF) of each 181
mutation was subsequently estimated according to the mutant VAFs after adjustment for the tumor 182
purities and local absolute copy number. Private, branch, trunk mutations were determined by the 183
extent of shared mutations among the different samples. 184
Phylogenetic tree construction 185
In principle, a phylogenetic tree was generated according to the numbers of shared mutations among 186
the samples. Of the sequenced samples, three were considered to be normal tissue samples (N1, N5, 187
and LN4) with few mutations, and another three (LN3, ME1, and N3) had a low tumor content (35%, 188
15%, and 17%, respectively). The four primary tumor samples (T1-T4) possessed a high tumor content 189
of over 80%, whereas the other two lymph node samples (LN1 and LN2) had a tumor cell proportion 190
of >50%. Utilizing low-tumor-content samples can introduce bias into the phylogenetic tree. Therefore, 191
we selected SNVs from six high-tumor-content samples to create the topological structure that 192
reflected the main evolutionary trajectory of this tumor. Then, we added the remaining samples to the 193
prototype tree according to the numbers of SNVs they shared with the existing taxon. Because ME1 194
and N3 had a relatively low tumor content, we were not able to estimate the precise divergence time 195
when they derived from the founding clone; thus, we suggested that they should be located at the 196
beginning of the trunk of the tree (Figure 1B). During tree construction, we also considered tumor 197
content, copy number changes, and intermixed clones that might have led to the misinterpretation 198
of our findings. By calculating branch/trunk lengths inferred from the mutation counts, the final trees 199
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
10
were drawn manually. 200
201
Results 202
Mutation profiles of primary mucosal melanoma 203
Whole exon-capture multiregion sequencing (WES) was performed on DNA from 12 samples derived 204
from esophagectomy specimens, including four samples obtained from primary tumor dissection 205
(T1-T4), three normal mucosal samples (N1, N3, and N5), one melanosis sample (ME1), and four 206
metastatic lymph node samples (LN1-LN4). Germline control DNA was extracted from the peripheral blood. 207
From 13 sequenced samples, high-coverage sequencing (mean, 200×) data were obtained. Overall, 208
386 nonsynonymous nucleotide substitutions and 134 indels (insertions and deletions) were detected 209
in at least one sample (Supplementary Table S1 and S2). Fifty single nucleotide variants (SNVs) were 210
randomly selected and validated in four primary tumor samples (T1-T4) by nucleotide mass spectrum 211
(Sequenom) The genotypes were successfully validated for 188 (94%) of the 200 (50 mutations per 212
sample) tested loci. Tumor purity of the primary tumor was estimated to be nearly 80% using 213
ABSOLUTE software according to the exome sequencing data (Table 1), indicating that our data were 214
sufficiently robust to detect subclonal mutations with low variant allele frequencies (VAFs). 215
To compare the differences in the mutation spectra between PMME and PCM, we acquired genomic 216
data for PCM from TCGA database (http://www.cbioportal.org/). A total of 281339 SNVs from 341 217
PCMs were used in this analysis. We also utilized WES data from 81 squamous esophageal carcinomas 218
(derived from unpublished data obtained in our laboratory) to compare similarities in mutational 219
signatures among ESCC, PCM, and PMME. As reported previously, the PCM genome exhibited a 220
specific predominance of C>T transitions at TpC dinucleotides, which were probably associated with 221
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
11
exposure to ultraviolet light. The ESCC genome was characterized by a mutational process associated 222
with an APOBEC-mediated and aging signature; thus, the mutations tended to be C>T or C>G 223
alterations. In this PMME genome, we found that 46.1% (178/386) of the somatic mutations were C>A 224
transitions, 41% (73/178) of which occurred in the CpCpA tri-nucleotide context (Figure 2A-B). 225
However, no similar signature was identified in the COSMIC data, suggesting that the CpCpA signature 226
found in this PMME genome may be caused by an unknown mutational process that needs to be 227
further investigated. In summary, although these three types of tumors shared similar anatomical or 228
pathological characteristics, they underwent distinct mutational processes, indicating that different 229
molecular mechanisms drive the tumorigenesis and progression of each tumor type. 230
Regional ploidy profiling and chromosomal aberration detection 231
SCNAs were investigated in this PMME. At the arm level, we observed copy number loss of 1p, 3p, 4p, 232
11p, 11q, and 14q and copy number gain of 1q, 6p, 8p, and 8q (Figure 3A). To identify focal 233
amplifications or deletions in these samples, we applied GISITC to determine copy numbers for these 234
13 regional samples and found 23 amplification and 21 deletion peaks across the genome, some of 235
which were associated with potential oncogenic genes, such as EIF3E, KRAS, and SOX5 and the tumor 236
