Study of Preanalytic and Analytic Variables for Clinical Next-Generation Sequencing of Circulating Cell-Free Nucleic Acid
2017; Elsevier BV; Volume: 19; Issue: 4 Linguagem: Inglês
10.1016/j.jmoldx.2017.03.003
ISSN1943-7811
AutoresMeenakshi Mehrotra, Rajesh R. Singh, Wei Chen, Richard S.P. Huang, Alaa A. Almohammedsalim, Bedia A. Barkoh, Crystal M. Simien, M. Rosario Hernandez, Carmen Behrens, Keyur P. Patel, Mark J. Routbort, Russell R. Broaddus, L. Jeffrey Medeiros, Ignacio I. Wistuba, Scott Kopetz, Rajyalakshmi Luthra,
Tópico(s)Single-cell and spatial transcriptomics
ResumoDetection of mutations in plasma circulating cell-free DNA (cfDNA) by next-generation sequencing (NGS) has opened up new possibilities for monitoring treatment response and disease progression in patients with solid tumors. However, implementation of cfDNA genotyping in diagnostic laboratories requires systematic assessment of preanalytical parameters and analytical performance of NGS platforms. We assessed the effects of peripheral blood collection tube and plasma separation time on cfDNA yield and integrity and performance of the Ion PGM, Proton, and MiSeq NGS platforms. cfDNA from 31 patients with diverse advanced cancers and known tumor mutation status was deep sequenced using targeted hotspot panels. Forty-five of 52 expected mutations and two additional mutations (KRAS p.Q61H and EZH2 p.Y646F) were detected in plasma through a custom bioinformatics pipeline. We observed comparable cfDNA concentration/integrity between collection tubes within 16 hours of plasma separation and equal analytical performance among NGS platforms, with 1% detection sensitivity for cfDNA genotyping. Detection of mutations in plasma circulating cell-free DNA (cfDNA) by next-generation sequencing (NGS) has opened up new possibilities for monitoring treatment response and disease progression in patients with solid tumors. However, implementation of cfDNA genotyping in diagnostic laboratories requires systematic assessment of preanalytical parameters and analytical performance of NGS platforms. We assessed the effects of peripheral blood collection tube and plasma separation time on cfDNA yield and integrity and performance of the Ion PGM, Proton, and MiSeq NGS platforms. cfDNA from 31 patients with diverse advanced cancers and known tumor mutation status was deep sequenced using targeted hotspot panels. Forty-five of 52 expected mutations and two additional mutations (KRAS p.Q61H and EZH2 p.Y646F) were detected in plasma through a custom bioinformatics pipeline. We observed comparable cfDNA concentration/integrity between collection tubes within 16 hours of plasma separation and equal analytical performance among NGS platforms, with 1% detection sensitivity for cfDNA genotyping. Determining the presence of molecular abnormalities in solid tumors is useful for diagnosis, selection of appropriate therapy, and monitoring tumor burden. Currently, diagnostic biopsy tissue specimens serve as the major source for tumor genotyping. However, this approach has substantial limitations because of intratumoral heterogeneity, cost, time, and risk associated with testing multiple biopsy specimens.1Diaz Jr., L.A. Bardelli A. Liquid biopsies: genotyping circulating tumor DNA.J Clin Oncol. 2014; 32: 579-586Crossref PubMed Scopus (1537) Google Scholar Furthermore, the impaired medical condition of many patients with advanced cancer, the inaccessible location of some tumors, and logistical considerations limit the feasibility of obtaining a biopsy in many circumstances. The so-called liquid biopsy, especially plasma, has emerged recently as an alternative to surgical biopsy that allows real-time assessment of molecular alterations in patients with solid tumors.2Cai X. Janku F. Zhan Q. Fan J.B. Accessing genetic information with liquid biopsies.Trends Genet. 