Acute myeloid leukemia (1) is a clonal malignant expansion of immature myeloid precursors due to block in differentiation. Mutation profiling is standard for routine baseline clinical evaluation of AML. Prior to 2008, there were two functional genetic groups for leukemic pathogenesis: class I (activated signaling genes such as FLT3, KIT and RAS mutations) that conferred the proliferative potential, and class II genes involved in transcription and differentiation such as CEBPA and RUNX1 (2). The development of high-throughput NGS sequencing platforms uncovered many somatic mutations in AML. In 2013, the Cancer Genome Atlas (TCGA) project expanded the functional genetic classes to nine families involved in the pathogenesis of the myeloid neoplasms. An average of 14 mutations were identified in the AML genome, ranging between five to 23 genetic mutations in each case (3). In 2016, a large cohort that enrolled 1540 AML patients described 11 subgroups of genomic alterations with different clinical outcomes using NGS. The additional separate categories included AML with mutations in genes encoding chromatin and RNA-splicing regulators, AML with TP53 mutations and/or chromosomal aneuploidies, and AML with IDH2 R172 mutations (4). Taken together, these findings supported the fact that AML phenotype is driven by underlying mutational signature ( Figure 1 ).

Accordingly, both the 2016 revisions to the WHO and 2017 European LeukemiaNet (5) incorporated gene mutations into the sub-classification and risk-stratification of AML (6, 7). Molecular testing plays a major role in the current World Health Organization (WHO) classification of myeloid neoplasms. Specific genetic abnormalities are regarded as disease-defining mutations and others have important prognostic and therapeutic implications. The WHO recognizes 3 distinct sub-categories of AML based on somatic mutations: AML with NPM1 mutation, AML with bi-allelic mutations of CEBPA, and AML with mutated RUNX1 (provisional entity). Genes such as TP53, RUNX1, IDH1/2 and FMS-related tyrosine kinase 3 (8), among others, are found to be altered in the different subcategories of AML with prognostic and/or therapeutic implications of utmost importance. The latest 2017 European Leukemia Net (ENL) guidelines for AML recommends molecular profiling for mutations in NPM1, CEBPA, FLT3-ITD, TP53, RUNX1, ASXL1, and BCR-ABL1. Therefore, the complete cytogenetic and molecular work-up is essential at the time of initial AML evaluation. The main goal of genome sequencing is the identification of actionable gene mutations and risk-stratification. Hence, multi-gene panel next-generation sequencing (NGS) based mutation analysis is the primary mode for assessment for multiple AML associated genes (9).

In this review, we will provide an overview of the diagnostic, prognostic and therapeutic data that can be obtained from routine clinical NGS data in AML. In addition, we will describe the implications of the vast amount of additional data that can be obtained from routine clinical NGS beyond prognostication and management, such as the ontogeny of AML, clues regarding underlying germline predisposition and clonal hematopoiesis, serial monitoring to assess the effectiveness of therapy and resistance mutations. We will use a few prototypic genes for each section, with emphasis on the recently described entities and actionable mutations, specifically pertaining to the Food and Drug Administration (10) approved target therapies and the ongoing clinical trials. AML-defining translocations are beyond the scope of this review.

Lindsley et al. evaluated the differences in gene mutation patterns between secondary AML, therapy-related AML, and de novo AML to decode the ontogeny of different subsets of AML. The authors identified a set of gene mutations that can provide an objective evidence of AML ontogeny irrespective of clinical information. The presence of a mutations in any of these genes: SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, or STAG2 was >95% specific for the diagnosis of secondary AML (88). Majority of these genes, SRSF2, SF3B1, U2AF1, ZRSR2, are genes encoding proteins that belong to the spliceosomal complex that encompasses a large number of members including small nuclear ribonucleoproteins and protein factors responsible for removing introns from a transcribed pre-mRNA (89). EZH2 and BCOR mutations are associated with worse OS in AML (55, 90, 91) while STAG2 is a part of the cohesion complex.