suppressor genes PTEN and CDKN2A. Interestingly, we observed characteristics of chromothripsis in 237
chromosome 3p in the primary tumor samples (T1-T4) and two metastasis lymph nodes (LN1 and LN2) 238
(Figure 3B). These observations were based on the following principals: 1) only one chromosome was 239
affected; 2) distinctive oscillation between two copy number states arose across the affected regions; 240
and 3) within the regions with ‘higher’ copy number states, retention of heterozygosity was observed. 241
Intra-tumor heterogeneity and spatial separation of subclones 242
To assess ITH, we examined 235 nonsynonymous point mutations from the primary tumor, including 243
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
12
the T1-T4 regions, to determine the regional mutation distribution. Only 20% (47/235) of the 244
mutations were detectable across all four sampled tumor regions, whereas 54.5% (128/235) of the 245
mutations existed in a single region. Among all of the sequenced spatial samples, N1, N5, and LN4 246
were confirmed to be in the normal mucosal region due to their infrequent mutations, and N3 and 247
ME1 showed pre-tumor clonogenic potential because they both shared pathogenic mutations with 248
other samples of high tumor content, suggesting that they were clonally related. As the tumor content 249
was relatively low in these two samples, the number of SNVs was not as high as that in the primary 250
tumor regions. T1-T4 and LN1-LN3 were considered to be tumor regions, and we found an average of 251
136 mutations per sample, with the number of SNVs ranging from 98 to 188, in regions with a tumor 252
content of over 50% (Table 1). 253
To characterize the ITH and evolutionary trajectories in the PMEE genome, a phylogenetic tree was 254
constructed based on the extent of shared somatic mutations identified in each of the tumor 255
cell-infected regions (T1-T4, LN1-LN3, ME1, and N3), and this tree was used to determine the clonal 256
relatedness of the tumor subclones that co-existed in this PMME. The results showed that this PMME 257
tumor had a typical branched rather than linear tumor evolution. Subclones of the primary tumor and 258
other regions shared genetic mutations inherited from a common ancestor. Allopatric clones evolved 259
and followed distinct evolutionary trajectories. ME1 and N3 remained at the original phylogenetic 260
position and slowly developed alterations, whereas the T1-T4 sample region acquired the necessary 261
genetic alterations to form the primary tumor clone. The original tumor clone continued to progress 262
and diversified into two major branches: clade1 and clade2 (Figure 1B). Interestingly, we observed 263
that clade1 seemed to undergo a trunk mutation accumulation stage, during which time dozens of 264
mutations emerged in this branch, which then shaped the main clonal constituents in clade1 that 265
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
13
evolved into new subclones with metastatic ability. Notably, we found that the tumor sample T4 266
shared mutations with both clades, suggesting that the subclones had originated from clade1 and 267
clade2 and that they had broken the spatial barriers of the solid tumor and intermixed in the T4 region. 268
We further divided T4 into T4a and T4b according to their respective subclones, and we then 269
compared the VAFs of the shared mutations between the spatial samples in clade1 and clade2. Our 270
results showed that nearly 20% and 60% of the tumor subclones of T4 came from clade1 and clade2, 271
respectively. Additionally, the VAFs of the mutations in T4a and T4b were lower than those of the 272
mutations in the other regions in each clade (Supplementary Figure S3A), which further highlights 273
the fact that T4 was derived from two progenitor. 274
At the copy number level, this PMME also exhibited ITH. Copy number loss of 1p, 3p, 11p, 11q, and 275
14q and gain of 1q, 8p, and 8q were ubiquitous and conserved across all tumor clones, and these 276
SCNAs were considered to have occurred as early evolutionary events (Figure 3A&3C). Focal 277
amplification of SOX5 and EIF3E, as well as homogenous deletion of CDKN2A were in the trunk clones. 278
Telomere-bounded 1q and 2q amplification, internal 4q deletion, neutral LOH of 7p, chr13 arm level 279
deletion and 12q gain were specific occurred in clade1; telomere-bounded 4q deletion, 7p 280
amplification, chr10 arm level deletion (Figure S3B), and chr18 arm level deletion were only identified 281
in clade2 (Figure 3A & Supplementary Figure S4A-C). Furthermore, we found that clade-specific SCNAs 282
intermixed in T4, such as deletions involving 4q, telomere-bounded amplification of 2q, chromosome 283
10 and 18 deletion, further confirming that T4 is intermixed clones from two different clades. For 284
example, 10q deletion involving PTEN (chr10q23.31) was detected in clade2 but absence in clade1, 285
whereas PTEN deletion of T4 showed a lower VAF (Supplementary Figure S3B). The phylogenetic 286
relationships between each of the spatial regions reflected by the SCNAs were highly consistent with 287
