ARTICLE Acute Myeloid Leukemia Ferrata Storti Foundation A novel deep targeted sequencing method for minimal residual disease monitoring in acute myeloid leukemia Esther Onecha,1,2 Maria Linares,1,2 Inmaculada Rapado,1,2,3 Yanira Ruiz-Heredia,1,2 Pilar Martinez-Sanchez,1 Teresa Cedena,1,2,3,4 Marta Pratcorona,5 Jaime Perez Oteyza,6 Pilar Herrera,7 Eva Barragan,4,8 Pau Montesinos,4,8 Jose Antonio Garcia Vela,9 Elena Magro,10 Eduardo Anguita,11 Angela Figuera,12 Rosalia Riaza,13 Pilar Martinez-Barranco,14 Beatriz Sanchez-Vega,1,2 Josep Nomdedeu,5 Haematologica 2019 Volume 104(2):288-296 Miguel Gallardo,2* Joaquin Martinez-Lopez1,2,3,4* and Rosa Ayala1,2,3,4* 1 Hematology Department, Hospital Universitario 12 de Octubre, Madrid; 2Hematological Malignancies Clinical Research Unit, CNIO, Madrid; 3Centro de Investigación Biomédica en Red Cáncer (CIBERONC), Madrid; 4Complutense University, Madrid; 5Hematology Department, Hospital Santa Creu i Sant Pau, Barcelona; 6Hematology Department, Hospital Universitario Sanchinarro, Madrid; 7Hematology Department, Hospital Universitario Ramon y Cajal, Madrid; 8 Hematology Department, Hospital Universitario La Fe, Valencia; 9Department of Hematology, Hospital Universitario de Getafe, Madrid; 10Hematology Department, Hospital Universitario Principe de Asturias, Madrid; 11Hematology Department, Hospital Clínico San Carlos, IdISSC, UCM, Madrid; 12Hematology Department, Hospital Universitario de la Princesa, Madrid; 13 Hematology Department, Hospital Universitario Severo Ochoa, Madrid and 14Hematology Department, Hospital Universitario Fundación Alcorcón, Madrid, Spain *MG, JM-L and RA contributed equally to this work. ABSTRACT A high proportion of patients with acute myeloid leukemia who achieve minimal residual disease negative status ultimately relapse because a fraction of pathological clones remains undetected by Correspondence: standard methods. We designed and validated a high-throughput sequenc- ing method for minimal residual disease assessment of cell clonotypes with rosam.ayala@salud.madrid.org mutations of NPM1, IDH1/2 and/or FLT3-single nucleotide variants. For clinical validation, 106 follow-up samples from 63 patients in complete remission were studied by sequencing, evaluating the level of mutations Received: April 2, 2018. detected at diagnosis. The predictive value of minimal residual disease sta- Accepted: August 8, 2018. tus by sequencing, multiparameter flow cytometry, or quantitative poly- Pre-published: August 9, 2018. merase chain reaction analysis was determined by survival analysis. The sequencing method achieved a sensitivity of 10-4 for single nucleotide vari- ants and 10-5 for insertions/deletions and could be used in acute myeloid doi:10.3324/haematol.2018.194712 leukemia patients who carry any mutation (86% in our diagnostic data set). Sequencing–determined minimal residual disease positive status was asso- Check the online version for the most updated ciated with lower disease-free survival (hazard ratio 3.4, P=0.005) and information on this article, online supplements, and information on authorship & disclosures: lower overall survival (hazard ratio 4.2, P<0.001). Multivariate analysis www.haematologica.org/content/104/2/288 showed that minimal residual disease positive status determined by sequencing was an independent factor associated with risk of death (haz- ard ratio 4.54, P=0.005) and the only independent factor conferring risk of ©2019 Ferrata Storti Foundation relapse (hazard ratio 3.76, P=0.012). This sequencing-based method simpli- Material published in Haematologica is covered by copyright. fies and standardizes minimal residual disease evaluation, with high appli- All rights are reserved to the Ferrata Storti Foundation. Use of cability in acute myeloid leukemia. It is also an improvement upon flow published material is allowed under the following terms and conditions: cytometry- and quantitative polymerase chain reaction-based prediction of https://creativecommons.org/licenses/by-nc/4.0/legalcode. outcomes of patients with acute myeloid leukemia and could be incorpo- Copies of published material are allowed for personal or inter- rated in clinical settings and clinical trials. nal use. Sharing published material for non-commercial pur- poses is subject to the following conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode, sect. 3. Reproducing and sharing published material for com- Introduction mercial purposes is not allowed without permission in writing from the publisher. Cytogenetic and molecular alterations at diagnosis and response to treatment are the most useful criteria for predicting the relative risk of relapse in acute myeloid leukemia (AML), and for guiding the choice between chemotherapy and 288 haematologica | 2019; 104(2) Deep sequencing method for MRD monitoring in AML hematopoietic stem cell transplantation in first complete sis, and availability of at least one follow-up