Acute myeloid leukemia is a category of diseases with a common aggressive clinical presentation but with a prognosis and management that is dependent upon the underlying genetic characteristics of the neoplasm. The purpose of this brief review is to update the practicing pathologist on the current standard of care in the genetic evaluation of acute myeloid leukemia and to highlight future directions in the classification, genetic assessment, and management of these devastating diseases.
Acute myeloid leukemia (AML) is a heterogeneous group of diseases, all exhibiting the common feature of a proliferation of immature myeloid cells (blasts and blast equivalents) in the bone marrow and/or blood (blasts generally constituting greater than 20% of the cells). The classification of AML has changed dramatically over the course of the last several decades, and perhaps more than in any other group of malignancies, the classification of AML is now predicated on identifying the genetic aberrations underlying an individual patient's disease. The purpose of this brief review is to provide an update on the current classification and standard-of-care genetic characterization of AML. We will then provide insight into where genetic testing will be headed in the near future in this rapidly evolving field.
CURRENT WORLD HEALTH ORGANIZATION AML CLASSIFICATION
The purpose of AML classification and genetic testing at diagnosis is largely to risk-stratify patients with AML, and thus help determine appropriate treatment modalities. Historically, AML was classified by the morphology and cytochemical (and later flow cytometric) phenotype of the tumor cells under the French-American-British (FAB) AML classification system.1 While very useful, particularly in its recognition of acute promyelocytic leukemia (AML FAB M3) as a distinct entity, the FAB system ultimately suffered from poor reproducibility2–4 and did not provide satisfactorily predictive prognostic information.5 The current World Health Organization (WHO) classification of AML has since moved toward a system based more on underlying genetics, first incorporating recurrent structural cytogenetic abnormalities,6 then specific gene mutations in the most recent 2008 edition.7
The current WHO classification organizes nonsyndromic AML in the following categories: AML with recurrent genetic abnormalities; AML with myelodysplasia (MDS)–related changes; therapy-related myeloid neoplasms; and AML, not otherwise specified. These are listed with their subcategories in Table 1. Classification independent of genetic findings can only happen in a specific instance. Therapy-related AML is diagnosed if the patient has a history of antecedent cytotoxic therapy. Any AML case without this history can only be adequately classified by integration of the genetic features (even in therapy-related AML, these cytogenetic findings are still important). If a patient has a recurrent chromosomal translocation as enumerated in Table 1, it is classified as AML with recurrent cytogenetic abnormalities. If a patient has specific chromosomal abnormalities as listed in Table 2, a history of antecedent MDS or myelodysplastic/myeloproliferative neoplasm, or a sufficient degree of morphologic dysplasia, AML with MDS-related changes should be diagnosed (though the latter criterion may be modified in the upcoming WHO revision).
IMPORTANCE OF CHROMOSOMAL STRUCTURAL ABNORMALITIES
The identification of cytogenetic abnormalities allows classification of AML in either the most appropriate WHO “AML with recurrent genetic abnormalities” subcategory or, often, as “AML with MDS-related changes.” In contrast with the limited prognostic ability of the FAB categorization, identifying chromosomal abnormalities in AML is, after patient age, the pretreatment factor most predictive of outcome.8 In general, the cytogenetic landscape of AML is characterized either by balanced chromosomal rearrangements, often involving hematopoietic transcription factors (TFs), or by chromosomal losses and gains similar to those present in MDS.
AML With Recurrent Cytogenetic Abnormalities
The focus of most of this discussion will be on single gene mutations in AML; however, the historically well-established category of AML with recurrent cytogenetic abnormalities deserves attention. This WHO category comprises AML cases with recurrent balanced translocations that lead to the creation of novel fusion genes. The most common subtypes in this category are acute promyelocytic leukemia (APL) with t(15;17)(q22;q12), PML-RARA; AML with t(8;21)(q22;q22), RUNX1-RUNX1T1; and AML with inv(16)(p13.1q22) or t(16;16)(p13.1;q22), CBFB-MYH11. These 3 cytogenetic abnormalities are unique in that their identification allows a diagnosis of AML in cases that have fewer than 20% blasts in the blood or bone marrow.