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
14
the mutation tree. 288
Potential parallel evolution was observed in this PMME at copy number level although no evidence of 289
parallel evolution was found at the mutational level. A mirrored subclonal allelic imbalance was 290
observed in chromosome 4. If the maternal allele is gained or lost in a subclone in one region, yet the 291
paternal allele is gained or lost in a different subclone in another region, it will result in a mirrored 292
subclonal allelic imbalance profile(30). Different SCNAs of chr4 deletion which affected maternal allele 293
of chr4:82-143M and paternal allele of chr4:93M-telomere were observed in clade1 and clade2 294
respectively (Figure 4), suggesting parallel evolution in branched stage. Moreover, even inside the 295
independent clades, such chromosomal instability persisted. In clade1, 27% subclones of T1 296
obtained a privately chr4 arm-level deletion, which coexisted with the clade1 common chr4 aberration 297
(chr4:82-143M); in clade2, 46% tumor subclones of T2 shared clade specific chr4 loss (93M-telomere) 298
with T3 in paternal allele, whereas the remaining 54% subclones of T2 carried chr4:103M-telomere 299
loss in maternal allele (Figure 4). 300
301
Mutation spectra of truncal and branch mutations 302
To further explore the dynamics of the mutational processes shaping this PMME genome over time, 303
the temporal spectra of the somatic mutations in the different evolutionary stages were analyzed. The 304
PMME genome exhibited a significant increase in the proportion of C>A transversions in branch 305
mutations compared with trunk mutations (P = 0.006) (Figure 2B), which was accompanied by a 306
decrease in C>T transitions. The private mutations in ME1 and N3 presented with similar signatures as 307
the founder mutation (Figure 2B). 308
Identification of truncal and branch driver mutations 309
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
15
To determine whether specific driver genes carried predominantly trunk or branch mutations, we 310
identified potential driver mutations according to the known cancer gene catalogs. A total of 22 311
altered genes were present in the catalogs, and each of them carried a single mutation, except CDH13 312
(Figure 4). Among the 26 potential driver mutations identified, 17 were predicted to be pathogenic by 313
SIFT/Polyphen2/MutationAssessor. Potential functional trunk mutations in BRAF, KRAS, DNAH7, and 314
CDH13 were detected in all of the primary tumor regions; therefore, these mutations were considered 315
to have occurred during the early stages of tumorigenesis. In addition, branch mutations in JAK, 316
ARHGEF4, and UMOD were observed. The JAK mutations belonged to clade2, and the mutations in the 317
other two genes were specific to clade1. Mutations in PIK3CA, GPRASP1, and ARHGAP26 occurred 318
exclusively in T1 and lymph node metastatic regions, whereas their cancer cell fraction (CCF) values 319
were increased in the metastatic regions, which suggested that the tumor cell populations that 320
harbored these mutations might have experienced subclonal expansion after exposure to the 321
metastatic lymph node environment. Activating mutations in TRPS1, MAPK3K13, AHNAK, DNAH3, 322
PCDHGA3, PRUNE2, and TAS1R2 were private and distributed throughout different spatial regions 323
(Figure 1B). However, their CCFs were much lower than those of the founding and branch mutations, 324
as they appeared to be intra-regional subclonal mutations (Figure 5). 325
Discussion 326
The genomic landscape of cutaneous melanoma has been well described through the use of 327
next-generation sequencing (NGS). New diagnostic and prognostic biomarkers, as well as therapeutic 328
approaches, are being developed based on known germline and somatic genetic alterations in PCM. 329
These advances will improve the treatment of PCM patients. However, the genetic nature of PMM 330
still has not been elucidated; thus, patients with PMM, which is an important subtype of melanoma, 331
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
16
may not benefit from the advances facilitated by the new technologies. Therefore, sequencing efforts 332
must be continued, and particular attention should be paid to rare melanoma subtypes to 333
comprehensively characterize the somatic mutational landscape of melanoma. In this study, we 334
sequenced a rare primary esophageal melanoma using a multiregion exome sequencing strategy to 335
characterize the genetic profile, ITH, and clonal evolution of this PMME. 336
Our data revealed that the four primary tumor samples harbored a total of 110 coding region nsSNVs, 337
whereas each tumor site carried an average of 55 nsSNVs (T1: 67, T2: 48, T3: 38, and T4: 67). The real 338
mutational burden was underestimated by one-fold when only one tumor sample was sequenced, 339
which suggests that high ITH existed in this PMME genome. The high somatic mutational burden in 340
melanoma has been previously linked to favorable outcomes in immune-based therapeutic 341
approaches(31); thus, it is necessary to consider the influence of ITH in the PMME genome when 342
