Figure. Schematic diagrams of mutations in selected head and neck squamous cell carcinoma genes. The features illustrated were identified using the Conserved Domain Architecture Retrieval Tool (CDART) and the Conserved Domain Database (CDD) at the National Center for Biotechnology Information (NCBI). A, Multiple mutations were observed in NOTCH2 and NOTCH3 in numerous cell lines. In contrast, only a single cell line exhibited a damaging single-nucleotide polymorphism (SNP) in PIK3CA. Epidermal growth factor (EGF) repeats, Lin-12/Notch repeats (LNR), and ankyrin (ANK) repeats are found in multiple copies within the NOTCH proteins. NOD and NODP are members of a superfamily found in most NOTCH proteins. The region rich in proline (P), glutamic acid (E), serine (S), and threonine (T) makes up the PEST domain. The p85-binding domain (p85), ras-binding domain (RBD) and catalytic domain (Kinase) of PIK3CA are indicated. The C2 and helical domains of PIK3CA are superfamily members with structural functions. B, Cell lines exhibited SNP mutations in TP53 and CDKN2A that were predicted to be damaging by altering the protein structure. Arrowheads indicate position of observed point mutations. Numbers indicate length in amino acids. The transcription activation domain (TAD) and tetramerization (TETRA) domain of p53 are illustrated.
Nichols AC, Yoo J, Palma DA, et al. Frequent mutations in TP53 and CDKN2A found by next-generation sequencing of head and neck cancer cell lines. Arch Otolaryngol Head Neck Surg. 10.1001/archoto.2012.1558.
eTable. Listing of mutations identified in the six cell lines
Nichols AC, Yoo J, Palma DA, Fung K, Franklin JH, Koropatnick J, Mymryk JS, Batada NN, Barrett JW. Frequent Mutations in TP53 and CDKN2A Found by Next-Generation Sequencing of Head and Neck Cancer Cell Lines. Arch Otolaryngol Head Neck Surg. 2012;138(8):732-739. doi:10.1001/archoto.2012.1558
Author Affiliations: Departments of Otolaryngology Head and Neck Surgery (Drs Nichols, Yoo, Fung, Franklin, and Barrett), Oncology (Drs Nichols, Yoo, Palma, Fung, Franklin, Koropatnick, and Mymryk), Pathology (Dr Nichols), and Microbiology and Immunology (Dr Mymryk), The University of Western Ontario, London, Ontario, Canada; London Regional Cancer Program, London, Ontario (Drs Nichols, Yoo, Palma, Fung, Franklin, Koropatnick, Mymryk, and Barrett); Lawson Health Research Institute, London, Ontario (Drs Nichols, Koropatnick, Mymryk, and Barrett); Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada (Dr Batada); and Genome Technologies Platform, Ontario Institute for Cancer Research, Toronto, Ontario (Dr Batada).
Objective To conduct high-throughput mutational analysis in 6 commonly used head and neck cancer cell lines. Comprehensive mutation analysis of primary head and neck squamous cell carcinoma (HNSCC) tumors has recently been reported, and mutations in the NOTCH receptors, TP53 and CDKN2A, were key findings. Established cell lines are valuable tools to study cancer in vitro. Similar high-throughput mutational analysis of head and neck cancer cell lines is necessary to confirm their mutational profile.
Design DNA was extracted from American Type Culture Collection (ATCC) cell lines Cal27, Detroit562, FaDu, SCC4, SCC15, and SCC25. Cell line identity was confirmed by short tandem repeat (STR) analysis, and human papillomavirus (HPV) infection status was assessed by real-time polymerase chain reaction. A total of 535 cancer-associated genes were sequenced through a limited exome capture on the Illumina HiSeq system.
Setting London Regional Cancer Program.
Results The identity of the 6 cell lines was confirmed by STR analysis, and all lines tested negative for HPV infection. We achieved an average of 129-fold coverage with paired-end 100 base-pair reads. Sequencing revealed an average of 38 damaging mutations in each cell line (range, 30-45). The TP53 mutations, predicted to confer loss of function, were noted in all cell lines, and damaging CDKN2A mutations were found in all lines except SCC15.
