Title: Enhanced Ratio of Signals Enables Digital Mutation Scanning for Rare Allele Detection
Abstract: The use of droplet digital PCR (ddPCR) for low-level DNA mutation detection in cancer, prenatal diagnosis, and infectious diseases is growing rapidly. However, although ddPCR has been implemented successfully for detection of rare mutations at pre-determined positions, no ddPCR adaptation for mutation scanning exists. Yet, frequently, clinically relevant mutations reside on multiple sequence positions in tumor suppressor genes or complex hotspot mutations in oncogenes. Here, we describe a combination of coamplification at lower denaturation temperature PCR (COLD-PCR) with ddPCR that enables digital mutation scanning within approximately 50-bp sections of a target amplicon. Two FAM/HEX-labeled hydrolysis probes matching the wild-type sequence are used during ddPCR. The ratio of FAM/HEX-positive droplets is constant when wild-type amplicons are amplified but deviates when mutations anywhere under the FAM or HEX probes are present. To enhance the change in FAM/HEX ratio, we employed COLD-PCR cycling conditions that enrich mutation-containing amplicons anywhere on the sequence. We validated COLD-ddPCR on multiple mutations in TP53 and in EGFR using serial mutation dilutions and cell-free circulating DNA samples, and demonstrate detection down to approximately 0.2% to 1.2% mutation abundance. COLD-ddPCR enables a simple, rapid, and robust two-fluorophore detection method for the identification of multiple mutations during ddPCR and potentially can identify unknown DNA variants present in the target sequence. The use of droplet digital PCR (ddPCR) for low-level DNA mutation detection in cancer, prenatal diagnosis, and infectious diseases is growing rapidly. However, although ddPCR has been implemented successfully for detection of rare mutations at pre-determined positions, no ddPCR adaptation for mutation scanning exists. Yet, frequently, clinically relevant mutations reside on multiple sequence positions in tumor suppressor genes or complex hotspot mutations in oncogenes. Here, we describe a combination of coamplification at lower denaturation temperature PCR (COLD-PCR) with ddPCR that enables digital mutation scanning within approximately 50-bp sections of a target amplicon. Two FAM/HEX-labeled hydrolysis probes matching the wild-type sequence are used during ddPCR. The ratio of FAM/HEX-positive droplets is constant when wild-type amplicons are amplified but deviates when mutations anywhere under the FAM or HEX probes are present. To enhance the change in FAM/HEX ratio, we employed COLD-PCR cycling conditions that enrich mutation-containing amplicons anywhere on the sequence. We validated COLD-ddPCR on multiple mutations in TP53 and in EGFR using serial mutation dilutions and cell-free circulating DNA samples, and demonstrate detection down to approximately 0.2% to 1.2% mutation abundance. COLD-ddPCR enables a simple, rapid, and robust two-fluorophore detection method for the identification of multiple mutations during ddPCR and potentially can identify unknown DNA variants present in the target sequence. In the era of personalized medicine, mutation detection methods that target mutations known to influence therapy response or clinical outcome are of great interest. Although real-time PCR methodologies have been described and are widely used for detecting mutations in clinical samples,1Bernard P.S. Wittwer C.T. Real-time PCR technology for cancer diagnostics.Clin Chem. 2002; 48: 1178-1185PubMed Google Scholar, 2Tyagi S. Kramer F.R. Molecular beacons: probes that fluoresce upon hybridization.Nat Biotechnol. 