evaluating mutational burden, particularly in clinical practice. 343
To better interpret the genomic ITH and tumor evolutionary history in this PMME, we selected 23 344
important mutations from among all of the altered genes as representative genetic evolutionary 345
events. The candidate driver genes for mucosal melanoma are currently unclear; thus, we employed 346
pan-cancer and cutaneous melanoma catalogs, as well as data from nine reported mucosal melanoma 347
genomes, to predict putative driver genes. Genomic ITH of this PMME was observed for both somatic 348
mutations and chromosomal imbalances. Among the 22 selected putative driver mutations, only four 349
were clonal events; the other mutations, including a canonical hotspot mutation, PIK3CAE545A, were 350
restricted to tumor subclones.In copy number level, ITH was also significant. Some SCNVs involving 351
important cancer genes such as homogeneous deletion of CDKN2A, and high level amplification of 352
KRAS and SOX5 (Figure 3C) were ubiquitous across the clade 1 and clade2.Each evolution braches had 353
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
17
its own specific SCNAs, such as PTEN focal deletion that was exclusively occurred in clade2 and chr13 354
arm-level deletion that only apperaed in clade1.The ITH status could have important implications for 355
targeted therapeutic strategies. Strikingly, we found that mutations in genes involved in the 356
PI3K-AKT-mTOR pathway, such as PIK3CA, tended to be present in subclones, and some reports have 357
suggested that PIK3CA mutations often lead to subclonal expansion, whereas mutations in genes 358
associated with RAS-MEK or cyclin-dependent kinase often result in founding clones. These results are 359
consistent with previous research(32). Therefore, our findings suggest that the current methods 360
utilized to identify cancer genes are likely biased to detect clonal drivers that occur at a higher 361
frequency in single samples; thus, it is necessary to use a more efficient strategy to obtain a full list of 362
clinically actionable genes, such as collecting spatial samples from the same individual and pooling 363
DNA for further analysis. 364
A phylogenetic tree was constructed for this PMME by analyzing the genetic variants identified in 365
tissues from each of the geographically separated regions. A branching evolution model was proposed, 366
as two of the tumor clades exhibited subclonal differentiation. Genetic diversity was greater in the T4 367
region than at the other sites; it seemed that T4 comprised two distinct subclones (T4a and T4b), 368
which belonged to clade1 and clade2, respectively. By comparison of VAFs of T4 mutations and 369
SCNAs with other primary tumor samples, it was very obviously that branch specific mutations or 370
SCNAs showed lower VAF value in T4 than others (Supplementary Figure S3A-B). These results 371
indicated that the subclones of clade1 and clade2 converged in T4 but that competitive subclones 372
with selective advantages had not yet adapted in that region. Recently, a common parallel evolution 373
pattern that involved chromosome 4 instability was suggested by Jamal‑Hanjani et al. in lung 374
cancer(30). In this study, we also observed chromosome 4 instability which performed diverse forms 375
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
18
in evolution branches (Figure 4). Ongoing dynamic chromosomal instability were associated with 376
intratumor heterogeneity and resulted in parallel evolution of driver somatic copy-number alterations, 377
which may possibly associated with poor survival in cancer patients(30). 378
Early (truncal) mutations likely occur either before or during tumor initiation or during early 379
development, whereas late (branch) mutations reflect mutational processes that shape the genome 380
during tumor maintenance and progression(33). By focusing on putative driver events, we found that 381
critical trunk mutations tended to be clonal, whereas branch mutations tended to be present in 382
subclones (Figure 4 & Supplementary Figure S5). Among the four trunk mutations, BRAF and KRAS, 383
particularly BRAF, are notorious cancer-promoting mutations in various cancers, and they account for 384
25% of melanoma cases in the Chinese population(34). We did not observe a hot spot mutation in 385
BRAFV600E in this case, but another pathogenic mutation, BRAFH574Q, was detected, which may also play 386
a significant role in tumorigenesis. KRAS gene mutations are rarely observed in cutaneous melanoma. 387
However, the hot spot mutation KRASK12S has been reported in one PMME(17). Notably, trunk 388
mutations in CDH13 and DNAH7 exhibited similar evolutionary patterns as those in BRAF and KRAS, 389
indicating that they might also function during tumor initiation. CDH13 encodes a member of the 390
cadherin superfamily, and it is hypermethylated in many types of cancers(35-37). DNAH7 is a 391
component of the inner dynein arm of ciliary axonemes and drives the sliding of microtubules, which 392
is associated with papillary craniopharyngioma(38). Thus, we suggest that CDH13 and DNAH7 might 393