2015; 31: 564-575Abstract Full Text Full Text PDF PubMed Scopus (95) Google Scholar More important, liquid biopsies such as plasma are minimally invasive, being easily obtained through a simple blood draw, which makes their use relatively inexpensive and readily scalable. Moreover, liquid biopsy can provide temporal measurements of tumor burden and can identify specific mutations that arise during therapy, provide early evidence of recurrence, and highlight mechanisms that underlie resistance to therapy.1Diaz Jr., L.A. Bardelli A. Liquid biopsies: genotyping circulating tumor DNA.J Clin Oncol. 2014; 32: 579-586Crossref PubMed Scopus (1537) Google Scholar, 3Esposito A. Bardelli A. Criscitiello C. Colombo N. Gelao L. Fumagalli L. Minchella I. Locatelli M. Goldhirsch A. Curigliano G. Monitoring tumor-derived cell-free DNA in patients with solid tumors: clinical perspectives and research opportunities.Cancer Treat Rev. 2014; 40: 648-655Abstract Full Text Full Text PDF PubMed Scopus (98) Google Scholar Potential sources of tumor genetic information in the circulation include cell-free circulating tumor DNA (cfDNA), circulating tumor RNA, circulating tumor cells, and exosomes. Among these, cfDNA offers an attractive option as a diagnostic, predictive, and prognostic biomarker for assessing tumor genetic information because of its stability and easy availability.4Gormally E. Caboux E. Vineis P. Hainaut P. Circulating free DNA in plasma or serum as biomarker of carcinogenesis: practical aspects and biological significance.Mutat Res. 2007; 635: 105-117Crossref PubMed Scopus (368) Google Scholar, 5Ulz P. Auer M. Heitzer E. Detection of circulating tumor DNA in the blood of cancer patients: an important tool in cancer chemoprevention.Methods Mol Biol. 2016; 1379: 45-68Crossref PubMed Scopus (16) Google Scholar Plasma cfDNA levels have been shown to correlate with tumor size, degree of tumor invasion, disease stage, survival, and disease progression under therapy.6Bettegowda C. Sausen M. Leary R.J. Kinde I. Wang Y. Agrawal N. et al.Detection of circulating tumor DNA in early- and late-stage human malignancies.Sci Transl Med. 2014; 6: 224ra24Crossref PubMed Scopus (2989) Google Scholar cfDNA is fragmented to an average length of 140 to 170 bp and is present in limited quantities per milliliter of peripheral blood, of which only a fraction may be tumor-derived DNA with diagnostically relevant mutations.7Frenel J.S. Carreira S. Goodall J. Roda D. Perez-Lopez R. Tunariu N. Riisnaes R. Miranda S. Figueiredo I. Nava-Rodrigues D. Smith A. Leux C. Garcia-Murillas I. Ferraldeschi R. Lorente D. Mateo J. Ong M. Yap T.A. Banerji U. Gasi Tandefelt D. Turner N. Attard G. de Bono J.S. Serial next-generation sequencing of circulating cell-free DNA evaluating tumor clone response to molecularly targeted drug administration.Clin Cancer Res. 2015; 21: 4586-4596Crossref PubMed Scopus (166) Google Scholar, 8El Messaoudi S. Mouliere F. Du Manoir S. Bascoul-Mollevi C. Gillet B. Nouaille M. Fiess C. Crapez E. Bibeau F. Theillet C.G. Mazard T. Pezet D. Mathonnet M. Ychou M. Thierry A.R. Circulating DNA as a strong multi-marker prognostic tool for metastatic colorectal cancer patient management care.Clin Cancer Res. 2016; 22: 3067-3077Crossref PubMed Scopus (122) Google Scholar Various methods have been developed to detect the presence of low-level tumor-associated mutations in cfDNA of cancer patients.1Diaz Jr., L.A. Bardelli A. Liquid biopsies: genotyping circulating tumor DNA.J Clin Oncol. 