Portal de Periódicos da CAPES

Combining Highly Multiplexed PCR with Semiconductor-Based Sequencing for Rapid Cancer Genotyping

2012; Elsevier BV; Volume: 15; Issue: 2 Linguagem: Inglês

10.1016/j.jmoldx.2012.09.003

1943-7811

Carol Beadling, Tanaya Neff, Michael C. Heinrich, Katherine Rhodes, Michael Thornton, John H. Leamon, Mark T. Andersen, Christopher L. Corless,

CRISPR and Genetic Engineering

There is growing demand for routine identification of actionable mutations in clinical cancer specimens. Genotyping platforms must provide rapid turnaround times and work effectively with limited amounts of formalin-fixed, paraffin-embedded (FFPE) tissue specimens that often yield poor quality DNA. We describe semiconductor-based sequencing of DNA from FFPE specimens using a single-tube, multiplexed panel of 190 amplicons targeting 46 cancer genes. With just 10 ng of input DNA, average read depths of 2000× can be obtained in 48 hours, with >95% of the reads on target. A validation set of 45 FFPE tumor specimens containing 53 point mutations previously identified with a mass spectrometry–based genotyping platform, along with 19 indels ranging from 4 to 63 bp, was used to evaluate assay performance. With a mutant allele ratio cutoff of 8%, we were able to achieve 100% sensitivity (95% CI = 97.3% to 100.0%) and 95.1% specificity (95% CI = 91.8% to 98.0%) of point mutation detection. All indels were visible by manual inspection of aligned reads; 6/9 indels ≤12 bp long were detected by the variant caller software either exactly or as mismatched nucleotides within the indel region. The rapid turnaround time and low input DNA requirements make the multiplex PCR and semiconductor-based sequencing approach a viable option for mutation detection in a clinical laboratory. There is growing demand for routine identification of actionable mutations in clinical cancer specimens. Genotyping platforms must provide rapid turnaround times and work effectively with limited amounts of formalin-fixed, paraffin-embedded (FFPE) tissue specimens that often yield poor quality DNA. We describe semiconductor-based sequencing of DNA from FFPE specimens using a single-tube, multiplexed panel of 190 amplicons targeting 46 cancer genes. With just 10 ng of input DNA, average read depths of 2000× can be obtained in 48 hours, with >95% of the reads on target. A validation set of 45 FFPE tumor specimens containing 53 point mutations previously identified with a mass spectrometry–based genotyping platform, along with 19 indels ranging from 4 to 63 bp, was used to evaluate assay performance. With a mutant allele ratio cutoff of 8%, we were able to achieve 100% sensitivity (95% CI = 97.3% to 100.0%) and 95.1% specificity (95% CI = 91.8% to 98.0%) of point mutation detection. All indels were visible by manual inspection of aligned reads; 6/9 indels ≤12 bp long were detected by the variant caller software either exactly or as mismatched nucleotides within the indel region. The rapid turnaround time and low input DNA requirements make the multiplex PCR and semiconductor-based sequencing approach a viable option for mutation detection in a clinical laboratory. Personalized cancer care is based on the principle that each patient's cancer is best treated by directing one or more targeted therapies to specific molecular defects in that specific tumor.1Druker B.J. Translation of the Philadelphia chromosome into therapy for CML.Blood. 2008; 112: 4808-4817Crossref PubMed Scopus (576) Google Scholar To fully realize the benefit of such a personalized approach, it is essential to incorporate screening for targetable mutations into routine clinical diagnostics. Next-generation sequencing methods are needed that are rapid and cost effective and that work well on DNA isolated from formalin-fixed paraffin-embedded (FFPE) clinical tissue specimens, keeping in mind that such DNAs are often of poor quality and available in very limited quantity.2Kerick M. Isau M. Timmermann B. Sültmann H. Herwig R. Krobitsch S. Schaefer G. Verdorfer I. Bartsch G. Klocker H. Lehrach H. Schweiger M.R. Targeted high throughput sequencing in clinical cancer settings: formaldehyde fixed-paraffin embedded (FFPE) tumor tissues, input amount and tumor heterogeneity.BMC Med Genomics. 