Double Drop-Off Droplet Digital PCR

In acute myeloid leukemia (AML), somatic gene mutations are important prognostic markers and increasingly constitute therapeutic targets. Therefore, robust, sensitive, and fast diagnostic assays are needed. Current techniques for mutation screening and quantification, including next-generation sequencing and quantitative PCR, each have weaknesses that leave a need for novel diagnostic tools. We established double drop-off digital droplet PCR (DDO-ddPCR) assays for gene mutations in NPM1, IDH2, and NRAS, which can detect and quantify diverse alterations at two nearby hotspot regions present in these genes. These assays can be used for mutation screening as well as quantification and sequential monitoring. The assays were validated against next-generation sequencing and existing ddPCR assays and achieved high concordance with an overall sensitivity comparable to conventional digital PCR. In addition, the feasibility of detecting and monitoring genetic alterations in peripheral blood cell-free DNA (cfDNA) of patients with AML by DDO-ddPCR was studied. cfDNA analysis was found to have similar sensitivity compared to quantitative PCR–based analysis of peripheral blood. Finally, the cfDNA-based digital PCR in several clinical scenarios was found to be useful in long-term monitoring of target-specific therapy, early response assessment during induction chemotherapy, and identification of mutations in patients with extramedullary disease. Thus, DDO-ddPCR–based cfDNA analysis may complement existing genetic tools for diagnosis and disease monitoring in AML. In acute myeloid leukemia (AML), somatic gene mutations are important prognostic markers and increasingly constitute therapeutic targets. Therefore, robust, sensitive, and fast diagnostic assays are needed. Current techniques for mutation screening and quantification, including next-generation sequencing and quantitative PCR, each have weaknesses that leave a need for novel diagnostic tools. We established double drop-off digital droplet PCR (DDO-ddPCR) assays for gene mutations in NPM1, IDH2, and NRAS, which can detect and quantify diverse alterations at two nearby hotspot regions present in these genes. These assays can be used for mutation screening as well as quantification and sequential monitoring. The assays were validated against next-generation sequencing and existing ddPCR assays and achieved high concordance with an overall sensitivity comparable to conventional digital PCR. In addition, the feasibility of detecting and monitoring genetic alterations in peripheral blood cell-free DNA (cfDNA) of patients with AML by DDO-ddPCR was studied. cfDNA analysis was found to have similar sensitivity compared to quantitative PCR–based analysis of peripheral blood. Finally, the cfDNA-based digital PCR in several clinical scenarios was found to be useful in long-term monitoring of target-specific therapy, early response assessment during induction chemotherapy, and identification of mutations in patients with extramedullary disease. Thus, DDO-ddPCR–based cfDNA analysis may complement existing genetic tools for diagnosis and disease monitoring in AML. In recent years, treatment of acute myeloid leukemia (AML) has advanced because of the development of novel therapeutics that target specific driver mutations (eg, in FLT3, IDH1, and IDH2).1DiNardo C.D. Stein E.M. de Botton S. Roboz G.J. Altman J.K. Mims A.S. et al.Durable remissions with ivosidenib in IDH1-mutated relapsed or refractory AML.N Engl J Med. 2018; 378: 2386-2398Crossref PubMed Scopus (704) Google Scholar, 2Perl A.E. Martinelli G. Cortes J.E. Neubauer A. Berman E. Paolini S. Montesinos P. Baer M.R. Larson R.A. Ustun C. Fabbiano F. Erba H.P. Di Stasi A. Stuart R. Olin R. Kasner M. Ciceri F. Chou W.