Nucleophosmin1 and isocitrate dehydrogenase 1 and 2 as measurable residual disease markers in acute myeloid leukemia

Monitoring measurable residual disease (MRD) in acute myeloid leukemia (AML) plays an important role in predicting relapse and outcome. The applicability of the leukemia-initiating nucleophosmin1 (NPM1) gene mutations in MRD detection is well-established, while that of isocitrate dehydrogenase1/2 (IDH1/2) mutations are matter of debate. The aim of this study was to investigate the stability of NPM1 and IDH1/2 mutations at diagnosis and relapse retrospectively in 916 adult AML patients. The prognostic value of MRD was evaluated by droplet digital PCR on the DNA level in a selected subgroup of patients in remission. NPM1 re-emerged at relapse in 91% (72/79), while IDH1/2 in 87% (20/23) of mutation-positive cases at diagnosis. NPM1 mutation did not develop at relapse, on the contrary novel IDH1/2 mutations occurred in 3% (3/93) of previously mutation-negative cases. NPM1 MRD-positivity after induction (n = 116) proved to be an independent, adverse risk factor (MRDpos 24-month OS: 39.3±6.2% versus MRDneg: 58.5±7.5%, p = 0.029; HR: 2.16; 95%CI: 1.25–3.74, p = 0.006). In the favorable subgroup of mutated NPM1 without fms-like tyrosine kinase 3 internal tandem duplication (FLT3-ITD) or with low allelic ratio, NPM1 MRD provides a valuable prognostic biomarker (NPM1 MRDpos versus MRDneg 24-month OS: 42.9±6.7% versus 66.7±8.6%; p = 0.01). IDH1/2 MRD-positivity after induction (n = 62) was also associated with poor survival (MRDpos 24-month OS: 41.3±9.2% versus MRDneg: 62.5±9.0%, p = 0.003; HR 2.81 95%CI 1.09–7.23, p = 0.032). While NPM1 variant allele frequency decreased below 2.5% in remission in all patients, IDH1/2 mutations (typically IDH2 R140Q) persisted in 24% of cases. Our results support that NPM1 MRD even at DNA level is a reliable prognostic factor, while IDH1/2 mutations may represent pre-leukemic, founder or subclonal drivers.

Introduction Acute myeloid leukemia (AML) is an aggressive hematological malignancy with a rapidly evolving treatment paradigm. Although the majority of patients remain incurable, long-term remissions can be achieved in roughly one-third of these patients. The identification of prognostic markers bears outstanding relevance for optimizing treatment strategy. Measurable residual disease (MRD) after induction therapy and before hematopoietic stem cell transplantation is an independent, post-diagnosis prognostic indicator of relapse and survival. The application of molecular genetics and multiparametric flow cytometry are recommended for monitoring. Requirements for a reliable molecular genetic MRD marker are the following: (i) mutation burden fluctuates in parallel with leukemic tumor burden: present at disease onset, disappearing in remission and re-emerging at relapse, (ii) available method with the capability of achieving high sensitivity [1][2][3].
Nucleophosmin1 (NPM1) mutations are among the most frequently detected genetic alterations in AML (present in 25-35% of primary AML) defining a separate disease entity. NPM1 frameshift mutations result in altered protein termination, loss of nuclear localization signals, and consequential abnormal cytoplasmic localization of the mutant protein [4][5][6]. Isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) mutations occur in 7-14% and 8-19% of AML cases respectively. The gain-of-function mutations result in the production of an oncometabolite with consequential hypermethylation, gene expression alterations and impaired hematopoietic differentiation [4,7].
NPM1 alterations were reported as definite leukemia-founder mutations and optimal MRD markers. On the other hand contradictory data exist, whether IDH1and IDH2 mutations represent pre-leukemic, or dominant clone mutations, therefore their value in MRD monitoring is not well established [3,8]. In our study, we aimed to correlate NPM1 and IDH1 and IDH2 mutational variant allele frequencies at diagnosis, remission and relapse to investigate the potential application of these mutations in MRD monitoring.

