Académique Documents
Professionnel Documents
Culture Documents
Original Investigation
IMPORTANCE Recurrent and/or metastatic head and neck cancer is usually incurable.
Supplemental content at
jamaoncology.com
Implementation of precision oncology for these patients has been limited by incomplete
understanding of the molecular alterations underlying advanced disease. At the same time,
the molecular profiles of many rare head and neck cancer types are unknown. These
significant gaps in knowledge need to be addressed to rationally devise new therapies.
OBJECTIVE To illuminate the distinct biology of recurrent and metastatic head and neck
cancers and review implementation of precision oncology for patients with advanced disease.
DESIGN, SETTING, AND PARTICIPANTS After exclusions, 151 patients with advanced,
treatment-resistant head and neck tumors, including squamous cell carcinoma (HNSCC),
adenoid cystic carcinoma (ACC), and other salivary and cutaneous cancers, whose tumors were
sequenced between January 2014 and July 2015 at Memorial Sloan Kettering were recruited.
Next-generation sequencing of tumors as part of clinical care included high-depth (median
600) exonic coverage of 410 cancer genes and whole-genome copy number analysis.
INTERVENTIONS Next-generation sequencing of tumors and matched normal DNA.
MAIN OUTCOMES AND MEASURES Feasibility, the frequency of actionable molecular
alterations, the effect on decision making, and identification of alterations associated with
recurrent and metastatic disease.
RESULTS Overall, 151 patients (95 men and 56 women; mean [range] age, 61.8 [17-100] years)
were included in the study. Next-generation sequencing ultimately guided therapy in 21 of 151
patients (14%) (13 of 53 [25%] of patients with HNSCC) by refining diagnoses and matching
patients to specific therapies, in some cases with dramatic responses on basket studies.
Molecular alterations were potentially actionable in 28 of 135 patients (21%). The genetic profiles
of recurrent and metastatic tumors were often distinct from primary tumors. Compared to
primary human papillomavirus (HPV)-positive tumors, many recurrent and metastatic HPVpositive tumors exhibited a molecular profile more similar to HPV-negative tumors, including
enriched frequencies of TP53 mutation (3 of 20 tumors [15%]), whole genome duplication
(5 of 20 tumors [25%]), and 3p deletion (11 of 20 tumors [55%]). There were high rates of TERT
promoter mutation in recurrent and metastatic HPV-negative HNSCC (13 of 30 tumors [43%]),
cutaneous SCC (11 of 21 tumors [52%]), basal cell carcinoma (3 of 4 tumors [75%]), and ACC (5 of
36 tumors [14%]). Activating NOTCH1 mutations were enriched in metastatic ACCs (8 of 36
tumors [22%]).
CONCLUSIONS AND RELEVANCE These findings reveal the molecular landscape of advanced
disease and rare cancer subtypes, both predominant challenges in head and neck oncology.
To understand the repertoire of targetable alterations in advanced cancers, it is necessary to
sequence recurrent and metastatic tumors. These data are important first steps toward
implementation of precision head and neck oncology.
JAMA Oncol. doi:10.1001/jamaoncol.2016.1790
Published online July 21, 2016.
(Reprinted) E1
Key Points
Question Can clinical implementation of next-generation
sequencing of tumors help guide treatment and expand our
understanding of the biology of advanced head and neck cancer?
Findings Molecular profiling of advanced (nearly all recurrent and
metastatic) head and neck cancers was able to identify actionable
alterations and guide treatment in a portion of patients, refine
diagnoses in certain rare cancers, and discover unique genetic
findings that are enriched in advanced cancers.
Meaning Implementation of precision oncology is beginning to
guide treatment and improve our knowledge of molecular
alterations relevant to advanced, treatment-resistant head and
neck cancers.
Results
Methods
The MSK-IMPACT (Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets) assay is a targeted next-generation sequencing (NGS) assay approved
through Clinical Laboratory Improvement Amendments.11 The
assay is optimized for DNA from formalin-fixed, paraffinembedded samples, and targets single nucleotide variants, indels, and structural variants in 410 genes functionally relevant to cancer or clinically actionable targets (eTable 1 in
Supplement 2), and genome-wide copy number.12,13
After receiving institutional review board approval from
the Memorial Sloan Kettering Cancer Center, we reviewed patients treated by our multidisciplinary head and neck oncology team whose tumors were sequenced after informed written consent to an institutional review boardapproved protocol
(NCT01775072). Of 224 patients enrolled, 151 with minimum
6-month follow-up, whose tumors were sequenced between
January 2014 and July 2015, were included. Treatment(s)
initiated after NGS were considered guided by molecular data
if results led to, or justified continuation with, specific
treatments. Molecular alterations were classified by evidence
indicating standard or investigational therapies (eTable 2 in
Supplement 2).
Copy number analyses with FACETS14 were performed to
identify whole-genome duplication, 3p deletion (eFigures 1 and
2 in Supplement 1), and subclonal population structure.
E2
Actionable Alterations
There are no standard-of-care biomarkers in head and neck
cancer, but 28 of 135 patients (21%) with tumors harboring alterations had findings with potential investigational implications7 patients (5%) with a US Food and Drug Administration (FDA)-approved biomarker and drug in another indication,
and 21 patients (16%) with clinical evidence linking the alteration to drug response (Figure 1) (eTable 2 in Supplement 2).
Ultimately, next-generation sequencing data guided therapy
in 21 of 151 patients (14%).