The findings have major clinical and diagnostic implications. When detected in “de novo” setting, the presence of secondary AML mutations can identify a subset of patients with worse outcome (88). The “secondary-type” mutations, as expected, are frequently present in MDS (92) and chronic myelomonocytic leukemia (CMML) (93). These findings imply these mutations are unlikely to promote the development of acute leukemia without a co-operating event. It is also to be noted that these mutations occur early in the disease course and persist during follow-up despite morphologic CR, hence should not be used for MRD follow up purposes (Figure 2 and Table 2 ).

Somatic NGS sequencing in AML can potentially uncover incidental germline mutations. The latest WHO recognizes myeloid neoplasms with germline predisposition syndromes as a distinct diagnostic category (57). Recognition of these conditions has major therapeutic implications (94) and different diagnostic and monitoring strategies for the patient himself, as well as the family members. When clinically suspicious, germline mutations testing and proper family screening are important, especially when an allogeneic bone marrow transplant is under consideration.

The new WHO classification introduced four broad categories of AML with germline predisposition: 1-nMyeloid neoplasms with germline predisposition without a pre-existing disorder or organ dysfunction, it includes AML with CEBPA and DDX41 mutations 2- Myeloid neoplasms with germline predisposition and pre-existing platelet disorders, which encompass RUNX1, ANKRD26 and ETV6 mutations 3- Myeloid neoplasms with germline GATA2 mutation and 4- Myeloid neoplasms with germline predisposition associated with inherited bone failure syndromes and telomere biology disorders. Hence, it is recommended to include these genes, CEBPA, DDX41, RUNX1, ANKRD26, ETV6, TERT and GATA2, within the standard NGS panel.

Since most of the mutated genes associated with germline predisposition disorders are also recurrently mutated in sporadic leukemia cases, attention should be given when interpreting the results of sequencing to identify clues suggesting a germline origin, such as double mutations, one with a near heterozygous or homozygous VAF (95). The presence of any of these mutations in a newly diagnosed leukemia is not sufficient to diagnose an AML with germline predisposition. Germline origin by testing constitutional DNA (skin fibroblast culture in most patients with active hematological malignancies) should be performed for confirmation (72, 88). It is important to highlight that targeted sequencing will not detect all the germline alterations such as intronic regions and large deletions spanning multiple exons; additional tests should be done to further investigate when clinically suspicious, such as array comparative hybridization for deletions and whole exome sequencing for novel mutations (96, 97). The confirmation of germline origin requires a prior genetic counseling since the results may cause significant disturbance of the affected individuals and families if not properly interpreted and handled ( Figure 2 and Table 2 ).

NGS is a quantitative assay that provides information on mutant allele burden, which is critical for management decisions and predict outcome. VAF is defined as the ratio of reads with mutation against total (mutant + wild-type) reads. Hence, current NGS reports are generally not limited to presence or absence of mutations but include the VAF of each alteration. VAF is important for management decisions and outcome prediction in AML and MDS. This was previously elaborated in the context of TP53 mutations in AML and MDS (81, 83). A high mutant allele burden at diagnosis can be a negative prognostic factor. Patel et al. demonstrated a negative prognostic effect of high NPM1 mutant allele burden at diagnosis in de novo AML cases (98). Sasaki et al. investigated the effect of the VAF of clonal hematopoiesis associated genes ASXL1, DNMT3A, JAK2, TET2, and TP53 mutations on survival in 421 newly diagnosed AML using NGS. Higher VAF (cut-off of 30%) was associated with worse survival in AML patients within intermediate-risk cytogenetic group (99).

Serial NGS is ideal for monitoring AML patients for mutational clearance and/or clonal evolution and relapse. Clearance of somatic mutations in non-preleukemia genes at the time of CR was associated with better OS and decreased risk for relapse (100). Sequential analysis for persistent mutations is particularly helpful to evaluate residual disease in patients treated using novel targeted therapeutic agents, such as IDH inhibitors, as these can pose diagnostic challenges on morphology and/flow cytometry. Therefore, NGS is particularly useful to objectively evaluate evidence of residual disease in these circumstances.