genomic (g)-DNA remission.1 The definition of complete remission for AML sample. includes criteria for the identification of patients with The MRD approach was applied to 51 (48%) follow-up sam- poor prognosis using cytomorphological methods,2 but ples taken after induction therapy and 55 (52%) taken after con- these studies do not have a good predictive value because solidation, corresponding to 63 patients diagnosed between 2006 most of the patients in complete remission relapse within and 2016 (for selection criteria see Online Supplement 6 and 3 years of diagnosis.3 Supplementary Table S1). Patients were treated according to Assessment of minimal residual disease (MRD) is critical PETHEMA (Programa Español de Tratamientos en Hematología) or in monitoring patients in morphological remission, to CETLAM (Grupo cooperativo de Estudio y Tratamiento de Leucemias inform decisions about further therapy.1 Indeed, several Agudas y Mielodisplasias) protocols. The study was conducted studies have reported MRD status as a stronger predictor according to Spanish law 14/2007 on biomedical research, and of relapse, because patients who are MRD negative have was approved by the Research Ethics Board of each participating a better prognosis than those who are MRD positive.4,5 In institution. All patients provided informed consent. The main support of this, recent non-randomized studies from clinical characteristics of the patients are summarized in Table 1. prospective multicenter trials suggested better outcomes All patients achieved complete remission according to cytomor- when leukemia therapy was selected based on the results phological criteria after induction therapy (<5% of bone marrow of MRD assessment.6-8 blasts). AML is, nevertheless, a biologically complex and het- To construct calibration curves, commercial (Horizon erogeneous disease, which makes MRD testing more chal- Discovery, UK) reference standard gDNA was used for somatic lenging in this condition than in other hematologic neo- SNV in IDH1 (R132C) and IDH2 (R172K). As a further source of plasms such as acute lymphoblastic leukemia or multiple gDNA, we used the OCI-AML3 cell line (ACC 582, DSMZ, myeloma. The detection of very low levels of MRD by Germany) with the NPM1 type A mutation conventional methods such as quantitative (q) polymerase (c.863_864insCCTG) to examine indels. As OCI-AML3 cells also chain reaction (PCR) or multiparameter flow cytometry present a SNV in DNMT3A (R882C), this was included only for (MFC) provides powerful independent prognostic infor- technical optimization. mation. Unfortunately, as described for cytomorphologi- cal complete remission, many patients who achieve MRD Deep targeted sequencing workflow negative status relapse as a result of the progression of The sequencing workflow included a first study at diagnosis undetected leukemic cells. The most common method for and a second study at follow-up. Mutational profile screening at MRD detection is MFC, which has intermediate applica- diagnosis was done with a customized NGS myeloid panel of 32 bility (70–80%) and limited sensitivity.9,10 However, there genes frequently mutated in myeloid diseases,13 (Online is no consensus on multi-antibody panels with regards to Supplementary Table S2) and NPM1 analysis was carried out with inter-laboratory performance, and the technique requires a qPCR.14 high level of expertise. The other principal method for The specific mutations detected at diagnosis were studied at fol- monitoring MRD, qPCR, has good sensitivity (10-4-10-6), low-up. We first tested a variety of experimental steps to define but its applicability is limited to the approximately 40% of optimal conditions (Online Supplement 1). We established an opti- patients who present with molecular alterations (RUNX1- mal protocol (Figure 1) that included DNA amplification, library RUNX1T1, CBFβ-MYH11 or NPM1) at diagnosis.11 preparation and sequencing as experimental steps (Online For the above reasons, new methods with higher sensi- Supplement 2). tivity, specificity, applicability and performance are need- Libraries were sequenced on the Ion Proton System platform ed for MRD assessment in AML. Against this background, (Life Technologies, Thermo Fisher Scientific Inc.) with an estimat- next-generation sequencing (NGS) and digital PCR (dPCR) ed depth ≥1,000,000 of reads, generating .fastq files. These files have recently emerged as potentially promising platforms were analyzed using a customized bioinformatic pipeline; which for the assessment of MRD.12 Here, we optimized and leads from the .fastq file and a .csv file that contains information clinically validated a new deep targeted NGS-based about name identifier, run and barcode identifier, chromosomal method, supported with dPCR technical validation, for position and the variant detected in the diagnosis to be evaluated the detection and quantification of MRD [both small in the follow-up