Introduction: RAS pathway mutations are common mechanisms of resistance to acute myeloid leukemia (AML) therapies. Trametinib, an oral MEK inhibitor, has been shown to have single-agent activity in relapsed/refractory AML and preclinical synergy with venetoclax. Methods: We conducted a single-center, open-label, phase 2 trial of the combination of azacitidine, venetoclax, and trametinib in patients with relapsed or refractory AML harboring a RAS pathway-activating mutation. Results: Sixteen patients were treated. The patients were heavily pretreated with a median number of 4 prior therapies; 13 (81%) had received a prior hypomethylating agent (HMA) with venetoclax, and 8 (50%) had undergone prior stem cell transplant. Four patients (25%) responded (CR, n = 1; CRi, n = 1; MLFS, n = 2). Two of the 3 patients (67%) who had not previously received HMA plus venetoclax responded; in contrast, only 2 of the 13 patients (15%) who had previously received HMA plus venetoclax responded. The median OS was 2.4 months, and the 6-month OS rate was 31%. Related grade 3–4 adverse events occurred in 50% of patients, and 50% of patients required a dose adjustment of trametinib. Conclusions: The combination of azacitidine, venetoclax, and trametinib had only modest activity in patients with relapsed/refractory AML, with a response rate that was similar to previous reports of trametinib monotherapy. Substantial toxicity was observed with this combination. Given the established role of RAS pathway mutations in mediating resistance to AML therapies, future studies of better tolerated, more active inhibitors of this pathway are still needed.
RAS pathway-activating mutations occur in approximately 10–25% of de novo acute myeloid leukemia (AML) and lead to constitutive activation of downstream prosurvival and anti-apoptotic pathways . Mutations in the RAS pathway are common mechanisms of resistance to AML-directed therapy, specifically to the BCL2 inhibitor venetoclax . The combination of a hypomethylating agent (HMA) with venetoclax is standard of care in patients with newly diagnosed AML who are unfit for intensive chemotherapy [3‒5]. However, RAS pathway mutations are associated with resistance and inferior survival with this combination [6‒8]. This pathway is also a well-described mechanism of clonal selection and relapse with FLT3, IDH1, and IDH2 inhibitors [9, 10]. Targeted inhibition of the RAS pathway may present an opportunity to more effectively treat these patients.
Trametinib, an oral MEK inhibitor, is approved for the treatment of BRAF-V600E-mutated metastatic melanoma, non-small-cell lung cancer, and anaplastic thyroid cancer. In relapsed/refractory AML with an NRAS or KRAS mutation, trametinib is active as monotherapy, with an overall response rate of 20% reported in one study . Preclinical in vitro and in vivo data suggest a MEK inhibitor in combination with venetoclax has synergetic antileukemia activity; inhibition of AML cell growth has been reported even in cell lines with single-agent resistance . Given the established role of RAS pathway-mediated resistance and the promising preclinical results of the combination of venetoclax and trametinib, we designed a phase 2 study of this combination in conjunction with azacitidine in patients with relapsed/refractory AML.
Adults 18 years and older with relapsed or refractory AML or intermediate- or high-risk (per the International Prognostic Scoring System) myelodysplastic syndrome or chronic myelomonocytic leukemia (CMML) with a RAS pathway-activating mutation were eligible. Eligible RAS pathway-activating mutations included a gain-of-function mutation in KIT, HRAS, NRAS, KRAS, BRAF, CBL, or PTPN11 or a loss-of-function mutation in NF1. Eligible patients were required to have an Eastern Cooperative Oncology Group performance status of ≤2, total serum bilirubin <2.5 times the upper limit of normal, ALT/AST <3 times the upper limit of normal, and a creatinine clearance >30 mL/min. This study was approved by the Institutional Review Board of the University of Texas MD Anderson Cancer Center and was registered at ClinicalTrials.gov (NCT04487106). All patients provided informed consent according to institutional guidelines and the Declaration of Helsinki.