-C. Podoltsev N. Recher C. Yokoyama H. Hosono N. Yoon S.-S. Lee J.-H. Pardee T. Fathi A.T. Liu C. Hasabou N. Liu X. Bahceci E. Levis M.J. Gilteritinib or chemotherapy for relapsed or refractory FLT3-mutated AML.N Engl J Med. 2019; 381: 1728-1740Crossref PubMed Scopus (420) Google Scholar, 3Pollyea D.A. Tallman M.S. de Botton S. Kantarjian H.M. Collins R. Stein A.S. Frattini M.G. Xu Q. Tosolini A. See W.L. MacBeth K.J. Agresta S.V. Attar E.C. DiNardo C.D. Stein E.M. Enasidenib, an inhibitor of mutant IDH2 proteins, induces durable remissions in older patients with newly diagnosed acute myeloid leukemia.Leukemia. 2019; 33: 2575-2584Crossref PubMed Scopus (105) Google Scholar Although many patients with AML achieve complete remission after intensive induction chemotherapy alone or in combination with molecularly targeted agents,4Stone R.M. Mandrekar S.J. Sanford B.L. Laumann K. Geyer S. Bloomfield C.D. Thiede C. Prior T.W. Döhner K. Marcucci G. Lo-Coco F. Klisovic R.B. Wei A. Sierra J. Sanz M.A. Brandwein J.M. de Witte T. Niederwieser D. Appelbaum F.R. Medeiros B.C. Tallman M.S. Krauter J. Schlenk R.F. Ganser A. Serve H. Ehninger G. Amadori S. Larson R.A. Döhner H. Midostaurin plus chemotherapy for acute myeloid leukemia with a FLT3 mutation.N Engl J Med. 2017; 377: 454-464Crossref PubMed Scopus (1077) Google Scholar relapse rates remain high, even in patients with intermediate or favorable risk group assignment.5Döhner H. Estey E. Grimwade D. Amadori S. Appelbaum F.R. Büchner T. Dombret H. Ebert B.L. Fenaux P. Larson R.A. Levine R.L. Lo-Coco F. Naoe T. Niederwieser D. Ossenkoppele G.J. Sanz M. Sierra J. Tallman M.S. Tien H.-F. Wei A.H. Löwenberg B. Bloomfield C.D. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel.Blood. 2017; 129: 424-447Crossref PubMed Scopus (2870) Google Scholar,6Herold T. Rothenberg-Thurley M. Grunwald V.V. Janke H. Goerlich D. Sauerland M.C. Konstandin N.P. Dufour A. Schneider S. Neusser M. Ksienzyk B. Greif P.A. Subklewe M. Faldum A. Bohlander S.K. Braess J. Wörmann B. Krug U. Berdel W.E. Hiddemann W. Spiekermann K. Metzeler K.H. Validation and refinement of the revised 2017 European LeukemiaNet genetic risk stratification of acute myeloid leukemia.Leukemia. 2020; 34: 3161-3172Crossref PubMed Scopus (58) Google Scholar In addition to baseline cytogenetic and molecular genetic risk stratification, persistence of leukemic cells during and after therapy [ie, measurable residual disease (MRD)] has emerged as an important predictor of relapse risk.7Schuurhuis GH M. Freeman S. Béné M.C. Buccisano F. Cloos J. Grimwade D. Haferlach T. Hills R.K. Hourigan C.S. Jorgensen J.L. Kern W. Lacombe F. Maurillo L. Preudhomme C. van der Reijden B.A. Thiede C. Venditti A. Vyas P. Wood B.L. Walter R.B. Döhner K. Roboz G.J. Ossenkoppele G.J. Minimal/measurable residual disease in AML: a consensus document from the European LeukemiaNet MRD Working Party.Blood. 2018; 131: 1275-1291Crossref PubMed Scopus (524) Google Scholar,8Freeman S.D. Hourigan C.S. MRD evaluation of AML in clinical practice: are we there yet?.Hematology. 2019; 2019: 557-569Crossref PubMed Scopus (16) Google Scholar Importantly, MRD-directed therapy may prevent hematologic relapse in AML. Currently, various techniques, including multiparameter flow cytometry and molecular assays, are used for MRD assessment.9Platzbecker U. Middeke J.M. Sockel K. Herbst R. Wolf D. Baldus C.D. Oelschlägel U. Mütherig A. Fransecky L. Noppeney R. Bug G. Götze K.S. Krämer A. Bochtler T. Stelljes M. Groth C. Schubert A. Mende M. Stölzel F. Borkmann C. Kubasch A.S. von Bonin M. Serve H. Hänel M. Dührsen U. Schetelig J. Röllig C. Kramer M. Ehninger G. Bornhäuser M. Thiede C. Measurable residual disease-guided treatment with azacitidine to prevent haematological relapse in patients with myelodysplastic syndrome and acute myeloid leukaemia (RELAZA2): an open-label, multicentre, phase 2 trial.Lancet Oncol. 