sample. Through Ensembl Perl API,15 the aligned insertion/deletions (indels) and single nucleotide variants mutated sequence and the aligned wild-type (wt) sequence are (SNV)] in AML patients, in an attempt to improve and/or presented in FASTA format (sequences of 40 bp). Finally, we complement the current techniques for MRD evaluation, obtained a .csv file containing the name identifier, run and barcode and to establish its potential as a predictor of patients’ out- identifier, chromosomal position, the variant, the specific target come. sequence in FASTA format (mutated forward, mutated reverse, wt forward and wt reverse), the counts of each and the ratio (mutat- ed/wt) in absolute values. Methods More detailed information can be found in the Online Results Supplementary Data (1–6). A high percentage of acute myeloid leukemia patients Patients and samples could benefit from deep sequencing minimal residual One hundred and ninety patients with de novo or secondary disease assessment non–M3 AML were included in mutational profile screening at In total, 211 (80%) SNV and 46 (20%) indels were diagnosis. We performed a new selection for retrospective MRD detected in the 190 patients analyzed at diagnosis using assessment using the following criteria: presence of the NPM1 the customized NGS panel. We detected one variant (SNV type A mutation, or SNV in FLT3, IDH1 and/or IDH2 at diagno- or indel) in 48 (25%) cases, two or more variants in 116 haematologica | 2019; 104(2) 289 E. Onecha et al. (61%) cases and no variants in 26 (14%) cases. In addition, Survival analysis revealed that positive MRD status we detected the NPM1 type A mutation in 53 (28%) (MRD levels >0.1%) after induction (n=35) was associat- patients by qPCR. Genes (TET2, ASXL1, or DNMT3A) ed with a significantly lower rate of overall survival with evidence of an association with clonal hematopoiesis [33% versus 78%; hazard ratio (HR): 3.5; 95% confidence of indeterminate potential (CHIP) were excluded from the interval (CI): 1.1–10.7; P=0.019], but a non-significant analysis.11 Consequently, 82% of patients in our cohort lower rate of disease-free survival (58% versus 78%; HR: could benefit from this approach. 2.18; 95% CI: 0.63–7.5; P=0.208) (Figure 3A,B). In post- Based on those genes reported as potential markers for consolidation samples (n=28), MRD positive status monitoring MRD,16 and also the availability of follow-up samples, we focused on IDH1/2 and FLT3-SNV. We iden- tified at diagnosis IDH1 mutations in 13 patients (7%), Table 1. Main characteristics of the patients with acute myeloid IDH2 mutations in 27 patients (14%) and FLT3-SNV leukemia included in the minimal residual disease study. mutations (18%) in 34 patients. Patients (n = 63) Deep sequencing minimal residual disease assessment Follow-up sample type has a sensitivity of 10-4 for single nucleotide variants Bone marrow 58 (92%) and 10-5 for insertions-deletions Peripheral blood 5 (8%) To establish the limit of quantification (LOQ) of the Sex method, we used 10-fold serial dilutions of mixed mutat- Male 21 (33%) ed and control DNA. To study prototype indels, we used Female 42 (67%) gDNA from OCI-AML3 cells (NPM1 type A) and to study Age at diagnosis, median 54 (IQR, 41.5–66.0) prototype SNV, we used both gDNA from OCI-AML3 cells (DNMT3A) and commercial reference gDNA Blasts at diagnosis, median count 69 (IQR, 51.0–81.0) (IDH1/IDH2). As a control, we used a pool of gDNA from ten individuals without somatic mutations in these chro- Leukocytes at diagnosis 15.7 (IQR,12.2–20.24) mosomal regions. In all cases, initial allele frequency was median count (×109/L) 50% and a total of six dilutions were tested to construct a Secondary AML calibration curve, covering a theoretical dynamic range No 59 (94%) from 10-1 to 10-7. Yes 4 (6%) As shown in Figure 2A,B, MRD NGS testing of NPM1 (indel) could quantify one mutated cell in the order of 10-5 Cytogenetic risk normal ones and in the case of SNV (IDH1, IDH2 and Favorable 25 (40%) Intermediate 36 (57%) DNMT3A) the LOQ was 10-4, which was reproducible for Adverse 2 (3%) all SNV tested. FLT3-ITD Next-generation sequencing is more sensitive than FLT3 negative 49 (78%) digital polymerase chain reaction analysis for minimal FLT3 positive 14 (22%) residual disease assessment FLT3-TKD We compared the sensitivity of the sequencing method FLT3 negative 60 (95%) with that of dPCR using the same LOQ dilution protocol. FLT3 positive 3 (5%) Clone frequency expressed as target concentration (mutat- NPM1 ed copies/mL in wt copies/mL) gradually decreased with NPM1 negative 6 (10%) each dilution, reaching a LOQ of 10-3 for NPM1, IDH1 and NPM1 positive 57 (90%) IDH2 (Figure 2C,D). While both methods showed similar Hematopoietic stem cell transplantation detection limits and good linearity, the LOQ for the No 42 (67%) sequencing method was one order of magnitude higher Allogeneic 7 (11%) than that for dPCR (IDH1 and IDH2), and two