2014; 32: 579-586Crossref PubMed Scopus (1537) Google Scholar However, clinical implementation of these cfDNA-based mutation tests has been impeded by a lack of robust preanalytical, analytical, and clinical validation studies. A comparison of frequently used next-generation sequencing (NGS)–based platforms in molecular diagnostic laboratories with respect to analytical sensitivity and specificity for cfDNA mutation analysis is lacking. In addition, variability in blood collection and handling can have substantial effects on quantitative measurement of cfDNA. For instance, cell lysis after venipuncture when using standard peripheral blood (PB) collection tubes necessitates separation of plasma within a short period after collection to reduce contamination of cfDNA with cellular DNA and increase the chances of detecting low-level tumor-associated mutations.9Toro P.V. Erlanger B. Beaver J.A. Cochran R.L. VanDenBerg D.A. Yakim E. Cravero K. Chu D. Zabransky D.J. Wong H.Y. Croessmann S. Parsons H. Hurley P.J. Lauring J. Park B.H. Comparison of cell stabilizing blood collection tubes for circulating plasma tumor DNA.Clin Biochem. 2015; 48: 993-998Crossref PubMed Scopus (77) Google Scholar, 10El Messaoudi S. Rolet F. Mouliere F. Thierry A.R. Circulating cell free DNA: preanalytical considerations.Clin Chim Acta. 2013; 424: 222-230Crossref PubMed Scopus (399) Google Scholar For these reasons, PB collection tubes with cell-stabilization agents that prevent cell lysis for several days and alleviate the need for immediate plasma preparation after PB collection are recommended. However, changing from routinely used EDTA-based PB collection tubes to cell-stabilizing tubes for cfDNA molecular analysis requires a process change in phlebotomy units and adds costs to the testing. A systematic analysis of the impact of preanalytical variables, such as PB collection tubes and time interval from collection of PB to plasma separation on downstream mutation testing, is required before cfDNA genotyping assays can be implemented in molecular diagnostic laboratories for patient care. In this study, we assessed the effect of different collection tubes on cfDNA yield and integrity and on downstream mutation analysis by gene panels on three NGS-based platforms (Ion PGM, Ion Proton, and MiSeq). We also compared the sensitivity and specificity of these three NGS platforms, which are widely used in molecular diagnostic laboratories, for detection of somatic mutations in cfDNA using a custom bioinformatics workflow in a clinical laboratory environment. To our knowledge, this is the first study to assess in parallel the impact of routinely used EDTA blood collection tubes versus cell-stabilizing collection tubes on cfDNA mutation analysis using NGS platforms for implementation in a Clinical Laboratory Improvement Amendments–certified laboratory environment. The study group included 31 patients with diverse advanced cancers and known tumor mutation status (Supplemental Table S1). The tumor mutations were identified using DNA derived from formalin-fixed, paraffin-embedded tumor tissue sections by using the semiconductor-based Ion PGM NGS platform with Ampliseq Cancer Hotspot Panel version 2 performed in our Clinical Laboratory Improvement Amendments–certified molecular diagnostic laboratory. The tumors included 28 carcinomas and 3 brain tumors. The carcinomas comprised eight endometrial adenocarcinomas (26%), seven colon adenocarcinomas (23%), three breast adenocarcinomas (10%), three prostate adenocarcinomas (10%), one ovarian carcinoma (3%), one lung adenocarcinoma (3%), one liver adenocarcinoma (3%), one esophagus signet ring adenocarcinoma (3%), one appendiceal mucinous adenocarcinoma (3%), one squamous cell carcinoma of the tongue (3%), and one parathyroid carcinoma (3%). The brain tumors were astrocytoma or ganglioneuroblastoma (10%). Tumor samples were obtained by resection (55%), biopsy (32%), or fine-needle aspiration (13%) from a variety of anatomical sites: lymph node (26%), liver (23%), abdominal tissue (10%), rectum (10%), brain (6%), uterus (6%), ovary (3%), prostate (3%), head and neck (3%), colon (3%), femur (3%), and esophagus (3%). There were 14 primary and 17 metastatic tumors. The patients had undergone various modalities of treatment before tumor specimens were obtained, including chemotherapy (n = 5), chemotherapy and radiation (n = 2), surgery and chemotherapy (n = 17), or surgery, chemotherapy, and radiation (n = 7). This study protocol was approved by the Institutional Review Board of MD Anderson Cancer Center and is consistent with international ethical standards on human subjects research. Informed consent was obtained from each study participant. Peripheral blood was drawn from each patient at the same time into two different blood collection tubes: a regular K3-EDTA tube (BD Vacutainer; Becton Dickinson, Franklin Lakes, NJ) and a Cell-Free DNA BCT tube (Streck, Inc., Omaha, NE), which contains cell-stabilizing agents that prevent cell lysis. Blood samples were mixed by inverting tubes 10 times and subjected to centrifugation at 2000 × g for 10 minutes at room temperature. Plasma was separated from blood at different time points (2 to ≥24 hours). The plasma layer was carefully removed without disturbing the buffy coat, transferred to a new vial, and subjected to centrifugation at 2000 × g for 10 minutes at room temperature to remove any residual cells. cfDNA was extracted from a 3-mL plasma sample using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Valencia, CA) by following the manufacturer's instructions. Elution was performed in 50 μL, and isolated cfDNA was kept at −20°C. DNA was quantified by using the Qubit dsDNA HS Assay (Life Technologies, Illkirch, France). The size distribution of the cfDNA was evaluated by Alu quantitative PCR (AluqPCR) that targets 300 bp long Alu repetitive sequences, which are randomly interspersed across the human genome.11Fawzy A. Sweify K.M. El-Fayoumy H.M. Nofal N. Quantitative analysis of plasma cell-free DNA and its DNA integrity in patients with metastatic prostate cancer using ALU sequence.J Egypt Natl Canc Inst. 2016; 28: 235-242Abstract Full Text Full Text PDF PubMed Scopus (41) Google Scholar Two primer sets were used to amplify Alu sequences of 115-bp DNA (Alu115) and 247-bp fragments (Alu247). Alu115 primers are nested within the region spanned by Alu247. The primers for amplification of Alu115 are as follows: forward, 5′-CCTGAGGTCAGGAGTTCGAG-3′; and reverse, 5′-CCCGAGTAGCTGGGATTACA-3′. The primers for amplification of Alu247 are as follows: forward, 5′-GTGGCTCACGCCTGTAATC-3′; and reverse, 5′-CAGGCTGGAGTGCAGTGG-3′. Alu247 amplifies the larger fragments (>247 bp), whereas Alu115 amplifies both short cfDNA fragments (140 bp) and large cellular DNA fragments (≥247 bp). Each AluqPCR was performed in a 20-μL reaction consisting of 1 × iTaq Universal SYBR Green Supermix (Bio-Rad, Hercules, CA), 200 nmol/L each forward and reverse primers, and 1 μL of DNA sample. PCR amplification was performed at 95°C for 3 minutes, followed by 35 cycles at 95°C for 5 seconds and 62°C for 30 seconds. Each run included fivefold dilution with an external standard (22, 2.2, 0.22, 0.022, and 0.0022 ng) of human genomic DNA (Promega, Fitchburg, WI) and a no template control. Each reaction was performed in duplicate on an ABI 7900 fast HT (Thermo Fisher Scientific, Santa Clara, CA). The DNA integrity index for each cfDNA sample was calculated as the ratio of AluqPCR results (Alu247/Alu115). For Ampliseq Cancer Hotspot Panel sequencing on the Ion PGM platform (Thermo Fisher Scientific), libraries were prepared from 10 ng cfDNA using the Ion Torrent Ampliseq 2.0 kit (Thermo Fisher Scientific), according to the manufacturer's instructions. Samples were barcoded and quantified by qPCR using the Ion Xpress Barcode