2018; 19: 1668-1679Abstract Full Text Full Text PDF PubMed Scopus (147) Google Scholar,10Köhnke T. Bücklein V. Rechkemmer S. Schneider S. Rothenberg-Thurley M. Metzeler K.H. Sauerland M.-C. Hiddemann W. Spiekermann K. Subklewe M. Response assessment in acute myeloid leukemia by flow cytometry supersedes cytomorphology at time of aplasia, amends cases without molecular residual disease marker and serves as an independent prognostic marker at time of aplasia and post-induction.Haematologica. 2019; 104: e510-e513Crossref PubMed Scopus (2) Google Scholar Molecular MRD measurements based on quantitative PCR (qPCR) are potentially more sensitive than multiparameter flow cytometry but are dependent on the presence of suitable target lesions, such as core binding factor fusion transcripts or mutated NPM1.11Hubmann M. Köhnke T. Hoster E. Schneider S. Dufour A. Zellmeier E. Fiegl M. Braess J. Bohlander S.K. Subklewe M. Sauerland M.-C. Berdel W.E. Büchner T. Wörmann B. Hiddemann W. Spiekermann K. Molecular response assessment by quantitative real-time polymerase chain reaction after induction therapy in NPM1-mutated patients identifies those at high risk of relapse.Haematologica. 2014; 99: 1317-1325Crossref PubMed Scopus (54) Google Scholar Absolute quantification of MRD levels by qPCR requires the use of external reference standards.12Hansen M.C. Haferlach T. Nyvold C.G. A decade with whole exome sequencing in haematology.Br J Haematol. 2020; 188: 367-382Crossref PubMed Scopus (6) Google Scholar Next-generation sequencing (NGS)–based detection of MRD is more broadly applicable than qPCR but currently is less sensitive than qPCR, costlier, and limited in availability.13Jongen-Lavrencic M. Grob T. Hanekamp D. Kavelaars F.G. Al Hinai A. Zeilemaker A. Erpelinck-Verschueren C.A.J. Gradowska P.L. Meijer R. Cloos J. Biemond B.J. Graux C. van Marwijk Kooy M. Manz M.G. Pabst T. Passweg J.R. Havelange V. Ossenkoppele G.J. Sanders M.A. Schuurhuis G.J. Lowenberg B. Valk P.J.M. Molecular minimal residual disease in acute myeloid leukemia.N Engl J Med. 2018; 378: 1189-1199Crossref PubMed Scopus (384) Google Scholar Both the development of new genetically targeted treatment options and the accumulating evidence supporting MRD-adapted treatment strategies result in an increasing need for targeted molecular diagnostic assays that are sensitive, provide rapid turnaround times, and are economically viable for serial disease monitoring. In addition, assays that use peripheral blood (pB) are associated with less patient discomfort compared with bone marrow (BM)–based tests and may be thus advantageous for serial MRD assessments and long-term disease monitoring. On the basis of this need for versatile and sensitive molecular diagnostic tools for the detection and monitoring of gene mutations in patients with AML, we established digital droplet PCR (ddPCR)–based assays for several common AML driver mutations, including established markers of residual disease (NPM1) and relevant therapeutic targets (IDH2). A novel type of ddPCR assay, the double drop-off (DDO)–ddPCR, was designed. DDO-ddPCR allows the detection of heterogeneous alterations in two neighboring mutational hotspot regions of a target DNA molecule, without prior knowledge of the precise sequence change that is present. In ddPCR, compartmentalization of the PCR reaction into microscopic droplets and subsequent readout by fluorescence measurement allows for