orders of Autologous 14 (22%) magnitude higher for indels (NMP1). Relapse No 42 (67%) Minimal residual disease status assessed Yes 21 (33%) by sequencing has prognosis value in acute Death myeloid leukemia No 40 (63%) The median depth coverage was 401,300 aligned reads Yes 23 (37%) (interquartile range 195,100–825,700) for the 88 NPM1 and Treatment* 18 SNV (9 IDH1, 7 IDH2, and 2 FLT3) follow-up samples 3+7 regimen 50 (80%) evaluated. We detected no mutated sequence in 13 (12%) Flugaza 8 (13%) samples, one to five mutated sequences in 19 (18%) sam- Mylotarg 2 (3%) ples, and more than ten in 74 (70%) samples. The ratio of Panobidara 3 (4%) mutated sequences to wt sequences defined MRD levels. AML: acute myeloid leukemia; ITD: internal tandem duplications; TKD: tyrosine kinase Considering MRD levels from the 106 samples evaluated domain; *3+7 regimen of chemotherapy: one or two induction cycles of cytarabine and idarubicin for 7 and 3 days, respectively; and two or three consolidation cycles of we established the optimal cutoff to classify MRD status high doses of cytarabine, twice a day for 3 alternate days followed by allogeneic or (positive versus negative) by receiver operating characteristic autologous hematopoietic stem cell transplantation. The remainder of the patients curves (Online Supplementary Figure S1) at each check-point were included in other clinical trials (Mylotarg, NTC0104104; Flugaza, NCT02319135; Panobidara, NCT00840346). Clinical data were collected in the following Spanish AML of MRD evaluation [post-induction (n=51), post-consolida- epidemiological registries: NCT01700413, NCT02006004, NCT00464217, NCT02607059, tion (n=55), or both together (n=106)]. NCT01041040 and NCT01296178. 290 haematologica | 2019; 104(2) Deep sequencing method for MRD monitoring in AML (MRD levels >0.025%) was associated with both signifi- shorter disease-free survival (57% versus 85%; HR: 2.9; cantly shorter overall survival (33% versus 81%; HR: 6.0; 95% CI: 0.9–7.6; P=0.052). Similar results were found 95% CI: 1.3–28.7; P<0.001) and significantly shorter dis- when we evaluated IDH1, IDH2 or FLT3-SNV as MRD ease-free survival (17% versus 94%; HR: 19.6; 95% CI: markers (n=11). Accordingly, MRD positive status was 2.5–155.6; P<0.001) (Figure 3C,D). Survival outcomes associated with both significantly shorter overall survival were also analyzed combining post-induction and post- (17% versus 100%; HR: not applicable; P=0.041), and consolidation (n=63) tests, in order to compare survival shorter disease-free survival (17% versus 75%; HR: 6.3; analysis with MFC and qPCR data sets. We observed 95% CI:0.7–54; P=0.058). that positive MRD status (MRD levels >0.035%) was In univariate Cox analysis (Table 2A), the risk of death associated with a higher risk of relapse (48% versus 81%; was significantly higher with increasing age (HR: 1.04; HR: 3.4; 95% CI: 1.4–8.5; P=0.005) and death (37% ver- P=0.013), in patients with FLT3-ITD (HR: 3.45; P=0.007), sus 81%; HR: 4.2; 95% CI: 1.6–10.7; P<0.001) (Figure and in those with MRD positive status as determined by 3E,F). In order to test the power of NPM1 and SNV as NGS (HR: 4.22; P=0.002). The risk of relapse was signifi- independent predictive markers, we performed the cantly higher only in those patients with MRD positive analysis separately. Evaluating NPM1 as an MRD marker status determined by NGS (HR: 3.4; P=0.008). In multi- (n=54), we found that MRD positive status was associat- variate analysis (Table 2B), the risk of death was signifi- ed with both significantly shorter overall survival (43% cantly higher with increasing age (HR: 1.05; P=0.004), in versus 78%; HR: 3.3; 95% CI: 1.2–8.8; P=0.011), and patients with mutated FLT3-ITD (HR: 8.87; P=0.001), and Figure 1. Workflow of the next-geneartion sequencing – minimal residual disease method. DNA amplification, library preparation and sequencing experimental workflow. Genomic DNA (gDNA) is amplified by quanti- tative (q) polymerase chain reaction (PCR) using specific primers. Libraries are pre- pared in four steps: end repair, adaptor ligation, size selection, and PCR amplifica- tion. The libraries are then sequenced. A customized bioinformatic pipeline ana- lyzes the sequences obtained. The results are expressed as a ratio of mutated sequences (mut) among wild-type (wt) sequences. haematologica | 2019; 104(2) 291 E. Onecha et al. in those with MRD positive status determined by NGS Discussion (HR: 4.54; P=0.005). The risk of relapse was higher only in patients who were MRD positive as determined by NGS We have optimized and validated a high sensitivity (HR: 3.76; P=0.012). NGS method for the detection and quantification of NPM1, IDH1, IDH2 and FLT3-SNV mutated sequences at Minimal residual disease assessment by sequencing very low allele frequency in follow-up gDNA samples. predicts overall survival and disease-free survival bet- NGS has demonstrated prognostic value for pre-treatment ter