Conclusions High-throughput sequencing of head and neck cancer cell lines revealed similar mutations to those observed in primary tumors. Thus, these lines reflect the tumor biology of HNSCC and can serve as valuable models to study HNSCC in vitro.
Next-generation sequencing (NGS) of head and neck squamous cell carcinoma (HNSCC) primary tumors has recently been reported in 2 seminal articles.1,2 Both studies highlight mutations deactivating the NOTCH receptor, as well as a very high incidence of TP53 mutations in tumors testing negative for human papillomavirus (HPV). The mutational landscape of HNSCC is dominated by alteration of tumor suppressors rather than the more easily targetable activating mutations in oncogenes. This situation presents a therapeutic challenge and necessitates the identification of new targets for therapy in downstream biochemical pathways that are dysregulated by tumor suppressor loss. Primary cell cultures and xenografts would ideally be used to identify these targets, but immortalized HNSCC cell lines have the advantage of being inexpensive, easy to manipulate, and widely available to the research community. To date, the mutational status of the key genes identified in the large-scale studies of primary tumors has not been reported in HNSCC cell lines. In this study, we report NGS analysis of 6 commonly used American Type Culture Collection (ATCC) HNSCC cell lines.
Cell lines Cal27, Detroit562, FaDu, SCC4, SCC15, SCC25 were obtained from the ATCC and grown in Dulbecco Modified Eagle Medium and Ham's F-12 in a 1:1 mix, supplemented with 10% fetal bovine serum (GIBCO; Invitrogen), 400-ng/mL hydrocortisone (Wisent), penicillin (100 IU/mL), and streptomycin (100 μg/mL) (Invitrogen). When cells reached 80% confluence, they were harvested with 0.25% trypsin/0.1% EDTA (ethylenediaminetetraacetic acid) (Wisent).
DNA was extracted from cultured cells using the AllPrep DNA/RNA/Protein kit (Qiagen) according to the instructions provided by the manufacturer. Minimum quality requirements for sequencing studies were 5 μg of genomic DNA with optical density 260/280 and 260/230 ratios of between 1.8 and 2.0.
For each cell line, a 200-ng sample of DNA was analyzed by short tandem repeat (STR) profiling at The Centre for Advanced Genomics (Toronto, Ontario, Canada). The lines were genotyped with a panel of 16 selected markers, including the 8 Combined DNA Index System (CODIS) core STR loci used by the ATCC.3 All STR profiling results for our 6 cell lines were compared with the ATCC STR database.
In an effort to screen clinical samples for the presence of HPV, we designed a multiplex quantitative polymerase chain reaction (qPCR) to identify those samples that were HPV positive and to confirm the genotype of the HPV subtype in any positive samples. We designed primer/probe sets (Table 1) against a 115-nucleotide (nt) fragment within the GAPDH exon 6 (internal control); a 110-nt region across E6-E7 of HPV-16; a 137-nt fragment across the HPV-18 E6-E7 region; and a 110-nt fragment in the E6-E7 region of all high-risk HPVs, including HPV-16, -18, -31, -33, -35, -39, -45, -51, -52, -56, -58, -59, and -68. The 4-plex PCR reactions were run on a Stratagene Mx3000P instrument using the conditions recommended in the Quantitech Multiplex handbook (Qiagen). Twenty-microliter reactions were heated for 15 minutes at 95°C to activate the amplification enzyme. This was followed by 40 cycles at 94°C for 60 seconds each and 40 cycles at 60°C for 90 seconds. CaSki cell DNA was used as a positive control for HPV-16, and HeLa genomic DNA was used as the HPV-18 control.
Next-generation sequencing was performed by Otogenetics, Tucker, Georgia, using custom probes (Roche/NimbleGen) to the 535 most common cancer genes listed in the Catalogue of Somatic Mutations in Cancer (COSMIC) database.4 Gene capture was performed using the standard exome protocol from NimbleGen. Sequencing was carried out on the Illumina HiSeq 2000 platform with 100-fold coverage and paired-end 100 base-pair (bp) reads.
Primary reads were compiled against reference human genome UCSC (University of California, Santa Cruz) assembly hg19 (HRCh37). The raw 100 bp × 2 paired-end reads were mapped onto the human reference genome (hg19; HRCh37) using the Novoalign Aligner (version 2.07.17; http://www.novocraft.com/). Nonuniquely mapping reads (ie, those with mapping scores <150) were discarded. The remaining reads were realigned, and base qualities were recalibrated using the Genome Analysis Toolkit (GATK; Broad Institute).5 All PCR artifacts were discarded using Picard's MarkDuplicates (http://picard.sourceforge.net/).