absolute quantification of variant alleles without the need for standard curves and with a high signal-to-noise ratio. These advantages allow analyses based on limited quantities of nucleic acids, such as cell-free DNA (cfDNA) isolated from blood and other body fluids. For solid tumors, mutational analyses of cfDNA (from so called liquid biopsies) are more representative of heterogeneous tumor compartments at different localizations throughout the body compared with biopsies of one individual tumor site and to be more sensitive in predicting therapy response.14Tarazona N. Gimeno-Valiente F. Gambardella V. Zuniga S. Rentero-Garrido P. Huerta M. Rosello S. Martinez-Ciarpaglini C. Carbonell-Asins J.A. Carrasco F. Ferrer-Martinez A. Bruixola G. Fleitas T. Martin J. Tebar-Martinez R. Moro D. Castillo J. Espi A. Roda D. Cervantes A. Targeted next-generation sequencing of circulating-tumor DNA for tracking minimal residual disease in localized colon cancer.Ann Oncol. 2019; 30: 1804-1812Abstract Full Text Full Text PDF PubMed Scopus (74) Google Scholar, 15Kruger S. Heinemann V. Ross C. Diehl F. Nagel D. Ormanns S. Liebmann S. Prinz-Bravin I. Westphalen C.B. Haas M. Jung A. Kirchner T. von Bergwelt-Baildon M. Boeck S. Holdenrieder S. Repeated mutKRAS ctDNA measurements represent a novel and promising tool for early response prediction and therapy monitoring in advanced pancreatic cancer.Ann Oncol. 2018; 29: 2348-2355Abstract Full Text Full Text PDF PubMed Scopus (66) Google Scholar, 16Wan J.C.M. Massie C. Garcia-Corbacho J. Mouliere F. Brenton J.D. Caldas C. Pacey S. Baird R. Rosenfeld N. Liquid biopsies come of age: towards implementation of circulating tumour DNA.Nat Rev Cancer. 2017; 17: 223-238Crossref PubMed Scopus (1170) Google Scholar Recent analyses in patients with lymphoma and myelodysplastic syndromes suggest that cfDNA analysis may also be useful for genetic profiling and therapy monitoring in hematologic neoplasms.17Kurtz D.M. Scherer F. Jin M.C. Soo J. Craig A.F.M. Esfahani M.S. Chabon J.J. Stehr H. Liu C.L. Tibshirani R. Maeda L.S. Gupta N.K. Khodadoust M.S. Advani R.H. Levy R. Newman A.M. Dührsen U. Hüttmann A. Meignan M. Casasnovas R.-O. Westin J.R. Roschewski M. Wilson W.H. Gaidano G. Rossi D. Diehn M. Alizadeh A.A. Circulating tumor DNA measurements as early outcome predictors in diffuse large B-cell lymphoma.J Clin Oncol. 2018; 36: 2845-2853Crossref PubMed Scopus (171) Google Scholar,18Yeh P. Dickinson M. Ftouni S. Hunter T. Sinha D. Wong S.Q. Agarwal R. Vedururu R. Doig K. Fong C.Y. Blombery P. Westerman D. Dawson M.A. Dawson S.J. Molecular disease monitoring using circulating tumor DNA in myelodysplastic syndromes.Blood. 2017; 129: 1685-1690Crossref PubMed Scopus (39) Google Scholar In this study, we report on the performance characteristics of DDO-ddPCR assays for clinically relevant AML driver mutations in comparison with current routine methods. In addition, we used these novel assays to conduct proof-of-principle experiments on the feasibility and clinical usefulness of cfDNA-based molecular analyses in myeloid malignancies. For assay development and validation, BM genomic DNA (gDNA) samples from patients with AML obtained through the Laboratory for Leukemia Diagnostics (Ludwig Maximilian University of Munich, Munich, Germany) were used. Liquid biopsy specimens of circulating tumor DNA were prospectively collected from patients with AML using cfDNA BCT blood collection tubes (Streck, La Vista, NE, USA) at the time of initial diagnosis as well as serially during treatment, including at time points of routine MRD assessment. All patients also underwent NGS-based mutation screening as previously described.19Metzeler K. Herold T. Rothenberg-Thurley M. Amler S. Sauerland M. Görlich D. SChneider S. Konstandin N. Dufour A. Bräundl K. Ksienzyk B. Zellmeier E. Hartmann L. Greif P. Fiegl M. Subklewe M. Bohlander S. Krug U. Faldum A. Spectrum and prognostic relevance of driver gene mutations in acute myeloid leukemia.Blood. 