Acute promyelocytic leukemia is the most critical subtype of AML to recognize at diagnosis owing to its propensity to be associated with devastating hemorrhagic events. Initial therapy for APL is different from that of other AML subtypes, incorporating all-trans retinoic acid during induction. Importantly, treatment should be initiated on the basis of the clinical and morphologic features of the case and not withheld until genetic confirmation. Therefore, it is absolutely critical for pathologists to be comfortable with the morphologic features of APL. In the typical case of APL, diagnosis is not difficult, as the neoplastic population consists of numerous cytologically atypical cells that resemble promyelocytes, frequently containing single or multiple Auer rods (Figure 1, A). The microgranular variant of APL is more challenging, as granules and Auer rods are rare. In this variant, the cells often exhibit nuclear bilobation, sometimes with “sliding-plate” or “butterfly-shaped” nuclei, with a thin nuclear isthmus connecting the more bulbous nuclear lobes (Figure 1, B and C).
Acute myeloid leukemia with t(8;21) and inv(16) are collectively referred to as core-binding factor (CBF) AML because RUNX1 (rearranged in t[8;21]) and CBFB (rearranged in inv) are both subunits of the CBF transcription complex, which is critical for myeloid differentiation.9 These AML subtypes exhibit different morphologic features. Acute myeloid leukemia with t(8;21) often has large blasts with a prominent perinuclear hof, occasional large cytoplasmic salmon-colored granules, and frequent fine Auer rods (Figure 1, D). Acute myeloid leukemia with inv(16), on the other hand, has myelomonocytic features and a population of eosinophils and precursors with abnormal basophilic granulation (Figure 1, E).
Other recurrent translocations also occur in AML, but at lower frequency. Of note, KMT2A (MLL) is often rearranged in AML with monoblastic features, particularly in the infant population. It may be joined with a variety of fusion partners: the t(9;11)(p22;q23) fusion with MLLT3 is most common and is codified in the WHO classification; other fusion partners should be explicitly noted as variants since the prognostic significance may be partner-dependent.10 Acute myeloid leukemias with more uncommon rearrangements of t(6;9) with DEK-NUP214 fusion, inv(3) or t(3;3) with RPN1-MECOM (EVI1) fusion, and t(1;22) with RBM15-MKL1 fusion are also considered to be discrete WHO subtypes.
In contrast with these recurrent chromosomal rearrangements that tend to occur in younger patients, older patients often have a pattern of unbalanced chromosomal gains and losses, similar to that present in MDS. As previously mentioned, identification of any of the abnormalities listed in Table 2 allows classification of AML with MDS-related changes if the blast percentage is sufficient and the patient does not have a history of chemotherapy.
The cytogenetic classification of AML is not a purely semantic exercise allowing us to give an individual case the most appropriate WHO-approved name; it also very importantly directs risk-adapted therapy for the individual patient. We can divide AML into cytogenetic risk groups: those with favorable, intermediate, or poor outcomes (Table 3).11 Those with favorable prognosis include CBF AML [t8:21] and inv) and APL with t(15;17); those with adverse prognosis include −5, −7, del(5q), 11q23 abnormalities other than t(9;11), abnormal 3q, and complex cytogenetics. Patients with other genetic findings, including the large group with a normal karyotype, have an intermediate prognosis.12 Leaving aside APL—which as previously discussed has a very different treatment from other AML subtypes—after standard induction chemotherapy, favorable-risk CBF AML would be treated with consolidative chemotherapy at the first complete remission. Allogeneic hematopoietic stem cell transplantation (HSCT) would be reserved in this population as an option in the event of subsequent relapse. Those patients with poor-risk cytogenetics, in contrast, will have poor outcomes with chemotherapy alone and are candidates for early HSCT. Appropriate therapy is less straightforward for those in the intermediate-risk category, however, as outcomes are heterogeneous. This is a significant clinical problem, given that many patients with AML fall under the cytogenetic intermediate-risk group. In the last decade, single-gene mutational testing has been incorporated to assist with prognostic stratification of this population.11
SINGLE GENE MUTATIONS
Testing of a number of single genes has been incorporated to further risk-stratify patients, thus providing useful data for therapeutic decisions for the patients with AML with intermediate-risk based on cytogenetics. Testing for these mutations allows reduction of the number of patients in the clinically heterogeneous intermediate-risk subgroup (Figure 2). The currently most important genes to evaluate include the nucleolar protein nucleophosmin (NPM1), CCAAT/enhancer-binding protein α (CEBPA), and fms-related tyrosine kinase 3 (FLT3). In addition, mutations of the v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog (KIT) modify the prognosis of otherwise favorable-risk CBF AML. This discussion will focus on the clinical implication of these gene mutations; a prior Archives article by Betz and Hess13 provides a useful overview of these genes' functions and testing platforms.