Landscape of HNSCC
The most frequent single nucleotide variants and copy number variants in HNSCC are depicted in Figure 2 and eTable 4
jamaoncology.com
HNSCC
All sites
1.0
1.0
ACC (94.1%)
NPC (83.3%)
0.8
0.6
Overall Survival
Overall Survival
0.8
HPV+ (70.9%)
0.4
HPV (60.5%)
0.2
0.4
0.2
P=.06
P=.58
0
0
0
18
26
9
14
6
10
2
7
No. at risk
ACC
NPC
Skin SCC
Other salivary
HNSCC
1
6
36
7
19
18
51
33
7
14
13
44
30
4
8
9
23
0.8
0.8
HPV+ (50.5%)
0.4
HPV (29.4%)
0.2
25
2
7
7
14
21
2
6
4
9
18
0
5
3
7
ACC (83.7%)
NPC (80.0%)
Cutaneous SCC (76.2%)
Other salivary (67.5%)
0.6
0.4
HNSCC (37.4%)
0.2
P=.008
P=.53
0
0
0
All sites
1.0
Overall Survival
Overall Survival
HNSCC
1.0
0.6
6
7
3
3
7
0
5
1
1
4
0
4
1
0
1
1
0
1
0
0
No. at risk
ACC
NPC
Skin SCC
Other salivary
HNSCC
34
7
19
18
49
22
4
12
10
13
14
2
7
3
6
10
0
6
2
2
20
40
60
80
Patients, %
A, Overall survival from time of index treatment and time of disease recurrence
and/or metastasis, for selected tumor types. Survival probabilities were
calculated with the Kaplan-Meier method, and tumor types were compared
using the log-rank test. B, Distribution of molecular alterations, classified by
levels of evidence for actionability. HNSCC, head and neck squamous cell
carcinoma. ACC, adenoid cystic carcinoma; HPV, human papillomavirus;
NPC, nasopharyngeal carcinoma; Other salivary, other salivary cancer types;
SCC, squamous cell carcinoma.
significantly more frequent than rates reported in primary HPVnegative HNSCCs18 (12 of 70 [17%]) (odds ratio (OR), 5.5;
P < .001), indicating enrichment for this mutation in advanced cancers, although we note that deep sequencing may
be more sensitive for single nucleotide variants. Remarkably,
10 of 11 (91%) HPV-negative tongue SCCs had TERT muta-
jamaoncology.com
E3
Figure 2. The Molecular Landscape of Recurrent and/or Metastatic Head and Neck Squamous Cell Carcinoma
50
Mutation Count
40
30
20
10
Nonsynonymous
Synonymous
HPV status
HPV Negative
HPV Positive
Unknown
Tumor location
Hypopharynx
Larynx
Oral cavity
Oropharynx
Chr3p loss
Yes
No
Yes
No
Sinonasal
Unknown
Unknown
Smoking status
Current
Former
Never
Site sequenced
Index primary
Distant metastasis
Local recurrence
Regional recurrence
Yes
No
Distant metastasis
100
Altered genome, %
0
Clones, No.
TP53 49.1%
TERT 32.1%
FAT1 15.1%
NOTCH1 17.0%
Mutation type
PIK3CA 24.5%
Missense
3q26.3 10.4%
Nonsense
Truncating
CDKN2A 24.5%
In-frame
CDKN2B 11.3%
Promoter
EGFR 13.2%
SCNA type
Amplification
ATR 13.2%
Deletion
KMT2D/MLL2 11.3%
KMT2C/MLL3 9.4%
NSD1 9.4%
11q13 10.4%
E4
Genetic alterations and clinical data for 53 recurrent and/or metastatic head and
neck squamous cell carcinomas. Copy number values are based on 48 cases
with informative copy number data; 11q13 includes CCND1, FGF3, FGF4, and
FGF19; 3q26.3 includes PIK3CA, SOX2, and DCUN1D1. Clones indicates number
of subclonal populations per tumor; HPV, human papillomavirus; SCNA, somatic
copy number alterations.
jamaoncology.com
like HPV-positive tumors trended toward poorer survival (hazard ratio [HR], 2.3; P = .19) although this comparison was limited by small sample size (eFigure 3C in Supplement 1).
The mutational profile of recurrent and metastatic HPVpositive tumors sequenced by MSK-IMPACT appeared intermediate between the HPV-positive and HPV-negative primary tumors in the TCGA. In comparison, the mutational
profile of HPV-negative recurrent and metastatic tumors closely
matched the primary HPV-negative cohort (P = 1.3 104)
(eFigure 4 and eFigure 16 in Supplement 1).
We examined enrichment for specific genetic alterations,
focusing on the most frequent: whole-genome duplication, 3p
deletion, TP53 mutation, and PIK3CA mutation. To allow for
potential differences between platforms, we determined enrichment by comparing HPV-negative:HPV-positive ORs between data sets (Figure 4A) (eTable 8 in Supplement 2). In the
recurrent and metastatic HPV-positive tumors, there was significant enrichment for whole-genome duplication (6.7-fold;
P = .024), 3p deletion (9.1-fold; P = .002), and 3p deletionTP53 mutation (7.6-fold; P = .03). Compared with primary HPVpositive tumors, recurrent and metastatic HPV-positive tumors had higher rates of TP53 mutation (3 of 20 tumors [15%]
vs 1 of 36 tumors [3%]; enrichment 4.6; P = .11) and lower rates
of PIK3CA mutation (2 of 20 tumors [10%] vs 13 of 36 tumors
[36%]; enrichment 0.21, P = .06).
We sought to confirm these findings in additional data sets
of recurrent and metastatic HPV-positive HNSCC. While there
are no comparable cohorts limited to recurrent and metastatic disease, the data sets of Chung et al23 and Chau et al21
included mutational data for HPV-positive tumors, a proportion of which were recurrent or metastatic. Consistent with our
data, each of the HPV-positive cohorts in the studies by Chung
et al and Chau et al exhibited higher frequencies of TP53 mutation (5%-13%) and lower frequencies of PIK3CA mutation
(13%-20%), compared with observed rates in HPV-positive primary tumors.2,3,5,6 For additional confirmation, we analyzed
TP53 status in primary uterine cervical cancers (TCGA, n = 170)
and recurrent and metastatic HPV-positive cervical cancers
(MSK-IMPACT, n = 16). The frequency of TP53 mutation was
markedly higher in metastatic cases than primary cases (4 of
16 tumors [25%] vs 5 of 170 [3%] tumors; OR, 8.3; P = .008)
(eTable 9 in Supplement 2).
TP53 mutations are uncommon (0%-3%) in HPV-positive
primary HNSCC.2,3,5,6 All TP53 mutations were truncating
(eTable 4 in Supplement 2). One patient was a current smoker
(35 pack-year), and 2 were former oligosmokers (1 and 5 packyear). Although p53 is inactivated by the HPV E6 oncoprotein,24
in vitro data reveal that complete inactivation of p53 in HPVpositive HNSCC cell lines induces resistance to radiotherapy and
more aggressive tumor behavior.25 Interestingly, TP53 mutations in HPV-positive tumors were much more likely to be subclonal than in HPV-negative tumors (3 of 4 mutations [75%] vs
4 of 27 mutations [15%]; P = .03) (eTable 7 in Supplement 2). Our
hypothesis generating data suggest that TP53 mutation in a subset of HPV-positive tumors may be associated with tobacco use,
occur later in tumor evolution, and reflect poorer prognosis.