Adapter 1–96 kit and the Ion Library TaqMan quantitation kit (Thermo Fisher Scientific), respectively. Libraries from four samples were pooled at different concentrations (range, 5 to 20 pmol/L) for ultra-deep sequencing on the Ion PGM 318 version 2 chip. Pooled libraries were clonally amplified on Ion Spheres (Thermo Fisher Scientific). Emulsion PCR using the Ion PGM Template OT2 kit version 2 and the Ion One Touch 2 system (Thermo Fisher Scientific) were performed per the manufacturer's instructions. Ion sphere particles were enriched by using the Ion Torrent OneTouch ES (Thermo Fisher Scientific). Enriched ion sphere particles were loaded on a PGM 318 chip and sequenced on the Ion PGM platform by using the Ion PGM sequencing kit. To increase scalability in terms of number of samples per run and to perform deep sequencing, we sequenced cfDNA libraries generated by using the Ion Torrent Ampliseq 2.0 kit and Ion Ampliseq Cancer Hotspot Panel on the Ion Proton platform (Thermo Fisher Scientific). The 5 pmol/L pooled libraries were clonally amplified, and emulsion PCR was performed by using the Ion PI HI-Q Template OT2 kit and Ion One Touch 2 system (Thermo Fisher Scientific), according to the manufacturer's instructions. Pooled libraries with 31 samples per run were sequenced on the Ion Proton using the Ion HI-Q PI Chip version 3 and Ion PI HI-Q sequencing 200 kit (Thermo Fisher Scientific). Sequencing reads were aligned with the reference genome and base calling by using Ion Torrent Suite software version 4.4.2 for Ion PGM and Ion Proton (Thermo Fisher Scientific). Human genome build 19 was used as the reference for alignment. Identification of sequence variants was facilitated by Ion Torrent Variant Caller Plugin software version 4.4-r76860, and coverage of each amplicon was determined by the Coverage Analysis Plugin software version 4.4-r77897. The integrative genomic viewer (IGV) was used to visualize the read alignment and the presence of variants against the reference genome as well as to confirm variant calls by checking for strand biases and sequencing errors.12Thorvaldsdottir H. Robinson J.T. Mesirov J.P. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration.Brief Bioinform. 2013; 14: 178-192Crossref PubMed Scopus (4916) Google Scholar Custom software developed in-house (OncoSeek) was used to interface the data generated by the Ion Torrent Variant Caller with the IGV, filter repeat errors due to nucleotide homopolymer regions, compare replicate samples, and annotate the sequencing information.13Singh R.R. Patel K.P. Routbort M.J. Reddy N.G. Barkoh B.A. Handal B. Kanagal-Shamanna R. Greaves W.O. Medeiros L.J. Aldape K.D. Luthra R. Clinical validation of a next-generation sequencing screen for mutational hotspots in 46 cancer-related genes.J Mol Diagn. 2013; 15: 607-622Abstract Full Text Full Text PDF PubMed Scopus (266) Google Scholar A cutoff of 300,000 reads with a quality score of AQ20 (one misaligned base per 100 bases) and minimum sequencing depth of 250× was used as a measure of successful sequencing of a sample. Sequencing results, mutations, and their respective allelic frequencies observed in cfDNA were compared to those identified in tumor tissue to establish concordance. For MiSeq (Illumina, San Diego, CA) sequencing, libraries were prepared from 20 ng cfDNA samples using the Accel amplicon 56G oncology targeted panel (Swift Biosciences, Ann Arbor, MI). This panel offers comprehensive and hotspot coverage of 56 clinically relevant oncology-related genes using a 263-amplicon design to generate multiplex libraries. The panel covers all hotspots covered by the Ion Torrent Ampliseq Cancer Hotspot Panel along with hotspots in six additional genes (MSH6, DNMT3A, DDR2, MAP2K1, TSC1, and FOXL2). A multiplex PCR was performed