Abstract
Acute myeloid leukemia (AML) is an aggressive malignancy characterized by clonal proliferation of neoplastic immature precursor cells. AML impacts older adults and has a poor prognosis. Despite recent advances in treatment, AML is complex, with both genetic and epigenetic aberrations in the malignant clone and elaborate interactions with its microenvironment. We are now able to stratify patients on the basis of specific clinical and molecular features in order to optimize individual treatment strategies. However, our understanding of the complex nature of these molecular abnormalities continues to expand the defining characteristics of high-risk mutations. In this review, we focus on genetic and microenvironmental factors in adverse risk AML that play critical roles in leukemogenesis, including those not described in an European LeukemiaNet adverse risk group, and describe therapies that are currently in the clinical arena, either approved or under development.
Introduction
Acute myeloid leukemia (AML) is an aggressive malignancy characterized by clonal proliferation of neoplastic immature precursor cells. AML impacts older adults, at a median age of 68 years, and has a poor prognosis with a 5-year overall survival (OS) of roughly 25%. 1 Approximately 35–40% of adults under 60 years of age attain complete remission (CR) and are deemed cured with treatment. Unfortunately, the prognosis in older adults remains poor, with those unable to receive intensive chemotherapy traditionally having a median overall survival of 5–10 months. The combination of a hypomethylating agent (HMA) and the small molecule BCL2 inhibitor venetoclax has improved the median OS to 15 months. 2 Nonetheless, AML remains a complex family of diseases characterized by a panoply of genetic and epigenetic aberrations in both the malignant clone and its microenvironment.
Intensive induction chemotherapy regimens combining cytarabine and anthracycline have been the standard of care for the last 40 years, and cause significant toxicities, especially in older patients. 3 Until the last decade, most clinical trials evaluated cytogenetic and molecular features through secondary or retrospective analysis. More recently, a select few of these aberrations have been targeted pharmacologically with resultant clinical success, 4 emphasizing the importance of upfront diagnostic testing to optimize treatment options. In 2017, the European LeukemiaNet (ELN) updated their risk classification groups by incorporating more mutations, including those associated with adverse clinical outcomes. The adverse risk group now recognizes FLT3-internal tandem duplication (ITD) with high expression (ITD high) with wild type NPM1, mutated RUNX1, mutated ASXL1 and mutated TP53 3 (Table 1) as contributing to worse outcomes.
High-risk features in newly diagnosed AML.
Residual disease defined by multiparameter flow cytometry or positive PCR for disease-specific genes describes a group of patients with significant risk of early recurrence after consolidation therapy, including consolidation in the form of allogeneic transplantation.
More recent data suggests that t(6;9)(p23;q34.1) should likely be re-classified within the intermediate risk group if treated intensively with allogeneic HSCT.5,6
Should not be used as an adverse prognostic marker if they co-occur with favorable-risk AML subtypes.
AML, acute myeloid leukemia; ELN, European LeukemiaNet; HSCT, hematopoietic stem cell transplantation; PCR, polymerase chain reaction.
As we have continued to dissect the molecular pathogenesis and pathophysiology of AML, we are able to stratify patients on the basis of specific clinical and molecular features in order to optimize individual treatment strategies. Herein, we describe selected factors associated with adverse risk AML, focusing on high risk mutations, including those not described in an ELN adverse risk group (Table 2). We discuss selected newer treatment options available for high-risk patients that target specific genetic and microenvironmental factors as well as selected treatments currently under development for future early phase clinical trials.
High-risk mutations, their functional class and mechanisms of leukemogenesis.
AML, acute myeloid leukemia; HSC, hematopoietic stem cells.
Adverse risk molecular factors in the 2017 ELN risk stratification
The clinical and molecular factors associated with a drug-resistant phenotype and overall poor prognoses are delineated in Table 1.