Similarly, alterations (mutations or deletions) in RB1 were
present in 3 of 20 (15%) recurrent and metastatic HPV-positive
(Reprinted) JAMA Oncology Published online July 21, 2016
E5
Figure 3. Mutation Distributions in PIK3CA, NOTCH1, TP53 and TERT in Head and Neck
Squamous Cell Carcinomas
No. of Mutations
PIK3CA
E545K
3
2
1
0
PI3..
0
PI3K_rbd
100
200
PI3K_C2
300
400
PI3Ka
500
PI3_PI4_kinase
600
700
800
900
1000 1068 aa
No. of Mutations
NOTCH1
HD-N domain
PEST domain
2
1
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400 2555 aa
No. of Mutations
TP53
3
2
1
0
P53..
0
P53
50
100
150
P53..
200
250
300
350
393 aa
TERT
1 295 250
1 295 228
1 295 162
1 295 104
TERT
TERT promoter
TSS
C250T
3 cases
ATG
C228T
11 cases
C228A
2 cases
HPV-positive HNSCC
9
8
3p- TP53
WGD
Chromosomal Instability
10
6
5
4
TP53 mutation
PIK3CA mutation
3
2
0.10
0.60
0.4
0.3
0.2
0.1
Frequency of alteration
0
0
0.5
1.0
1.5
2.0
2.5
3.0
log10 P value
E6
P = .009
Mutant
Wild-type
jamaoncology.com
Cutaneous Carcinomas
Twenty-seven advanced nonmelanoma skin carcinomas of
the head and neck were sequenced: 21 squamous, 2 skin
adnexal, and 4 basal cell carcinomas (BCCs) (Figure 7) (eFigure 18 in Supplement 1 and eTable 4 in Supplement 2). A
predominating UV light mutational signature was evident in
18 of 21 (86%) cutaneous SCCs and was associated with
higher average mutation counts than the cutaneous SCCs
lacking a UV light mutational signature (37.9 mutations vs
4.3 mutations; P = .008). Deletion of 3p in combination with
TP53 mutation is a strong negative prognostic factor in
mucosal HNSCC,32 but these co-occurring alterations have
not been described in cutaneous SCC. We found 3p deletion
and TP53 mutation in 6 of 19 cases (32%).
Three of 4 advanced BCCs exhibited the UV light mutational signature, including 1 hypermutated tumor with 122 mutations. In this case, the finding of high mutational load led to
a decision to treat with immunotherapy.
One BCC lacking the UV light signature had only 2 nonsynonymous mutations: a patient who had scalp irradiation
in childhood. Three of four BCCs responded initially to hedgehog pathway inhibitors (prior to NGS) and were subsequently
confirmed to have PTCH1 mutations.
Nasopharyngeal Carcinomas
Eight recurrent and metastatic nasopharyngeal carcinomas
were sequenced (eFigure 19 in Supplement 1 and eTable 4 in
Supplement 2). Four patients with nasopharyngeal carcinoma (50%) were enrolled on clinical trials, of which one (13%)
was NGS-guided: a PIK3CA mutation led to enrollment on a
PI3K inhibitor trial.
Odontogenic Carcinomas
Two odontogenic carcinomas were profiled. Unexpectedly,
EWSR1 gene fusions were found in both (eFigure 9 in
Supplement 1). One case, initially diagnosed as ameloblastic carcinoma, had an EWSR1-FLI1 fusion, confirmed with a polymerase chain reaction assay. Fluorescence in situ hybridization analysis performed earlier had not identified rearrangement
of EWSR1, indicating the superior sensitivity of NGS in this case.
This changed the diagnosis to Ewing sarcoma with epithelial dif(Reprinted) JAMA Oncology Published online July 21, 2016
E7
Mutation Count
6
5
4
3
2
1
Nonsynonymous
Synonymous
Tumor location
Histology
(growth pattern)
Cribiform or tubular
Solid
Unknown
Index primary
Distant metastasis
Site sequenced
Local recurrence
Regional recurrence
Yes
No
Distant metastasis
Current
Former
Smoking status
Altered
genome, %
Tracheobronchial
100
Never
Present
Absent
Myb/NFIB staining
Not tested
NOTCH1 33.3%
KDM6A 19.4%
ARID1A 13.9%
Mutation type
TERT 13.9%
Missense
BCOR 11.1%
Nonsense
Truncating
MLL2 11.1%
In-frame
Promoter
4q12 11.1%
SCNA type
RASA1 8.3%
Amplification
Deletion
RUNX1 8.3%
TBX3 5.6%
YAP1 5.6%
CDKN2A/2B 5.6%
Genetic and clinical data for 36 recurrent and/or metastatic adenoid cystic carcinomas. SCNA indicates somatic copy number alterations.
Germline Analysis
One HNSCC patient had a germline heterozygous deletion in
FANCA (eFigure 10 in Supplement 1) and a tumor with a somatic
FANCA stopgain mutation. While homozygous loss of FANCA in
the germline causes Fanconi anemia, this double-hit affecting
FANCA is an interesting finding in HNSCC,33 similar to a report
in a prostate cancer, where it conferred sensitivity to cisplatin.34
The ability of NGS to profile both somatic and germline alterations is an additional source of potentially actionable data.
Complete genomic and clinical data from this study are
available in searchable form at http://www.cbioportal.org
/study?id=hnc_mskcc_2016.
jamaoncology.com
Figure 6. Mutational Landscape of Recurrent and/or Metastatic Other Salivary Carcinoma Types
30
Mutation Count
25
20
15
10
5
0
Histology
Location
Site sequenced
Distant metastasis
Smoking status
ETV6-NTRK3 fusion
Altered
genome, %
100
0
Nonsynonymous
Synonymous
Salivary duct carcinoma
Salivary ADC, NOS
Mucoepidermoid carcinoma
Acinic cell carcinoma
Epithelial-myoepithelial carcinoma
Mammary analog secretory carcinoma
Local recurrence
Regional recurrence
Local and
regional recurrence
Yes
No
Current
Former
Unknown
Yes
No
Fraction of genome with log2 copy number
value >0.2 or <0.2
TP53 50%
ERBB2 25%
PIK3CA 25%
Mutation type
HRAS 20%
Missense
Nonsense
CDKN2A 15%
Truncating
In-frame
FGFR1 10%
Promoter
MYC 15%
SCNA type
Amplification
NF1 15%
Deletion
CDK12 10%
FAT1 10%
MLL3 10%
Genetic and clinical data for 20 recurrent and/or metastatic other salivary carcinoma types. ADC indicates adenocarcinoma; SCNA, somatic copy number alterations.