Only targeted genomic regions with at least 10-fold coverage and Phred-scaled base quality of 30 or higher were considered. A custom mutation caller6 and GATK were used to identify variants. All mutations were analyzed with a bioinformatics pipeline that has been demonstrated to report only mutations that can be independently validated.6 Running the pipeline involved subjecting candidate mutations to a series of filters: (1) candidate mutations were discarded if they were present exclusively in the latter third of the reads; (2) candidate mutations were ignored if the mutant allele was flanked by a homopolymeric region (defined as repeats of 2 or longer); (3) candidate mutations were discarded if BLAST (Basic Local Alignment Search Tool) alignment of reads containing them did not concur with the Novoalign mapping. Functional annotation of single-nucleotide polymorphisms (SNPs) into groups of either nonsynonymous or synonymous changes were further filtered based on a prediction of damaging mutations using SeattleSeq Annotator online tool (http://gvs.gs.washington.edu/SeattleSeqAnnotation/ [registration required]). Furthermore, GATK's UnifiedGenotyper was used as an independent tool to identify variants, and these were recalibrated and filtered using GATK-recommended parameters.
Genes identified with predicted nonsynonymous mutations that have already been reported in the same cell lines from the COSMIC database or that matched those previously reported in patient samples from the large-scale sequencing studies were selected for further analysis. We validated a subset of these mutations (n = 80) by either mass spectrometry or PCR amplification followed by Sanger sequencing. Mutations were also confirmed by visual inspection using the DNAnexus genome browser (DNAnexus Inc).
Primers were designed around a subset of the mutations identified by the initial coverage (≥10-fold) and quality screen (Phred-scaled base quality of 30), as noted. Multiplex genotyping was carried on the MassARRAY platform by the Clinical Genomics Centre, University Health Network, Toronto, Ontario, Canada.
An additional subset of mutations was validated by conventional PCR followed by Sanger sequencing. The primers used are listed in Table 2. The PCR reactions (20 μL) were prepared using 0.2 μL of template DNA extracted from the appropriate cell line and amplified with Phusion DNA polymerase (Fisher Scientific) following a hot start (98°C for 30 seconds followed by 37 cycles at 98°C for 10 seconds, 55°C for 10 seconds, and 72°C for 30-seconds). Amplified products were resolved on 1% agarose gels. Bands of interest were extracted from agarose gel slices, purified on columns (Bio Basic), and sequenced with the appropriate primer at the London Regional Genomics Centre, Robarts Research Institute, London, Ontario, Canada.
We confirmed the identities of all 6 cell lines by STR analysis, and they exactly matched the STR profile available from the ATCC and recent publications (Table 3).7 The 6 cell lines all tested negative by qPCR for HPV, as expected, including HPV-16 and HPV-18 (data not shown).
The limited-exome NGS provided excellent coverage of the exons of the 535 targeted genes (Table 4). The average coverage of targeted exon bases was 129-fold, with over 90% of bases sequenced more than 30 times.
The bioinformatics pipeline identified 99 unique variants across the 6 lines. An additional 125 mutations were confirmed in genes that have been reported in the large-scale exome-sequencing articles or in the COSMIC database (eTable). The observed mutations included 157 nonsynonymous SNPs, 1 multiple-nucleotide polymorphism (MNP), 31 deletions, 8 insertions, 20 frameshifts, 2 splicing mutations, and 5 stop-gain mutations. We attempted to validate 80 mutations by Sequenom MassArray genotyping and/or PCR and sequencing and confirmed 77 (96%). The 3 discordances are described herein.
Of the mutations reported in the COSMIC database,4 we were able to validate 16 of 19 (84%) by NGS and Sequenom MassArray genotyping and/or PCR (Table 5). One splicing mutation in TP53 in cell line SCC15 was determined to be a c.672 + 1G→A change instead of a G→T change. The 2 other discordances were observed in the FaDu line and included a whole gene deletion of SMAD4 and splice site mutation in TP53 c.376-1G→A that we could not confirm. Very low coverage of SMAD4 in the NGS implied a mutation in agreement with COSMIC; however, the region was present by Sequenom MassArray genotyping of the region in question. There was no evidence of a TP53 splice site mutation in any of the 72 NGS reads that covered that site.