Mutations in NPM1 are involved in 25% to 35% of patients with AML and 45% to 64% of patients with AML with normal cytogenetics.14 NPM1 mutations are generally mutually exclusive with other WHO-defined recurrent cytogenetic abnormalities, and AML with NPM1 mutation is considered to be a provisional entity in the 2008 WHO classification.7 The prognostic importance of an NPM1 mutation is dependent upon the mutational status of a second gene, FLT3. Internal tandem duplication (ITD) mutations of FLT3 are present in approximately 20% of patients with AML, in 28% to 34% of patients with AML with normal cytogenetics, and in 40% of patients with AML with concurrent NPM1 mutation.15 The presence of a FLT3-ITD confers an adverse prognosis for those with intermediate-risk cytogenetics and for patients with NPM1 mutations,11,15 and so testing of these 2 genes must be performed together to provide accurate prognostic information.
In the absence of FLT3-ITD mutation, an NPM1 mutation confers a favorable prognosis for those with normal cytogenetics.12,16 Though HSCT is effective in patients with a normal karyotype and NPM1+/FLT3-ITD− genotype, retrospective analysis suggests the benefits of HSCT do not outweigh its associated significant morbidity and mortality in this population, while HSCT should be considered at first remission in other genotypes.15 Recent studies have suggested that HSCT could still be considered in this population17,18 ; however, patients with NPM1 mutations are often able to have effective salvage treatment at relapse, and in general HSCT is initially withheld in this cohort.
Mutations in CEBPA are present in 10% to 18% of patients with AML with normal cytogenetics.19 Whereas a single mutation in CEBPA does not affect prognosis, biallelic mutations confer a favorable prognosis,20 constitute another provisional WHO AML subtype,7 and direct consolidative chemotherapy rather than HCST at remission.
The scenario for KIT testing is different from the other single gene mutations mentioned above. Mutations in KIT occur in 20% to 30% of CBF AML cases, which is typically considered to be favorable-risk. However, the presence of KIT mutations confers an increased risk of relapse.21,22 Therefore, CBF AML with KIT mutations is considered to be at intermediate-risk in the most recent National Comprehensive Cancer Network guidelines.12
CURRENT STATE OF PRACTICE
So what, then, is the current standard of care for patients with AML? At the time of this writing, the minimum genetic workup of a new AML case is not substantially different than it was even half a decade ago.13 Metaphase karyotyping is absolutely necessary, and for those patients with intermediate-risk cytogenetics, molecular assessment of FLT3, NPM1, and CEBPA should be performed. If the patient is found to have CBF AML, KIT should be assessed for mutations.
Supplementation of these approaches with fluorescence in situ hybridization (FISH) is controversial, and current clinical practice varies. The turnaround time of FISH testing is often much better than that of routine karyotyping, and FISH is useful for expedited confirmation of abnormalities such as PML-RARA fusion in APL. Some laboratories routinely perform a fairly comprehensive panel of FISH probes to assess for recurrent translocations and chromosomal copy number changes at the time of diagnosis, while others do not. The yield of such an unselected approach is not clear. If the laboratory that initially receives the specimen has local access to a high-quality cytogenetics facility, it is likely that FISH is of limited utility if an adequate number of metaphases are present for conventional karyotyping.23 However, karyotypic failure is not uncommon,24 and cryptic translocations do occur that are missed by metaphase analysis.25 FISH panels have clear utility in the setting of an inadequate karyotypic analysis, and given the critical importance of identifying recurrent rearrangements in AML, a reasonable argument could be made for its general application in the workup of new AML cases. If the morphologic features of a case suggest the presence of a recurrent cytogenetic abnormality that is not detected by cytogenetics, FISH testing should be performed, as some abnormalities such as MLL rearrangements and inv(16) may be karyotypically subtle. The most appropriate approach to FISH analysis is best established at the local level, in consultation with the testing cytogenetic laboratory.