1996; 14: 303-308Crossref PubMed Scopus (3711) Google Scholar, 3Li J. Wang F. Mamon H. Kulke M.H. Harris L. Maher E. Wang L. Makrigiorgos G.M. Antiprimer quenching-based real-time PCR and its application to the analysis of clinical cancer samples.Clin Chem. 2006; 52: 624-633Crossref PubMed Scopus (30) Google Scholar, 4Amicarelli G. Shehi E. Makrigiorgos G.M. Adlerstein D. FLAG assay as a novel method for real-time signal generation during PCR: application to detection and genotyping of KRAS codon 12 mutations.Nucleic Acids Res. 2007; 35: e131Crossref PubMed Scopus (50) Google Scholar interest in digital PCR5Vogelstein B. Kinzler K.W. Digital PCR.Proc Natl Acad Sci U S A. 1999; 96: 9236-9241Crossref PubMed Scopus (1388) Google Scholar is rising in view of the unique aspects of the technology and the emergence of commercial droplet digital PCR (ddPCR) platforms.5Vogelstein B. Kinzler K.W. Digital PCR.Proc Natl Acad Sci U S A. 1999; 96: 9236-9241Crossref PubMed Scopus (1388) Google Scholar, 6Taly V. Pekin D. Benhaim L. Kotsopoulos S.K. Le Corre D. Li X. Atochin I. Link D.R. Griffiths A.D. Pallier K. Blons H. Bouche O. Landi B. Hutchison J.B. Laurent-Puig P. Multiplex picodroplet digital PCR to detect KRAS mutations in circulating DNA from the plasma of colorectal cancer patients.Clin Chem. 2013; 59: 1722-1731Crossref PubMed Scopus (416) Google Scholar, 7Day E. Dear P.H. McCaughan F. Digital PCR strategies in the development and analysis of molecular biomarkers for personalized medicine.Methods. 2013; 59: 101-107Crossref PubMed Scopus (160) Google Scholar, 8Hindson C.M. Chevillet J.R. Briggs H.A. Gallichotte E.N. Ruf I.K. Hindson B.J. Vessella R.L. Tewari M. Absolute quantification by droplet digital PCR versus analog real-time PCR.Nat Methods. 2013; 10: 1003-1005Crossref PubMed Scopus (1047) Google Scholar ddPCR has been implemented in a variety of fields such as cancer biomarker and viral load detection, fetal screening, or library quantification for next-generation sequencing.9Huggett J.F. Foy C.A. Benes V. Emslie K. Garson J.A. Haynes R. Hellemans J. Kubista M. Mueller R.D. Nolan T. Pfaffl M.W. Shipley G.L. Vandesompele J. Wittwer C.T. Bustin S.A. The digital MIQE guidelines: Minimum Information for Publication of Quantitative Digital PCR Experiments.Clin Chem. 2013; 59: 892-902Crossref PubMed Scopus (634) Google Scholar One of the most common ddPCR applications is in the detection of known DNA variants present within a large excess of wild-type DNA, for instance in DNA from heterogeneous samples that harbor subclonal populations of mutated tumor cells.10Diehl F. Schmidt K. Choti M.A. Romans K. Goodman S. Li M. Thornton K. Agrawal N. Sokoll L. Szabo S.A. Kinzler K.W. Vogelstein B. Diaz Jr., L.A. Circulating mutant DNA to assess tumor dynamics.Nat Med. 2008; 14: 985-990Crossref PubMed Scopus (2036) Google Scholar In ddPCR, the amplification reaction is compartmentalized into microscopic emulsion-based droplets containing at most a few target molecules per droplet. By segregating the interrogated sample, the effect of in-droplet target competition is reduced, which translates into increased assay discrimination and facile determination of wild-type versus mutant status.11Hindson B.J. Ness K.D. Masquelier D.A. Belgrader P. Heredia N.J. Makarewicz A.J. et al.High-throughput droplet digital PCR system for absolute quantitation of DNA copy number.Anal Chem. 2011; 83: 8604-8610Crossref PubMed Scopus (1907) Google Scholar However, as currently applied, ddPCR can only be used to detect mutations at known sequence positions. ddPCR incorporates two reporter probes, one mutant-specific and one wild-type, because of the requirement to account for PCR-amplification variability among droplets. This approach, by design, allows only the detection of previously known mutations. In cancer, tumor suppressor genes such as TP53 harbor mutations that are scattered throughout the gene as opposed to oncogenes that usually carry mutations located in specific hotspots.12Marsh D.J. Theodosopoulos G. Howell V. Richardson A.L. Benn D.E. Proos A.L. Eng C. Robinson B.G. Rapid mutation scanning of genes associated with familial cancer syndromes using denaturing high-performance liquid chromatography.Neoplasia. 