using 20 ng cfDNA, according to the manufacturer's instructions. After multiplex PCR, cleanup with SPRI select beads (Beckman Coulter, Indianapolis, IN) was followed by indexing for 20 minutes at 37°C, according to the manufacturer's instructions. The indexing step was followed by polyethylene glycol sodium chloride cleanup and elution in 20 μL low Tris-EDTA buffer. The library was quantified by qPCR with the KAPA library quantification kit (KAPA Biosystems, Wilmington, MA) for the Illumina platform, according to the manufacturer's instructions. Quantified libraries from 12 cfDNA samples were pooled, and the 20 pmol/L library was loaded on MiSeq standard flow cell for paired-end sequencing using the MiSeq 300 Cycles version 2 kit (Illumina). A custom bioinformatics workflow was developed using open-source software, including cutadapt version 1.8.1, BWA-MEM version 0.7.12, Fastqc version 0.10.1, Picard version 1.119, SAMtools version 0.1.19, BEDtools version 2.23.0, LoFreq version 2.1.2, GATK version 3.4-46, and Pindel version 0.2.4t and in-house scripts.14Ye K. Schulz M.H. Long Q. Apweiler R. Ning Z. Pindel: a pattern growth approach to detect break points of large deletions and medium sized insertions from paired-end short reads.Bioinformatics. 2009; 25: 2865-2871Crossref PubMed Scopus (1396) Google Scholar, 15Wilm A. Aw P.P. Bertrand D. Yeo G.H. Ong S.H. Wong C.H. Khor C.C. Petric R. Hibberd M.L. Nagarajan N. LoFreq: a sequence-quality aware, ultra-sensitive variant caller for uncovering cell-population heterogeneity from high-throughput sequencing datasets.Nucleic Acids Res. 2012; 40: 11189-11201Crossref PubMed Scopus (682) Google Scholar, 16McKenna A. Hanna M. Banks E. Sivachenko A. Cibulskis K. Kernytsky A. Garimella K. Altshuler D. Gabriel S. Daly M. DePristo M.A. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data.Genome Res. 2010; 20: 1297-1303Crossref PubMed Scopus (14878) Google Scholar Briefly, the fastqc files from the cfDNA runs on MiSeq were run through cutadapt to remove primer sequences and then through BWA-MEM to align with the hg19 human reference genome. The sequences were sorted; read groups were added and indexed. GATK was used for insertion/deletion realignment and base recalibration. Both GATK and BED tools were used for coverage analysis. Variant calling was performed by GATK, LoFreq, and Pindel. Quality control reports were generated using a modified version of Fastqc with related modules only. Analyses were performed on a high-performance computing cluster, with workflows built in IBM Platform Project Manager and published on the IBM Platform Application Center version 9.1.3 interface (IBM, Armonk, NY). The workflow was deployed on a Redhat Enterprise Linux cluster (RHEL 6.4) with Intel 2.6 GHz compute nodes of 24 CPU cores and 384 GB of RAM each (Red Hat, Raleigh, NC). Trimmomatic version 0.33 was used to trim any possible existing universal Illumina adapters, and snpEff to functionally annotate the variants and predict the effect of the proteins.17Bolger A.M. Lohse M. Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data.Bioinformatics. 2014; 30: 2114-2120Crossref PubMed Scopus (28313) Google Scholar The settings for Trimmomatic, cutadapt, and snpEff were adapted from Swift Bioscience's bioinformatics pipeline, according to manufacturer guidelines. The settings of most software used for analysis were default, except for the GATK haplotype caller, for which optimized settings (-dt none -stand_call_conf 10 -stand_emit_conf 10 -mbq 10) were used. Specifically, no duplication was marked or removed before the insertion/deletion realignment, base recalibration, or variant calling. dbsnp_138.hg19.vcf and the BED file were included for on-target variant calling. Other customized settings were -dt none, -stand_call_conf 