2011; 4: 68Crossref PubMed Scopus (150) Google Scholar, 3Schweiger M.R. Kerick M. Timmermann B. Albrecht M.W. Borodina T. Parkhomchuk D. Zatloukal K. Lehrach H. Genome-wide massively parallel sequencing of formaldehyde fixed-paraffin embedded (FFPE) tumor tissues for copy-number- and mutation-analysis.PLoS One. 2009; 4: e5548Crossref PubMed Scopus (148) Google Scholar Both primer extension–based single-nucleotide polymorphism multiplexing assays (eg, SNaPshot; Life Technologies, Foster City, CA) and mass spectrometry–based multiplexing assays have been successfully used to detect somatic mutations in FFPE-derived clinical specimens.4Beadling C. Heinrich M.C. Warrick A. Forbes E.M. Nelson D. Justusson E. Levine J. Neff T.L. Patterson J. Presnell A. McKinley A. Winter L.J. Dewey C. Harlow A. Barney O. Druker B.J. Schuff K.G. Corless C.L. Multiplex mutation screening by mass spectrometry evaluation of 820 cases from a personalized cancer medicine registry [Erratum appeared in J Mol Diagn 2012, 14:41].J Mol Diagn. 2011; 13: 504-513Abstract Full Text Full Text PDF PubMed Scopus (71) Google Scholar, 5MacConaill L.E. Campbell C.D. Kehoe S.M. Bass A.J. Hatton C. Niu L. Davis M. Yao K. Hanna M. Mondal C. Luongo L. Emery C.M. Baker A.C. Philips J. Goff D.J. Fiorentino M. Rubin M.A. Polyak K. Chan J. Wang Y. Fletcher J.A. Santagata S. Corso G. Roviello F. Shivdasani R. Kieran M.W. Ligon K.L. Stiles C.D. Hahn W.C. Meyerson M.L. Garraway L.A. Profiling critical cancer gene mutations in clinical tumor samples [Erratum appeared in PLoS One 2010, 5 (doi: 10.1371/annotation/3a0c8fee-57ef-43ed-b6c2-55b503e6db5e) and in PLoS One 2010, 5 (doi: 10.1371/annotation/613c7509-e4c9-42ac-82fb-fc504400d9e0)].PLoS One. 2009; 4: e7887Crossref PubMed Scopus (305) Google Scholar, 6Su Z. Dias-Santagata D. Duke M. Hutchinson K. Lin Y.L. Borger D.R. Chung C.H. Massion P.P. Vnencak-Jones C.L. Iafrate A.J. Pao W. A platform for rapid detection of multiple oncogenic mutations with relevance to targeted therapy in non-small-cell lung cancer.J Mol Diagn. 2011; 13: 74-84Abstract Full Text Full Text PDF PubMed Scopus (159) Google Scholar However, these approaches are limited by the inability of each assay to interrogate more than a limited number of common variants. Broader screening methods are needed to detect more complex and heterogeneous mutations, and variants for which a priori knowledge may not be available. Semiconductor-based DNA sequencing is well suited to this purpose.7Rothberg J.M. Hinz W. Rearick T.M. Schultz J. Mileski W. Davey M. et al.An integrated semiconductor device enabling non-optical genome sequencing.Nature. 2011; 475: 348-352Crossref PubMed Scopus (1548) Google Scholar This method is based on detection of hydrogen ions released during DNA synthesis as nucleotides are incorporated into the nascent strand. Sequencing runs of up to 200 nucleotides can be completed in approximately 2 hours, and the throughput of 80 to 100 Mb/hour compares favorably against the 60 Mb/hour output of sequencing platforms that use optical detection of labeled nucleotides.8Loman N.J. Misra R.V. Dallman T.J. Constantinidou C. Gharbia S.E. Wain J. Pallen M.J. Performance comparison of benchtop high-throughput sequencing platforms [Erratum in Nat Biotechnol 2012, 30:562].Nat Biotechnol. 2012; 30: 434-439Crossref PubMed Scopus (1013) Google Scholar Two target enrichment strategies are commonly used for next-generation sequencing. One, the hybrid capture method, uses hybridization of fragmented genomic DNA to oligonucleotide baits complementary to the target regions of interest. This method enables scaling to a broad range of target regions, as well as detection of copy number changes and structural rearrangements.9Gnirke A. Melnikov A. Maguire J. Rogov P. LeProust E.M. Brockman W. Fennell T. Giannoukos G. Fisher S. Russ C. Gabriel S. Jaffe D.B. Lander E.S. Nusbaum C. Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing.Nat Biotechnol. 2009; 27: 182-189Crossref PubMed Scopus (1039) Google Scholar, 10Lonigro R.J. Grasso C.S. Robinson D.R. Jing X. Wu Y.M. Cao X. Quist M.J. Tomlins S.A. Pienta K.J. Chinnaiyan A.M. Detection of somatic copy number alterations in cancer using targeted exome capture sequencing.Neoplasia. 