Cytogenetics
Cytogenetic findings are classified according to favorable, intermediate, and unfavorable risk categories. 3 Unfavorable cytogenetics define adverse ELN risk and thus provide critical prognostic information that can inform treatment options. 7 Nonetheless, −7, −5/del(5q), monosomal karyotypes, and complex cytogenetics with at least three abnormalities carry an adverse prognosis independent of treatment type. 8 Adverse risk cytogenetics often accompany secondary AMLs, including myelodysplasia-related (MDS/AML) and therapy-related (t-AML) variants, older age, high risk molecular pathways implicated in leukemogenesis (e.g., TP53), and multidrug resistance.8,9
Of the high-risk cytogenetic translocations, we have chosen to highlight mixed lineage leukemia (MLL)-rearranged AML. AMLs with MLL-based translocations are associated with poor outcomes in adults and frequent relapses, despite initial response to standard induction chemotherapy and allogeneic hematopoietic stem cell transplantation (HSCT). 10 The MLL gene, located on chromosome 11q23, encodes the histone-lysine-N- methyltransferase 2A (KMT2A), which plays a critical role in differentiation and homeostasis through chromatin remodeling and regulation of homeobox (HOX) genes. MLL-rearranged leukemia is characterized by balanced translocation and fusion with over 80 different partner genes including AF4, Afadin, AF9, ELL, and ENL, with the resultant translocations t(4;11)(q21;q23), t(6:11)(q27:q23), t(9;11)(q22;q23), t(11;19)(q23;p13.1), and t(11;19)(q23;p13.3). 11 For leukemia to develop, MLL fusion proteins must interact with the protein menin, a tumor suppressor protein responsible for regulating cell growth for endocrine organs (encoded on the MEN 1 gene). 12 When co-factor menin and MLL fusion proteins interact, there is an upregulation of HOXA9 and MEIS1 genes, which ultimately promotes leukogenesis and proliferation. In fact, when menin is blocked in MLL transformed leukemic blasts, HOX gene upregulation and cell differentiation arrest ceases, supporting menin’s crucial role for oncogenesis. 13 MLL-rearrangement is found more frequently in t-AML (9.4%) than in de novo AML (2.6%, p < 0.0001), with particular occurrence in the setting of agents that freeze the topoisomerase II-DNA complex, such as anthracyclines and epipodophyllins. 14
The histone H3K79 methyltransferase DOT1L is recruited in MLL-rearranged AML, leading to methylation of oncogenic downstream targets HOXA9 and Meisi1. 10 In a murine model, inhibition of DOT1L led to the suppression of downstream MLL target genes with significant tumor regression. The DOT1L inhibitor Pinometostat – a potent and selective small molecule inhibitor of DOT1L methyltransferase activity – has the ability to abrogate HOX cluster gene expression in AML cells, which leads to leukemia cell apoptosis. A phase I study of Pinometostat in MLL-rearranged relapsed/refractory (R/R) myeloid malignancy patients demonstrated tolerability and modest including morphologic changes in the bone marrow consistent with myeloid differentiation. 15 An ongoing phase Ib/II open-label, single-arm trial enrolling R/R MLL-rearranged AML patients will evaluate the tolerability and early efficacy of pinometostat in combination with azacitidine. 16
New therapeutic agents targeting the menin-KMT2A protein–protein interaction are being investigated in early phase clinical trials. 7 Preliminary results of KO-539 in a phase I/IIA trial of adults with R/R disease, show that the drug is generally well tolerated with no dose interruptions or discontinuations due to drug-related adverse events. There is also some suggestion of good anti-leukemic activity. 17
Genetic mutations
FLT3-ITD high+ and wild type NPM1
FLT3 is a transmembrane ligand-activated receptor tyrosine kinase and is expressed on hematopoietic progenitor cells. 18 Mutations in FLT3 occur in 25–30% of all AMLs and result in aberrant activation of RAS/RAF/MEK/mammalian target of rapamycin (mTOR) pathways, as well as through phosphatidylinositol 3 kinase (PI3K)/AKT pathways, all of which lead to cell growth and survival. Higher allele frequencies/ratios, have been associated with poorer outcomes, especially with wild type NPM1.
Prior to ELN 2017, all FLT3 mutations irrespective of allelic ratio were considered to be high risk. A low ITD allelic ratio is considered <0.5, whereas a high allelic ratio is over ⩾0.5. ELN now lists patients with wild-type NPM1 without FLT3-ITD or with FLT3-ITDlow (without adverse-risk genetic lesions) and mutated NPM1 and FLT3-ITDhigh as intermediate risk. 3 Patients with mutated NPM1 with FLT3-ITDlow are listed as having a favorable risk. We do not have prospective data that suggests FLT3 mutated patients in favorable and intermediate risk groups have better outcomes without allogeneic HSCT. Several tyrosine kinase inhibitors (TKIs) have been studied either as monotherapy or in combination with various chemotherapies. Midostaurin is now approved in combination with 7+3 as front-line treatment for newly diagnosed adults, who can tolerate induction treatment. 19 This phase III trial found significantly improved event-free survival (EFS) and OS in the combination midostaurin and daunorubin/cytarabine (7+3) arm compared with the 7+3 arm alone (Table 3). Interestingly, the CR rates were similar in both arms, suggesting that the improvement in OS could reflect a greater depth of CR and/or an increased number of patients able to go to HSCT.
Efficacy of drugs used in AML patients with high-risk mutations.
Composite remission includes CR or CRi.
Investigator-reported ORR = CR+CRi+MLFS.
7 + 3, induction treatment with daunorubicin and cytarabine; AML, acute myeloid leukemia; BSC, best supportive care; CI, confidence interval; CIR, cumulative incidence of relapse; CMML, chronic myelomonocytic leukemia; CR, complete remission; CRi, complete remission with incomplete hematologic recovery; DFS, disease-free survival; EFS, event free survival; GO, Gemtuzumab ozogamicin; HIDAC, high dose Ara-C; LDAC, low dose Ara-C; MDS, Myelodysplastic syndrome; MPN, myeloproliferative neoplasms; ORR, overall response rate; OS, overall survival; RFS, Relapse free survival; R/R, relapsed/refractory.
Sorafenib has been most successful in the post-transplant setting for patients in CR after allogeneic HSCT. 41 Sorafenib was instituted between 30 days after induction/consolidation and 120 days post-transplant once patients achieved count recovery. The OS at 24, 36, and 48 months was 76% [95% confidence interval (CI), 63–91%], 76% (95% CI, 63–91%), and 57% (95% CI, 31–91%), respectively. The EFS at 24, 36, and 48 months was 74% (95% CI, 62–90%), 64% (95% CI, 48–85%), and 64% (95% CI, 48–85%), respectively. Sorafenib in this study population was well tolerated and did not impair engraftment, with a non-relapse mortality rate at 3 years post-transplant of 10% (95% CI, 1–20%). The phase II SORMAIN trial (N = 83) evaluated the use of sorafenib for 24 months in adults with FLT3-ITD positive AML after obtaining complete hematologic remission with HSCT. 42 At a median follow up of 41.8 months, they found a 24-month relapse free survival of 53.3% (95% CI, 0.36–0.68) in the placebo group versus 85.0% (95% CI 0.70–0.93) with hazard ratio (HR) of 0.26 (95% CI 0.10–0.65, p = 0.002).