also be critical during initial tumorigenesis and that more attention should be paid to these genes. 394
However, more biological evidence is needed. 395
From the phylogenetic topology, we were also able to infer the possible evolutionary trajectories of 396
the tumor metastases. Among the four tested lymph nodes, three were found to be infiltrated by 397
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
19
metastatic tumor cells. All of the metastatic tumor subclones were nested in clade1, and T1 was their 398
most recent common ancestor. Metastatic subclones containing pathogenic mutations in ARHGAP26 399
and PIK3CA, as well as Chr13 arm-level deletion experienced subclonal expansion in metastatic 400
environments (Supplementary Figure S6). The subclone that carried ARHGEF4 and UMOD gene 401
mutations was detected in T1, T4, LN1, and LN2, and it maintained subclonal dominance in LN1 and 402
LN2. We employed TCGA data to test whether above PMME metastatic molecular changes are also 403
metastatic features in other melanoma (supplementary file, Table S3& Figure S7). It is noticeable that 404
chr13 arm-level deletion is significantly occurred in metastasis TCGA samples (Figure S7), which 405
suggested that this chromosome aberration could be a common metastatic marker in melanoma 406
regardless the sample type. The microenvironment in the lymph nodes might have facilitated positive 407
selection for metastatic tumor cells with such mutations, providing them with growth advantages. 408
Spatially separated lymph nodes had similar subclones and belonged to the same monophyletic group, 409
although most VAFs in the LN3 region were very low. Our results highlight the fact that spatially 410
segregated metastatic subclones can follow very similar evolutionary trajectories. 411
Another interesting finding was that two of the “normal” mucosal samples were related with the 412
primary tumor clone and may derived in the very early stage of tumorigenesis. ME1 and N3 shared a 413
certain number of SNVs with the primary tumors. A total of 41.3% of the SNVs (19/46) in N3 were 414
shared with the tumor clones, and 36.2% (29/80) of those in ME1 were also present in the primary 415
tumor (Supplementary Figure S8). Nevertheless, no significant gene mutations were identified in ME1 416
or N3. Intensive SCNA analysis could further confirm this relationships. Since ME1 (CCF=11%) and N3 417
(CCF=7%) had extremely low tumor cell content, only SCNAs with BAF> 0.535 based on ASCAT 418
algorithm (Figure S9) could be detected. Two chromosome aberrations which located in chr7q and 419
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
20
chr8 (Figure 3A) were identified in ME1, and chr8 amplification could also be observed in N3. Chr7q 420
and Chr8q amplification were trunk SCNAs and across both clade1 and clade2 samples, which 421
indicated that N3 and ME1 had a common ancestor with the primary tumor. Moreover, chr15 422
(60M-telomere) LOH was absence in ME1 or N3, which was another ubiquitous SCNA and with a high 423
detection threshold (BAF> 0.535). These results indicated that ME1 and N3 diverged from other tumor 424
samples in very early stage. It is reported that PMME has a tendency for vertical growth within the 425
esophageal wall, and has a tendency to disseminate by the submucosal lymphatic route(15). In this 426
case, ME1, N3 and the primary tumor distributed vertically along the esophageal wall (Supplementary 427
Figure S1), so it was most probably that early metastatic clade3 (ME1 and N3) was formed by 428
dissemination through the submucosal lymphatic route. Findings of this study illustrated the 429
mechanisms of early intramural metastasis, which might be a common cause for bad survival in cancer 430
patients(30). This finding prompt the clinicians pay more attention to improve the surgical strategy of 431
PMME to decrease the risk of recurrence or metastasis. Technically, it also warned that using “normal” 432
tissues as germline control is risky for missing some important molecular aberrations. 433
In this PMME, we observed significant shifts in the mutational signatures during different tumor 434
progression stages. The decreased proportion of C>T mutations was accompanied by an increase in 435
C>A mutations over time. In skin melanoma, C>T is the predominant nucleotide base substitution that 436
may be caused by the misrepair of ultraviolet-induced covalent bonds between adjacent 437
pyrimidines(39). C>T mutations may also be caused by another inductive event, such as the 438
spontaneous deamination of 5-methyl-cytosine, which predominantly occurs at NpCpG 439
trinucleotides(40). We observed a relatively high proportion of C>T transitions at CpG sites in the 440
trunk of the phylogenetic tree; however, within the branches, C>A changes became the most common 441
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
21
alteration (Table 1 and Figure 2B). Some studies have suggested that smoking may be a risk factor for 442
C>A substitutions(24,41). However, this patient was not a smoker, so it is unlikely that tobacco caused 443