2016; 128: 686-698Crossref PubMed Scopus (318) Google Scholar This study was approved by the local ethics committee in accordance with the Declaration of Helsinki (approval number 18-539). All patients provided written informed consent. For the NPM1 drop-off assay, primers published by Nakamura et al were used.20Nakamura SY K. Shimizu E. Yusa N. Kondoh K. Ogawa M. Takei T. Kobayashi A. Ito M. Isobe M. Konuma T. Kato S. Kasajima R. Wada Y. Nagamura-Inoue T. Yamaguchi R. Takahashi S. Imoto S. Miyano S. Tojo A. Prognostic impact of circulating tumor DNA status post–allogeneic hematopoietic stem cell transplantation in AML and MDS.Blood. 2019; 133: 14Crossref Scopus (37) Google Scholar All other primers and all probes for DDO-ddPCR and drop-off ddPCR assays were designed using Primer3 Plus version 2.4.221Untergasser A. Cutcutache I. Koressaar T. Ye J. Faircloth B.C. Remm M. Rozen S.G. Primer3—new capabilities and interfaces.Nucleic Acids Res. 2012; 40: e115Crossref PubMed Scopus (5425) Google Scholar and purchased from TIB MolBiol (Berlin, Germany). Target regions and sequences of primers and probes are given in Table 1. Mutation-specific IDH1 R132H and DNMT3A R882H assays were purchased from Bio-Rad (Hercules, CA, USA).Table 1Overview of Assay Designs for NPM1, IDH2, and NRAS MutationsAssayPCR target region (GRCh37/hg19)Annealing temperature, °CLeft primerRight primerProbe 1Probe 2NPM1 double drop-off5:170837474-170837645595'-TTGATGTCTATGAAGTGTTGTGGTTC-3'5'-GGACAGCCAGATATCAACTGTTACA-3'5'-HEX-TTC+AAGATCT+CT+G+GC-BHQ1-3'5'-FAM-TGG+A+G+GAAGTCTCTT-BHQ1-3'IDH2 double drop-off15:90631789-90631974575'-GTGGAAAAGTCCCAATGG-3'5'-AGGTCAGTGGATCCCCTCT-3'5'-HEX-C+CGGA+AC+AT+CC+TGGG-BHQ1-3'5'-FAM-CC+ATTGGCAG+GCACG-BHQ1-3'NRAS drop-off1:115258626-115258772605'-TACAAACTGGTGGTGGTTGG-3'5'-TGAGAGACAGGATCAGGTCA-3'5'-HEX-CCCA+C+CATA+GAGGT-BHQ1-3'5'-FAM-CAG+GTGG+T+GTTGGG-BHQ1-3' Open table in a new tab Plasma was separated by centrifuging whole blood specimens for 10 minutes at 1600 × g. The plasma layer was transferred to a new tube and centrifuged for 10 minutes at 16.100 × g, and the supernatant was stored at −80 °C. The cfDNA was isolated using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The gDNA was quantified using the Invitrogen Qubit dsDNA HS Assay Kit (Invitrogen, Paisley, UK), and cfDNA samples were quantified using the Agilent 2100 Bioanalyzer with the Agilent High Sensitivity DNA Kit (Agilent, Santa Clara, CA, USA), according to the manufacturers' instructions. For ddPCR, the Bio-Rad QX200 droplet digital PCR System (Bio-Rad) was used. Samples were not randomized and data acquisition was not blinded. Analysis of ddPCR results was performed using QuantaSoft version 1.7.4 (Bio-Rad). In brief, wgates were manually set for each droplet population to distinguish negative and positive droplets. Wells were excluded and limits of detection were set according to the manufacturer's instructions.22BioradDroplet Digital PCR Droplet Digital PCR Applications Guide. Bulletin 6407.2018: 1-145Google Scholar All statistical analyses were performed in R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Graphs were made using GraphPad Prism version 9.0.0 for Windows (GraphPad Software, San Diego, CA, USA). Sample sizes were not statistically determined before experimentation. All materials and kits used are approved for investigational use only. Standard ddPCR assays use probes specific for a particular sequence