Multiple mutations were noted in 36 genes (Table 6). Of these, 26 recurrent specific mutations were noted in 2 to 6 cell lines making up a total of 92 mutations (eTable). We tested 13 of the 26 mutations by either PCR or Sequenom MassArray genotyping, and all were confirmed to be valid. Of note, identical mutations were noted at 3 sites in the NOTCH2 in all cell lines except SCC15 (Figure, A). Similarly, recurrent identical mutations occurred in NOTCH3 (6668C→T) in CAL27, Detroit562, FaDu, and SCC4. As well, there was a unique NOTCH3 mutation, 3692G→A, in SCC15 (Figure, B). No NOTCH1 mutations were noted.
In contrast to the specific recurrent mutations noted in the NOTCH receptors, unique mutations, representing a range of genetic changes, were observed in TP53 in all cell lines (Figure, B). There was an additional known polymorphism in TP53 (215C→G), not thought to be oncogenic, noted in CAL27, SCC15, and SCC25.8 All cell lines except SCC15 possessed CDKN2A mutations (Figure, B). The mutation in SCC4 involved a 10-bp deletion at the last splice site immediately followed by an SNP that resulted in a splicing mutation.
Mutations were noted in known oncogenes NRAS and PIK3CA. The mutation in PIK3CA (H1047R) is a known activating hotspot (Figure, A).9,10 The consequences of these NRAS mutations are unknown because they do not occur at known activating sites.
High-throughput sequencing of cancer genomes has recently provided tremendous insights into the genetic changes that drive cancer biology. Two large-scale multi-institutional sequencing efforts in HNSCC have recently been reported.1,2 Identification of a frequent activating mutation in an oncogene that could be simply targeted with molecular agents—similar to the epidermal growth factor receptor mutations in a subset of non–small cell lung cancer or the BRAF V600E mutation in melanoma—would have been ideal results in these 2 studies. However, the main findings were the identification of inactivating mutations in tumor suppressors, including TP53, CDKN2A, NOTCH 1, NOTCH 2, and NOTCH 3.1,2
Targeting tumor suppressors is more complicated than targeting activating mutations. It may be possible to design small molecules against select point mutations, to cause the protein to refold and restore function as has been demonstrated for TP53.11 For nonsense mutations, insertions, and deletions, alternative strategies will need to be used. Loss of tumor suppressor function may lead to upregulation and downregulation of numerous pathways, each of which can potentially be targeted with molecular agents. However, for most tumor suppressors, these pathway alterations have not been elucidated. We endeavored to identify identical changes in cell lines to those seen in primary tumors that will allow in vitro manipulation to further investigate these downstream changes.
Cell lines are not perfect models of cancer. Studies have demonstrated that cell lines acquire epigenetic changes and additional mutations compared with the primary tumor or xenografts.12,13 These changes may confer a survival advantage in culture conditions or result from clonal replication of a subset of cells in contrast to a heterogeneous primary tumor.12,13 Regardless, cell lines have been invaluable tools to study cancer on a molecular level. The ease and low cost with which they can be manipulated has greatly contributed to our understanding of carcinogenesis and the identification of new therapeutic agents.
In our study, we elected to study 6 of the 7 available HNSCC cell lines available from the ATCC because these lines are readily available to all researchers and have been characterized with STR profiling. We excluded 1 ATCC line, SCC9, because there was a significant component of fibroblast-like and other stromal-like cells in addition to the tumor epithelial cells within the culture. We have since learned that this is an accepted feature of this cell line (ATCC personal communication, Ana Pereira, BSc, January 31, 2012).
The limited exome sequencing provided by Otogenetics was selected because the cancer-specific gene panel included a large number of the genes identified in the large-scale genome studies at very high coverage. The high quality of the sequencing results and bioinformatics is supported by the very high rate of our independent validation (96%) by mass spectrometry and/or PCR. This raises the question of whether independent validation is necessary for future sequencing studies, provided high coverage is obtained along with a robust bioinformatics pipeline.