2001; 3: 236-244Abstract Full Text PDF PubMed Scopus (27) Google Scholar Although mutation scanning methods based on amplicon fluorescent melting analysis following real-time PCR have been developed,13Wittwer C.T. Reed G.H. Gundry C.N. Vandersteen J.G. Pryor R.J. High-resolution genotyping by amplicon melting analysis using LCGreen.Clin Chem. 2003; 49: 853-860Crossref PubMed Scopus (1045) Google Scholar, 14Li J. Berbeco R. Distel R.J. Janne P.A. Wang L. Makrigiorgos G.M. s-RT-MELT for rapid mutation scanning using enzymatic selection and real time DNA-melting: new potential for multiplex genetic analysis.Nucleic Acids Res. 2007; 35: e84Crossref PubMed Scopus (23) Google Scholar, 15Rice J.E. Reis Jr., A.H. Rice L.M. Carver-Brown R.K. Wangh L.J. Fluorescent signatures for variable DNA sequences.Nucleic Acids Res. 2012; 40: e164Crossref PubMed Scopus (30) Google Scholar, 16Milbury C.A. Li J. Makrigiorgos G.M. COLD-PCR-enhanced high-resolution melting enables rapid and selective identification of low-level unknown mutations.Clin Chem. 2009; 55: 2130-2143Crossref PubMed Scopus (64) Google Scholar monitoring fluorescent melting within individual droplets during ddPCR is not available at this time. Accordingly, a ddPCR approach that could be implemented in a mutation scanning format is desirable. Here, we demonstrate that incorporation of coamplification at lower denaturation temperature PCR (COLD-PCR),17Carotenuto P. Roma C. Cozzolino S. Fenizia F. Rachiglio A.M. Tatangelo F. Iannaccone A. Baron L. Botti G. Normanno N. Detection of KRAS mutations in colorectal cancer with Fast COLD-PCR.Int J Oncol. 2012; 40: 378-384PubMed Google Scholar, 18Pritchard C.C. Akagi L. Reddy P.L. Joseph L. Tait J.F. COLD-PCR enhanced melting curve analysis improves diagnostic accuracy for KRAS mutations in colorectal carcinoma.BMC Clin Pathol. 2010; 10: 6Crossref PubMed Scopus (27) Google Scholar, 19Kristensen L.S. Daugaard I.L. Christensen M. Hamilton-Dutoit S. Hager H. Hansen L.L. Increased sensitivity of KRAS mutation detection by high-resolution melting analysis of COLD-PCR products.Hum Mutat. 2010; 31: 1366-1373Crossref PubMed Scopus (33) Google Scholar, 20How Kit A. Mazaleyrat N. Daunay A. Nielsen H.M. Terris B. Tost J. Sensitive detection of KRAS mutations using enhanced-ice-COLD-PCR mutation enrichment and direct sequence identification.Hum Mutat. 2013; 34: 1568-1580Crossref PubMed Scopus (40) Google Scholar, 21Li J. Wang L. Mamon H. Kulke M.H. Berbeco R. Makrigiorgos G.M. Replacing PCR with COLD-PCR enriches variant DNA sequences and redefines the sensitivity of genetic testing.Nat Med. 2008; 14: 579-584Crossref PubMed Scopus (332) Google Scholar within the ddPCR workflow in conjunction with two fluorescently labeled probes matching the wild-type amplicon, provides a simple and robust method for mutation scanning of target amplicons. COLD-PCR suppresses wild-type sequences and enables preferential amplification of mutation-containing droplets, for any mutation along the amplicon.21Li J. Wang L. Mamon H. Kulke M.H. Berbeco R. Makrigiorgos G.M. Replacing PCR with COLD-PCR enriches variant DNA sequences and redefines the sensitivity of genetic testing.Nat Med. 2008; 14: 579-584Crossref PubMed Scopus (332) Google Scholar, 22Li J. Milbury C.A. Li C. Makrigiorgos G.M. Two-round coamplification at lower denaturation temperature-PCR (COLD-PCR)-based Sanger sequencing identifies a novel spectrum of low-level mutations in lung adenocarcinoma.Hum Mutat. 