2011; 13: 1019-1025Abstract Full Text PDF PubMed Scopus (68) Google Scholar, 11Levin J.Z. Berger M.F. Adiconis X. Rogov P. Melnikov A. Fennell T. Nusbaum C. Garraway L.A. Gnirke A. Targeted next-generation sequencing of a cancer transcriptome enhances detection of sequence variants and novel fusion transcripts.Genome Biol. 2009; 10: R115Crossref PubMed Scopus (162) Google Scholar This approach has been validated with FFPE specimens,12Wagle N. Berger M.F. Davis M.J. Blumenstiel B. DeFelice M. Pochanard P. Ducar M. Van Hummelen P. MacConaill L.E. Hahn W.C. Meyerson M. Gabriel S.B. Garraway L.A. High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing.Cancer Discov. 2012; 2: 82-93Crossref PubMed Scopus (424) Google Scholar although the procedure is labor-intensive and time consuming. The other method uses PCR to generate target-specific amplicons, and uses nanofluidic arrays or single-droplet emulsions containing unique primer pairs.13Kiss M.M. Ortoleva-Donnelly L. Beer N.R. Warner J. Bailey C.G. Colston B.W. Rothberg J.M. Link D.R. Leamon J.H. High-throughput quantitative polymerase chain reaction in picoliter droplets.Anal Chem. 2008; 80: 8975-8981Crossref PubMed Scopus (278) Google Scholar, 14Jones M.A. Bhide S. Chin E. Ng B.G. Rhodenizer D. Zhang V.W. Sun J.J. Tanner A. Freeze H.H. Hegde M.R. Targeted polymerase chain reaction-based enrichment and next generation sequencing for diagnostic testing of congenital disorders of glycosylation.Genet Med. 2011; 13: 921-932Abstract Full Text Full Text PDF PubMed Scopus (80) Google Scholar This approach requires dedicated, costly equipment that has limited sample throughput. We report here the genotypic analysis of 45 FFPE-derived cancer specimens using a multiplex panel of 190 PCR assays that encompass frequently mutated regions in BRAF, KRAS, EGFR, and 43 other cancer genes (Ion AmpliSeq cancer panel; Ion Torrent, Life Technologies, Guilford, CT). Amplicons were generated from 10 ng of DNA in a single multiplex PCR run on a standard cycler, and samples were multiplexed for sequencing on an Ion Torrent Personal Genome Machine (PGM) sequencer. Sequence alignment, target region coverage analysis, and variant calling were all performed with Torrent Suite software version 2.0 (Ion Torrent) optimized for use with the PGM system. Blocks of FFPE tumor or normal tissue, or unstained sections of FFPE tissue, were obtained from the pathology archives of Oregon Health & Science University. The study was conducted under full Institutional Review Board approval and according to federal and institutional guidelines. The diagnosis in each case was confirmed by a single pathologist (C.L.C.). Tumor-rich areas (40% to 90%) were dissected from unstained sections (5 μm thick ) by comparison with an H&E-stained slide, and genomic DNA was extracted using a QIAamp DNA mini kit (Qiagen, Valencia, CA). Twenty-nine samples had been previously genotyped on a MassArray panel (Sequenom, San Diego, CA).2Kerick M. Isau M. Timmermann B. Sültmann H. Herwig R. Krobitsch S. Schaefer G. Verdorfer I. Bartsch G. Klocker H. Lehrach H. Schweiger M.R. Targeted high throughput sequencing in clinical cancer settings: formaldehyde fixed-paraffin embedded (FFPE) tumor tissues, input amount and tumor heterogeneity.BMC Med Genomics. 2011; 4: 68Crossref PubMed Scopus (150) Google Scholar Laser capture microdissection (LCM) was performed using an Applied Biosystems Arcturus XT instrument (Life Technologies, Foster City, CA). Sections (7 μm thick) were prepared from the paraffin block, deparaffinized, stained with methyl green, and thoroughly dehydrated in xylene. Laser capture was performed with Arcturus CapSure macro LCM caps (Life Technologies). An estimated 1000 cells were collected per cap per extraction. Caps were incubated at 65°C with proteinase K and buffer for approximately 16 hours, and DNA was extracted using an Arcturus PicoPure DNA extraction kit (Life Technologies). The Ion AmpliSeq cancer panel (Ion Torrent) was used to generate target amplicon libraries. Briefly, 10 ng of DNA derived from FFPE tissue was amplified by PCR using the premixed Ion AmpliSeq cancer