The discovery of NOTCH receptor mutations was perhaps the most novel finding of the large-scale studies, with NOTCH1, - 2, and - 3 mutations occurring in 14%, 5%, and 4% of tumors, respectively.2 In contrast to the activating NOTCH mutations in the heterodimerization and to those found in the PEST domain (proline, glutamic acid, serine, threonine-rich) seen in T-cell acute lymphoblastic leukemia,14NOTCH1 mutations in HNSCC are largely dispersed elsewhere in the gene or produce truncated products.1,2 Thus, NOTCH1 mutations in HNSCC are thought to represent loss of function changes. Importantly, this suggests that NOTCH1 is a tumor suppressor in SCC, and this is supported by research in knockout mice models.15
Given the observed rate of NOTCH1 mutations (14%), it is not entirely surprising that a NOTCH1 variant was not noted in our 6 lines. In contrast, we noted numerous identical mutations in NOTCH2 and NOTCH3. However, all but 1 of the mutations (3692G→A in SCC15) were recurrent specific mutations seen in multiple lines (Figure, A). In NOTCH2, all the variants occurred in or just downstream of the signal sequence at the N-terminus, while the recurrent mutation in NOTCH3 occurred just upstream of the C-terminus. The high rate of these mutations is unexpected given the fresh tumor results, particularly since recurrent specific mutations were rare.2
The NOTCH2 112G→A mutation was considered “damaging,” while all other NOTCH2 and - 3 mutations were considered “tolerated” by SIFT (sorting intolerant from tolerant) analysis.16 However, SIFT analysis is not error proof. The clearest example of the inaccuracy of SIFT predictions in our study is the 3140A→G mutation seen in PIK3CA. Our SIFT analysis predicted that this is a tolerated variant, however in vivo studies have determined that this may be the most potent activating mutation in this gene.9,17
The recurrent variants seen in our study may represent polymorphisms or mutations that confer a survival advantage in cell culture. Ultimately, detailed interrogation of the relevant signaling pathways and their downstream targets will be necessary to understand the importance of the genetic changes observed. An important question that remains to be addressed will be whether restoration of the function of any of these mutated tumor suppressor genes will be sufficient to reverse some or all of the cancerous properties of these cells. Confirmation of the importance of some of these pathways will allow future work to focus on developing therapies against those targets with the greatest likelihood of having strong antitumor effects.
Although this study provides information on the key mutations in these cell lines, other sequencing techniques including whole exome, whole genome, and RNA seq analysis can be used in primary tumors and cell lines to fully understand the genetic and expression profile of HNSCC. This type of comprehensive analysis will be necessary to develop new therapies and identify robust biomarkers of treatment response.
Correspondence: Anthony C. Nichols, MD, Department of Otolaryngology–Head and Neck Surgery, Victoria Hospital, London Health Science Centre, Room B3-431A, 800 Commissioners Rd E, London, ON N6A 5W9, Canada (Anthony.Nichols@lhsc.on.ca).
Submitted for Publication: April 4, 2012; final revision received May 19, 2012; accepted June 10, 2012.
Author Contributions: Drs Nichols and Barrett had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Nichols, Yoo, Fung, Mymryk, and Barrett. Acquisition of data: Nichols, Fung, Franklin, and Barrett. Analysis and interpretation of data: Nichols, Palma, Fung, Koropatnick, Mymryk, Batada, and Barrett. Drafting of the manuscript: Nichols, Yoo, Mymryk, and Barrett. Critical revision of the manuscript for important intellectual content: Nichols, Yoo, Palma, Fung, Franklin, Koropatnick, Mymryk, and Batada. Statistical analysis: Nichols and Batada. Obtained funding: Nichols, Yoo, Palma, and Mymryk. Administrative, technical, and material support: Nichols, Yoo, Palma, Fung, Franklin, Koropatnick, and Mymryk. Study supervision: Yoo and Mymryk.
Financial Disclosure: None reported.
Funding/Support: This research was supported by the Translational Head and Neck Cancer Program Fund, Western University, London Regional Cancer Program small grant (Drs Nichols and Mymryk).
Additional Information: Drs Nichols and Yoo contributed equally to this work; Drs Batada and Barrett contributed equally to this work.