2009; 30: 1583-1590Crossref PubMed Scopus (56) Google Scholar Detecting changes to the ratio of COLD-PCR–enhanced signals caused by mutations anywhere within the probed region enables mutation scanning with high selectivity. This novel enhanced ratio of signal-based mutation scanning COLD-ddPCR enables a rapid method for the detection of mutations during ddPCR without prior knowledge of the specific DNA variant present in the target sequence. We demonstrate in this paper the application of COLD-ddPCR to the detection of multiple mutations present in TP53 exon 8, as well as for the T790M resistance mutation in EGFR exon 20 in DNA from mutated cell lines and cell-free circulating DNA (cfDNA) from clinical cancer samples. Missense mutations in several positions of the TP53 exon 8 and mutation p.T790M in EGFR exon 20 were assessed in this study. Human genomic DNA from cancer cell lines DU-145 (ATCC HTB-81D), HCC2218 (ATCC CRL-2343), and MDA-MB-231 (ATCC HTB-26D) was purchased from ATCC (Manassas, VA). Genomic DNA from commercial cell lines SW480 (ATCC no. CCL-228) and H1975 (ATCC no. CRL-5908) was extracted using the DNeasy Blood and Tissue kit (Qiagen, Valencia, CA) following the manufacturer's protocol (Table 1). Human genomic DNA (Promega, Madison, WI) was used as wild-type control DNA and for creating dilutions of gradually decreasing mutation abundances. All experiments were replicated at least three independent times for assessing the reproducibility of the results.Table 1Summary of Cell Lines UsedTarget regionCell lineMutation (nt)Mutation (aa)EGFR exon 20H1975c.2369C>Tp.T790MTP53 exon 8SW480c.818G>Ap.R273HDU-145c.820G>Tp.V274FMDA-MB-231c.839G>Ap.R280KHCC2218c.847C>Tp.R283Caa, amino acid; nt, nucleotide. Open table in a new tab aa, amino acid; nt, nucleotide. To evaluate the efficacy of this assay in characterizing specimens from different origins, we analyzed a colorectal tumor sample, known to harbor a G>A (p.R273H) missense mutation present at a low frequency (approximately 1%). This mutation had been previously identified and validated using different methods [COLD-PCR sequencing, denaturing high-pressure liquid chromatography, restriction endonuclease-mediated selective PCR, and differential strand separation at critical temperature (Tc)16Milbury C.A. Li J. Makrigiorgos G.M. COLD-PCR-enhanced high-resolution melting enables rapid and selective identification of low-level unknown mutations.Clin Chem. 2009; 55: 2130-2143Crossref PubMed Scopus (64) Google Scholar, 23Guha M. Castellanos-Rizaldos E. Liu P. Mamon H. Makrigiorgos G.M. Differential strand separation at critical temperature: a minimally disruptive enrichment method for low-abundance unknown DNA mutations.Nucleic Acids Res. 2013; 41: e50Crossref PubMed Scopus (23) Google Scholar]. We also evaluated cfDNA isolated from patients with lung adenocarcinomas at different stages of disease progression and treatment, and a cfDNA sample from an esophageal cancer case. Mutations in these cfDNA samples had been previously identified using a ddPCR allele-specific approach with hydrolysis probes specific either to the wild-type or the mutant allele.24Oxnard G.R. Paweletz C.P. Kuang Y. Mach S.L. O'Connell A. Messineo M.M. Luke J.J. Butaney M. Kirschmeier P. Jackman D.M. Janne P.A. Noninvasive detection of response and resistance in EGFR-mutant lung cancer using quantitative next-generation genotyping of cell-free plasma DNA.Clin Cancer Res. 2014; 20: 1698-1705Crossref PubMed Scopus (660) Google Scholar