than multiparameter flow cytometry or quantitative status in patients with AML,17 and may also be a useful polymerase chain reaction analysis tool for detecting MRD.18,19 We first studied the mutational A positive correlation was found when comparing MRD profile of patients with AML using a customized NGS assessment by NGS versus MFC (r=0.47, P=0.005, n=75), panel to ensure a high applicability (82% of patients). This and NGS versus qPCR (r=0.62, P<0.001, n=80) (Online approach is also a useful screening method for detecting Supplementary Figure S2). There were differences between all potential MRD markers and choosing those most rele- positive MRD and negative MRD groups of patients test- vant. The combination of several markers is possible and ed by MFC, but they were not significant for either overall recommended to overcome limitations of MRD assess- survival (P=0.193) or disease-free survival (P=0.117) (n=46) ment due to sub-clonal heterogeneity of AML and CHIP.11 (Figure 4A). Similarly, differences were observed between Accordingly, our method has the capacity to evaluate mul- positive MRD and negative MRD groups defined by qPCR tiple markers simultaneously and, considering that 61% of of NPM1, although statistical significance was not reached patients in our cohort had two or more genetic alterations, for either overall survival (P=0.212) or disease-free survival this approach is sufficiently robust to monitor MRD even (P=0.086) (n=46) (Figure 4B). in patients with clonal evolution. A B C D Figure 2. Calibration curve of minimal residual disease in serial dilutions. (A,B) Ten-fold dilution curves for the assessment of the sensitivity of next-generation sequencing (NGS) in (A) insertions-deletions (InDel), using OCI-AML3 gDNA with 50% NPM1 type A mutation (R2 = 0.98); and (B) single nucleotide variabts (SNV), using OCI-AML3 gDNA with 50% mutated DNMT3A (R2 = 0.98), and gDNA with 50% mutated IDH1 or IDH2 from a commercial standard (R2 = 0.91 and R2 = 0.98, respectively). (C,D) The same 10-fold dilution curves for the assessment of sensitivity of digital polymerase chain reaction (dPCR) in (C) InDel (R2 = 0.98); and (D) SNV (R2 = 0.91 for IDH1 and R2 = 0.98 for IDH2). The vertical red bars indicate the limit of quantification (LOQ) according to the sample. Clone frequency is expressed as target concentration as mutated copies/mL in wild-type copies/mL. Negative controls are included in the calibration curves and had levels below the corresponding LOQ values. 292 haematologica | 2019; 104(2) Deep sequencing method for MRD monitoring in AML Reported variants associated with CHIP are frequently prognosis because in these cases the genetic alterations located in DNMT3A, TET2 or ASXL1 genes, and are have been acquired in the leukemic clone and not before.24 detected in the preleukemic phase and during complete The sensitivity of this method equates to one mutated AML remission.20-23 Indeed, any gene could carry both cell per 100,000 cells (LOQ 10-5) for NPM1 and one mutat- CHIP and non-CHIP variants, and these should be evalu- ed cell per 10,000 cells (LOQ 10-4) for IDH1, IDH2 and ated for each patient. Moreover, studies have shown that FLT3-SNV. This difference in sensitivity is related to the genes related to CHIP (IDH1/2) are useful for predicting fact that the NPM1 type A mutation (insCCTG) is rarely A B C D Figure 3. Analysis of overall sur- vival and disease-free survival in patients with acute myeloid leukemia stratified according to minimal residual disease lev- els determined by sequencing. E F Analysis of overall survival for (A) the induction data set, (C) the consolidation data set, and (C) both together. Analysis of disease-free survival for (B) the induction data set, (D) the con- solidation data set, and (F) both together. The cutoff used for overall and disease-free survival was 0.001 at the post–induc- tion check-point (n=35), 0.00026 at the post-consolida- tion check–point (n=28) and 0.00035 for both check-points (all data set) (n=63). The num- bers of censored patients with respect to the stratified groups and the numbers at risk are indicated. Statistically signifi- cant values: *P<0.05, **P<0.01. haematologica | 2019; 104(2) 293 E. Onecha et al. Table 2. Cox regression analyses. A. Risk of death Risk of relapse HR (95%CI) P value HR (95%CI) P value Sex (female vs. male) 1.20 (0.50–2.83) 0.682 0.94 (0.37–2.44) 0.906 Age per year 1.04 (1.00–1.07) 0.013 * 1.03 (0.99–1.06) 0.069 Blasts at diagnosis (%) 1.00 (0.99–1.02) 0.667 1.01 (0.99–1.03) 0.532 Leukocytes at diagnosis (×109/L) 1.01 (0.99–1.01) 0.418 1.00 (0.99–1.01) 0.508 Favorable vs. adverse ELN risk 0.67 (0.08–5.43) 0.714 0.75 (0.09–6.00) 0.786 Intermediate vs. adverse ELN risk 1.03 (0.13–7.86) 0.976 1.02 (0.13–7.82) 0.988 Mutated FLT3-ITD 3.45 (1.40-8.52) 0.007 * 2.37 (0.86–6.51) 0.095 Allo-HSCT vs. intensive chemotherapy 1.35 (0.40–4.57) 0.634 1.78 (0.41–7.78) 0.44 Allo-HSCT vs. auto-HSCT 0.29 (0.05–1.74) 