primer pools and Ion AmpliSeq HiFi master mix (Ion AmpliSeq kit version 2.0β). Primer sequences were manufactured specifically for use with the Ion AmpliSeq kits and contained proprietary modifications. The resulting 190 multiplexed amplicons were treated with FuPa reagent (Ion Torrent) to partially digest primer sequences and phosphorylate the amplicons. The amplicons were then ligated to adapters from the Ion Xpress barcoded adapters 1–16 kit according to the manufacturer's instructions (Ion Torrent). After ligation, the amplicons underwent nick-translation and additional library amplification by PCR to complete the linkage between adapters and amplicons. An Agilent 2100 Bioanalyzer high-sensitivity DNA kit (Agilent, Santa Clara, CA) was used to visualize the size range and determine the library concentration. Both individual libraries and multiplexed barcoded libraries were amplified by emulsion PCR on Ion Sphere particles (ISPs) at a 1:1 ratio of total library molecules to ISPs (280 × 106 molecules per reaction) (Ion Xpress Template kit version 2.0; Ion Torrent). The templated ISPs were recovered from the emulsion, and the ratio of templated ISPs to empty ISPs was determined by a fluorometric assay using fluorescently labeled oligonucleotides complementary to adapter sequences. The optimal templated signal ratio was determined to be between 10% and 40%. Positive templated ISPs were biotinylated during the emulsion PCR process, so that samples with an optimal templated signal ratio were then enriched with Dynabeads MyOne streptavidin C1 beads (Life Technologies). Individual nonbarcoded samples were sequenced on an Ion 314 chip, and four barcoded samples were multiplexed on an Ion 316 chip. Sequencing was performed on a PGM sequencer (Ion Torrent) using the Ion PGM 100 sequencing kit according to the manufacturer's instructions. Torrent Suite software version 2.0 (Ion Torrent) was used to parse barcoded reads, to align reads to the reference genome, and to generate run metrics, including chip loading efficiency and total read counts and quality. Variants were identified with Variant Caller software version 2.0, and target coverage was evaluated with Coverage Analysis software version 2.0 (both from Ion Torrent). Amino acid predictions were performed using the SIFT algorithm15Kumar P. Henikoff S. Ng P.C. Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm.Nat Protoc. 2009; 4: 1073-1081Crossref PubMed Scopus (5001) Google Scholar and PolyPhen2 software tools.16Adzhubei I.A. Schmidt S. Peshkin L. Ramensky V.E. Gerasimova A. Bork P. Kondrashov A.S. Sunyaev S.R. A method and server for predicting damaging missense mutations.Nat Methods. 2010; 7: 248-249Crossref PubMed Scopus (9279) Google Scholar Linear regression analysis and paired t-tests were performed with StatView software version 5.0 (SAS Institute, Cary, NC). Receiver operating characteristics (ROC) analysis was performed using pROC software.17Robin X. Turck N. Hainard A. Tiberti N. Lisacek F. Sanchez J.C. Müller M. pROC: an open-source package for R and S+ to analyze and compare ROC curves.BMC Bioinformatics. 2011; 12: 77Crossref PubMed Scopus (6058) Google Scholar A premixed pool (Ion AmpliSeq cancer primer pool; Ion Torrent) of 190 primer pairs (Supplemental Table S1) was used to create sequencing libraries from 45 FFPE-derived cancer specimens. Amplicons were generated from 10 ng of DNA in one multiplex PCR run on a standard cycler (Figure 1). Amplicons were phosphorylated, ligated to barcoded adapters, and quantified. Libraries from up to four specimens were combined and amplified onto ISPs by emulsion PCR. Templated particles were enriched, and then subjected to single-end 100-nt sequence analysis on Ion 316 chips (PGM sequencer; Ion Torrent). The specimens, which had previously been genotyped on a mass spectrometry–based platform (Sequenom MassArray),4Beadling C. Heinrich M.C. Warrick A. Forbes E.M. Nelson D. Justusson E. Levine J. Neff T.L. Patterson J. Presnell A. McKinley A. Winter L.J. Dewey C. Harlow A. Barney O. Druker B.J. Schuff K.G. Corless C.L. Multiplex mutation screening by mass spectrometry evaluation of 820 cases from a personalized cancer medicine registry [Erratum appeared in J Mol Diagn 2012, 14:41].J Mol Diagn. 