Samples were obtained from patients after informed consent and Dana Farber-Cancer Institute Institutional Review Board approval (Table 2).Table 2List of Clinical Samples Used for Validation PurposesTarget regionSample nameTypeTumor/originMutation (nt)Mutation (aa)Estimated frequency∗Mutation frequencies were estimated by previous analysis or conventional digital PCR using allele-specific probes.TP53 exon 8CT20gDNAColorectal cancerc.818G>Ap.R273H∼1%FC39–4cfDNAEsophageal cancerc.847C>Tp.R283C∼5%EGFR exon 20PT-04–09cfDNALung adenocarcinomac.2369C>Tp.T790M∼0.1%PT-21cfDNALung adenocarcinomac.2369C>Tp.T790MWild typePT-04–11cfDNALung adenocarcinomac.2369C>Tp.T790M∼2.1%PT-10–06cfDNALung adenocarcinomac.2369C>Tp.T790M∼0.43%PT-07–16cfDNALung adenocarcinomac.2369C>Tp.T790M∼2.03%PT-10–33cfDNALung adenocarcinomac.2369C>Tp.T790M∼0.38%PT-04–17cfDNALung adenocarcinomac.2369C>Tp.T790M∼4.08%Circulating cell-free tumor DNA samples (cfDNA).aa, amino acid; DNA, genomic DNA; nt, nucleotide.∗ Mutation frequencies were estimated by previous analysis or conventional digital PCR using allele-specific probes. Open table in a new tab Circulating cell-free tumor DNA samples (cfDNA). aa, amino acid; DNA, genomic DNA; nt, nucleotide. Gene-specific primers were used to pre-amplify selected targets from genomic or cfDNA (approximately 20 ng and approximately 10 ng, respectively) before performing ddPCR experiments. Twenty cycles of pre-amplification of EGFR exon 20 were done using the GoTaq Flexi DNA polymerase system (Promega) and 0.2 μmol/L forward and reverse primers. Inconsistent with previous studies, the forward primer was engineered to encompass the A/G single nucleotide polymorphism present within this amplicon to generate a uniform template for downstream analysis.25Sherry S.T. Ward M.H. Kholodov M. Baker J. Phan L. Smigielski E.M. Sirotkin K. dbSNP: the NCBI database of genetic variation.Nucleic Acids Res. 2001; 29: 308-311Crossref PubMed Scopus (5145) Google Scholar, 26Guha M. Castellanos-Rizaldos E. Makrigiorgos G.M. DISSECT method using PNA-LNA clamp improves detection of T790m mutation.PLoS One. 2013; 8: e67782Crossref PubMed Scopus (30) Google Scholar The initial 30 cycles of pre-amplification of TP53 exon 8 were conducted using the Phusion polymerase system (New England Biolabs, Ipswich, MA) and 0.2 μmol/L forward and reverse primers (Tables 3 and 4). Amplified DNA was then diluted and used as a template for ddPCR experiments using nested primers (Table 3).Table 3List of Primers and Probes Sequences Used in This StudyTarget regionPCR roundPrimer and probe sequencesSize (bp)TP53 exon 8Genomic DNA pre-amplificationF: 5′-GCTTCTCTTTTCCTATCCTG-3′167R: 5′-CTTACCTCGCTTAGTGCT-3′Digital PCRF: 5′-TGGTAATCTACTGGGACG-3′87R: 5′-CGGAGATTCTCTTCCTCT-3′Hydrolysis probes5′-FAM-TGGGAGAGACCGGCGCA-BHQ_1-3′∗These hydrolysis probes had sequences complementary to the wild-type allele.5′-HEX-TTTGAGGTGCGTGTTTGTGCC-BHQ_1-3′∗These hydrolysis probes had sequences complementary to the wild-type allele.EGFR exon 20Genomic DNA pre-amplificationF: 5′-GCTGGGCATCTGCCTCACCTCCACCGTGCAACT-3′91R: 5′-GTCTTTGTGTTCCCGGACATAG-3′Digital PCRF: 5′-GCTGGGCATCTGCCTCA-3′67R: 5′-CAGGAGGCAGCCGAAGG-3′Hydrolysis probes5′-FAM-ATGAGTTGCACGGTGGA-BHQ_1-3′∗These hydrolysis probes had sequences complementary to the wild-type allele.5′-HEX-CTCATCACGCAGCTCATG-BHQ_1-3′∗These hydrolysis probes had sequences complementary to the wild-type allele.5′-FAM-CTCATCATGCAGCTCATG-BHQ_1-3′†Hydrolysis probes specific to the mutant (FAM) or wild-type (HEX) allele used for precise quantification of mutational abundances.5′-HEX-CTCATCACGCAGCTCATG-BHQ_1-3′†Hydrolysis probes specific to the mutant (FAM) or wild-type (HEX) allele used for precise quantification of mutational abundances.F, forward; R, reverse.