Next-generation TKIs (crenolanib, quizartinib, and gilteritinib) have more specific activity with fewer off-target effects. 43 Of these TKIs, gilteritinib demonstrates the greatest clinical benefit to date, as seen in the R/R AML ADMIRAL trial, 44 and is approved for R/R AML. Compared with salvage chemotherapy, gilteritinib had a higher CR rate (34% versus 15.3%) and OS of 9.3 months versus 5.6 months (p < 0.001). Post hoc analyses of phase II and III combination trials suggest improved efficacy in the subgroup of patients with FLT3 mutations relative to standard approaches. Combinations of TKIs with diverse chemotherapies (CPX-351, azacitidine, 7 + 3, cladribine) and targeted agents such as venetoclax are in clinical trials, with over 20 open and recruiting trials within the United States for FLT3-positive AML. Other studies with subgroup analysis of FLT3 positive patients are listed in Table 3.
RUNX1
RUNX1, located on chromosome 21q22, is an important transcription factor involved in hematopoietic stem cell (HSC) growth, differentiation, and homeostasis. Different types of RUNX1 mutations are found in AML, including missense mutations, deletions, truncation mutations, and frameshift mutations in the “Runt” homology domain.45,46 These diverse mutations lead to loss-of-function mutations with attendant chemotherapy resistance and poor prognosis. 47 RUNX1 mutations are typically associated with older age, male gender, more immature morphology, and MDS/AMLs. 48
RUNX1 mutations are seen in approximately 8–16% of AML patients and are typically associated with ASXL1 mutations and other epigenetic modifiers (IDH2, KMT2A, EZH2) as well as spliceosome mutations. 48 They are typically inversely associated with NPM1 and CEBPA mutations, and are associated with lower CR rates, shorter disease-free survival (DFS), EFS, and OS. 49 Co-mutation with ASXL1, SRSF2, or PHF6 confers a significantly worse prognosis relative to pairing with other mutations such as IDH2. 48
To date, there are no approved targeted agents for mutated RUNX1 (mtRUNX1), but preclinical studies demonstrate that the depletion of RUNX1 by short-hairpin RNA as well as editing-out RUNX1 eR1 (CRISPR/Cas9-mediated) leads to AML cell death. 46 In vitro treatment with bromodomain and extraterminal (BET) protein antagonists has also proven efficacious in causing mtRUNX1 AML cell death. BET proteins recognize acetylated lysine moieties on histones and act as a scaffold to recruit promoters and enhancers to co-activate expression of genes involved in cell growth and survival. AML mtRUNX1 cell engrafted in mice treated with a BET protein inhibitor exhibited increased apoptosis and subsequently improved survival. A phase I trial of the BET inhibitor OTX015 in 41 patients yielded CR or CR with incomplete recovery (CRi) in 3, but did not detect a correlation between response and mtRUNX1. 50 Other early trials with other BET inhibitors have shown similar findings.51,52 The first generation BET protein inhibitor, ABBV-075, in combination with venetoclax, was found to significantly reduce AML cell-burden and prolong survival in AML engrafted immune depleted mice, 53 which may provide a springboard for BET inhibitor/venetoclax combination trials.
ASXL1
Additional sex comb-like 1 (ASXL1) is a chromatin-binding polycomb protein required for normal embryogenesis through epigenetic activation and repression of gene transcription, and is located on chromosome band 20q11. 54 ASXL1 mutations are detected in 10–20% of AMLs and consist predominantly of heterogenous nonsense/frameshift mutations that appear to result in loss of function.55,56 Nonetheless, gain-of-function mutations have also been suspected with homozygous mutations. Similar to RUNX1, they are found in older patients, those with secondary AML, and are often co-mutated with RUNX1 and spliceosome mutations. 54 They are inversely associated with FLT3 ITD mutations, mutually exclusive with NPM1 mutations, and associated with lower rates of CR and overall poorer prognosis when compared with wild type ASXL1.55,57 There are currently no targeted agents that have shown improved outcomes in this subgroup of patients. Nonetheless, BET inhibitors may afford complementary effects on chromatin remodelling.51,52
Tumor protein 53
The tumor protein 53 (TP53) gene, located on chromosome 17p13.1, is a prototypical tumor suppressor gene, and is mutated in up to 50% of cancers. 58 The encoded TP53 protein is involved in transcriptional regulation of downstream pathways crucial in tumor suppression through apoptosis, DNA repair, and cell cycle arrest/senescence. Multiple forms of TP53 mutations are detected in AML, including missense mutations, deletions, insertions, and nonsense mutations. 59 TP53 gain-of-function mutations can also lead to proliferation and survival of tumor cells, as well as angiogenesis and metastasis. 60 The most common TP53 mutations are missense alterations occurring in the DNA-binding domain. 61 Though mutations in other domains also impact TP53 protein function, the implication of these mutations remains uncertain.