This was a phase 2, single-center, open-label study. In cycle 1, patients received azacitidine 75 mg/m2 subcutaneously or intravenously on days 1–7, venetoclax on days 1–28 (100 mg on day 1, 200 mg on day 2, 400 mg orally on days 3–28), and trametinib 2 mg orally on days 1–28. Bone marrow examination was performed on day 21, and if blasts were <5% or if the marrow was aplastic, then venetoclax was held to allow for count recovery. For consolidation cycles, azacitidine 75 mg/m2 subcutaneously or intravenously was given on days 1–7, venetoclax was given daily on days 1–21, and trametinib 2 mg was given daily. Dose reductions were permitted for grade 3–4 drug-related toxicities. Venetoclax dose adjustments were made for concomitant use of CYP3A4 inhibitors. Bone marrow aspiration for response assessment was performed on days 21 and 28 of cycle 1. After cycle 1, repeat bone marrows were performed at least every 2–4 cycles.
Genomic samples were obtained from bone marrow aspirates. Libraries were created from these samples utilizing hybridization and enrichment of targeted genomic regions. Bidirectional paired-end sequencing was performed using our next-generation sequencing platform (MiSeq; Illumina, San Diego, CA, USA) that covers 81 recurrently mutated genes in myeloid malignancies. Covered genes were those with a coverage depth of 250 reads or more. The lower limit of detection of this assay (analytical sensitivity) for recurrent RAS pathway mutations is ∼2% (one mutant allele in the background of 49 wild-type alleles).
The primary objective of this phase 2 study was to determine the composite rate of complete remission (CR), CR with incomplete hematologic recovery (CRi), and morphological leukemia-free state (MLFS). Secondary objectives included measurable residual disease status, relapse-free survival (RFS), and overall survival (OS). Responses were assessed according to the European LeukemiaNet 2017 criteria . Kaplan-Meier estimates were used to calculate RFS and OS. RFS was calculated from the time of response until relapse or death, censored at last follow-up if still alive. OS was calculated from the time of treatment initiation until death, censored at last follow-up if still alive. GraphPad Prism version 9 was used for all statistical analyses.
Between August 2020 and May 2021, 16 patients were treated. Baseline characteristics are shown in Table 1. The median age was 67 years (range, 28–84 years); 6 patients (38%) were >75 years of age. This was a very heavily pretreated population with a median number of prior therapies 4 (range, 1–7). All but one patient (94%) had received prior HMA therapy, including 13 patients (81%) who had received prior HMA with venetoclax. Eight patients (50%) had undergone prior hematopoietic stem cell transplant (HSCT). Fourteen patients (88%) had adverse-risk cytomolecular disease, including 6 (38%) with complex karyotype and 3 (19%) with a TP53 mutation. RAS pathway mutations included NRAS in 8 patients (50%), NF1 in 4 patients (25%), KRAS and PTPN11 in 3 patients each (19%), and CBL in 1 patient (6%). Four patients had more than 1 RAS pathway mutation: 1 patient had 2 NRAS mutations; 1 patient had 2 PTPN11 mutations; 1 patient had an NRAS and a PTPN11 mutation; and 1 patient had 2 NRAS, 1 KRAS, and 2 CBL mutations. All patients had at least 1 co-occurring, non-RAS pathway mutations. The most common co-mutations were ASXL1 (50%), TET2 (31%), RUNX1 (25%), and WT1 (25%) (Fig. 1).
Overall, 4 patients (25%) responded, with 1 CR, 1 Cri, and 2 MLFS. All patients responded after 1 cycle. Two of the 3 patients (67%) who had not previously received HMA plus venetoclax responded (1 CR and 1 MLFS); in contrast, only 2 of the 13 patients (15%) who had previously received HMA plus venetoclax responded (1 CRi and 1 MLFS).