∗ These hydrolysis probes had sequences complementary to the wild-type allele.† Hydrolysis probes specific to the mutant (FAM) or wild-type (HEX) allele used for precise quantification of mutational abundances. Open table in a new tab Table 4Cycling Conditions UsedTarget regionPCR roundCycling conditionsTP53 exon 8Genomic DNA pre-amplificationInitial denaturation98°C for 30 secondsThermocycling: 30 cycles98°C for 10 seconds58°C for 20 seconds72°C for 10 secondsConventional Digital PCRInitial denaturation95°C for 10 minutesThermocycling: 40 cycles94°C for 30 seconds58°C for 60 secondsHold98°C for 10 minutesDigital COLD-PCRInitial denaturation95°C for 10 minutesThermocycling: 5 cycles94°C for 30 seconds58°C for 60 secondsThermocycling: 40 cycles78°C (Tc) for 30 seconds58°C for 60 secondsHold98°C for 10 minutesEGFR exon 20Genomic DNA pre-amplificationInitial denaturation95°C for 120 secondsThermocycling: 20 cycles95°C for 15 seconds55°C for 30 seconds72°C for 30 seconds72°C for 60 secondsConventional Digital PCRInitial denaturation95°C for 10 minutesThermocycling: 40 cycles94°C for 30 seconds52°C for 60 secondsHold98°C for 10 minutesDigital COLD-PCRInitial denaturation95°C for 10 minutesThermocycling: 5 cycles94°C for 30 seconds52°C for 60 secondsThermocycling: 35 cycles79.9°C (Tc) for 30 seconds52°C for 60 secondsHold98°C for 10 minutesCOLD-PCR, coamplification at lower denaturation temperature PCR. Open table in a new tab F, forward; R, reverse. COLD-PCR, coamplification at lower denaturation temperature PCR. To apply COLD-ddPCR approach, two TaqMan hydrolysis probes (FAM and HEX labeled, respectively) were designed to match the wild-type sequence and to bind to different regions on the target amplicon (Figure 1). Conventional ddPCR reactions were performed following the manufacturer's indications (Bio-Rad Laboratories, Hercules CA). Amplifications were performed in a 20-μL volume containing 2× ddPCR supermix for probes (Bio-Rad Laboratories), 900 nmol/L forward and reverse primers (synthesized by Integrated DNA Technologies, Coralville, IA), 250 nmol/L FAM and HEX probes (Integrated DNA Technologies), and pre-amplified DNA template (using 1:4000 final dilution for the EGFR exon 20 experiments and 1:100,000 for the TP53 exon 8) (Table 3). We started from a pre-amplified template, given the nature (very low number of target copies) of the samples used for the validation stage (cfDNA). Droplets were then generated using the DG8 droplet generator cartridges by mixing the aqueous phase with 70 μL of droplet generation oil (DG; Bio-Rad Laboratories). Samples were transferred to a 96-well reaction plate and then sealed using the PX1 PCR plate sealer (Bio-Rad Laboratories) for 10 seconds at 180°C before thermal cycling. Cycling conditions for conventional ddPCR are summarized in Table 4. In its simplest form, COLD-PCR (fast-COLD-PCR22Li J. Milbury C.A. Li C. Makrigiorgos G.M. Two-round coamplification at lower denaturation temperature-PCR (COLD-PCR)-based Sanger sequencing identifies a novel spectrum of low-level mutations in lung adenocarcinoma.Hum Mutat. 2009; 30: 1583-1590Crossref PubMed Scopus (56) Google Scholar) is performed by replacing the conventional denaturation temperature during the PCR cycling by a lower denaturation temperature. This results in a preferential amplification of mutations that decrease the melting temperature of the amplicon (G:C>A:T and G:C>T:A, or melting temperature–decreasing insertion-deletions). These mutation types comprise the majority (70% to 95%) of somatic mutations reported in breast, lung, gastric, colorectal, renal, ovarian, glioma, and melanoma cancers, as well as germline polymorphisms.27Greenman C. Stephens P. Smith R. Dalgliesh G.L. Hunter C. Bignell G. et al.Patterns of somatic mutation in human cancer genomes.Nature. 