At the end of this approach, if no recurrent cytogenetic abnormality is identified, and if there is no reason to classify the case as therapy-related AML or AML with MDS-related changes, the WHO suggests classification as AML, not otherwise specified, a category that is then further subclassified, essentially recapitulating the FAB classification (Table 1). Not surprisingly, this approach provides little to no additional prognostic value.26 As discussed further below, this remaining group of patients is still genetically heterogeneous, and more extensive genetic profiling can yield additional prognostic information.
The Genomic Landscape of AML
High-throughput genomic technologies have led to the identification of many genes newly known to be associated with AML. These studies have shown that the historic model of AML pathogenesis that postulated 2 classes of gene mutations, those that block differentiation and those that lead to proliferation, is overly simplified.27 The Cancer Genome Atlas' analysis of AML using whole genome or whole-exome sequencing, RNA and microRNA sequencing, and DNA methylation analysis recently identified recurrent mutations organized into 9 functional groups (TF gene fusions, NPM1, tumor suppressor genes, DNA methylation–related genes, chromatin modifying genes, signaling genes, myeloid TF genes, cohesin-complex genes, and spliceosome-complex genes; see Table 4).28 The genetic abnormalities in these 9 groups exist in a complex network of either cooperation or mutual exclusivity. The TF fusions, MLL fusions, and mutations of NPM1, CEBPA, RUNX1, or TP53 appear to define mutually exclusive AML groups. In contrast, some mutations, such as DNMT3A, NPM1, and FLT3, exhibit a strong tendency to co-occur. This study is useful in that it gives us a blueprint of the genes most likely to have an impact upon AML pathobiology and prognosis.
Prognostic Implications of Newly Identified Genetic Mutations Associated With AML
Understanding the prognostic implications of these recently identified genetic abnormalities in AML is critical for the translation of these new findings into clinical practice. A high-profile recent study by Patel and colleagues29 has provided us with an updated scheme to further risk-stratify younger patients with AML. They evaluated 18 genes in 398 patients with AML who were younger than 60 years. The authors then generated a prognostic model that was validated in an independent set of 104 patients. They found that FLT3-ITD, partial tandem duplications of MLL, ASXL1, and PHF6 mutations associated with adverse prognosis, that mutations in CEBPA and IDH2 associated with a better prognosis, and that the favorable prognosis conferred by NPM1 mutations is restricted to those with co-occurring IDH1 or IDH2 mutations. The authors built a model that risk-stratified patients with intermediate-risk cytogenetics, based first upon FLT3-ITD status, and then incorporated additional mutations that allowed them to reduce their intermediate-risk group from 63% of their cohort to 35% of their cohort, with most of the patients transitioning to poor-risk status after incorporation of the molecular results.29
Complexities of Interpreting New AML Genetic Data
In addition to the somewhat overwhelming number of genes now known to be associated with AML, there are other sources of complexity as we seek to refine our ability to choose the most appropriate therapy for our patients.