damage to the genomic DNA. We presume that specific processes promoted the generation of genetic 444
diversity in this PMME and that the spontaneous deamination mutational process might have played 445
an important role in founding clone formation; subsequently, some unknown mutational processes 446
resulted in C>A alterations and eventually promoted subsequent subclonal expansions. Patterns of 447
mutation spectrum could also confirmed our presumption that metastatic event of ME1 and N3 448
started at early stage, because the mutation spectrum of ME1 and N3 were as similar as the founding 449
clone. 450
As in any case study, the present study has notable limitations. It involves only one patient, and thus 451
further studies are needed to determine whether the principles discovered here apply to other 452
patients. Despite such limitations, this case provides evidence for tumor evolution models and tumor 453
cell disseminate route in molecular level, which to our knowledge has not yet been demonstrated in 454
patients with PMME. 455
In conclusion, by sequencing multiregional samples obtained from a PMME patient, we determined 456
the evolutionary history of this tumor. The presence of ITH and regionally separated driver events, in 457
addition to the relentless and dynamic nature of genomic instability processes, were observed in this 458
case. These findings could help us better understand the processes of PMME tumorigenesis, 459
maintenance, and metastasis. Although large-scale sequencing efforts must be continued for mucosal 460
melanoma and PMME to comprehensively characterize their genomic landscapes, our results have 461
highlighted the importance of viewing tumor development as a dynamic evolutionary process. We 462
have revealed valuable genetic information that contributes to the understanding of the biological 463
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
22
mechanism of PMME and that also provides insights that may facilitate the development of 464
biomarkers and therapeutic strategies. 465
466
Acknowledgement: 467
The authors thank the patient and her family to participate this study. 468
469
470
References 471
1. Mihajlovic M, Vlajkovic S, Jovanovic P, Stefanovic V. Primary mucosal melanomas: a 472
comprehensive review. International journal of clinical and experimental pathology 473
2012;5:739-53 474
2. Zhang T, Dutton-Regester K, Brown KM, Hayward NK. The genomic landscape of cutaneous 475
melanoma. Pigment cell & melanoma research 2016 476
3. Shain AH, Yeh I, Kovalyshyn I, Sriharan A, Talevich E, Gagnon A, et al. The Genetic Evolution of 477
Melanoma from Precursor Lesions. N Engl J Med 2015;373:1926-36 478
4. Wong SQ, Behren A, Mar VJ, Woods K, Li J, Martin C, et al. Whole exome sequencing 479
identifies a recurrent RQCD1 P131L mutation in cutaneous melanoma. Oncotarget 480
2015;6:1115-27 481
5. Shain AH, Garrido M, Botton T, Talevich E, Yeh I, Sanborn JZ, et al. Exome sequencing of 482
desmoplastic melanoma identifies recurrent NFKBIE promoter mutations and diverse 483
activating mutations in the MAPK pathway. Nat Genet 2015;47:1194-9 484
6. Krauthammer M, Kong Y, Bacchiocchi A, Evans P, Pornputtapong N, Wu C, et al. Exome 485
sequencing identifies recurrent mutations in NF1 and RASopathy genes in sun-exposed 486
melanomas. Nat Genet 2015;47:996-1002 487
7. Arafeh R, Qutob N, Emmanuel R, Keren-Paz A, Madore J, Elkahloun A, et al. Recurrent 488
inactivating RASA2 mutations in melanoma. Nat Genet 2015;47:1408-10 489
8. Furney SJ, Turajlic S, Stamp G, Nohadani M, Carlisle A, Thomas JM, et al. Genome sequencing 490
of mucosal melanomas reveals that they are driven by distinct mechanisms from cutaneous 491
melanoma. J Pathol 2013;230:261-9 492
9. Curtin JA, Fridlyand J, Kageshita T, Patel HN, Busam KJ, Kutzner H, et al. Distinct sets of 493
genetic alterations in melanoma. N Engl J Med 2005;353:2135-47 494
10. Omholt K, Grafstrom E, Kanter-Lewensohn L, Hansson J, Ragnarsson-Olding BK. KIT pathway 495
alterations in mucosal melanomas of the vulva and other sites. Clin Cancer Res 496
2011;17:3933-42 497
11. Sekine S, Nakanishi Y, Ogawa R, Kouda S, Kanai Y. Esophageal melanomas harbor frequent 498
NRAS mutations unlike melanomas of other mucosal sites. Virchows Arch 2009;454:513-7 499
12. Spencer KR, Mehnert JM. Mucosal Melanoma: Epidemiology, Biology and Treatment. Cancer 500
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
23
Treat Res 2016;167:295-320 501
13. Jiang W, Zou Z, Liu B. Primary malignant melanoma of the esophagus: A case report and 502
review of the literature. Oncology letters 2015;9:2036-40 503
14. Zhou YB, Yuan Y, Hu B, Che GW. Primary Multifocal Malignant Melanoma of Esophagus 504
Co-Occurs With Esophagogastric Junction Adenocarcinoma. Am J Gastroenterol 2016;111:312 505
15. Bisceglia M, Perri F, Tucci A, Tardio M, Panniello G, Vita G, et al. Primary malignant melanoma 506
of the esophagus: a clinicopathologic study of a case with comprehensive literature review. 507
Advances in anatomic pathology 2011;18:235-52 508
16. Sanlorenzo M, Ribero S, Osella Abate S, Mariani S, Strignano P, Salizzoni M, et al. Genetic 509