Discussion
Critical gaps in knowledge must be addressed to rationally devise new therapies for head and neck cancers. To date, genomics studies have been limited to cohorts of primary tumors. The
molecular landscape of the most clinically challenging tumors
treatment-resistant, recurrent and metastatic diseasehas not
been defined. Furthermore, many uncommon cancer types
have not been molecularly profiled.
Here, we review our implementation of precision oncology for patients with advanced head and neck cancers. This
effort has been useful in 3 respects: (1) illuminating the unique
molecular landscape of advanced or understudied cancers;
jamaoncology.com
E9
Figure 7. Mutational Landscape of Advanced Cutaneous Carcinomas of the Head and Neck
Mutation Count
1000
100
10
Nonsynonymous
Synonymous
Squamous cell carcinoma
Basal cell carcinoma
Skin adnexal carcinoma
Yes
Unknown
No
Yes
Unknown
No
Yes
No
Index primary
Local recurrence
Local and
Distant metastasis
Regional recurrence regional
recurrence
Yes
Histology
Ultraviolet light
signature
Chr3p loss
Immunosuppressed
Site sequenced
Distant metastasis
Altered
genome, %
No
Fraction of genome with log2 copy number
value >0.2 or <0.2
100
0
TP53 85.2%
TERT 51.9%
Mutation type
NOTCH1 48.1%
Missense
CDKN2A 55.6%
Nonsense
Truncating
NOTCH2 48.1%
In-frame
FAT1 44.4%
Promoter
MLL2 37%
SCNA type
Amplification
NOTCH3 33.3%
Deletion
ROS1 33.3%
ASXL1 29.6%
static ACC. We also identified genetic alterations defining recurrent and metastatic HPV-positive cancers: many tumors had
genetic profiles closer to HPV-negative tumors, a finding that
has potential significance to current studies of targeted therapies and chemoradiotherapy deintensification in this population. An important caveat is that these findings are based on a
small cohort, and when comparing results to other data sets,
it is possible that differences in patient characteristics, prior
therapy, sequencing or analytic methodologies may contribute to differences observed. These results require further
confirmation.
Many head and neck cancers are understudied due to
their rarity. We report the first molecular data in mucoepidermoid carcinomas, acinic cell carcinomas, epithelialmyoepithelial carcinomas, MASC, salivary adenocarcinoma,
skin adnexal carcinomas, odontogenic carcinomas, and
head and neck neuroendocrine carcinomas. While sample
sizes were too limited to draw broad conclusions, many
potentially actionable findings were identified and several
uncommon diagnoses refined.
Finally, this analysis demonstrated the potential of targeted NGS to measure intratumor heterogeneity and identify
certain mutational signatures: those of carcinogenesis driven
by UV light, tobacco, AP OBEC, and mismatch repair
E10
deficiency.15,36,37 Emerging data implicate each of these features as potential predictive biomarkers for immunotherapy:
ITH, mutational load, and the tobacco and UV light mutational signatures have each been associated with response to
immune checkpoint inhibitors in lung, skin, and colorectal
cancers.38-42
Conclusions
To advance precision head and neck oncology, a necessary first
step is describing the complete catalog of molecular alterations in incurable and rare cancers. As demonstrated here, the
molecular profile of recurrent and metastatic tumors is quite
distinct from primary tumors. In these understudied diseases, no single institution will be able to profile a large number of patients. Pooling genomic data will be critical to devising rational therapies. Finally, a major therapeutic limitation
in head and neck cancer has been the lack of matched trial options. This shows the value of basket studies, which accounted for half of clinical trial enrollments in this cohort. As
understanding of molecular drivers and biomarkers expands, we anticipate the clinical value added by routine NGS
will continue to increase.
jamaoncology.com
ARTICLE INFORMATION
REFERENCES
jamaoncology.com
E11
E12
jamaoncology.com
eAppendix.
Supplement 2 Table Legends.
eReferences.
eFigures 1-19.
This supplementary material has been provided by the authors to give readers
additional information about their work.
eAppendix
SUPPLEMENTARY METHODS
Patients and tumor samples
The MSK-IMPACT (Memorial Sloan Kettering Integrated Mutation Profiling of Actionable Cancer
Targets) assay is a next-generation sequencing (NGS) assay approved for clinical use through CLIA
(Clinical Laboratory Improvement Amendments) by the Centers for Medicare and Medicaid Services1.
The majority of patients have recurrent or metastatic disease but patients are otherwise unselected.
Since 2014, over 9,000 solid tumors have been profiled using MSK-IMPACT.
We reviewed all patients treated by the MSK multidisciplinary head and neck oncology team, whose
tumors were sequenced with MSK-IMPACT in the context of an Institutional Review Board (IRB)approved study (NCT01775072). Of 224 patients enrolled, 151 with a minimum of 6 month clinical
followup, whose tumors were sequenced between January 2014 and July 2015, were included. This
analysis was approved by the MSK IRB. Correlative clinical data are recorded prospectively in an IRBapproved database. Human papillomavirus (HPV) status was determined by p16 immunohistochemistry
and HPV in situ hybridization as part of routine care. Treatment(s) initiated after ordering NGS were
considered to be guided by molecular data if results led to, or justified continuation with, a specific
treatment.