2011; 13: 504-513Abstract Full Text Full Text PDF PubMed Scopus (71) Google Scholar collectively harbored 53 confirmed point mutations. An additional 19 indels ranging in length from 4 to 63 bp were also evaluated. Included were known actionable mutations in several cancer-related genes, including BRAF, EGFR, KIT, PDGFRA, KRAS, NRAS, and PIK3CA. Sequence coverage was assessed from the number and distribution of reads across target amplicons. An average of 4.0 million of the total 6.3 million addressable wells in the Ion 316 chip were consistently loaded with ISPs, and 3.7 million (92%) of these particles contained library templates. After subtraction of multiple-templated beads and poor-quality sequence reads, an average of 1.7 million reads were obtained across four samples per chip. The individual samples averaged 428,241 mapped sequence reads (range, 172,825 to 710,397). The mean read length was 76 bp, constituting an average of approximately 33 Mb of sequence per sample. Multiplex PCR–mediated target capture was very effective, as an average of 95% of the sequence reads mapped to targeted gene regions. The distribution of reads across the 190 amplicons was consistent across samples. With normalization to 400,000 reads per specimen, there was an average of 1941 reads per amplicon (range, 88 to 4948) (Figure 2A), and 181/190 (95%) amplicons averaged at least 400 reads (≥20% of the mean) (Figure 2B). Only one amplicon averaged <200 reads. Overall, 13,370 nt were interrogated in each of 45 FFPE tumor specimens, for a total of 601,650 nt (Figure 3A). A total of 1225 variant calls with an allele ratio of ≥1% were predicted to cause nonsynonymous coding changes; of these, 344 variants were unique. Notably, 54 of the 344 unique variants were identified recurrently at low variant allele ratios in multiple tumor and normal specimens (Supplemental Figure S1 and Supplemental Table S2). These 54 variants were present in at least 6 of the 45 tumors and/or at least 1 of 7 unmatched normal FFPE tissue specimens, and exhibited a mean variant allele ratio of 6.5 ± 4.2%. Collectively, these 54 recurrent variants accounted for 854 of the 1225 total variant calls. After these recurrent variants (0.4% of the total 13,370-bp target region) had been filtered out, 371 variant calls remained. Such filtering should not interfere with the detection of true variants at these positions. For example, a PIK3CA H0147R mutant with 44% G allele was readily detected, despite a recurrent T call at the same position (chr3:178952085; T mean allele ratio, 5.3% ± 1.6% in 31/45 tumor specimens and 5.8% ± 3.5% in 6/7 unmatched normal specimens) (Supplemental Table S2). An ROC curve was used to determine the optimal variant allele threshold for variant calling among the 371 filtered variants (Figure 3B). Using a variant allele ratio cutoff of 8%, 236 low-frequency variants were excluded, leaving 135 variant calls. This filter retained 92 confirmed variants and 34 germline polymorphisms, for 100% sensitivity (95% CI = 97.3% to 100.0%). The area under the curve was 0.955 (95% CI = 0.943 to 0.966). The selection of an 8% cutoff was deliberately biased toward exclusion of false-negative variant calls. An additional nine variants passed the 8% variant ratio filter, but were not confirmed by Sanger sequence analysis. These nine variants included one that was detected in four specimens within a homopolymeric stretch of three T nucleotides, and two variants in PIK3CA that likely resulted from pseudogene interference (Supplemental Table S3). All of the recurrent variants and false-positive calls involved single-nucleotide substitutions, with the exception of a single 3-nt insertion (Supplemental Table S2). All of the 53 known point mutations that had been detected by mass spectrometry–based assays were correctly identified, including mutations in two DNA samples prepared from LCM tumor cells. The mass spectrometry assays and the Ion AmpliSeq cancer panel showed a linear correlation ranging from 8% to 96% mutant allele ratio (Figure 3C and Supplemental Table S4). Bias was minimal, with the mass spectrometry assays exhibiting a mean allele ratio 3.4% greater