Patients
The study included 916 adult patients (449 males/467 females, median age at diagnosis 54 years; range: 16-94), consecutively diagnosed with AML between January 2001 and June 2020 in our Institute (Department of Hematology and Stem Cell Transplantation, Central Hospital of Southern Pest National Institute of Hematology and Infectious Diseases, Budapest, Hungary). In this patient cohort 253 patients were NPM1, 68 IDH1 and 94 IDH2 mutations positive (74 patients carried both NPM1 and IDH1/2 mutations). A significant proportion of patients 81% (n = 746/916) received curative treatment, out of which 26% (n = 176/746) was treated by allogeneic hematopoietic stem cell transplantation (HSCT). MRD monitoring was retrospectively evaluated in a selected subgroup of 116 NPM1 (51 male/65 female, median age at diagnosis 48 years), and 62 IDH1/2 positive patients (23 male/39 female, median age was 49 years). The inclusion criteria for the MRD monitored subgroup were the following: (i) curative chemotherapy; (ii) morphologic leukemia-free state (MLFS) after induction [1]; (iii) available DNA sample at diagnosis, after induction, and/or before HSCT. Patients with palliative therapy, death in aplasia or death from indeterminate cause, no MLFS after 2 courses of intensive induction treatment; unavailable DNA sample, or patients with rare undetectable NPM1 or IDH1/2 mutation were excluded from MRD evaluation. MRD was determined after induction and one month before HSCT if DNA samples were available. Data from AML patients diagnosed between 2001 and 2009 have already been reported in an earlier study [9] and IDH1/2 data between 2001-2018 were presented in a Hungarian report [10]. Data collection was performed retrospectively. Definitions of fms-like tyrosine kinase 3 internal tandem duplication (FLT3-ITD) low and high allelic ratio, MLFS, overall survival (OS) and relapse-free survival (RFS) were described by European LeukemiaNet (ELN) 2017 recommendations [1]. The study was in accordance of the Declaration of Helsinki and was approved by the Institutional Review Board of Central Hospital of Southern Pest National Institute of Hematology and Infectious Diseases. Written informed consent was provided by all patients.

Molecular genetic methods
Genomic DNA and RNA were isolated from bone marrow samples drawn at diagnosis, remission and relapse. Screening for hotspot mutations were performed from genomic DNA, at the time points of diagnosis and repeatedly at relapse by fragment analysis in case of NPM1 (NPM1 diagnosis n = 916, relapse n = 161 if DNA was available); [11], and by high-resolution melting (HRM) or allele specific PCR in case of IDH1/2 (diagnosis n = 842, relapse n = 116 if DNA was available) [9]. Positive cases were monitored with droplet digital PCR (ddPCR) after induction therapy (NPM1 n = 116; IDH1/2 n = 62; double positive = 33), before HSCT (1-30 days before; NPM1 n = 38; IDH1/2 n = 22), if DNA was available at that time point. Mutant NPM1 RNA expression was also tested at diagnosis and after induction therapy (n = 39).
Diagnosis and follow-up samples of NPM1 as well as IDH1 and IDH2 positive AML patients were investigated by ddPCR. For NPM1 mutation detection primer and probe sequences are summarized in S1 Table [ [12][13][14]. NPM1 type-A (c.860_863dupTCTG, p.Trp288CysfsTer12) specific reverse primer was described by Gorello et al. [13,15]. A degenerate R primer (type-N) reported by Mencia-Trinchant et al. [14] was applied to detect NPM1 mutations at the same position with different nucleotide insertions (c.860_863dupNNNN, p.Trp288Cysf-sTer12, referred as NPM1 type-N mutation in this study). GAPDH was used as the reference gene for the assay for DNA [16], and ABL1 for RNA [17]. Reactions were performed using Supermix for Probes (no dUTP) (BioRad), 900 nM of each primer, 250 nM of each probe, 100 ng DNA or 240 ng cDNA per well. For genomic DNA, assays were designed by Bio-Rad for the detection of the most common IDH1/2 mutations (IDH1 R132C ID: dHsaMDV2010053, R132H ID: dHsaMDV2010055 and IDH2 R140Q ID: dHsaMDV2010057, R172K ID: dHsaMDV2010059). The PCR program started with an initial denaturation at 98˚C for 10 min, 40 cycles of denaturation at 94˚C for 30 sec, annealing at 55˚C (for DNA) and at 60˚C (for RNA) for 60 sec followed by enzyme deactivation at 98˚C for 10 min. The QX200 Droplet Digital PCR System and QuantaSoft Software (Version 1.7.4.0917, BioRad) were used for the evaluation of the results.
RFS [1]. After induction OS were calculated from the time point of diagnosis, RFS from remission irrespective from performing HSCT. Regarding pre-transplant MRD monitoring, comparisons of OS and RFS were performed from the time point of HSCT. Following univariate analysis, age, cytogenetics, FLT3-ITD allelic ratio [1], NPM1, white blood cell count (WBC) at diagnosis, and MRD status were included in a Cox proportional hazard model for OS and RFS. Hazard ratios (HR) and 95% confidence interval (95%CI) values were calculated. In order to identify the cut off discriminating between low and high MRD burden groups, HRs for OS were compared at six different limits (0.05%; 0.1%; 0.2%; 0.5%; 1% and 2%) for NPM1 type-A and type-N separately and combined [21]. P values below 0.05 were considered as statistically significant. For the statistical analysis SPSS Statistics version 22 (Armonk, NY) was applied.