Sequencing platform and variant calling
MSK-IMPACT is optimized for DNA extracted from low-input formalin-fixed, paraffin embedded (FFPE)
samples. The assay is designed to detect single nucleotide variants (SNVs), indels, copy number variants
(CNVs) and structural variants in genes that are functionally relevant to cancer and/or clinically
actionable targets. The current assay uses hybrid capture technology (Nimblegen SeqCap EZ library
custom oligo) to perform deep (>200x) sequencing (Illumina HiSeq 2500) of all 5781 exons and selected
introns of 410 cancer genes, including canonical and selected non-canonical transcripts, the TERT
promoter region, and 33 introns of 14 rearranged genes (eTable 1). The panel includes 1042 tiling
probes covering single nucleotide polymorphisms (SNPs), allowing genotyping to ensure tumor-normal
matching, identify contaminating DNA, and serve as a low-density SNP array for CNV analysis. MSKIMPACT has been extensively validated. Specific details of panel design, capture protocol, sequencing,
quality control, read alignment, and variant calling have been described2,3. Molecular alterations were
classified according to MSK levels of evidence that support standard or investigational therapeutic
indications (eTable 2).
Bioinformatic analyses
Additional copy number analyses were performed using FACETS4, which performs segmentation of total
and allelic copy ratio, and estimates allele-specific copy number, in targeted capture sequencing data.
We used FACETS to identify whole genome duplication (WGD) and deletion of 3p in SCC (eFigures 1-2).
The fraction of the genome with copy number alteration was defined with a log2 copy number ratio
threshold of .2 or -.2. For comparison, tumor ploidy and allele-specific copy number in The Cancer
Genome Atlas (TCGA) HNSCC cases5 were determined using SNP6 copy number data input into ASCAT6,
as previously described7.
Analyses of intratumor heterogeneity were also performed using FACETS, which estimates the cancer
cell fraction of somatic variants, by incorporating variant allelic fraction and adjusting for tumor purity
and underlying copy number4. The cancer cell fractions of somatic variants are expressed with intervals
representing full width at half maximum boundaries. Mutations with an upper boundary overlapping 1.0
were considered clonal; those with an upper boundary <0.95 were considered to be subclonal, similar to
the approach described by McGranahan et al8. Mutations with overlapping confidence intervals were
considered to comprise an individual subclonal or clonal population, and the total number of subclonal
populations, SNVs, and subclonal SNVs was determined for each sample. These parameters were
compared among subsets using stepwise logistic regression to control for HPV status, site sequenced,
purity, and ploidy.
Network analyses of SNVs were performed in Cytoscape 3.2.1 using CyniTools to infer significant
positive and negative correlations9. Functional annotations were analyzed by examining enrichment of
Gene Ontology terms with BINGO10.
Nonsynonymous and synonymous SNVs were used to delineate certain mutational signatures. The
APOBEC enrichment score, previously described, represents the enrichment of mutated C (or G) within
TCW (or WGA) motifs, in context of all mutated C (or G) 20 nucleotides upstream/downstream11. For the
tobacco signature, the number of transversions per tumor was calculated. The UV signature was based
on a predominance of C to T (or G to A) transitions in dipyrimidine contexts12.
Somatic alteration profiles of HNSCC were examined using a binary classifier to determine if profiles
most closely resembled HPV-positive or HPV-negative tumors. A support vector machine (SVM)
algorithm (in GenePattern13) maximizes the hyperplane separating HPV-positive and HPV-negative
samples14. The SVM model was first trained and cross-validated in the TCGA cohort of 298 cases with
somatic alteration (mutations and high-level CNVs in MSK-IMPACT genes) and HPV status information.
The model had 88% accuracy (p<10-6) in the validation set. The model was then applied to MSK-IMPACT
samples, and HPV status predicted for each case. A posterior probability (confidence) for each sample
classification represents distance from the hyperplane.
To determine if certain frequent genetic alterations were more common than expected in
recurrent/metastatic tumors, the MSK-IMPACT and TCGA cohorts were compared using enrichment
analysis. To allow for potential differences in sequencing platforms, enrichment was calculated as the
log-ratio of odds ratios: the odds ratio (OR) for alterations occurring in HPV-negative tumors, compared
to HPV-positive tumors, was calculated within each cohort, and ORs compared between cohorts.
Significance was tested with logistic regression using an interaction term:
logit (probability of variant) = E0 + E1(cohort) + E2(HPV status) + E3 (cohort x HPV status)
Where eE3 is the enrichment for an alteration in MSK-IMPACT HPV-positive tumors, compared to TCGA
tumors.
Germline variant analysis
After IRB approval for additional germline analyses, these were performed as previously described3.
Briefly, SNVs and indels identified in anonymized normal DNA were analyzed in Ingenuity Variant
Analysis software, categorizing variants by pathogenicity based on 2015 ACMG guidelines15. Pathogenic
and likely pathogenic variants were manually confirmed. We also included 93 genes in Online Mendelian
Inheritance in Man (OMIM) that have been associated with susceptibility to any cancer type3, and 26
genes recommended by the ACMG16.
eTable 1. List of 410 genes (entire exonic regions included) and 36 additional intronic regions included in the MSKIMPACT next-generation sequencing panel.
eTable 2. Actionable molecular alterations, categorized by MSK levels of evidence.
eTable 3. Clinical and tumor characteristics of 151 patients with recurrent/metastatic and treatmentresistant head and neck cancers. HNSCC, head and neck squamous cell carcinoma. ACC, adenoid cystic
carcinoma. NPC, nasopharyngeal carcinoma. ONB, olfactory neuroblastoma. CT, chemotherapy. RT,
radiation therapy.
eTable 4. Matrix of clinical data and genetic alterations found in 151 advanced head and neck cancers,
subdivided by cancer type. Each alteration is described as: mutation genomic coordinates, nucleotide
variant, amino acid variant, variant allelic frequency, category of mutation | copy number alteration.
eTable 5. Gene ontology annotations enriched in the cluster of genes more frequently altered in HPVpositive HNSCC (clusters of green nodes in Figure 2D), with statistical testing corrected using BenjaminiHochberg false discovery rate.
eTable 6. FACETS estimates of tumor purity, ploidy, number of subclonal populations and number of
subclonal mutations (see Methods) for HNSCC cases.
eTable 7. Multivariable logistic regression models of the associations between intratumor heterogeneity
and (A) HPV status and (B) tumor site (primary vs. recurrence/metastasis); (C) (sub)clonal status of TP53
mutations in HPV-positive and HPV-negative HNSCC samples.
eTable 8. Odds ratios for common somatic alterations in HPV-negative compared to HPV-positive
tumors, in both recurrent/metastatic (MSK-IMPACT) and primary (TCGA) HNSCC datasets. Enrichment is
calculated as the ratio of odds ratios. Significance testing as indicated.
eTable 9. TP53 mutations in 16 recurrent/metastatic HPV-associated cervical carcinoma sequenced on MSK-IMPACT.
eTable 10. REMARK (Reporting recommendations for tumor marker studies) checklist.