The implications of NGS include assessment of effectiveness of therapy and detection of resistance mutations which can have implications on management and therapeutic regimen choices. Most importantly, sequencing of blast cells can detect TP53 mutations which are independent predictors of poor survival and treatment resistance (7). Moreover, mutations that trigger resistance to FLT3 inhibitors can be identified in AML cases including emergence of leukemic clones harboring mutations that activate RAS/MAPK pathway signaling (37). On the other hand, emergence of FLT3-TKD mutation during treatment can engender resistance to TKI; therefore, sequential mutational analysis is mandatory for early detection of potential treatment resistance and the choice of alternative drugs. Finally, the acquisition of IDH1, WT1, ASXL1 variants in certain AML clones, either present at diagnosis or gained at relapse confer chemotherapy resistance (101).

While NGS facilitates identification of disease evolution and treatment resistance mutations, there are several caveats. First, the sensitivity of standard NGS panels used in clinical laboratories, currently limited to 2-5% VAF (102) is less or comparable to standard flow cytometry techniques for detection of residual disease at the time of morphologic remission. Yet, NGS results are more helpful than latter in settings where blasts show monocytic differentiation or in settings with limited sample quality. The identification of the mutations is partly limited by the intrinsic error rates (0.1% to 1%), that can be potentially overcome using the error correction methodologies such as molecular barcoding (103). Alternately, ultra-deep sequencing NGS with an ultra-high depth of coverage or individual gene assays such as ddPCR assays can be helpful (104). In one study, a high-throughput deep sequencing NGS method showed a detection sensitivity of 10^-4 for SNVs and 10^-5 for insertions/deletions (105).

Second, persistence of “pre-leukemia” mutation signature cannot be used for MRD detection. NGS enables the identification of background clonal hematopoiesis. Clonal hematopoiesis of indeterminate potential (CHIP), defined as the presence of a somatic alterations (either a somatic mutation associated with myeloid malignancy present at least 2% VAF or cytogenetic abnormality) in apparently healthy individuals. These patients have an increased risk of developing a hematologic malignancy, higher risk of developing therapy related myeloid neoplasms following chemotherapy for a solid tumor and increased frequency of adverse cardiovascular events (106).

Most of the CHIP alterations belong to genes DNMT3A, TET2 and ASXL1 (DTA) (106, 107). All three genes are involved in epigenetic regulation of myeloid differentiation and are considered to be within preleukemic hematopoetic stem cells (108). DTA gene mutations cannot induce leukemia without a co-operating second hits (109). The mutations in either one of the DTA genes impart negative prognostic outcomes in AML (110–112). The 2017 ELN classifies ASXL1 mutations as adverse risk category (7), and its presence was >95% specific for AML of secondary origin (88).

Mutations in all three genes can be seen within the entire coding region, and include missense, frameshift, nonsense and splice-site mutations leading to a non-functional protein. The only exception being DNMT3A, which has a hotspot point mutation in codon R882 (113, 114). Loss of TET2 function can also occur via mutations in IDH1, IDH2 and WT1 (115, 116), that explain the mutual exclusivity between IDH1/2-TET2-WT1 mutations in AML (116). Hence, NGS is an ideal technique for detection of all these mutations.

Importantly, DTA gene mutations likely persist at the time of remission of AML in pre-leukemic clones, hence they cannot be used to detect MRD (109). DNMT3A R882 mutation persists in 75% of AML patients during remission without any negative impact on outcomes (117). Interestingly, ascorbate supplementation can restore methylation patterns and minimize proliferation of blasts (118). A clinical trial evaluating the efficacy of azacitidine and high dose ascorbic acid in AML with mutated TET2 is ongoing (NCT03397173).

9- Subclonal evolution in AML using single-cell technology:

All malignancies are genetically heterogeneous, composed of mutationally-defined subclonal cell populations characterized by distinct phenotypes. Precise identification of clonal and sub-clonal architecture is mandatory to understand the temporal evolution of tumor and the emergence of treatment resistance. Bulk sequencing cannot definitively resolve the actual complex clonal composition of neoplasms generally and AML specifically. Therefore, there is great interest in understanding the genetic alterations at a single-cell level using the newly designed sequencing platforms. Multiple reports published recently highlighted the subclonal selection during the treatment journey. Morita et al. reported the sub-clonal complexity in 37, clonal architecture and mutational histories of 123 AML patients. The authors explored single cell-level mutation co-occurrence and mutual exclusivity revealing novel clonal relationships; and emergence and selection of resistant subclones under therapies by longitudinal analysis (119). Interestingly, current emerging single-cell multi-omics technology, aid in profiling simultaneously single-cell mutations and cell surface proteins in AML cases, allowing correlation of genetic and phenotypic heterogeneity (120). Using the novel features of this technology, Petti et al., using a high-throughput platform to distinguish tumor and non-tumor cells in AML, identified tumor cells showing phenotypic aberrancies and lineage infidelity, evaluated the sub-clonal progression of tumor samples with time with a molecular signature for each sample, and cell-surface markers that could be used to isolate specific cells for downstream studies (121). Another study showed that FLT3-ITD mutation was present in the primitive cells, whereas FLT3-TKD mutation was present in the more differentiated cells within the same tumor. The study demonstrated that FLT3 variants differentially affected AML differentiation that explained the worse prognosis associated with certain alleles (122).

We believe that NGS is an exciting tool that has helped pathologists and oncologists tremendously to improve their understanding of AML pathogenesis and clonal evolution of the disease. It has played a major role in designing therapies targeting the disease and control relapse. Genomic analysis of cases at diagnosis and relapse have uncovered the alterations and clonal evolution of the genetic profile of the tumor cells during disease progression. The cited studies and clinical trial results highlight the unique genetic signature of every patient’s disease, not only with respect to different combinations of mutations, but also in terms of clonal burden of different mutations, sequence of mutations, and concurrent chromosomal gain and losses. As a consequence of this, 1- the prognostic outcome of each patient can be unique and cannot be simply generalized based on the presence or absence of mutations; 2- the therapeutic molecules should be targeted to attack both the primary clone and emerging sub-clones with potential resistance mechanisms. Hence, a better understanding of each patient’s AML genome by NGS is mandatory to make decisions related to appropriate personalized therapy.

Nevertheless, a lot more standardization is needed for implementing NGS in daily clinical practice; for instance, so far there is no universal consensus, yet on which target genes should be included in the sequencing panels. Moreover, the sequencing depth of coverage is still a subjective number chosen by each laboratory for the validation of their sequencer. Furthermore, there is no consistent practice for quality assessment of sequencing data; the accuracy of results is compromised at genome locations with highly repetitive sequences, or in GC-rich locations, common problems encountered during sequencing and data analysis. Finally, results interpretation can be time-consuming and requires specific expertise from bioinformatics and pathology (or equivalent degree). Further standardizations are needed when NGS is used for serial follow-up to assess measurable residual disease. It is important to keep in mind that bulk NGS represents the “average” findings per cell, and hence the data on genomic complexity is only inferred. On the other hand, single-cell NGS can accurately provide concrete information on sub-clonal architecture, but it has not reached the mainstream clinical work-flow yet (123).

At the same time, while NGS is an important component of AML workup, other testing needs to be performed to accurately sub-classify the disease, specify the prognosis, and determine the best targeted therapy. A comprehensive workup for AML should include a karyotype to identify the disease defining chromosomal translocations including t(8;21), t(15;17), t(16;16) and the new WHO entity AML with BCR/ABL (57); however, some fusions can be cryptic (57), and other testing is required to highlight these aberrations including Fluorescence in situ hybridization (FISH), RT-PCR, RNAseq, and other targeted fusions assays. Moreover copy-neutral loss of heterozygosity (cnLOH) has prognostic significance in patients with acute leukemia (124); Walker et al. showed that LOH mediated by uniparental disomy (UPD) is a common finding in cytogenetically normal AML. Also, UPD involving 13q and 11p are important for genetic risk stratification in these cases (125). FISH and karyotyping can be used to depict copy number aberrations; however, they are limited by low resolution or restriction to targeted assessment. The alternative method of testing is Chromosomal microarray which can characterize chromosomal copy number changes and cnLOH in myeloid malignancies (126).

Overall, NGS has enabled phenomenal advances in understanding of molecular genetics of AML and opened up new horizons for development of highly effective therapeutic molecules and protocols for individualized treatment and monitoring that are completely reshaping the management of the different subtypes of AML.

All authors contributed to writing the manuscript, reviewing the final version and preparing the figures and tables. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.