While germline TP53 mutations define the archetypal familial cancer syndrome known as Li-Fraumeni, only 1.1% of AML cases have a germline TP53 mutation On the other hand, abnormalities in TP53 arise as somatic mutations in 10% of de novo AMLs 20% in t-AMLs, and up to 90% in erythroleukemias. 58 Prognosis with TP53 mutations is dismal, with high rates of disease refractory to both chemotherapy and HSCT with CR rates <30% and 3-year OS <15%.58,62 A recent study showed that patients with TP53 mutations with variant allele frequency (VAF) >40% had significantly worse cumulative incidence of relapse (p = 0.030), relapse-free survival (RFS [p = 0.001]) and OS (p = 0.003) than patients with a VAF ⩽40%. 63 In patients treated with a cytarabine-based regimen, the median OS of patients with a VAF >40% was 4.7 months versus 7.3 months for patients with a VAF ⩽40% (p = 0.006). In patients treated with HMA, the VAF did not affect OS significantly. Interestingly, patients with VAF ⩽40% treated with cytarabine based regimen, had improved OS compared with those treated with HMA (1-year OS rates of 44% and 31%, respectively; p = 0.04) whereas patients with VAF >40% had poorer OS regardless of treatment choice (1-year OS rate <25% in all groups). Similarly, Sasaki et al. assessed VAF in various common driver mutations in 421 patients with newly diagnosed AML. 64 TP53 mutations, found in 20% of their cohort, were the major contributor to decreased OS with each increasing increment in VAF associated with a 1% higher risk of death. The median VAF in their cohort was 45.7% (range 1.15–93.74%).
APR-246 (PRIMA-1MET) is a small molecule that restores wild-type TP53 transcriptional activity of unfolded wild-type or mutant TP53 protein, resulting in apoptosis of TP53 mutated cancer cells. Preliminary results from a phase II trial of APR-246 in combination with azacitidine in 45 treatment naïve patients with intermediate/high/very high risk MDS, myeloproliferative neoplasms (MPNs) or oligoblastic AML (blasts ⩽30%) harboring TP53 mutations yielded an overall response rate (ORR) of 87% and CR rate of 53%. 29 The median duration of response was 6.5 months with intent-to-treat OS of 11.6 months (95% CI 9.2–14). The median VAF of TP53 (25% at the start of the trial) was serially assessed during treatment using next generation sequencing (NGS). At a VAF cut off of 5%, 39% achieved NGS negativity with treatment, which was associated with improvement in OS (12.8 versus 9.2 months, p = 0.02). Additionally, the median VAF at maximum mutation clearance was 0.63% (0.0–5%), with 5 (11%) becoming minimal residual disease (MRD) negative. There is an ongoing phase III trial with APR-246/azacitidine [ClinicalTrials.gov identifier: NCT03745716]. In addition, APR-246 is being studied in a phase I trial in combination with azacitidine and venetoclax [ClinicalTrials.gov identifier: NCT04214860].
Magrolimab (Hu5F9-G4) is a blocking antibody directed against CD47, which is also known as the macrophage immune checkpoint or the “don’t eat me” signal on cancer cell surfaces. 30 This antibody induces phagocytosis and AML cell death, particularly in TP53 mutated patients. A phase Ib study combining magrolimab and azacitidine in 34 treatment-naïve AML patients unfit for intensive chemotherapy resulted in transfusion independence in 56%, ORR 65%, CR 44%, and CRi 12%, with a median time to response of 2.04 months. 65 Of the 19 patients achieving CR/CRi, 37% (7/19) became MRD negative by flow cytometry. Importantly, patients with a TP53 mutation had a 71% objective response (15/21) with CR 48% (10), and 1 CRi, with a median duration of response of 9.9 months. The median OS for the TP53 mutated patients was 12.9 months (95% CI: 6.24 months–not reached) compared with 18.9 months for TP53 wild-type patients (95% CI: 4.34 months–not reached).
Pevonedistat (MLN4924) is a first-in-class small molecule inhibitor of Nedd8-activating enzyme (NAE), thereby inactivating E3 ubiquitin ligases known as ring ligases (CRLs), which in turn leads to accumulation of CRL protein substrates such as c-MYC, leading to eventual cell death though activation of pro-apoptosis pathways. CRLs are crucial for cell proliferation and survival. Based on clinical efficacy as a single agent in the R/R setting, a phase Ib study combining pevonedistat with azacitidine in adults aged ⩾60 years with newly diagnosed AML showed an ORR of 50% in the intent-to-treat analysis, with a median duration of response of 8.3 months in 44% who were able to receive at least six cycles of the combination. 66 Of the five patients with TP53 mutation identified at baseline, CR or PR was achieved in four. A total of six out of eight (75%) total TP53-mutated patients achieved CR/CRi. There is an ongoing phase Ib trial in adults ⩾50 years with de novo or secondary AML including those with adverse cytogenetics and/or TP53 mutations, combining venetoclax and azacitidine with escalating doses of pevonedistat (PAVE) [ClinicalTrials.gov identifier: NCT04172844].