Pfeifer GP, You YH, Besaratinia A. Mutations induced by ultraviolet light. Mutation research. Apr
1 2005;571(1-2):19-31.
13. Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP. GenePattern 2.0. Nature genetics.
May 2006;38(5):500-501.
Rifkin R, Mukherjee S, Tamayo P, et al. An Analytical Method for Multiclass Molecular Cancer
Classification. SIAM Review. 2003;45(4):706-723.
15. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence
variants: a joint consensus recommendation of the American College of Medical Genetics and
Genomics and the Association for Molecular Pathology. Genetics in medicine : official journal of
the American College of Medical Genetics. May 2015;17(5):405-424.
16. Green RC, Berg JS, Grody WW, et al. ACMG recommendations for reporting of incidental findings in clinical
exome and genome sequencing. Genetics in medicine : official journal of the American College of Medical
Genetics. Jul 2013;15(7):565-574.
14.
eFigure 1.
HNSCC tumor with single copy loss of 3p and single copy gain of 3q
eFigure 2A.
HNSCC tumor with whole genome duplication and copy-neutral loss of heterozygosity at 3p
eFigure 2. Analysis of whole-genome copy number via targeted capture next-generation sequencing, as
depicted in eFigure 1, showing whole genome duplication. (A) A representative case of a tumor (HPVnegative HNSCC) with whole genome duplication and copy-neutral loss of heterozygosity at 3p. (B) Ploidy
in HPV-positive and HPV-negative HNSCC tumors. (C) Whole genome duplication and 3p deletion in
HNSCC tumors.
eFigure 2B.
Ploidy and whole-genome duplication (WGD) in HPV-positive and HPV-negative HNSCC tumors
eFigure 2C.
Whole genome duplication (WGD) in 3p-deleted HNSCC tumors
WGD in 3p-deleted vs. 3p-normal tumors: p=.0001
Chromosome 3p deletion
eFigure 3.
TP53
p=4.4x10-5
TERT
NOTCH1
PIK3CA
3q26
CDKN2A
Overall survival
EGFR
ATR
SVM: HPV-positive
SVM: HPV-negative
MLL2
MLL3
NSD1
eFigure 3. (A) Posterior probabilities of Support Vector Machine (SVM) algorithm, used to
classify HPV status based on genetic alterations in MSK-IMPACT HNSCC cases. X-axis
reflects actual HPV status. Y-axis reflects the confidence level of the classification, described in
Methods. (B) Alterations present in the HPV-positive (MSK-IMPACT) HNSCC cases that
were classified as HPV-negative or HPV-positive by SVM. (C) Overall survival of HPVpositive HNSCC cases in MSK-IMPACT cohort, based on SVM categorization as HPVpositive or HPV-negative.
eFigure 4.
eFigure 5.
APOBEC signature
Tobacco signature
eFigure 5. Distribution of tobacco signature (transversions per tumor) by HNSCC subsite, indicating that
transversion-high tumors were mostly laryngeal cancers.
HPV
HPV +
eFigure 6.
eFigure 7.
NF1 intragenic deletion
eFigure 7. Structural variants observed in salivary duct carcinomas: Integrative Genomics Viewer (IGV) panels.
eFigure 9.
t(22;12)(q12.2;q13.12)
eFigure 10.
eFigure 11
Odontogenic
Olfactory neuroblastoma
H&N neuroendocrine
Nasopharyngeal carcinoma
Other salivary
H&N cutaneous
Adenoid cystic carcinoma
H&N squamous cell carcinoma
2
2
3
8
20
27
36
53
10
20
30
40
50
60
eFigure 12
eFigure 12. Scatterplot of log (mutation count) and fraction of genome with copy number alteration,
by tumor type
HNSCC, head and neck squamous cell carcinoma. ACC, adenoid cystic carcinoma. NPC,
nasopharyngeal carcinoma.
50
eFigure 13
Number of Mutations
40
30
Subgroup
HPVpos
HPVneg
20
10
0.00
0.25
0.50
% Genome Altered
0.75
1.00
eFigure 13. (Above) Scatterplot of mutation counts and percentage of genome with copy number alteration, by
human papillomavirus (HPV) status in head and neck squamous cell (HNSCC) samples.
eFigure 14
HPV
More frequent in
HPV+
eFigure 15
HPV +
p=.007
HPV
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVpos
HPVneg
HPVneg
HPVneg
HPVneg
HPVpos
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVpos
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVpos
HPVpos
HPVneg
HPVpos
HPVpos
HPVpos
HPVneg
HPVneg
HPVneg
HPVpos
HPVneg
HPVneg
HPVpos
HPVneg
HPVneg
HPVpos
HPVpos
HPVneg
HPVpos
HPVpos
HPVpos
57% match
MSK-IMPACT
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVneg
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
HPVpos
91% match
Actual | Predicted
-1
-1
-2
eFigure 16
eFigure 16. Column graph representing the most frequent mutations in recurrent/metastatic
tumors, and the similarity of mutational frequencies to those in primary HPV-positive/negative
tumors. Similarity score is defined as: [ -log10(|TCGA%HPVsame -IMPACT%|) - -log10(|TCGA%
HPVdifferent - IMPACT%)], for all mutations with average frequency >.02.
eFigure 17
HD-N domain
PEST domain
TERT
eFigure 18
PEST domain
SCC (43%)
eFigure 18. NOTCH1 and TERT Mutations in Cutaneous Squamous (SCC) and Basal
Cell (BCC) Carcinomas of the Head and Neck. Green circles indicate missense
mutations; red, truncating; black, inframe insertions or deletions.