variant (eg, a specific single-base pair exchange) and thus require prior knowledge of the variant to be detected (Figure 1A). To circumvent this limitation, drop-off assays were developed. The drop-off ddPCR uses a fluorescent reference probe designed to anneal to the PCR product in the vicinity of a known mutational hotspot and a reporter probe complementary to the wild-type sequence of the hotspot region. Thus, a double-positive fluorescence signal from both the reference and reporter probes indicates the presence of wild-type DNA, whereas a fluorescence signal for only the reference probe indicates the presence of a DNA sequence variant in the target region that abrogates binding of the reporter probe (Figure 1B). However, several recurrently mutated oncogenes in myeloid malignant tumors have multiple mutational hotspots that are in relative proximity on the cDNA or gDNA level but cannot be covered by a single reporter probe. Therefore, an extension of the drop-off assay principle was used to detect mutations at two distinct mutational hotspots co-located within a region spanned by a single PCR amplicon. In these DDO assays, one probe binds to the wild-type DNA sequence of one hotspot region (eg, IDH2 codon p.R140), whereas the other probe binds to the wild-type sequence of the second, neighboring hotspot (eg, IDH2 codon p.R172). Hence, each probe simultaneously serves as a reporter probe for one and as reference probe for the other mutation hotspot. Amplification of wild-type DNA will lead to a double-positive fluorescence signal. A single-positive fluorescence signal indicates a sequence variant at the site covered by the nonbinding probe (Figure 1C). Mutations in NPM1 and IDH2 are clinically relevant for prognostication and selection of targeted therapies, respectively, in patients with AML. In both genes, mutations cluster at two neighboring hotspots co-located within one exon, but there is a wide variety of specific nucleotide insertions (for NPM1) or exchanges (for IDH2). To demonstrate the clinical utility of the DDO-ddPCR assay principle, DDO-ddPCR assays were established for NPM1 and IDH2 mutations and a standard drop-off assay established for NRAS mutations (Table 1). To determine the detection threshold and assess analytical specificity, ≥20 samples that were known to be wild type for the target gene variant based on NGS analysis were analyzed. For each assay, the detection threshold was defined as the mean number of positive droplets plus two times the SD. All assays achieved a high analytical specificity (99.89% to 99.975%) (Table 2), thus avoiding false-positive results.Table 2Specificity and Limit of Detection of Each Assay as Determined by Measuring 20 Negative Samples for Each MutationAssayTarget siteThreshold by dropletThreshold by ratio, %Analytical specificity, %Sensitivity, %NPM1c.86330.03799.9630.050c.87730.02599.975Not determinedIDH2c.14030.03599.9650.048c.17230.03599.9650.048NRASc.12/1350.11099.8900.068 Open table in a new tab Next, the analytical sensitivity of each assay was determined using serial dilutions of mutated samples in the wild-type DNA. For NPM1, DDO-ddPCR achieves a sensitivity of 0.037% for mutations at the CDS position c.863 (ie, type A/B/D mutations) (Figure 2A) , and 0.25% for rare NPM1 mutations at the second hotspot (data not shown). For IDH2 mutations, the analytical sensitivity for both codon p.R140 and codon p.R172 mutations (Figure 2, C and D) was 0.035%. For NRAS mutations, a conventional drop-off ddPCR achieved an analytical sensitivity of 0.11% for codon 12 and codon 13 mutations (Figure 2B). Thus, DDO-ddPCR can achieve an analytical sensitivity similar to established drop-off and conventional ddPCR