Flotetuzumab, a bispecific DART (dual-affinity retargeting agent) antibody-based molecule to CD3ε and CD123, is being investigated as monotherapy in the R/R setting, and early findings suggest that the 42 patients with TP53 mutations have higher rates of CD+ T cell infiltration, expression of immune checkpoints, and IFN-γ signalling than patients with other risk defining mutations such as ASXL1, TET2, and DNMT3A. 67 CR was achieved in 47% (7/15) of patients with R/R AML and TP53 abnormalities, with 2 remaining in CR >6 months. Responders exhibited significantly higher tumor inflammation signature at baseline defined by FOXP3, CD8, inflammatory chemokine, and PD1 gene expression scores, compared with nonresponders. Immune infiltration in the tumor microenvironment is an important predictor of treatment outcomes with immunotherapy, and is discussed further in the section Microenvironmental Targets: Immunotherapy below. Patients with TP53 mutations who achieved CR had a median OS of 10.3 months (range 3.3–21.3 months), suggesting that flotetuzumab immunotherapy could improve treatment outcomes in this high-risk group compared with standard of care – a hypothesis that needs to be tested. Other studies with TP53 mutations analysis are listed in Table 3.
Mutations not included in the 2017 ELN risk stratification
KIT
Some chromosomal abnormalities seen in AML are t(8;21)(q22;q22) and inv(16)(p13;q22), known as core binding factor-AML (CBF-AML), which produce corresponding abnormal fusion genes RUNX1-RUNX1T1 and CBFB-MYH1.68,69 CBF-AML is seen in approximately 15–20% of newly diagnosed AML cases and generally has a better prognosis except in the presence of a c-kit mutation; seen in approximately 60–80% of AML patients, with activating mutations seen in 20% of CBF-AML adult patients.45,68 KIT is located at chromosome band 4q12 and encodes a transmembrane glycoprotein that activates downstream signaling pathways involved in cell proliferation, differentiation, and survival. 70 Discrepancies exist in terms of outcomes in patients with c-kit mutations, with some studies showing worse OS and decreased remission duration, while others show similar outcomes in those with and without the mutation.68,71–75
Clinical trials have evaluated the effects of adding a TKI to standard treatment options in CBF-AML. Dasatinib in combination with7 + 3 induction followed by consolidation and maintenance treatment was studied in the CALGB 10801 phase II trial. 32 This trial enrolled 61 adult patients with newly diagnosed CBF-AML, of whom 19% harbored a c-kit mutation (Table 3). In a post hoc analysis of this limited cohort, patients with c-kit mutations had similar outcomes to those with wild-type KIT with 3-year rates of DFS (67% versus 75%) and OS (73% versus 76%), and there were no differences related to the magnitude of KIT wildtype expression.
The phase II AMLSG 11-08 trial also examined the effects of dasatinib in combination with intensive induction and consolidation chemotherapy in newly diagnosed CBF-AML. 33 In contrast to CALGB 10801, patients with c-kit mutation (n = 19) had inferior outcomes compared with c-kit wild type (n = 70). There is an ongoing phase III trial of intensive chemotherapy with or without dasatinib in newly diagnosed adult patients with CBF AML with RUNX1-RUNX1T1 (or variant form) or CBFB-MYH11 fusion transcripts [ClinicalTrials.gov identifier: NCT02013648]. Additionally, there is a phase II study combining midostaurin with standard induction, consolidation, and maintenance therapy in adult patients ⩽65 years of age with de novo AML harboring rearrangements in CBF-AML genes (RUNX1-RUNX1T1 and CBFB-MYH11); [ClinicalTrials.gov identifier: NCT03686345] and [ClinicalTrials.gov identifier: NCT01830361].
RAS
RAS genes encode a family of proteins that are crucial in cell signaling networks that regulate cell function across multiple tissue types. 76 RAS oncogenes are the most common somatic mutations in human cancer and occur in 12–27% of patients with AML. In AML, the incidence of NRAS mutations is about 11%, compared with approximately 5% in KRAS mutations. The RAS pathway is activated by mutations in upstream receptor tyrosine kinases (FLT3 or c-KIT) or mutations/overexpression of downstream effector pathways in RAS/RAF/MEK/ERK kinase pathways. 77 Therefore, patients without RAS mutations but with FLT3, KIT or PDGRF mutations have activation of RAS-dependent pathways, thereby increasing proliferation and preventing apoptosis of leukemic cells. The exact prognostic implication of RAS mutations is unclear, particularly in the newly diagnosed setting.78–84 Nonetheless, a retrospective study found decreased median EFS in AML patients that harbored NRAS or KRAS mutations (4.9 months) compared with RAS wild-type patients (11.4 months; p < 0.01) at a median follow up of 25 months. 85 RAS was associated independently with increased risk of death with HR of 1.85 (p = 0.016). Interestingly, relapse occurred despite ongoing RAS mutation clearance in 6 out of 10 patients.
Acquisition of RAS or FLT3 ITD mutations at the time of progression from MDS to AML clearly leads to worse outcomes. 86 The median OS after leukemia transformation in those with a detectable RAS and/or FLT3-ITD mutations was 2.4 months compared with 7.5 months in patients with wild type RAS and FLT3 (HR: 3.08, 95% CI 1.9–5.0, p < 0.0001). For RAS mutation alone, the median survival after leukemic transformation was 3.6 months compared with 7 months in patients with wild-type RAS (p = 0.0008).