Mutation Count
20
15
10
5
0
Viral Status
WHO Type
Site Sequenced
Distant Metastasis
100
% Altered Genome 50
0
MLL2 50%
MDC1 33.3%
ARID1A 16.7%
AXIN1 16.7%
BCL6 16.7%
BMPR1A 16.7%
CARD11 16.7%
ERCC4 16.7%
FBXW7 16.7%
FOXP1 16.7%
JAK1 16.7%
MAX 16.7%
MLL 16.7%
NFE2L2 16.7%
NRAS 16.7%
PIK3CA 16.7%
PTPRD 16.7%
TP53 16.7%
YAP1 16.7%
TGFBR2
16.7%
Downloaded
From: http://oncology.jamanetwork.com/
by a Texas A&M University User on 08/29/2016
eFigure 19
Non-synonymous
Synonymous
HPV+ve
EBV+ve
WHO III
WHO I
WHO II
Index Primary
Local Recurrence
Distant Metastasis
Regional Recurrence
Yes
No
Fraction of Genome with log2
Copy Number Value > 0.2 or < -0.2
eFigure 19. (Above) Mutational landscape of recurrent/metastatic nasopharyngeal carcinomas. EBV, EpsteinBarr virus; HPV, Human papillomavirus.
eTable 1. List of genes and additional intronic regions included in MSK-IMPACT next-generation sequencing panel.
Gene Symbol
Gene Name
ABL1
ACVR1
AKT1
AKT2
AKT3
ALK
ALOX12B
ANKRD11
APC
AR
ARAF
ARID1A
ARID1B
ARID2
ARID5B
ASXL1
ASXL2
ATM
ATR
ATRX
AURKA
AURKB
AXIN1
AXIN2
AXL
B2M
BAP1
BARD1
BBC3
BCL10
BCL2
BCL2L1
BCL2L11
BCL6
BCOR
BIRC3
BLM
BMPR1A
BRAF
BRCA1
BRCA2
BRD4
BRIP1
BTK
CALR
CARD11
CASP8
CBFB
CBL
CCND1
CCND2
CCND3
CCNE1
CD274
CD276
CD79A
CD79B
CDC73
CDH1
CDK12
CDK4
CDK6
CDK8
CDKN1A
CDKN1B
CDKN2A
CDKN2B
CDKN2C
CEBPA
CENPA
CHEK1
CHEK2
CIC
CREBBP
CRKL
CRLF2
CSF1R
CSF3R
CTCF
CTLA4
CTNNB1
CUL3
CXCR4
DAXX
DCUN1D1
DDR2
DICER1
DIS3
DNAJB1
DNMT1
DNMT3A
DNMT3B
DOT1L
E2F3
EED
EGFL7
EGFR
EIF1AX
EIF4A2
EIF4E
EP300
EPCAM
EPHA3
EPHA5
EPHA7
EPHB1
ERBB2
ERBB3
ERBB4
cyclin D3
cyclin E1
CD274 molecule
CD276 molecule
CD79a molecule, immunoglobulin-associated alpha
CD79b molecule, immunoglobulin-associated beta
cell division cycle 73, Paf1/RNA polymerase II complex component, homolog (S. cerevisiae)
cadherin 1, type 1, E-cadherin (epithelial)
cyclin-dependent kinase 12
cyclin-dependent kinase 4
cyclin-dependent kinase 6
cyclin-dependent kinase 8
cyclin-dependent kinase inhibitor 1A (p21, Cip1)
cyclin-dependent kinase inhibitor 1B (p27, Kip1)
cyclin-dependent kinase inhibitor 2A (melanoma, p16, inhibits CDK4)
cyclin-dependent kinase inhibitor 2B (p15, inhibits CDK4)
cyclin-dependent kinase inhibitor 2C (p18, inhibits CDK4)
CCAAT/enhancer binding protein (C/EBP), alpha
centromere protein A
CHK1 checkpoint homolog (S. pombe)
CHK2 checkpoint homolog (S. pombe)
capicua homolog (Drosophila)
CREB binding protein
v-crk sarcoma virus CT10 oncogene homolog (avian)-like
cytokine receptor-like factor 2
colony stimulating factor 1 receptor
colony stimulating factor 3 receptor (granulocyte)
CCCTC-binding factor (zinc finger protein)
cytotoxic T-lymphocyte-associated protein 4
catenin (cadherin-associated protein), beta 1, 88kDa
cullin 3
chemokine (C-X-C motif) receptor 4
death-domain associated protein
DCN1, defective in cullin neddylation 1, domain containing 1 (S. cerevisiae)
discoidin domain receptor tyrosine kinase 2
dicer 1, ribonuclease type III
DIS3 mitotic control homolog (S. cerevisiae)
DnaJ (Hsp40) homolog, subfamily B, member 1
DNA (cytosine-5-)-methyltransferase 1
DNA (cytosine-5-)-methyltransferase 3 alpha
DNA (cytosine-5-)-methyltransferase 3 beta
DOT1-like, histone H3 methyltransferase (S. cerevisiae)
E2F transcription factor 3
embryonic ectoderm development
EGF-like-domain, multiple 7
epidermal growth factor receptor
eukaryotic translation initiation factor 1A, X-linked
eukaryotic translation initiation factor 4A2
eukaryotic translation initiation factor 4E
E1A binding protein p300
epithelial cell adhesion molecule
EPH receptor A3
EPH receptor A5
EPH receptor A7
EPH receptor B1
v-erb-b2 erythroblastic leukemia viral oncogene homolog 2, neuro/glioblastoma derived oncogene homolog (avian)
v-erb-b2 erythroblastic leukemia viral oncogene homolog 3 (avian)
v-erb-a erythroblastic leukemia viral oncogene homolog 4 (avian)
ERCC2
ERCC3
ERCC4
ERCC5
ERG
ERRFI1
ESR1
ETV1
ETV6
EZH2
FAM123B
FAM175A
FAM46C
FANCA
FANCC
FAT1
FBXW7
FGF19
FGF3
FGF4
FGFR1
FGFR2
FGFR3
FGFR4
FH
FLCN
FLT1
FLT3
FLT4
FOXA1
FOXL2
FOXO1
FOXP1