Very often, studies exhibit contradictory results regarding the prognostic significance of individual genes. The IDH genes and DNMT3A are relevant examples of this phenomenon. In the case of IDH1 and IDH2, multiple studies30–33 have suggested that the presence of IDH1 mutations confer a worse prognosis in AML with a NPM1+/FLT3-ITD− genotype. However, Patel et al29 described a favorable prognosis with any IDH mutation in the same genetic context. This finding appears to have been driven primarily by IDH2 R140 mutations,34,35 and British Medical Research Council studies have also supported that the IDH2 R140 mutation confers a favorable prognosis.36 Other studies30 have not found prognostic significance for IDH2 R140 mutations, while several30,31,36 have suggested that IDH2 R172 mutations are associated with an adverse prognosis. The bulk of the current evidence would suggest that IDH1 mutations are associated with a somewhat worse prognosis, while IDH2 mutant effects are codon-dependent.37 Clearly, this is important when a clinician is determining whether to offer an NPM1+, IDH1+ patient HSCT at first remission or to proceed with consolidative chemotherapy. This area needs more study, and additional meta-analyses of large, uniformly treated groups would be ideal. The case of IDH mutations illustrates 2 additional dimensions of uncertainty in integrating genomic findings into clinical practice: each specific mutation within a single gene may have differing effects on prognosis; and even very well-designed, large clinical trials may be underpowered to detect the true biologic effects of each possible mutational combination.
In the case of DNMT3A, mutations were initially reported to be associated with poor survival.38 Marcucci and coworkers39 then reported that specific mutations had differing importance in different age groups: non-R882 mutations were adverse in younger patients (<60 years) in their cohort, while R882 mutations were adverse in older patients. Gaidzik and colleagues40 arrived at an opposite conclusion, finding that R882 mutations were unfavorable, while non-R882 mutations were favorable. The patient group that Patel and colleagues29 studied had been enrolled in the ECOG E1900 trial that had previously demonstrated a survival benefit to dose intensification of daunorubicin during induction.41 Molecular analysis revealed that this survival benefit was restricted to patients with DNMT3A or NPM1 mutations or with MLL translocations, suggesting that any prior indication that DNMT3A mutations were prognostically unfavorable was no longer relevant with intensified induction. This finding has been subsequently supported by another group.42 The issues surrounding the prognostic importance of DNMT3A mutations again illustrates the difficulty in interpretation of conflicting published data, and also highlights that mutations could have different effects on different age groups, and that the significance of a mutation under one treatment paradigm may vanish when therapies change.
Another factor that appears to be important in AML prognostication is the allele burden of the mutation in question. It is well established that AML specimens contain multiple disease clones at diagnosis,28 and mutations in signaling genes, such as FLT3 and KIT, are frequently present in only a subset of the malignant blasts at diagnosis. Therefore, it is plausible that a FLT3-ITD mutation present in a small minority of the malignant blasts may not be as prognostically significant as that present in the entirety of the blast population. Indeed, this seems to be the case, as the adverse effect of FLT3-ITD mutations on cytogenetically normal AML appears to increase with increasing allele burden.43,44 In some cases of FLT3-ITD–positive AML, the mutant allele to wild-type FLT3 ratio is actually greater than 0.5, indicative of loss of heterozygosity at the FLT3 locus, often secondary to duplication of the mutant FLT3 allele. These cases may have a particularly poor prognosis.43,44 However, not all studies have confirmed a strong correlation between prognosis and allelic burden, and some recent data suggest that even low-level FLT3-ITD mutant clones at diagnosis may indicate aggressive disease, which would be consistent with the concept that these subclones may be especially chemoresistant.45,46 Other data suggest that KIT mutations may be most prognostically important if present at a relatively high level.47 Further studies are necessary to definitively answer these questions, though some clinical laboratories have begun reporting FLT3-ITD mutant allelic ratio.