For patients with IDH mutations, co-mutations with RAS have been associated with higher rates of resistance to monotherapy with targeted treatments. 87 Proposed mechanisms include constitutive activation of RAS pathway, bypassing the differentiation induced by IDH1/2 inhibitors, or RAS mutation being a marker of higher mutational burden, which increases rates of resistance. These associations have not been consistently identified, and, more recently, the AG221-AML-005 trial reported that co-mutations with RAS/FLT3 have not affected response rates to combination treatment with the IDH 2 inhibitor enasidenib plus azacitidine in newly diagnosed patients being treated with this regimen in the front line setting. 88 However, a recent phase Ib/II trial combining the IDH1 inhibitor ivosidenib, venetoclax +/− azacitidine in advanced myeloid malignancies, found that active signaling co-mutations (including RAS) were seen in 66% of patients who had no response or had relapsed. 89
The emergence of AML subclones with new RAS-MAPK pathway mutations in patients who relapse after the FLT3 inhibitor gilteritinib has also been detected.90,91 At least in theory, combinations of FLT3 inhibitors and inhibitors of MEK inhibitors or other downstream intermediaries might prevent or overcome FLT3 inhibitor resistance. In this context, preclinical data have supported the notion that inhibition of downstream pathways such as the mitogen-activated protein kinase (MAPK) and/or PI3K/AKT pathway could abrogate net RAS signalling. 77
RNA spliceosomes
Normal cells require functional RNA splicing of mRNA precursors by a multiprotein spliceosome complex for complete gene expression. Mutations in RNA spliceosomes leads to mis-splicing of mRNA precursors leading to abnormal epigenetic regulation, transcription, and genome integrity, ultimately leading to cancer.92,93 RNA splicing mutations (e.g., SRSF2, SF3B1, U2AF1, and ZRSR2) are frequently identified in MDS and chronic myelomonocytic leukemia (CMML), with about 10% of AML patients harboring these mutations. 45 These mutations are usually exclusive of one another and are expressed with its wild type allele, suggesting the requirement of the wild-type allele function for survival of the cancer cell. These mutations are typically found in older patients with less proliferative disease, and are associated with refractory disease.
H3B-8800 – a small molecule agent that binds to SF3b complex leading to alternative splicing changes – has been studied in patients with MDS, AML, or CMML. 34 Pharmacodynamic correlates detected dose-dependent alterations in mature mRNA transcripts in blood mononuclear cells with good overall patient tolerability (mostly grade 1 or 2 toxicities). While there were no objective responses seen in the 84 patients (24 with AML), there were measurable improvements in red blood cell (RBC) transfusion requirements.
In addition to preclinical studies looking at small molecule inhibitors to target the core spliceosome, RNA-binding proteins (RBPs) have also become targets of interest in myeloid leukemias with RNA splicing mutations. 94 Specifically, RBM39 is a suspected RBP molecular target of sulfonamide compounds (such as indisulam, E7820, and chloroquinoxalin). RBM39 is an RBP, which has important functions in transcriptional coactivation and pre-mRNA splicing. Hypothetically, depletion or inhibition of RBM39 should lead to splicing alterations that, in turn, lead to cell death. Indisulam, a sulfonamide compound, has been shown to recruit RBM39 to the CUL4-DCAF15 E3-ubiquitin ligase, leading to its degradation and ultimately aberrant splicing. Finally, anti-sense oligonucleotide based-therapies hybridize with RNA and may thereby affect splicing events of pre-mRNA or promote RNA degradation. Early trials in this field are currently underway.
DNMT3A
DNMT3A is a pivotal regulator of the epigenetic processes of DNA methylation and chromatin modification and is mutated in about 20% of de novo AML cases. 95 DNMT3A found on human chromosome 2p23, which encodes a 130-kDa protein that is expressed in two major forms: a long isoform known as DNMT3A1, and a short isoform known as DNMT3A2. Both isoforms catalyze the methylation of cytosine residues in cytosine guanine dinucleotide islands with resultant silencing of diverse genes involved in the processes of differentiation and self-renewal. DNMT3A mutations are detected mainly in older patients, and are associated with higher white blood cell counts, normal cytogenetics, and myelomonocytic or monocytic morphology. DNMT3A is often co-mutated with NPM1, IDH1, and FLT3-ITD and, although not yet incorporated into the ELN risk score, confers a poor prognosis.35,95–97 The DNMT3A-R882 mutation has been associated with poor prognosis in patients over age 60 years old, while other DNMT3A mutations have been associated with adverse prognosis in younger patients. 98 A recent Chinese study of 870 adults with AML, found that prognosis of DNMT3A R882 was associated with a mutant-allele ratio, with higher allele ratios predicting a relatively poor prognosis. 99
The frequency and adverse prognostic features of DNMT3A mutations suggest that the ability to target these mutations directly or through downstream intermediaries could have significant clinical impact. In this regard, the H3 lysine methyltransferase DOT1L as a critical downstream mediator of growth and survival-promoting HOX genes, is a potential therapeutic target for the treatment of DNMT3A-mutated AML,35,100 with interest in DOTL1 inhibitor Pinometostat (discussed above in regards to MLL-rearranged leukemia).
Variants in the DNMT3A gene have also been associated with clonal hematopoiesis of indeterminate potential (CHIP), and clonal cytopenias of undetermined significance (CCUS), which have been identified as pre-malignant conditions, often associated with older age.101–103 Though several variants in DNMT3A have been identified in CCUS and CHIP, not every mutation has been later linked to the development of AML, and the recommendation has been to monitor such patients.