assays for common oncogenic driver mutations. To compare the performance characteristics of these assays to currently used targeted NGS testing, BM gDNA samples that harbored a spectrum of known mutations at NPM1 positions c.863 and c.877, IDH2 codons p.R140 and p.R172, and NRAS codons p.G12 and p.G13 were studied. For each hotspot, the variant allele frequencies (VAFs) were compared for at least 10 BM gDNA samples as measured by a diagnostic NGS panel and by ddPCR. In NPM1 position c.863 mutated samples, a high correlation was observed between VAFs determined by NGS or DDO-ddPCR (r2 = 0.95; 95% CI, 0.88–0.99) (Figure 3A). Correlation with NGS was slightly lower for rare NPM1 types occurring at mutations at position c.877 (types U, AB, AE, AF and AT; r2 = 0.79; 95% CI, 0.69–0.89) (Figure 3A). For longer insertions (type AF, 12 nt and type AE, 9 nt), NGS sometimes resulted in lower VAF estimates compared with DDO-ddPCR, possibly reflecting inefficient mapping and variant calling for these mutations. DDO-ddPCR was closely correlated with NGS-determined VAFs for IDH2 mutations at c.140 (r2 = 0.94; 95% CI, 0.88–0.99) and c.172 (r2 = 0.94; 95% CI, 0.88–0.99) (Figure 3, B and C). Similar results were obtained for NRAS mutations measured by single reporter probe (drop-off ddPCR correlation with NGS: r2 = 0.86; 95% CI, 0.504–1.22) (Figure 3D). Overall, similar to conventional drop-off assays, mutation loads measured by DDO-ddPCR correlated well with NGS-based measurements. One potential use case for DDO-ddPCR assays is residual disease monitoring. The use of ddPCR allows one to perform MRD monitoring not only on pB or BM specimens but also using cfDNA. We hypothesized that cfDNA-based MRD testing may have higher sensitivity than pB cell–based assays, thereby facilitating serial MRD monitoring while avoiding the patient discomfort associated with repeated BM sampling. The results of cfDNA DDO-ddPCR–based mutation detection were therefore compared to a standard cDNA-based RT-qPCR test used on pB and BM specimens in routine patient care. Fifty-two pB circulating tumor DNA samples from nine patients with NPM1 mutations, obtained at time points when BM and/or pB cells were also collected were analyzed. Patient characteristics are given in Supplemental Table S1. Results of DDO-ddPCR testing of cfDNA isolated from pB were compared with qPCR results from pB and, if available, from BM mononuclear cells (MNCs). Paired cfDNA-ddPCR and pB qPCR measurements were available for 25 time points (Figure 4). At most time points, cfDNA-ddPCR and pB qPCR had concordant results. However, at two time points cfDNA ddPCR results were positive, whereas pB qPCR results were negative. For one of these two time points, a matched BM sample was also available and confirmed our positive measurement from cfDNA. On the other hand, there was one time point where pB qPCR results were positive but cfDNA ddPCR results were negative. No BM sample was available from this time point. For 49 time points, cfDNA and BM qPCR measurements were available (Figure 4). At 38 time points (78%) results of cfDNA-ddPCR and BM qPCR were concordant, whereas at 11 time points (22%) during follow-up a NPM1 mutation was detected in BM but not in pB cfDNA. For the 22 time points where all three measurements (qPCR on BM and pB cells and cfDNA-ddPCR) were available, cfDNA-ddPCR and pB cell–based qPCR were equally accurate in detecting positive samples (Figure 4). In summary, detection of NPM1 mutations and disease monitoring by DDO-PCR is less sensitive than BM qPCR but at least as sensitive as pB qPCR. Next, the clinical utility of cfDNA-based ddPCR testing was studied in different clinical scenarios. Serial MRD monitoring