2007; 446: 153-158Crossref PubMed Scopus (2459) Google Scholar In conventional ddPCR, once the DNA is compartmentalized, PCR amplification occurs simultaneously in every droplet. By contrast, COLD-ddPCR amplification takes place at a denaturation temperature lower than that of the denaturation temperature of the wild-type amplicon. As a result, amplification occurs preferentially within those droplets where the mutant allele is present, resulting in a mutant-biased amplification (Figure 2). To determine the optimal critical denaturation temperature (Tc) for the COLD-ddPCR reactions, a gradient of Tc temperatures was tested using the Eppendorf Mastercycler (Eppendorf, Hamburg, Germany). Series of wild-type and 3% to 5% mutation-containing samples were used in ddPCR reactions using the temperature gradient feature of the thermal cycler. We first tested a broad temperature window and then narrowed it to a 1°C temperature window to find a suitable Tc (Supplemental Figures S1 and S2). As in bulk-solution COLD-PCR, the Tc chosen was the lowest temperature that generates substantial inhibition, yet reproducible amplification, for the wild-type sample.28Galbiati S. Brisci A. Lalatta F. Seia M. Makrigiorgos G.M. Ferrari M. Cremonesi L. Full COLD-PCR protocol for noninvasive prenatal diagnosis of genetic diseases.Clin Chem. 2011; 57: 136-138Crossref PubMed Scopus (55) Google Scholar At this temperature, the FAM/HEX differences between wild-type and mutant samples were maximized. Despite the wild-type inhibition, amplification still occurs in a fraction of wild-type DNA–containing droplets, hence providing positive signals in both HEX and FAM channels. The threshold in the QuantaSoft software (Bio-Rad Laboratories) is set such that for wild-type control samples, FAM/HEX = 1. The same threshold was then applied to all interrogated samples. Samples with the FAM/HEX ratio deviating substantially from 1 (P < 0.01) are considered mutant. ddPCR reactions were then analyzed and processed using the droplet reader and the QuantaSoft software version 1.3.2.0. To validate mutational abundances in the cfDNA samples, two hydrolysis (TaqMan) probes, binding to the same DNA region, were designed for allele-specific ddPCR. Reactions were run under conventional PCR cycling conditions per manufacturer recommendations: FAM-labeled, complementary to the mutant allele, and HEX-labeled, complementary to the wild-type allele. Hydrolysis probes (Table 3) were synthesized by Integrated DNA Technologies. Calculation of absolute number of positive events for a given channel (FAM or HEX), and the ratio and the fractional abundance of mutation for each sample were performed by the QuantaSoft software. The ratio was calculated as the number of copies per microliter obtained in the FAM channel (A), divided by copies per microliter in the HEX channel (B): A/B. The fractional abundance of mutant allele was obtained by dividing the number of copies per microliter of mutant allele (FAM channel) by the total copies per microliter of wild-type allele (HEX) plus mutant (FAM): [A/(A + B)]. The determination of number of target copies per droplet (number of copies of target molecule) was adjusted by the software to fit a Poisson distribution model with a 95% confidence level. The t-test was conducted to evaluate the statistical significance of the FAM/HEX ratio differences between wild-type and mutant alleles. Results with a two-tailed P value below 0.01 were considered significant. During COLD-ddPCR, the presence of a mutation-generated mismatch anywhere in the DNA sequence under