Number of driver mutations: combining cytogenetics and mutations
While complex karyotypes or combined chromosomal abnormalities are present in approximately 10–12% of AML and are associated with adverse prognosis, older age, and treatment resistance in AML, little is known regarding the impact of specific co-mutations on prognosis.3,104,105 The tumor-mutational burden is relatively low in AML relative to solid tumors, and, as such, there may be less neoantigen presentation and T cell mediated immune response compared with other malignancies. 106 Nonetheless, data suggest that increasing number of mutations is associated with poorer prognosis. 105
Metzeler et al. sequenced the entire coding sequences of 37 genes and recurrently mutated regions in 664 untreated AML patients who received intensive induction chemotherapy between 1999 and 2012. 107 The majority (97%) of patients had at least one identified driver mutation, and an additional 15 patients had recurrent balanced chromosomal translocations. The number of mutated genes per patient was 4 (range 0–10). The median number of driver gene mutations increased with age (p < 0.001). Patients with intermediate-risk cytogenetics had a higher number of mutated driver genes (median of 4) compared with patients with favorable or adverse cytogenetics (median of 1 and 2, respectively). Additionally, they found that patients with secondary AML or t-AML had fewer mutations in genes covered, compared with those with de novo AML.
In a prospective study looking at 28 frequently mutated genes in 271 patients with de novo AML, patients in the intermediate cytogenetic risk group had an average of 2.76 mutations, with co-mutations associated with shorter OS (p = 0.006). Mutations in NPM1 (p < 0.0001), DNMT3A (p < 0.0001), FLT3-ITD (p < 0.0001), TET2 (p = 0.0001), and IDH1/2 (p = 0.0048) were found in higher rates in complex molecular genetic abnormalities involving three or more genes (CMGAs) in the intermediate cytogenetic risk group. Mutations in WT1 and KMT2A-PTD were also seen at higher rates in CMGAs at a high frequency and were mutually exclusive with prognostically favorable CEBPA double mutations (dm) (p = 0.0019). CMGAs were detected in 63.0% of patients within intermediate cytogenetic risk and they carried a shorter 5-year OS in CMGA patients (18.1% versus CMGA-negative, 45.9%; p = 0.0006) and higher 5-year cumulative incidence of relapse (CIR) (CMGA-positive, 83.2% versus CMGA-negative, 52.6%; p = 0.0052). CMGAs were also associated with significantly worse OS and CIR in patients with normal cytogenetics. These trends persisted in patients aged ⩽65 years who were also FLT3-ITD negative (OS: p = 0.0010; CIR: p = 0.1800). The increasing use of NGS (discussed below) will likely allow CMGA positivity to become a strong prognostic marker, which could be incorporated in the adverse risk ELN group.
Microenvironmental targets: immunotherapy
The multifaceted bone marrow microenvironment exerts a critical influence on leukemia pathogenesis, pathophysiology, and the ability to eradicate the leukemic clone. 108 A key determinant in leukemia generation versus eradication is the multicompartmental immune system. Dysfunction of this system can lead to immune escape of AML cells throughout the course of the disease. Additionally, dysfunction within CD8+ T cells, and increased regulatory T cells in the peripheral blood has pointed towards important dysregulation of T cells in AML.
Both phenotype and genotype features of exhaustion and senescence are seen in CD8+ T cells identified in AML patients. 109 Furthermore, different phenotypes are also seen in patients who respond and those who do not respond to induction treatment. Patients who do respond to treatment have upregulation of co-stimulatory T cell signaling pathways, and downregulation of apoptotic T cell signaling pathways. Additionally, increased levels of regulatory T cells have been associated with inferior outcomes. Preclinical studies have shown that primary AML cells express inducible T-cell co-stimulator (ICOS), which is a member of the CD28 family of co-stimulatory molecules, and maintains durable immune reactions when binding to ICOS ligand. 110 ICOS/ICOS ligand binding plays an important role in Treg cell function and differentiation. High levels of ICOS+ regulatory T cells secrete high interleukin (IL)-10, which promotes AML cell proliferation. Pomalidomide in an immunomodulatory imide drug (IMiD) that induces T cell proliferation and enhances IL-2 and interferon γ (IFNγ) production through its interaction with cereblon – a substrate receptor for E3 ubiquitin ligase complex. 111 This second generation IMiD also inhibits regulatory T cells and enhances both natural killer (NK) cell mediated cytotoxicity and antibody-dependent cellular cytotoxicity (ADCC) due to increased IL-2 production. In an attempt to modulate this immune response, investigators have done early trials with pomalidomide after timed sequential therapy induction treatment with AcDVP16 (cytarabine, daunorubicin and etoposide) in newly diagnosed AML patients and high risk MDS, at the beginning of early lymphocyte recovery (days 14–21). In 43 patients (AML n = 39, remainder MDS), the overall CR (CR + CRi) rate was 75% with a median OS of 27.1 months, and DFS of 20.6 months. Within the AML group, there was an overall 86% CR/CRi rate in AML patients with unfavorable-risk cytogenetics.