in AML was explored, both during long-term targeted therapy as well as during intensive induction chemotherapy. First, serial blood samples were collected from a patient with AML carrying an IDH2 p.R172 mutation treated with the IDH2 inhibitor enasidenib for >400 days (Supplemental Table S1). The patient achieved a complete response after approximately 4 months of enasidenib treatment. In parallel, decreases were observed for both the total cfDNA concentration and the IDH2 p.R172 VAF measured by ddPCR in pB circulating cfDNA and in pB and BM cell gDNA. On day 155, the results of ddPCR on pB MNCs were negative for the IDH2 mutation, whereas the results were still positive for pB cfDNA, and on day 168 all three materials were ddPCR negative. On day 223 mutated IDH2 was again detected in cfDNA by ddPCR, accompanied by an increase in total pB cfDNA concentration, whereas the mutation remained undetectable in pB MNC DNA. This molecular relapse initially detected by cfDNA analysis was confirmed on day 306 when ddPCR pB MNC results also turned positive and was followed by hematologic relapse verified by BM aspiration on day 330. Although a dose increase of enasidenib led to a short-term therapeutic response and decline of total cfDNA concentration and IDH2 p.R172 VAF, clinically refractory disease was accompanied by another increase of IDH2 p.R172 VAF in pB cfDNA and BM and led to discontinued use of the drug at day 407 (Figure 5). Second, to explore the utility of our assay for early response assessment during intensive induction chemotherapy, cfDNA samples were serially collected from intensively treated patients carrying NPM1 mutations during and after induction chemotherapy (Figure 6). NPM1 mutated cfDNA remained detectable during aplasia when white blood cell counts were low, whereas NPM1 mutated VAF had good correlation with BM blast count (Figure 6). On the other hand, an increasing total cfDNA concentration was also observed in patients with nonleukemic increase in leukocyte counts, as shown in Figure 6B where a white blood cell increase due to a fungal infection coincided with a spike in cfDNA concentration. This finding cautions against the use of total cfDNA concentration, instead of measuring specific VAFs, as a marker of residual disease in hematologic malignant neoplasms. Another important point is the very tight correlation between our NPM1 DDO assay and a commercially available mutation-specific ddPCR assay for IDH1 p.R132H (Figure 6D), which further validates our overall approach. The persistence of the DNMT3A mutation in the same patient probably indicates persistence of a DNMT3A mutated preleukemic clone. A third clinical situation in which cfDNA analysis may be clinically useful in patients with AML is the presence of extramedullary disease (EMD), in particular when BM involvement is minimal or absent (ie, myelosarcoma). EMD frequently involves anatomical locations where biopsy specimens are difficult to obtain, such as the central nervous sytem. At the same time, the availability of targeted therapeutic agents creates a clinical need to screen for druggable mutations in patients with EMD. It was therefore tested whether genotyping AML EMD through pB cfDNA analysis is feasible. A patient who presented with isolated myelosarcoma was studied (Supplemental Table S1) and an EMD biopsy sample (obtained from a paraaortic chloroma), BM MNC DNA, and pB cfDNA (Figure 7A) was comparatively analyzed. NGS testing of the EMD biopsy specimen revealed an IDH2 p.R140Q mutation. The same mutation was identified in pB cfDNA by ddPCR, with a VAF similar to the EMD NGS result, whereas routine NGS analysis from BM failed to detect the IDH2 mutation. Another patie