FUBP1
FYN
GATA1
GATA2
GATA3
GLI1
GNA11
GNAQ
GNAS
GPS2
GREM1
GRIN2A
GSK3B
H3F3A
H3F3B
H3F3C
HGF
HIST1H1C
HIST1H2BD
HIST1H3A
HIST1H3B
HIST1H3C
HIST1H3D
HIST1H3E
HIST1H3F
HIST1H3G
HIST1H3H
HIST1H3I
HIST1H3J
HIST2H3C
HIST2H3D
HIST3H3
HLA-A
HNF1A
HOXB13
HRAS
ICOSLG
ID3
IDH1
IDH2
IFNGR1
IGF1
IGF1R
IGF2
IKBKE
IKZF1
IL10
IL7R
INHA
INHBA
INPP4A
INPP4B
INSR
IRF4
IRS1
IRS2
JAK1
JAK2
JAK3
JUN
KDM5A
KDM5C
KDM6A
KDR
KEAP1
KIT
KLF4
KRAS
LATS1
LATS2
LMO1
MALT1
MAP2K1
MAP2K2
MAP2K4
MAP3K1
MAP3K13
MAP3K14
MAPK1
MAPK3
MAX
MCL1
MDC1
MDM2
MDM4
MED12
MEF2B
MEN1
MET
MGA
MITF
MLH1
MLL
MLL2
MLL3
MPL
MRE11A
MSH2
MSH6
MST1
MST1R
MTOR
MUTYH
MYC
MYCL1
MYCN
MYD88
MYOD1
NBN
NCOA3
NCOR1
NEGR1
NF1
NF2
NFE2L2
NFKBIA
NKX2-1
NKX3-1
NOTCH1
NOTCH2
NOTCH3
NOTCH4
NPM1
NRAS
NSD1
NTRK1
NTRK2
NTRK3
NUP93
PAK1
PAK7
PALB2
PARK2
PARP1
PAX5
PBRM1
PDCD1
PDGFRA
PDGFRB
PDPK1
PGR
PHOX2B
PIK3C2G
PIK3C3
PIK3CA
PIK3CB
PIK3CD
PIK3CG
PIK3R1
PIK3R2
PIK3R3
PIM1
PLCG2
PLK2
PMAIP1
PMS1
PMS2
PNRC1
POLD1
POLE
PPM1D
PPP2R1A
PPP6C
PRDM1
PRKAR1A
PTCH1
PTEN
PTPN11
PTPRD
PTPRS
PTPRT
RAB35
RAC1
RAD21
RAD50
RAD51
RAD51C
RAD51L1
RAD51L3
RAD52
RAD54L
RAF1
RARA
RASA1
RB1
RBM10
RECQL4
REL
RET
RFWD2
RHEB
RHOA
RICTOR
RIT1
RNF43
ROS1
RPS6KA4
RPS6KB2
RPTOR
paired-like homeobox 2b
phosphoinositide-3-kinase, class 2, gamma polypeptide
phosphoinositide-3-kinase, class 3
phosphoinositide-3-kinase, catalytic, alpha polypeptide
phosphoinositide-3-kinase, catalytic, beta polypeptide
phosphoinositide-3-kinase, catalytic, delta polypeptide
phosphoinositide-3-kinase, catalytic, gamma polypeptide
phosphoinositide-3-kinase, regulatory subunit 1 (alpha)
phosphoinositide-3-kinase, regulatory subunit 2 (beta)
phosphoinositide-3-kinase, regulatory subunit 3 (gamma)
pim-1 oncogene
phospholipase C, gamma 2 (phosphatidylinositol-specific)
polo-like kinase 2
phorbol-12-myristate-13-acetate-induced protein 1
PMS1 postmeiotic segregation increased 1 (S. cerevisiae)
PMS2 postmeiotic segregation increased 2 (S. cerevisiae)
proline-rich nuclear receptor coactivator 1
polymerase (DNA directed), delta 1, catalytic subunit 125kDa
polymerase (DNA directed), epsilon
protein phosphatase, Mg2+/Mn2+ dependent, 1D
protein phosphatase 2, regulatory subunit A, alpha
protein phosphatase 6, catalytic subunit
PR domain containing 1, with ZNF domain
protein kinase, cAMP-dependent, regulatory, type I, alpha (tissue specific extinguisher 1)
patched 1
phosphatase and tensin homolog
protein tyrosine phosphatase, non-receptor type 11
protein tyrosine phosphatase, receptor type, D
protein tyrosine phosphatase, receptor type, S
protein tyrosine phosphatase, receptor type, T
RAB35, member RAS oncogene family
ras-related C3 botulinum toxin substrate 1 (rho family, small GTP binding protein Rac1)
RAD21 homolog (S. pombe)
RAD50 homolog (S. cerevisiae)
RAD51 homolog (RecA homolog, E. coli) (S. cerevisiae)
RAD51 homolog C (S. cerevisiae)
RAD51-like 1 (S. cerevisiae)
RAD51-like 3 (S. cerevisiae)
RAD52 homolog (S. cerevisiae)
RAD54-like (S. cerevisiae)
v-raf-1 murine leukemia viral oncogene homolog 1
retinoic acid receptor, alpha
RAS p21 protein activator (GTPase activating protein) 1
retinoblastoma 1
RNA binding motif protein 10
RecQ protein-like 4
v-rel reticuloendotheliosis viral oncogene homolog (avian)
ret proto-oncogene
ring finger and WD repeat domain 2
Ras homolog enriched in brain
ras homolog gene family, member A
RPTOR independent companion of MTOR, complex 2
Ras-like without CAAX 1
ring finger protein 43
c-ros oncogene 1 , receptor tyrosine kinase
ribosomal protein S6 kinase, 90kDa, polypeptide 4
ribosomal protein S6 kinase, 70kDa, polypeptide 2
regulatory associated protein of MTOR, complex 1
RUNX1
RYBP
SDHA
SDHAF2
SDHB
SDHC
SDHD
SETD2
SF3B1
SH2B3
SH2D1A
SHQ1
SMAD2
SMAD3
SMAD4
SMARCA4
SMARCB1
SMARCD1
SMO
SOCS1
SOX17
SOX2
SOX9
SPEN
SPOP
SRC
SRSF2
STAG2
STAT3
STAT5A
STAT5B
STK11
STK40
SUFU
SUZ12
SYK
TBX3
TCEB1
TCF3
TCF7L2
TERT
TET1
TET2
TGFBR1
TGFBR2
TMEM127
TMPRSS2
TNFAIP3
TNFRSF14
TOP1
TP53
TP63
TRAF2
TRAF7
TSC1
TSC2
TSHR
U2AF1
VEGFA
VHL
VTCN1
WT1
XIAP
XPO1
XRCC2
YAP1
YES1
ZFHX3
ZRSR2