10, -stand_emit_conf 10, and -mbq 10 for no downsampling, standard call confidence, standard emit confidence, and minimum base quality, respectively. For variant calling a low- and high-stringency variant frequency filters as ≥1%, ≥2%, and ≥5%, a minimum 250× sequencing depth and 25× variant coverage were applied to current bioinformatics workflow. The sensitivity of Ion PGM, Ion Proton, and MiSeq was determined by sequencing serially diluted cfDNA samples positive for KRAS p.G12D (allele frequency, 36%) and SMAD4 p.R361C (allele frequency, 45.09%). The positive cfDNA was diluted with a cfDNA sample negative for these mutations to obtain 50%, 25%, 12.5%, 6.25%, 3.15%, 1.5%, or 0.75% positive (mutated) DNA in wild-type DNA. Samples were tagged with different barcodes and sequenced on the same chip on Ion PGM (12 or 4 samples per run on Ion PGM 318 chip), Ion Proton (31 samples per run on Ion HI-Q PI Chip), or MiSeq (12 samples per run on MiSeq standard flow cell). Statistical analysis of correlation between a pair of selected data for mutant allelic frequency in cfDNA was performed using Pearson correlation coefficient, and P < 0.05 calculated by two-tailed test was considered as significant. A paired t-test and Wilcoxon rank sum test were also performed for statistical analysis to compare results among data sets. Peripheral blood samples were collected from 31 patients with advanced cancer of various types and known tumor mutation status (Supplemental Table S1). The median age for the patient cohort was 57.5 years (range, 32 to 84 years), and 20 (65%) were women. At last follow-up, 24 (77%) of patients were alive and 7 (23%) had deceased. The overall median survival was 66.3 months (range, 7 to 172 months). The median time from tumor biopsy to plasma collection was 284 days (range, 0 to 1490 days). Fifty-two mutations in a number of genes were identified in the tumor tissue by using the Ion Ampliseq Cancer Hotspot Panel on the Ion PGM platform (Supplemental Table S2). The median number of altered mutations per tumor was 2 (range, 1 to 4). The most frequently mutated genes in this cohort were TP53 (42.3%), KRAS (17.0%), PIK3CA (13.2%), and APC (5.7%). The yields of cfDNA in plasma from PB collected in Streck and K3-EDTA tubes separated at different time intervals (2, 4, 16, and ≥24 hours from collection of PB) were comparable until 16 hours. The mean cfDNA yields from Streck tubes were 25.1 ng/mL (range, 7.8 to 65.5 ng/mL) at 2 hours, 24.04 ng/mL (range, 6.8 to 87.8 ng/mL) at 4 hours, 34.2 ng/mL (range, 8.7 to 89.8 ng/mL) at 16 hours, and 22 ng/mL (range, 8.8 to 40.7 ng/mL) at ≥24 hours. The yields from K3-EDTA tubes were 20.3 ng/mL (range, 7.9 to 54.5 ng/mL) at 2 hours, 23.78 ng/mL (range, 7.8 to 81.3 ng/mL) at 4 hours, 41.74 ng/mL (range, 9.8 to 95.3 ng/mL) at 16 hours, and 37.03 ng/mL (range, 17.9 to 56.7 ng/mL) at time points of ≥24 hours (Figure 1A and Supplemental Table S3). The cfDNA yields for the two collection tubes showed good correlation (R2 = 0.8200, P < 0.0001) (Figure 1B). The cfDNA yield was slightly higher for the K3-EDTA tubes than for the Streck tubes at ≥24 hours (P < 0.001). cfDNA yield varied among different tumor types tested, with a median value of 43 ng/mL (range, 10.6 to 191.7 ng/mL) (Supplemental Table S4). cfDNA quantitated using Alu115qPCR revealed similar quantity of cfDNA yield as obtained by qubit (Supplemental Table S5). DNA integrity of plasma cfDNA, calculated as the ratio of AluqPCR values (Alu247/Alu115), revealed slightly higher levels of larger fragments in the cfDNA from K3-EDTA tubes (range, 0.39 to 0.55) than in the cfDNA from Streck tubes (range, 0.35 to 0.43; P < 0.01) (Figure 1C). Qualitative analysis of cfDNA extracted from different blood collection tubes at different time intervals by the