mutations in primary malignant melanoma of the esophagus: case report and literature 510
review. Giornale italiano di dermatologia e venereologia : organo ufficiale, Societa italiana di 511
dermatologia e sifilografia 2015 512
17. Langer R, Becker K, Feith M, Friess H, Hofler H, Keller G. Genetic aberrations in primary 513
esophageal melanomas: molecular analysis of c-KIT, PDGFR, KRAS, NRAS and BRAF in a series 514
of 10 cases. Mod Pathol 2011;24:495-501 515
18. Schuchter LM, Green R, Fraker D. Primary and metastatic diseases in malignant melanoma of 516
the gastrointestinal tract. Curr Opin Oncol 2000;12:181-5 517
19. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. 518
Bioinformatics 2009;25:1754-60 519
20. Ye K, Schulz MH, Long Q, Apweiler R, Ning Z. Pindel: a pattern growth approach to detect 520
break points of large deletions and medium sized insertions from paired-end short reads. 521
Bioinformatics 2009;25:2865-71 522
21. Ramos AH, Lichtenstein L, Gupta M, Lawrence MS, Pugh TJ, Saksena G, et al. Oncotator: 523
cancer variant annotation tool. Human mutation 2015;36:E2423-9 524
22. Davoli T, Xu AW, Mengwasser KE, Sack LM, Yoon JC, Park PJ, et al. Cumulative 525
haploinsufficiency and triplosensitivity drive aneuploidy patterns and shape the cancer 526
genome. Cell 2013;155:948-62 527
23. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Jr., Kinzler KW. Cancer genome 528
landscapes. Science 2013;339:1546-58 529
24. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, Golub TR, et al. Discovery 530
and saturation analysis of cancer genes across 21 tumour types. Nature 2014;505:495-501 531
25. Vaser R, Adusumalli S, Leng SN, Sikic M, Ng PC. SIFT missense predictions for genomes. 532
Nature protocols 2016;11:1-9 533
26. Adzhubei I, Jordan DM, Sunyaev SR. Predicting functional effect of human missense 534
mutations using PolyPhen-2. Current protocols in human genetics / editorial board, Jonathan 535
L Haines [et al] 2013;Chapter 7:Unit7 20 536
27. Reva B, Antipin Y, Sander C. Predicting the functional impact of protein mutations: application 537
to cancer genomics. Nucleic Acids Res 2011;39:e118 538
28. Van Loo P, Nordgard SH, Lingjaerde OC, Russnes HG, Rye IH, Sun W, et al. Allele-specific copy 539
number analysis of tumors. Proc Natl Acad Sci U S A 2010;107:16910-5 540
29. Amarasinghe KC, Li J, Hunter SM, Ryland GL, Cowin PA, Campbell IG, et al. Inferring copy 541
number and genotype in tumour exome data. BMC genomics 2014;15:732 542
30. Jamal-Hanjani M, Wilson GA, McGranahan N, Birkbak NJ, Watkins TBK, Veeriah S, et al. 543
Tracking the Evolution of Non-Small-Cell Lung Cancer. N Engl J Med 2017 544
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
24
31. Van Allen EM, Miao D, Schilling B, Shukla SA, Blank C, Zimmer L, et al. Genomic correlates of 545
response to CTLA-4 blockade in metastatic melanoma. Science 2015;350:207-11 546
32. McGranahan N, Favero F, de Bruin EC, Birkbak NJ, Szallasi Z, Swanton C. Clonal status of 547
actionable driver events and the timing of mutational processes in cancer evolution. Science 548
translational medicine 2015;7:283ra54 549
33. Gerlinger M, Horswell S, Larkin J, Rowan AJ, Salm MP, Varela I, et al. Genomic architecture 550
and evolution of clear cell renal cell carcinomas defined by multiregion sequencing. Nat 551
Genet 2014;46:225-33 552
34. Si L, Kong Y, Xu X, Flaherty KT, Sheng X, Cui C, et al. Prevalence of BRAF V600E mutation in 553
Chinese melanoma patients: large scale analysis of BRAF and NRAS mutations in a 432-case 554
cohort. Eur J Cancer 2012;48:94-100 555
35. Sato M, Mori Y, Sakurada A, Fujimura S, Horii A. The H-cadherin (CDH13) gene is inactivated in 556
human lung cancer. Hum Genet 1998;103:96-101 557
36. Widschwendter A, Ivarsson L, Blassnig A, Muller HM, Fiegl H, Wiedemair A, et al. CDH1 and 558
CDH13 methylation in serum is an independent prognostic marker in cervical cancer patients. 559
Int J Cancer 2004;109:163-6 560
37. Moelans CB, Verschuur-Maes AH, van Diest PJ. Frequent promoter hypermethylation of 561
BRCA2, CDH13, MSH6, PAX5, PAX6 and WT1 in ductal carcinoma in situ and invasive breast 562
cancer. J Pathol 2011;225:222-31 563
38. Jones RT, Abedalthagafi MS, Brahmandam M, Greenfield EA, Hoang MP, Louis DN, et al. 564
Cross-reactivity of the BRAF VE1 antibody with epitopes in axonemal dyneins leads to 565
staining of cilia. Mod Pathol 2015;28:596-606 566
39. Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, et al. Mutational 567
heterogeneity in cancer and the search for new cancer-associated genes. Nature 568
2013;499:214-8 569
40. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, Behjati S, Biankin AV, et al. Signatures of 570
mutational processes in human cancer. Nature 2013;500:415-21 571
41. Axell T, Hedin CA. Epidemiologic study of excessive oral melanin pigmentation with special 572
reference to the influence of tobacco habits. Scand J Dent Res 1982;90:434-42 573
574
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
25
Table 1. Summary of the exome sequencing information for a multiregion-sequenced PMME. 575
Region Origin Tumor purity
SNVs nsSNVs Subclonal
mutations* C>T