Most of the data that indicate that individual gene testing is prognostically helpful applies to younger patients, younger than 60 years. This is a major clinical problem, in that AML is actually a disease of the elderly, with a median age at diagnosis of approximately 67 years (Figure 3). Elderly patients are often not candidates for intensive chemotherapy or HSCT, and many published risk models are not applicable in the elderly population.29,18 Older patients are more likely to have adverse, MDS-like cytogenetic features or antecedent MDS. Recent studies48 suggest that even in elderly patients without MDS-like cytogenetics or antecedent MDS, the mutational profile of AML is often MDS-like, with frequent mutations of spliceosome genes, ASXL1, and EZH2 in these elderly patients with genetically MDS-like, but clinically de novo AML. This may be of more than academic interest, as the identification of an MDS-like genetic signature in elderly patients has been associated with induction failure and decreased event-free survival versus patients with de novo (NPM1, CBF, or MLL rearrangements) or pan-AML mutational signatures.48 This may allow for selection of elderly patients who are candidates for more intensive therapy and allow prediction of those patients for whom alternative approaches are indicated. Elderly patients with AML are also more likely to have TP53 mutations than are younger patients. As previously mentioned, TP53-mutant AML likely represents a discrete AML subtype as it co-occurs with very few other single gene mutations.28 It is, however, strongly associated with complex cytogenetic abnormalities, so its prognostic importance is largely already indirectly reflected by this variable in risk stratification algorithms. There are emerging data, however, that its importance is even greater than that of the general genomic instability that it causes; several studies suggest that TP53 mutations further risk-stratify patients with poor-risk cytogenetics,49–51 identifying patients with exceedingly poor prognosis with current therapeutic strategies.52
While improved risk stratification of AML has been one of the most tangible benefits of genetic analysis, the ultimate goal remains to identify therapies that specifically target the genetic events that give rise to leukemia. With the exception of all-trans retinoic acid therapy for APL and the introduction of HSCT, disease-directed clinical therapy for AML has not substantively changed in 4 decades.53 Even the “favorable”-risk CBF AML subtypes are associated with a 5-year overall survival of only approximately 60%; there is much room for improvement in AML treatment. Tyrosine kinase inhibitors have been exceptionally effective in other tumor types dependent upon aberrant kinase signaling. Multiple tyrosine kinase inhibitors have been tested in clinical trials to target FLT3, with largely disappointing results. Newer, more specific FLT3 inhibitors are currently being tested, in the hope that these will lead to improved patient outcomes.54 Preliminary data suggest that the addition of dasatinib to standard induction chemotherapy is potentially able to abrogate the negative effect of a KIT mutation in CBF leukemia,55 a finding that if confirmed would have testing implications for clinical laboratories, as confirmation of CBF leukemia would need to be expedited. Unlike the situation with a disease such as chronic myelogenous leukemia, tyrosine kinase inhibitors are unlikely to be a viable monotherapy in AML given that tyrosine kinase mutations are frequently secondary, subclonal events that may not be present at relapse.56–58 Compounds that target mutant IDH have shown initial promising early clinical results, with the IDH2 inhibitor AG-221 eliciting durable complete remissions in some patients.59 Retrospective analysis suggests that IDH mutations may also predict clinical response to hypomethylating agents, which are already used in the clinic for patients who are unfit for intensive chemotherapy.60 Given the many epigenetic regulators that are altered in AML, other epigenetic-modifying therapies may hold promise. Recent work has suggested strategies to target MLL-associated AML by inhibiting the function of DOT1L or MLL-menin interaction, both critical to oncogenic MLL functions.61,62
Genetic analysis of AML is essential, and every effort should be made to obtain adequate material for testing. Conventional karyotyping is the standard of care and is critical in patient management. Supplementation by FISH analysis may be helpful in certain scenarios such as failed cytogenetics analysis, normal karyotype but with morphologic suspicion of a recurrent cytogenetic abnormality such as inv(16), need for a rapid turnaround time for PML-RARA detection, or for detection of recurrent rearrangements. Molecular testing for NPM1, FLT3-ITD, and CEBPA is necessary in intermediate-risk AML, as is testing for KIT in CBF AML.
These genetic assessments currently impact prognosis and HSCT decision making, and in the future, assessment of these genes and others may direct targeted therapy. With the ever-increasing understanding of the genetics of AML, more comprehensive genetic profiling of AML will likely be standard of care in the future. This testing is currently readily clinically available in the United States; however, there is no real consensus yet regarding which additional genes should be tested: combined College of American Pathologists/American Society of Hematology recommendations are being formulated that may provide more guidance, and a revision of the current WHO classification is underway. Further studies will be needed to guide interpretation of complex issues such as the interactions between mutations, differences between specific mutations within genes, and the importance of allele burden and subclonal mutational architecture in the management of AML.
The authors have no relevant financial interest in the products or companies described in this article.
Presented in part at the New Frontiers in Pathology: An Update for Practicing Pathologists meeting; University of Michigan; September 4–6, 2014; Ann Arbor, Michigan.