than that detected by the amplicon sequencing (slope = 0.782, r = 0.92, P < 0.0001). Forty-three of the point mutations were evaluated both in singleplex analyses on Ion 314 chips and in fourplexes of samples on Ion 316 chips (Figure 3D). A good correlation was observed between the allelic ratios detected in samples sequenced individually or in fourplexes, indicating reproducibility of library preparation and variant detection (slope = 0.985, r = 0.94, P = 0.37; mean difference between groups, 0.7%). In addition to the known sequence variants, 27 new nonsynonymous mutations with >8% mutant allele frequency were detected in gene regions not covered by the mass spectrometry–based panel. These included 24 single-nucleotide substitutions and three deletions of 2 bp. All 27 mutations were confirmed by Sanger sequence analysis, including one variant detected in both a tumor sample and a LCM specimen prepared from the same tumor (Supplemental Table S5). All of the 19 known insertions and deletions were visible under manual inspection of the sequence alignments. However, only two were identified exactly by the variant caller software, and five others were called as mismatched nucleotides (Supplemental Table S6). A 9-bp insertion in the EGFR gene and a 6-bp insertion in the KIT gene that were not called correctly are located at the edge of amplicons, so that inclusion of additional flanking bases in these amplicons should improve performance. Fourteen of the indels were ≥9 bp long, including five indels ≥33 bp long. Difficulties with automated detection of deletions this size is a known problem,12Wagle N. Berger M.F. Davis M.J. Blumenstiel B. DeFelice M. Pochanard P. Ducar M. Van Hummelen P. MacConaill L.E. Hahn W.C. Meyerson M. Gabriel S.B. Garraway L.A. High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing.Cancer Discov. 2012; 2: 82-93Crossref PubMed Scopus (424) Google Scholar and manual inspection remains necessary for gene regions known to harbor indels. Our results demonstrate that sequencing libraries can be prepared by a simple, multiplexed amplicon approach that can be combined with semiconductor-based sequencing to provide rapid identification of cancer-related variants in FFPE-derived cancer specimens. Only small amounts (10 ng) of input DNA are required, and the system works well with DNA isolated from LCM tumor tissue. Target preparation, sequencing, and data analysis may all be performed within 48 hours. This workflow is considerably faster than the combination of hybrid capture-based target enrichment and optical detection-based sequencing that is currently in wide use, and is well suited to the rapid turnaround that is critical in the clinical laboratory. The speed of sequence analysis with the PGM system has been appreciated in clinical detection of microbial isolates,18Mellmann A. Harmsen D. Cummings C.A. Zentz E.B. Leopold S.R. Rico A. Prior K. Szczepanowski R. Ji Y. Zhang W. McLaughlin S.F. Henkhaus J.K. Leopold B. Bielaszewska M. Prager R. Brzoska P.M. Moore R.L. Guenther S. Rothberg J.M. Karch H. Prospective genomic characterization of the German enterohemorrhagic Escherichia coli O104:H4 outbreak by rapid next generation sequencing technology.PLoS One. 2011; 6: e22751Crossref PubMed Scopus (594) Google Scholar, 19Vogel U. Szczepanowski R. Claus H. Jünemann S. Prior K. Harmsen D. Ion Torrent Personal Genome Machine sequencing for genomic typing of Neisseria meningitidis for rapid determination of multiple layers of typing information.J Clin Microbiol. 2012; 50: 1889-1894Crossref PubMed Scopus (52) Google Scholar and coupling with the multiplex PCR cancer panel affords the same advantage to oncology testing. The percentage of on-target reads achieved with the amplicon panel is substantially greater than that reported with hybrid capture in FFPE specimens (95% versus 41%),12Wagle N. Berger M.F. Davis M.J. Blumenstiel B. DeFelice M. Pochanard P. Ducar M. Van Hummelen P. MacConaill L.E. Hahn W.C. Meyerson M. Gabriel S.B. Garraway L.A. High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing.Cancer Discov. 