transitions CG>TG
alterations C>A
transitions CC>AA
alterations Chromothripsis
T1 Primary tumor 0.85 149 122 0.61 0.36 0.52 0.40 0.47 3p T2 Primary tumor 0.77 136 108 0.56 0.32 0.57 0.38 0.41 3p T3 Primary tumor 0.79 98 79 0.41 0.38 0.59 0.30 0.34 3p T4 Primary tumor 0.77 188 150 0.69 0.28 0.55 0.49 0.43 3p
LN1 Lymph node 0.52 140 116 NA 0.38 0.53 0.35 0.49 3p
LN2 Lymph node 0.59 138 114 NA 0.38 0.53 0.36 0.48 3p
LN3 Lymph node 0.36 104 85 NA 0.39 0.54 0.30 0.48 NA
LN4 Lymph node NA 5 5 NA 0.40 0.50 0.60 0.33 NA
ME1 Mucosal
melanosis 0.11 70 54 NA 0.47 0.61
0.23 0.31 NA
N1 Normal mucosa 0.07 7 6 NA 0.57 0.50 0.00 0.00 NA
N3 Normal mucosa NA 30 27 NA 0.47 0.57 0.23 0.43 NA
N5 Normal mucosa NA 4 3 NA 0.33 0.00 0.25 0.00 NA
* Percentage of subclonal mutations in each sample. 576 nsSNV: nonsynonymous single nucleotide variant. 577 NA: Data not available. 578
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
26
Figure legends 579
Figure 1. Geographical sampling approach and evolution tree of PMME. (A) Geographic distribution 580
of multiregional samples obtained from the dissection of esophagectomy and metastasectomy PMME 581
specimens. (B) Phylogenetic trees were generated by the clustering of somatic mutations identified in 582
multiregional PMME samples. Relative branch lengths were determined by the proportion of 583
mutations in each branch. The most recent common ancestor of ME1 and N3 could not be inferred 584
precisely due to their low tumor content; therefore, they were placed at the distal end of the trunk. 585
T4a and T4b were parallel subclones in the T4 region. Chr4del* and chr4del# represents branch-specific 586
deletions Chr4: (82-143M) deletion and chr4: (93M to telomere) deletion in clade1 and clade2., 587
respectively. 588
589
Figure 2. Mutation spectrum of PMME. (A) Mutation spectrum for nsSNVs in 81 ESCC, 341 PCM, and 1 590
PMME sample. The numbers of SNVs and indels in spatial PMME samples were determined. (B) 591
Temporal and spatial dissection of the mutation spectrum in PMME samples. The proportion of C>A 592
transitions increased with tumor progression, accompanied by a decrease in C>T substitutions. Gray 593
indicates insufficient mutations for analysis. 594
595
Figure 3. Copy number alterations in PMME. (A) BAF (B allele frequencies) profile across the genome 596
of PMME. Chr13 arm level deletion is metastatic specific which only occurred in clade1 (Green 597
rectangle). SCNAr. (B) Chromothripsis was observed in chromosome 3. (C) High-level focal 598
amplification was detected in chromosome 12. The SOX5 and KRAS genes are located in this amplified 599
region. 600
Figure 4. Diverse evolutionary patterns within chromosome 4. Copy number analyses reveal diverse 601
evolutionary patterns within chromosome 4 during tumor progression. Blue rectangle represents 602
branch-specific deletion Chr4: 82-143M deletion in clade1 and Chr4: 93M to telomere deletion in 603
clade2; gray and green rectangle represent chr4:93M-telomere loss and chr4:103M-telomere loss, 604
which account for 46% and 54% tumor cells of T2 respectively. 605
Figure 5. Heatmap showing the CCF of putative driver genes in each spatial region in PMME. 606
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
27
Putative driver genes were classified into six tiers. CM: cutaneous melanoma-related genes; PMM: 607
recurrent gene mutation in mucosal melanoma. TUSON: suggested cancer gene by TUSON Explore; Vo: 608
suggested cancer gene by Vogelstein et al; Lau: suggested significant cancer gene by Lawrence et al; 609
and Consensus: cancer-related gene in the Cancer Gene census catalog. Genes marked with an 610
asterisk indicate that the mutation in the gene was predicted to be benign with SIFT, PolyPhen-2, or 611
MutationAssessor. For each region, red or yellow indicates the presence of mutations, whereas blue 612
indicates the absence of mutations. The values in the squares are the CCF (red) or VAF (yellow) for 613
each specific SNV. The VAF was used instead of the CCF when the CCF could not be estimated due to 614
low tumor purity or conflict with the SCNA data. 615
616
617
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938
Published OnlineFirst September 29, 2017.Cancer Res Jingjing Li, Shi Yan, Zhen Liu, et al. esophagusclonal dynamics in primary malignant melanoma of the Multiregional sequencing reveals genomic alterations and
Updated version
10.1158/0008-5472.CAN-17-0938doi:
Access the most recent version of this article at:
Material
Supplementary
http://cancerres.aacrjournals.org/content/suppl/2017/09/29/0008-5472.CAN-17-0938.DC1
Access the most recent supplemental material at:
Manuscript
Authorbeen edited. Author manuscripts have been peer reviewed and accepted for publication but have not yet
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)
.http://cancerres.aacrjournals.org/content/early/2017/09/29/0008-5472.CAN-17-0938To request permission to re-use all or part of this article, use this link
Research. on January 28, 2020. © 2017 American Association for Cancercancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on September 29, 2017; DOI: 10.1158/0008-5472.CAN-17-0938