0.176 0.64 (0.11–3.77) 0.629 MRD+ by MFC 2.10 (0.67–6.62) 0.203 2.40 (0.77–7.46) 0.130 MRD+ by qPCR 2.51 (0.56–11.2) 0.228 5.01 (0.64–38.8) 0.123 MRD+ by NGS 4.22 (1.66–10.7) 0.002 ** 3.41 (1.37–8.48) 0.008 ** B. Risk of death Risk of relapse HR (95%CI) P value HR (95%CI) P value Age per year 1.05 (1.02–1.09) 0.004 * 1.03 (0.99–1.07) 0.061 Sex (female vs. male) 0.84 (0.33–2.17) 0.720 1.25 (0.44–3.52) 0.671 Leukocytes at diagnosis (×109/L) 1.01 (0.99–1.03) 0.219 1.07 (0.99–1.02) 0.481 Favorable vs. adverse ELN risk 13.75 (0.84–226.1) 0.067 7.09 (0.37–134.15) 0.192 Intermediate vs. adverse ELN risk 11.22(0.82–154.2) 0.071 5.86 (0.39–86.84) 0.203 Mutated FLT3-ITD 8.87 (2.54–30.95) 0.001 ** 4.18 (1.11–15.69) 0.034* MRD+ by NGS 4.54 (1.58–13.03) 0.005 ** 3.76 (1.34–10.54) 0.012* Cox regression analyses of prognostic factors influencing the risk of relapse and risk of death of patients with acute myeloid leukemia. (A) Univariate Cox regression analysis of each prognostic factor. (B) Multivariate Cox regression analysis evaluating the most relevant factors detected in the univariate analyses. HR: hazard ratio; 95% CI: 95% confi- dence interval; ITD: internal tandem duplication; ELN: European LeukemiaNet; allo-HSCT: allogeneic hematopoietic stem cell transplantation; auto-HSCT: autologous hematopoi- etic stem cell transplantation; MFC, multiparametric flow cytometry; MRD, minimal residual disease; NGS, next-generation sequencing. Statistically significant values: *P<0.05, **P<0.01. generated erroneously by NGS, and the quantification is (10-2).30 Compared with NGS, dPCR is a faster measure- precise. ment technique but, as it is focused, it requires allele-spe- Our method, as with any NGS method, has an intrinsic cific primers that can complicate the experimental proce- error rate that limits its sensitivity for most SNV to 1–2% dure, and a high number of parallel experiments are of all reads. This limitation can nevertheless be overcome needed to raise the sensitivity, which increases the cost by virtue of the scalable nature of NGS.16 Thus, we boost- of the assay. Additionally, although it is possible to mul- ed NGS sensitivity by increasing the amount of DNA by tiplex dPCR, unfortunately only a few targets can be PCR prior to sequencing, which increased the depth of monitored simultaneously within each sample.29 Another coverage to one million reads. By also optimizing the advantage of NGS technology is that it does not require bioinformatic analysis, we focused the search for the pre- calibration curves in each assay, and the results are cise variant in order to eliminate random sequencing reported in absolute values, facilitating its standardiza- errors, enhancing the specificity of the technique and tion. reducing the computational time. To the best of knowl- The NGS method described here showed comparable edge, our NGS method presents possibly the highest sen- sensitivities (10-4 for SNV and 10-5 for indels) to those of sitivity reported for NGS in AML.18,19,24-27 MFC methods in those cases with immunophenotyphi- dPCR is a relatively novel technique for precise and cally aberrant populations.10,31 Although our method had a absolute quantification of nucleic acids, which is based similar sensitivity to that of qPCR, it does not require on limiting partitions of the PCR volume and Poisson sta- oligonucleotides that hybridize specifically to a particular tistics.28 It is also an extremely sensitive technique, with sequence, so all nucleotides in the amplified region can be a high specificity due to the detection of mutant alleles.29 studied. Consequently, the NGS test is capable of detect- However, when we compared the same standard dilu- ing all NPM1 subtype mutations in the same assay. tions in NGS and dPCR, NGS afforded a 2-log increment We found positive correlations when MRD levels were in LOQ for indels (NPM1) and a 1-log increment for SNV evaluated by NGS versus MFC and versus qPCR, but not (IDH1/2), with the sensitivity of dPCR for indels being with the expected results. In the case of MFC, this could be similar to that reported in a previously published study explained, in part, by the fact that NPM1 mutations are 294 haematologica | 2019; 104(2) Deep sequencing method for MRD monitoring in AML A B C D Figure 4. Analysis of overall survival and disease-free survival in patients with acute myeloid leukemia stratified according to minimal residual disease levels deter- mined by conventional methods. Kaplan-Meier plots of (A) overall survival and (B) disease-free survival according to minimal residual disease (MRD) assessment by multiparametric flow cytometry (MRC) and (C) overall survival and (D) disease-free survival according to MRD assessment by quantitative polymerase chain reac- tion (qPCR) analysis. The numbers of censored patients with respect to each stratified group and numbers at risk are indicated. Statistically significant values: *P<0.05, **P<0.01. usually