In the NPM1-positive cohort, 211 patients were treated with curative intent, out of which remission (MLFS) was reached in 174 cases (Fig 1). The stability of NPM1 mutation during disease evolution was studied with 79 paired NPM1 mutant samples drawn at diagnosis and relapse. The NPM1 mutation re-emerged at relapse in 91% of NPM1 positive cases (n = 72/79). Time period from diagnosis till relapse was not significantly longer in cases where NPM1 was undetectable at relapse compared to cases with persistent NPM1 mutation at relapse [median 7.1 month (range: 0.1-172.2 month) versus 6.6 month (range: 2.2-152.9 month) respectively, p = 0.46]. All seven patients with clonal NPM1 regression had normal karyotype at the time of diagnosis; one patient out of five with karyotyping available at relapse had clonal evolution (trisomy 8). None of our NPM1 negative AML cases gained NPM1 mutation positivity at relapse, In the IDH1/2-positive cohort (n = 162), 132 patients were treated with curative intent, out of which remission (MLFS) was reached in 90 cases (Fig 2). IDH1/2 mutations were undetectable at relapse in 13% of the IDH1/2-positive cohort with available DNA (n = 3/23, 1 IDH1 R132C, 1 IDH2 R140Q and 1 IDH2 R172K). Time from diagnosis till relapse was not proven to be significantly longer in cases where IDH1/2 was undetectable at relapse compared to cases with persistent IDH1/2 mutation at relapse [median 7.4 month (range: 2.2-11.4 month) versus 8.6 month (range: 0.83-57.2 month) respectively, p = 0.65]. Interestingly in three (IDH1 R132H: n = 1; IDH2 R140Q: n = 2) out of 93 IDH1/2 negative AML cases where diagnosis and relapse samples were both available, IDH1/2 mutations appeared only at relapse. These cases were re-evaluated by the more sensitive ddPCR method at diagnosis and VAF (0-0.23%) was under the detection limit of HRM and/or allele specific PCR in each case.

NPM1 MRD monitoring
The LoD for NPM1 type-A ddPCR was lower than type-N ddPCR both in DNA and RNA settings (S6 Table). NPM1 mutant VAF values in DNA and NPM1 mutant expression levels in RNA were considered as MRD negative if below 0.01% (type-A) or below 0.05% (type-N). In case of NPM1, 174 NPM1 positive cases reached MLFS after induction, MRD monitoring could not be performed in 5 cases due to technical limitations (NPM1 mutation could not be detected by type-A or type-N primers), and in 53 cases due to non-available DNA. Basic characteristics such as gender, age at diagnosis, induction therapy, HSCT, and outcome (death in aplasia or in indeterminate cause, remission, relapse, cytogenetic and molecular genetic data) of NPM1 positive and MRD monitored patients were included in S3 Table. In 116 NPM1 MRD monitored patients, NPM1 mutant VAF was reduced below 2.5% in all patients in MLFS after induction.
Similarly to genomic DNA two or three log reduction was observable in mutant NPM1 RNA expression (NPM1/ABL1, n = 39 patients) [

Discussion
The serial acquirement of somatic mutations in myeloid clone(s) was described as the multistep pathogenesis of AML. Several lines of evidence prove that NPM1 mutations are responsible for the definitive acute leukemic transformation, therefore considered as leukemia founder mutations: (i) NPM1 mutations are completely absent in the population without hematological malignancies even at a higher age [22-24]; (ii) NPM1 mutations cannot be detected in AML patients months or years before the manifestation of AML [25, 26]; (iii) NPM1 mutations occur rather rarely (approximately 2-3%) in preleukemic myeloid malignancies such as myelodysplastic syndrome (MDS) or in myelodysplastic/myeloproliferative neoplasm  In line with previous studies less than 10% of our NPM1 mutation positive cases relapsed as wild type NPM1 AML. The loss of NPM1 mutation in our patient cohort was not associated with longer remission before relapse, which was suggested by several previous studies [31, 32]. Although our study did not investigate the spectrum of preleukemic mutations, the persistence of IDH2 R140Q mutation was observed in a single case with NPM1mutation loss relapse 13 month after diagnosis. Contradictory data exist in the literature, whether IDH1 and IDH2 mutations are preleukemic or AML founder mutations. Several studies suggest IDH1/2 mutations as epigenetic modifiers as preleukemic events. (i) In large scale populational screening studies for clonal hematopoiesis of indeterminate potential (CHIP) mutations, IDH2 R140 mutations were extreme rarely detected in elderly individuals (IDH2 R140Q/W: 0.014%, four out of 29562 individuals) [22]. (ii) IDH1 and IDH2 mutations