2012; 2: 82-93Crossref PubMed Scopus (424) Google Scholar and an average read depth of 1941 was attained. Of the 190 amplicons in the panel, 181 averaged at least 400 reads, sufficient to detect a variant allele ratio of approximately 5%. All but four amplicons averaged at least 250 reads, sufficient to detect an 8% mutant allele ratio. Using an 8% variant allele cutoff, we were able to achieve 100% sensitivity (95% CI = 97.3% to 100.0%) and 95.1% specificity (95% CI = 91.8% to 98.0%) in the detection of 53 known point mutations in 45 FFPE tumor specimens. The detection of variant allele ratios of <8% is obscured by the presence of apparently erroneous variant calls; 54 such recurrent variants were observed, with a mean variant allele ratio of approximately 6%. Of these recurrent calls, 19 were located in homopolymeric regions of at least three consecutive nucleotides, including 8 in regions of ≥5 consecutive nucleotides. Such errors in homopolymeric tracts have been noted on this platform.19Vogel U. Szczepanowski R. Claus H. Jünemann S. Prior K. Harmsen D. Ion Torrent Personal Genome Machine sequencing for genomic typing of Neisseria meningitidis for rapid determination of multiple layers of typing information.J Clin Microbiol. 2012; 50: 1889-1894Crossref PubMed Scopus (52) Google Scholar, 20Elliott A.M. Radecki J. Moghis B. Li X. Kammesheidt A. Rapid detection of the ACMG/ACOG-recommended 23 CFTR disease-causing mutations using Ion Torrent semiconductor sequencing.J Biomol Tech. 2012; 23: 24-30Crossref PubMed Scopus (58) Google Scholar However, these recurrent variants represent only approximately 0.4% of the total target region, and their presence should not prevent identification of true variants at these positions. Although the evaluation of matched normal specimens may aid in identification of these recurrent variants, it should not be necessary once the filter has been established. In addition, these recurrent calls are expected to become less prevalent with the adoption of improved reagent chemistries and paired-end read protocols. Thus, although LCM may at present be needed for low-purity tumor samples, the ability to detect lower-frequency variants can be expected to continue to improve, thereby reducing the requirement for high-purity specimens. Nineteen known insertions and deletions were all visible on manual inspection of aligned reads, although only two were identified exactly by the variant caller software. It is expected that manual inspection will be needed to evaluate common gene regions known to harbor large indels, such as in KIT and EGFR. Notably, 27 variants were observed in genes that had not previously been tested in these samples, and all were confirmed. These included variants in the tumor suppressor genes APC, ATM, PTEN, RB1, and TP53, which illustrates the power of this approach to extend beyond detection of oncogenic hotspot mutations. Finally, the amplicon approach is readily scalable, so that custom panels may be designed for any desired target regions. The flexibility, speed, and simplicity of this approach are well suited to a clinical diagnostic laboratory. Coupling multiplex amplicon-based library preparation with semiconductor sequencing thereby provides an important advance toward the implementation of routine clinical specimen genotyping and personalized cancer care. Download .pdf (.1 MB) Help with pdf files Supplemental Figure S1Variant allele ratios and prevalences across all tumor specimens. Libraries were prepared by multiplex PCR of 190 amplicons, and variants were detected by fourplex emulsion PCR and sequence analysis. Variants with allele ratios of ≥1% were tabulated across 45 FFPE tumor specimens, and the average variant allele frequencies were calculated for variants common to multiple specimens. Each data point represents a unique variant, and variants detected in at least one of seven unrelated FFPE normal specimens are marked in red. Download .xlsx (.03 MB) Help with xlsx files Supplemental Table S1 Download .xls (.03 MB) Help with xls files Supplemental Table S2 Download .xls (.02 MB) Help with xls files Supplemental Table S3 Download .xls (.03 MB) Help with xls files Supplemental Table S4 Download .xls (.03 MB) Help with xls files Supplemental Table S5 Download .xls (.02 MB) Help with xls files Supplemental Table S6