associated with monocytic subtype-AML, which Accordingly, mutated DNA is more representative of the frequently presents more difficulties for identifying MRD tumor burden than is overexpression of mutated RNA.35 It by MFC. Indeed, Salipante et al.27 described that the level of should be noted that the prediction of survival and progres- success of MFC depends greatly on the immunophenotype sion of AML using MRD NGS was better than that of the of the abnormal blasts and how to discriminate them from other methodologies employed, at least in the cohorts background regenerative blasts. Moreover, due to the lack evaluated. of standardization, MFC shows substantial variability Finally, survival analysis showed that MRD positive sta- across laboratories, including that of sample processing, tus determined by NGS was associated with a higher risk instrument configuration, number of events, and training of relapse and death and that MRD negative status in post- of pathologists.32 The lack of a strong correlation between consolidation ssmples was associated with longer overall NGS and qPCR could be explained by the nature of the and disease-free survival, in accordance with recently pub- sample (sequencing uses gDNA whereas qPCR uses lished studies.23 Supporting these findings, previous studies cDNA). Although RNA overexpression allows a higher reported that an MRD check-point after consolidation sensitivity of detection, RNA levels do not correlate with could be the best moment for analysis because it afforded the number of tumor cells, in contrast to mutated DNA. better prediction.8,34-37 Cox regression multivariate analyses haematologica | 2019; 104(2) 295 E. Onecha et al. confirmed that MRD positive status determined by Acknowledgments sequencing was the only statistically significant predictor This study was supported by the Subdirección General de of the risk of relapse (P=0.012). Investigación Sanitaria (Instituto de Salud Carlos III, Spain) grants In conclusion, we have optimized a new targeted PI13/02387 and PI16/01530, and the CRIS against Cancer foun- sequencing method with high sensitivity for MRD evalua- dation grant 2014/0120. ML holds a postdoctoral fellowship of the tion with applicability in a high percentage of AML Spanish Ministry of Economy and Competitiveness (FPDI-2013- patients, improving the capacity, over MFC or qPCR, to 16409). PRP holds a postdoctoral fellowship of the Spanish predict the survival outcomes of the AML patients in our Instituto de Salud Carlos III: Contrato Predoctoral de Formación en cohort. Investigación en Salud i-PFIS (IFI 14/00008). residual disease in acute myeloid leukemia: Transplant. 2017;23(7):1064-1071. References which platforms are ready for "prime time"? 26. Debarri H, Lebon D, Roumier C, et al. Hematology Am Soc Hematol Educ Program. IDH1/2 but not DNMT3A mutations are 1. Dohner H, Estey E, Grimwade D, et al. 2014;2014(1):222-233. suitable targets for minimal residual disease Diagnosis and management of AML in 13. Cedena MT, Rapado I, Santos-Lozano A, et monitoring in acute myeloid leukemia adults: 2017 ELN recommendations from an al. Mutations in the DNA methylation path- patients: a study by the Acute Leukemia international expert panel. Blood. way and number of driver mutations predict French Association. Oncotarget. 2015;6(39): 2017;129(4):424-447. response to azacitidine in myelodysplastic 42345-42353. 2. Cheson BD, Bennett JM, Kopecky KJ, et al. syndromes. Oncotarget. 2017;8(63):106948- 27. Salipante SJ, Fromm JR, Shendure J, et al. Revised recommendations of the 106961. Detection of minimal residual disease in International Working Group for Diagnosis, 14. Gorello P, Cazzaniga G, Alberti F, et al. NPM1-mutated acute myeloid leukemia by Standardization of Response Criteria, Quantitative assessment of minimal residual next-generation sequencing. Mod Pathol. Treatment Outcomes, and Reporting disease in acute myeloid leukemia carrying 2014;27(11):1438-1446. Standards for Therapeutic Trials in Acute nucleophosmin (NPM1) gene mutations. 28. Brunetti C, Anelli L, Zagaria A, et al. Droplet Myeloid Leukemia. J Clin Oncol. Leukemia. 2006;20(6):1103-1108. digital PCR is a reliable tool for monitoring 2003;21(24):4642-4649. 15. McLaren W, Gil L, Hunt SE, et al. The minimal residual disease in acute promyelo- 3. Dohner H, Weisdorf DJ, Bloomfield CD. ensembl variant effect predictor. Genome cytic lseukemia. J Mol Diagn. 2017;19(3): Acute myeloid leukemia. N Engl J Med. Biol. 2016;17(1):122. 437-444. 2015;373(12):1136-1152. 16. Tomlinson B, Lazarus HM. Enhancing acute 29. Roloff GW, Lai C, Hourigan CS, et al. 4. Ivey A, Hills RK, Simpson MA, et al. myeloid leukemia therapy - monitoring Technical advances in the measurement of Assessment of minimal residual disease in response using residual disease testing as a residual disease in acute myeloid leukemia. J standard-risk AML. N Engl J Med. guide to therapeutic decision-making. Clin Med. 2017;6(9). 