Agilent bioanalyzer showed the average size of cfDNA fragments ranged from 100 to 120 bp at 2, 4, and 16 hours, with no high-molecular-weight cellular DNA (Supplemental Figure S1). cfDNA libraries prepared by the Ampliseq Cancer Hotspot Panel were subjected to deep sequencing on the Ion PGM sequencer by running four samples/Ion PGM 318 chip. Barcoded cfDNA libraries were pooled and sequenced at 20, 12, 10, and 5 pmol/L concentrations to optimize the library concentration required for high-depth sequencing. We observed 5,423,763 total reads at a 5 pmol/L library loading concentration and 2,726,743 at 20 pmol/L, and the polyclonal frequency at 5 pmol/L (32%) was low compared to that at 20 pmol/L (67%) (Supplemental Figure S2, A and B). Polyclonal frequency reflects the percentage of the total ion sphere beads that are polyclonal (ie, having two or more distinct DNA templates). These noninformative reads are filtered by bioinformatics analysis pipeline used for variant calling and hence resulting in reduced number of total reads and mean depth. A total of 1,298,334 mapped reads with an average mean sequencing depth of 5918× per sample were obtained at 5 pmol/L (Supplemental Figure S2C). A comparison of coverage and allelic frequency for four variants in four samples at 20, 12, 10, and 5 pmol/L demonstrated 2.5-fold greater coverage at 5 pmol/L, with no effect on variant allelic frequency (VAF) (Supplemental Figure S2, D and E). Four plasma cfDNA libraries prepared from samples collected in K3-EDTA and Streck tubes were pooled at 5 pmol/L and sequenced per run. An average depth of 3000× per sample was obtained. Thirteen variant calls present in tumor tissues were compared and correlated across eight corresponding patient plasma samples (Supplemental Figure S3A and Supplemental Table S6). Mutant allelic frequency in the two collection tube types showed 100% concordance, and correlation analysis yielded an R2 = 0.95 (P < 0.0001) (Supplemental Figure S3B). cfDNA libraries were generated by the Ion Ampliseq Library kit 2.0 and Ampliseq Cancer Hotspot Panel. cfDNA libraries (10, 16, 24, or 31 samples per run) were pooled at 5 and 10 pmol/L and sequenced on the Ion Proton platform. Total reads for runs with different numbers of samples did not vary significantly. An average of 2,349,325 reads per sample with a quality score of AQ20 (1 misaligned base per 100 bases) was achieved at 31 samples per run. The Ion Proton platform yielded an average 15.7-fold more total reads than the Ion PGM platform, with less polyclonal frequency (Supplemental Figure S4, A and B) and a 2.3-fold greater average mean depth (8860× on the Ion Proton, 31 samples per run versus 3893× on Ion PGM, four samples per run) (Figure 2A, Supplemental Figure S4C, and Supplemental Table S7). A total of 78,106,120 and 4,258,166 reads were obtained on Ion Proton and Ion PGM, respectively (Supplemental Figure S4A). Correlation of allelic frequencies of the calls between the two platforms yielded an R2 = 0.9972 (P < 0.0001) with 100% concordance (Figure 2B). Twelve cfDNA libraries were pooled at 20 pmol/L and sequenced per flow cell at an average cluster density of 600 K/mm2. Of the clusters, 97% passed the filter from which the sequencing information was obtained. The average total sequencing output for the runs was 2.91 Gb, with a sequencing quality score >Q30 (1 misaligned base per 1000 bases). The average total sequencing reads obtained from the runs was 9,745,069. The identified sequencing reads or average reads passing the filter was 9,297,949, indicating that 83.4% of the reads were identified with the barcode indexes. The average sequencing depth per sample in these runs was 3000×. To measure the mutation detection sensitivity, we serially diluted a cfDNA sample positive for
Referência(s)