In addition to T cell dysfunction, AML cells in the relapsed setting have been shown to have high levels of PD-1 expression also contributing to immune evasion. 112 Interestingly, PD-1 expression in CD4+ and CD8+ T cells in newly diagnosed patients is significantly less, and often not observed. It is, however, seen more commonly during therapy, after alloHSCT, and at disease relapse. Having higher PD-1 expression is also associated with poor prognosis.
Early findings in the R/R AML setting, as seen in a single center azacitidine and nivolumab non-randomized phase II study, found moderate responses with an ORR of 33% (23/70 patients), of which there were 22% overall CRs (CR/CRi). 113 A phase II trial of pembrolizumab given post HIDAC salvage chemotherapy in R/R AML (n = 37), had an overall response (ORR: CR+CRi+PR+MLFS) of 46% and composite CR (CR+CRi) rates of 38%. 114 In the newly diagnosed setting, a large randomized phase II trial evaluated the efficacy and safety of azacitidine and durvalumab versus azacitidine alone in both high risk MDS and older and unfit AML patients not eligible for intensive chemotherapy. 115 Of the 129 enrolled AML patients, the median OS and median PFS were not statistically different between the two groups at 13.0 versus 14.4 months and 8.1 versus 7.2 months, respectively. Though treatment was well tolerated, there was no clinically relevant benefit to adding immunotherapy to treatment.
There are several large trials currently underway evaluating pembrolizumab, nivolumab, and atezolizumab in combination with various backbone treatments including venetoclax, HMA, or induction chemotherapy in both R/R and newly diagnosed AML patients, largely in the post-transplant and post remission setting, where PD 1 expression is likely to be higher than at baseline. 112
Measuring the depth of AML cell kill: a critical part of assessing efficacy
MRD is defined as the presence of residual leukemia that was not detected through traditional histopathologic methods and in the presence of <5% blasts. 116 MRD positivity correlates with an increased risk of relapse either at the end of treatment or prior to allogeneic BMT. 117 Nonetheless, MRD has not yet been standardized quantitatively or qualitatively, which will be necessary for widespread clinical application. 118 MRD can be identified using multiparametric flow cytometry (MPFC), digital PCR, allele-specific oligonucleotide quantitative polymerase chain reaction (PCR), and NGS, depending on resources available and the genomic abnormality being followed (Table 4).116,119 MPFC can identify the presence of leukemia cells down to levels of 1 × 104–1 × 106. It is fast and widely available, and provides an absolute quantification; however, MPFC has less sensitivity than other methods, and variable antigen expression may yield false negative results. Quantitative PCR methods detect aberrant mutations in leukemic cells compared with a housekeeper gene down to levels of 1 × 106–1 × 107, with variability depending on the mutation. The problem with this technology is that there is often no PCR target, especially in the elderly.
Technologies for MRD detection.
ASO, allele-specific oligonucleotide; dPCR, digital polymerase chain reaction; MPFC, multiparametric flow cytometry; MRD, minimal residual disease; NGS, next generation sequencing.
By comparison, NGS can detect the presence of leukemia cells down to levels of <1 × 106, has a high degree of sensitivity, and does not need patient-specific primers or probes as compared with allele-specific olignonucleotide PCR. 120 The downside to NGS is that it remains expensive, requires a high degree of expertise, and is not yet standardized. Extensive discussion of these techniques and upcoming technology is not the focus of this review and is therefore not included.
All the methods listed above allow for detection of significantly smaller amounts of residual disease compared with standard morphology. Evidence now suggests that identifying patients with MRD-positive disease can lead to early intervention to prevent relapse with post remission therapy. 118 For patients who are MRD positive prior to transplant, outcomes remain poor. As of yet, however, MRD is not a standardized tool used in clinical setting, despite its crucial role in prognosis and outcome. Current and future prospective trials are crucial to examining the role of specific post-induction strategies, including maintenance therapies in HSCT or patients with MRD positivity.
Conclusion
The molecular dissection of AML has uncovered the genomic architecture and the interactive pathways that are essential for leukemia cell growth and survival. Indeed, AML pathogenesis and pathophysiology represent a complex symphony with diverse components that harmonize to create and sustain malignant cells that are programmed to survive a multiplicity of existential threats including cytotoxic therapies. These components exist not only in the leukemia cells themselves but also in diverse components of the microenvironment.
In this review, we have highlighted a few selected molecular lesions where the development of targeted therapies to modulate those targets is underway. In particular, we have focused on specific high-risk genetic mutations that play critical roles in leukemogenesis and for which there are drugs that are currently in the clinical arena, either approved or under rigorous clinical investigation. There are numerous pivotal genomic lesions and molecular pathways that we have not addressed, for instance the intermediaries that regulate cell cycle progression and the diverse pathways that are activated by DNA damage and orchestrate the repair of such damage. In this regard, optimal anti-AML therapy requires complementary strategies, including diagnostic standards, aimed at interdicting multiple aberrations and at preventing those mechanisms, whether primary or in response to initial therapy, that confer net resistance.
Footnotes
Author contributions
CL developed the idea and format for the review. KD constructed all the tables. CL, KD, and JK all contributed equally to the writing and approval of the final manuscript.
Conflict of interest statement
CL has served on an advisory board for Abbvie, Agios, Daiichi-Sankyo, Macrogenics, and Jazz Pharmaceuticals and as a speaker for Astellas and Jazz Pharmaceuticals. KD and JK declare that they have no competing interests.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