were detectable as premalignant, high-risk gene mutations years before the diagnosis of AML, but not in age-matched controls (8%; n = 15/188; three IDH1 R132C/H/G and 12 IDH2 R140 positive individuals with a median of 7 years before AML diagnosis) [26]. (iii) IDH1 and IDH2 mutations are also rarely present in preleukemic myeloid malignancies: 0.8-4% in chronic phase MPN, 4-14% in MDS, but its frequency increases up to 20-25% in blast phase transformation [40-43]. (iv) IDH1 and IDH2 mutations were detectable in preleukemic hematopoietic stem cells [28,29]. The comparison of VAF values suggested that IDH1 and IDH2 mutations were more likely to develop before NPM1 mutations [6]. The persistence of IDH1/2 mutations (especially IDH2 R140Q) in remission was observed in 7-39% of AML cases in the literature, [19,[44][45][46] which is in line with our study (24% of IDH1 or IDH2 mutations were detectable in complete remission with a higher than 2.5% VAF, 67% of persisting mutations was IDH2 R140Q). The high mutational load in remission is a direct proof of preleukemic origin of the somatic mutation. This phenomenon in case IDH1/2 mutations is not as frequent as in case of DNMT3A, TET2, ASXL1 gene mutations, where reported rates vary between 51-82% [2,[47][48][49], (v) At relapse both IDH1/2 gene mutations showed a relatively high stability (86-88% reported in publications, 87% in our study) similar that of NPM1 mutation [31]. (vi) In IDH1/2 mutation negative AML, the emergence of IDH1 or IDH2 mutations at relapse was observed in 10% in our study, which suggests the subclonal, late origin of these mutations. Interestingly, there is a usual mutation order in AML pathogenesis, but some mutations might appear both early and late events [31,50].
A recent meta-analysis proved that lower MRD was consistently associated with improved outcome independently from applied method, sample source or sampling time of the assessment [51]. Regarding molecular genetic detection techniques, like quantitative PCR, digital PCR and next generation sequencing are extensively applied for MRD detection. High assay precision and reproducibility make ddPCR particularly suitable for MRD monitoring, which was reported in connection with several oncohematological drivers [52][53][54][55][56]. Individual assay designs make the quantification of multiple NPM1 mutations challenging, but the application of degenerated primers allows the simultaneous detection of multiple NPM1 mutations affecting the same localization (c.860_863dupNNNN) [14]. Our data and other studies also supported, that less than 5% of NPM1 mutations affects nucleotides at different positions [11,38,57].
Although consensus exists about the importance of NPM1 MRD; broad range of heterogeneity was displayed concerning thresholds discriminating between low-and high-risk MRD. Studies comparing mutant NPM1 transcript levels parallel in bone marrow (BM) and peripheral blood (PB) samples identified strong correlation, but an average of 1-log higher sensitivity in BM [21,34,38,[57][58][59][60][61]. In line with this observation, 3-log reduction of NPM1 mut /ABL1 transcript level was pointed as favorable prognostic indicator in BM, [21, 59, 60] but 4-log reduction was required in PB after induction therapy [57,58]. In our study, bone marrow samples were processed. As NPM1 mut expression is highly abundant, greater sensitivity (median: 1.3 log, range: 0-2.78 log in our study) was achieved on RNA level than on DNA. NPM1 mut RNA expression level detection for MRD monitoring is recommended in the literature [13,38]. Shayegi et al. investigated that 1% NPM1 mut /ABL1 expression corresponds to 0.016% NPM1 mut VAF or 1 in 32000 cells (1.8 log difference between RNA and DNA levels) [60]. These data suggested that NPM1 mut MRD screening should be performed on RNA expression, but in case of RNA unavailability, highly sensitive DNA methods can substitute. The applied cut-off for MRD negativity in our study (NPM1 type-A: 0.01% and type-N: 0.05% on DNA level) corresponds approximately to 1% NPM1 mut /ABL1 expression level. We were unable to test large number of RNA samples, which is a major limitation of our retrospective study. Ivey et al. [57] demonstrated that RNA-MRD positivity in PB after induction (2 cycles) corresponded to higher cumulative incidence of relapse (MRC17 trial 3-year CIR: 82% versus 30%), similarly Balsat et al. [58] (ALFA-0702 trial: 2-year CIR: 55% versus 21%); Hubmann et al. [62] less than 3log-reduction in BM RNA-MRD (AMLCG 1999, 2004 and 2008 trial: 2-year CIR 77.8% versus 26.4%,); Kapp-Schwoerer et al.