2016;374(5):422-433. Expert Rev Hematol. 2017;10(6):563-574. 30. Minervini A, Francesco Minervini C, Anelli 5. Kronke J, Schlenk RF, Jensen KO, et al. 17. Patel JP, Gonen M, Figueroa ME, et al. L, et al. Droplet digital PCR analysis of Monitoring of minimal residual disease in Prognostic relevance of integrated genetic NOTCH1 gene mutations in chronic lym- NPM1-mutated acute myeloid leukemia: a profiling in acute myeloid leukemia. N Engl phocytic leukemia. Oncotarget. 2016;7(52): study from the German-Austrian Acute J Med. 2012;366(12):1079-1089. 86469-86479. Myeloid Leukemia Study Group. J Clin 18. Thol F, Kolking B, Damm F, et al. Next-gen- 31. Ossenkoppele G, Schuurhuis GJ. MRD in Oncol. 2011;29(19):2709-2716. eration sequencing for minimal residual dis- AML: does it already guide therapy deci- 6. Rubnitz JE, Inaba H, Dahl G, et al. Minimal ease monitoring in acute myeloid leukemia sion-making? Hematology Am Soc Hematol residual disease-directed therapy for child- patients with FLT3-ITD or NPM1 muta- Educ Program. 2016;2016(1):356-365. hood acute myeloid leukaemia: results of tions. Genes Chromosomes Cancer. 32. Cruz NM, Mencia-Trinchant N, Hassane the AML02 multicentre trial. Lancet Oncol. 2012;51(7):689-695. DC, et al. Minimal residual disease in acute 2010;11(6): 543-552. 19. Kohlmann A, Nadarajah N, Alpermann T, et myelogenous leukemia. Int J Lab Hematol. 7. Balsat M, Renneville A, Thomas X, et al. al. Monitoring of residual disease by next- 2017;39 (Suppl 1):53-60. Postinduction minimal residual disease pre- generation deep-sequencing of RUNX1 33. Duployez N, Nibourel O, Marceau-Renaut dicts outcome and benefit from allogeneic mutations can identify acute myeloid A, et al. Minimal residual disease monitoring stem cell transplantation in acute myeloid leukemia patients with resistant disease. in t(8;21) acute myeloid leukemia based on leukemia with NPM1 mutation: a study by Leukemia. 2014;28(1):129-137. RUNX1-RUNX1T1 fusion quantification on the Acute Leukemia French Association 20. Metzeler KH, Herold T, Rothenberg- genomic DNA. Am J Hematol. 2014;89(6): Group. J Clin Oncol. 2017;35(2):185-193. Thurley M, et al. Spectrum and prognostic 610-615. 8. Zhu HH, Zhang XH, Qin YZ, et al. MRD- relevance of driver gene mutations in acute 34. Jacobsohn DA, Tse WT, Chaleff S, et al. directed risk stratification treatment may myeloid leukemia. Blood. 2016; 128(5):686- High WT1 gene expression before improve outcomes of t(8;21) AML in the 698. haematopoietic stem cell transplant in chil- first complete remission: results from the 21. Bullinger L, Dohner K, Dohner H. Genomics dren with acute myeloid leukaemia predicts AML05 multicenter trial. Blood. of acute myeloid leukemia diagnosis and poor event-free survival. Br J Haematol. 2013;121(20):4056-4062. pathways. J Clin Oncol. 2017;35(9):934-946. 2009;146(6):669-674. 9. Freeman SD, Virgo P, Couzens S, et al. 22. Koeffler HP, Leong G. Preleukemia: one 35. Hourigan CS, Gale RP, Gormley NJ, et al. Prognostic relevance of treatment response name, many meanings. Leukemia. 2017; Measurable residual disease testing in acute measured by flow cytometric residual dis- 31(3):534-542. myeloid leukaemia. Leukemia. 2017;31(7): ease detection in older patients with acute 23. Jongen-Lavrencic M, Grob T, Hanekamp D, 1482-1490. myeloid leukemia. J Clin Oncol. et al. Molecular minimal residual disease in 36. Rossi G, Carella AM, Minervini MM, et al. 2013;31(32):4123-4131. acute myeloid leukemia. N Engl J Med. Optimal time-points for minimal residual 10. Terwijn M, van Putten WL, Kelder A, et al. 2018;378(13):1189-1199. disease monitoring change on the basis of High prognostic impact of flow cytometric 24. Klco JM, Miller CA, Griffith M, et al. the method used in patients with acute minimal residual disease detection in acute Association between mutation clearance myeloid leukemia who underwent allogene- myeloid leukemia: data from the after induction therapy and outcomes in ic stem cell transplantation: a comparison HOVON/SAKK AML 42A study. J Clin acute myeloid leukemia. JAMA. 2015;314 between multiparameter flow cytometry Oncol. 2013;31(31):3889-3897. (8):811-822. and Wilms' tumor 1 expression. Leuk Res. 11. Schuurhuis GJ, Heuser M, Freeman S, et al. 25. Getta BM, Devlin SM, Levine RL, et al. 2015;39(2):138-143. Minimal/measurable residual disease in Multicolor flow cytometry and multigene 37. Jourdan E, Boissel N, Chevret S, et al. AML: a consensus document from the next-generation sequencing are complemen- Prospective evaluation of gene mutations European LeukemiaNet MRD Working tary and highly predictive for relapse in and minimal residual disease in patients Party. Blood. 2018;131(12):1275-1291. acute myeloid leukemia after allogeneic with core binding factor acute myeloid 12. Grimwade D, Freeman SD. Defining minimal transplantation. Biol Blood Marrow leukemia. Blood. 2013;121(12): 2213-2223. 296 haematologica | 2019; 104(2)