a hydrolysis probe decreases the overall fluorescent signal generated from that probe compared to the signal contributed by the second probe that remains fully matched (Figure 1). Accordingly, the ratio of FAM/HEX signals either decreases relative to the wild type or increases, depending on whether there is a mutation under the FAM or HEX, respectively. This principle was first applied on a EGFR exon 20 amplicon where the p.T790M resistance-causing mutation lies.29Kuang Y. Rogers A. Yeap B.Y. Wang L. Makrigiorgos M. Vetrand K. Thiede S. Distel R.J. Janne P.A. Noninvasive detection of EGFR T790M in gefitinib or erlotinib resistant non-small cell lung cancer.Clin Cancer Res. 2009; 15: 2630-2636Crossref PubMed Scopus (213) Google Scholar We initially used conventional PCR cycling conditions using serial dilutions of p.T790M mutant DNA into wild-type DNA. Under conventional cycling, the FAM-positive versus HEX-positive droplets changed by about approximately 10-fold in the presence of 100% p.T790M mutant sequence (Figure 3A). Serial dilutions of mutant into wild-type indicated that the lowest mutation dilution for which the FAM/HEX ratio is different from the wild-type samples is 3% (P < 0.01), given the SD of the signals from three replicates. Two-dimensional plots of the signals from droplets following ddPCR are depicted in Supplemental Figure S3 (one replicate of each sample). For wild-type samples, the vast majority of droplets generate both FAM- and HEX-positive signals, consistent with both TaqMan probes binding efficiently to the target sequence. A 100% mutant probe shifts the majority of droplets with positive signals toward FAM, whereas reducing the mutant abundance below 3% yields results statistically indistinguishable from the wild-type (t-test P > 0.01). Changing the threshold that separates FAM/HEX-positive from negative droplets in these two-dimensional plots changes the absolute number of droplets within each quadrangle but does not affect significantly the FAM/HEX-positive ratio or the lowest mutation abundance detectable (not shown). By replacing conventional PCR with COLD-PCR cycling conditions, the overall number of droplets generating signals (positive events) for FAM and HEX are sharply reduced, consistent with inhibition of the wild-type amplification (Figure 3B). Furthermore, the shift in the ratio of FAM- to HEX-positive events is more pronounced, and the serial dilutions indicate that approximately 0.2% mutational abundance can be discriminated from wild type (with approximately 1.7× higher FAM/HEX ratio) (Figure 3B). Supplemental Figure S4 depicts two-dimensional plots of FAM versus HEX amplitude following COLD-ddPCR for wild-type, 5%, 1.25%, and 0.3% mutational abundance (one replicate of each sample). The sample with 0.3% p.T790M mutational abundance was distinguishable from that of the wild type according to t-test with two-tailed P < 0.01. The smear-like appearance is an unavoidable aspect of the technique; however, because the method preferentially inhibits the amplification of the wild-type amplicon, it does not affect the power of discrimination of the mutant amplicon. Experiments have been repeated multiple times, and these smears are reproducible between replicates. Changing by 10% to 20% the threshold that separates FAM/HEX-positive from negative droplets in these two-dimensional plots does not affect significantly the FAM/HEX-positive ratio (data not shown). Finally, to compare the ratio-of-signals method using real-time PCR instead of ddPCR, we performed the same experiment using real-time PCR on an ECO thermocycler (Illumina, San Diego, CA). We calculated the FAM/HEX ratio of fluorescence signals at cycle 40, as used in ddPCR, and