The co-occurrence with FLT3-ITD was recognized as an adverse factor in NPM1 mutant AML, due to the highly proliferative nature of the leukemic clone with ITD [38,63]. Although NPM1 mutation was referred as favorable or intermediate ELN prognostic categories depending on the presence of FLT3-ITD with high mutational load [1]. Recently, the reclassification of ELN prognostic criteria identified high FLT3-ITD load as adverse risk irrespective of NPM1 mutation status [64]. Allogeneic HSCT in first complete remission is not recommended in favorable risk AML, on the other hand relapsed NPM1-positive cases have adverse outcome [65]. We observed that the measurement of NPM1 MRD was capable to identify high risk patients even in the favorable risk NPM1 positive AML without high ITD load. NPM1 MRD negativity (NPM1 mut VAF<0.01-0.05% after induction) with high FLT3-ITD allele burden at diagnosis showed similarly adverse survival to NPM1 MRD positive patients.
Molecular MRD measurements serve not only prognostic, but may influence therapy. In case of persistent MRD, HSCT consolidation improved survival over chemotherapy [66]. In ELN 2017 favorable risk NPM1 mut AML subgroup, molecular failure (defined as NPM1 mut / ABL1 >0.05% after consolidation or NPM1 mut reappearance after molecular response; which affected 40% of NPM1 mut cases) served as indication for allogeneic HSCT in first complete remission. MRD-guided approach involving early intervention resulted in improved outcome (two-year OS: 85% for HSCT-treated patients with molecular failure and 39% for patients with hematological relapse) [67]. For elderly or unfit patients, azacitidine was reported to prevent or delay hematological relapse in MRD-positive AML [68].
Our data investigating pre HSCT NPM1 mut MRD are in good concordance with other studies with similar MRD time-point assessment: pre HSCT MRD negativity predicts favorable outcome after HSCT [21, 66,[69][70][71]. Detection of MRD-positivity before HSCT guide therapeutic choices during conditioning and graft versus host disease prevention, e. g. preferably Tcell repleted versus T-cell depleted transplant [21]; preferably myeloablative versus reduced intensity conditioning [72]. MRD measurements can even guide targeted FLT3-inhibitor therapy identifying patients who benefit mostly [73].
The role of MRD-monitoring is well-documented in case of NPM1, but data are scarce about IDH1 and IDH2 mutations. We applied BioRad-designed mutation detection reagents on BioRad QX200 Droplet Digital PCR System, but interestingly we were not able to reach as high sensitivity as in case of NPM1. Similar technical limits (LoD: 0.2%) were reported in a previous study applying the same detection [19]. Our data also supported the preleukemic nature of IDH1/2 mutations, but the persistence of IDH1/2 mutations (VAF>2.5%) in complete remission was associated with adverse outcome, higher chance of relapse or the development of myelodysplasia [19,44]. The presence of a preleukemic clone in morphologic leukemia-free remission was generally reported to associate with inferior survival compared to patients without persisting oncogenic mutations [74,75]. On the other hand, persistent DNMT3A, TET2, ASXL1 mutations were not connected with higher relapse rate and several reports described long-term remission even with high VAF [2,[47][48][49]. The frequency of persistent IDH1/2 mutations in remission was reported as high as 7-39% depending on the VAF cut -off (1-5%) or on the applied chemotherapy [2,19,44,45], which was similar to our observation (24%). In line with previous publications [19,44,45], our data also indicated that persisting IDH1/2 mutations in remission were associated with adverse prognostic impact. Currently no guidelines exist whether pre-emptive therapeutic interventions (such as HSCT or IDH1/2 inhibitors) could reduce relapse rate or improve survival in case of persisting IDH1/2 mutations in remission. The combination of IDH1 or IDH2 inhibitors with intensive chemotherapy in newly diagnosed AML might improve mutation clearance, although no comparative data exist with or without the inhibitors [76].
In summary, we investigated a considerably large number of AML patients systematically over a long time, the limitation of our study is the retrospective study design and the heterogeneous treatment protocols applied during the observational period. Our results support that NPM1 MRD even at DNA level is a reliable prognostic factor. On the other hand, IDH1/2 mutations may represent pre-leukemic, founder or subclonal drivers, still IDH1/2 MRD may also identify high risk AML. As MRD represents a biological continuum, special detailed guidelines are required to establish proper thresholds for the initiation of pre-emptive therapies.
Supporting information S1