Among the patients with response, 1 patient had CMML, and 3 had AML. The only patient enrolled who had CMML achieved CR and still remains on therapy at last follow-up. This patient had diploid cytogenetics, an NRAS G12D mutation, and had received only 1 prior therapy (azacitidine monotherapy). Interestingly, despite good response, this patient continued to have a detectable NRAS mutation with relatively stable variant allele frequency (VAF) (49% to 41% after 10 cycles of therapy). Two other responding patients had diploid AML with at least 1 NRAS mutation and achieved CRi. One of these patients had an NRAS G13R mutation and reduction in NRAS VAF from 51% to 2% after 2 cycles of treatment. The other patient had multiple RAS pathway mutations including 2 NRAS mutations (G12A and G12D), a KRAS mutation, and 2 CBL mutations. The NRAS and KRAS VAF decreased from 15% to 5% (NRAS G12A), 2% to undetectable (NRAS G12D), and 2% to undetectable (KRAS). The fourth patient had a complex karyotype and an NF1 frameshift mutation and achieved MLFS but did not have serial VAF assessment performed.
In addition to the 4 patients with formal responses, 4 patients (25%) had a bone marrow blast reduction of ≥50%, including 1 patient with prior treatment with HMA plus venetoclax who had blast reduction from 26% to 6% and who continued on study for 5.5 months. Three of these patients had an NRAS mutation and 1 patient had a KRAS mutation.
The median duration of follow-up was 11.9 months. At last follow-up, 14 patients had died, and 2 were still alive. All 12 nonresponding patients died. Among the 4 responders, 1 patient relapsed while on study after a remission of 3.6 months and subsequently died, 1 patient proceeded to HSCT while in MLFS and then relapsed 8 months after HSCT and subsequently died, and 2 patients remain on study in continuous remission, with response durations of 9 and 13 months. For the entire cohort, the median OS was 2.4 months, and the 6-month OS rate was 31% (Fig. 2a). The median OS for those who responded versus did not respond to the combination therapy was 12.9 months and 1.9 months, respectively (Fig. 2b).
Safety and Early Mortality
All patients experience at least one adverse event of any grade, and a grade 3–4 adverse event occurred in 14 of the 16 treated patients (88%) (Table 2). The most common nonhematological grade 3–4 events included: pneumonia (44%), febrile neutropenia (31%), mucositis (25%), heart failure (25%), sepsis (25%), and diarrhea (19%). Two patients died while on study due to sepsis, both in cycle 2.
Nonhematologic adverse events of any grade that were considered at least possibly related to trametinib occurred in 13 (81%) patients. These adverse events included diarrhea (75%), nausea (56%), mucositis (50%), central serous retinopathy (6%), and decreased ejection fraction (6%) (Table 3). Grade 3–4 possibly related adverse events occurred in 50% of patients. A total of 8 patients (50%) required dose adjustments of trametinib due to a treatment-emergent adverse event: 1 patient had temporary interruption of trametinib, 5 patients had dose interruptions followed by dose reductions of trametinib, and 2 patients permanently discontinued trametinib. The most common reason for trametinib dose modification was mucositis and/or diarrhea, which required dose adjustment in 5 patients (31%). One patient developed grade 2 central serous chorioretinopathy in cycle 3; after temporary cessation of trametinib for 4 weeks, the vision improved, and trametinib was restarted at a lower dose (1.5 mg) without further visual changes. In 1 patient, grade 3 decrease in left ventricular ejection fraction (from 54% at baseline to 34%) was incidentally noted on day 10 of cycle 1 when an echocardiogram was repeated for an unrelated issue. This patient had no clinical signs of heart failure, and trametinib was held for the rest of cycle 1 and resumed in cycle 2 at a dose of 1 mg. Follow-up echocardiogram 4 weeks later showed recovery of ejection fraction to the baseline level.
The 30-day mortality rate was 19%, and the 60-day mortality rate was 31%. All deaths within 60 days of treatment were due to progressive leukemia or uncontrolled infection.
Despite the preclinical synergy with venetoclax and trametinib, we observed limited activity of the combination of azacitidine, venetoclax, and trametinib in this study. Initially, 20 patients were planned to be enrolled; however, the study was closed early due to low efficacy and concerns for toxicity. This was a very heavily pretreated population, with a median of 4 prior therapies and 81% of patients having previously received HMA plus venetoclax, a population known to have very poor outcomes. In one study of patients with relapsed/refractory AML with prior HMA plus venetoclax exposure, the response rate to salvage therapy was 21%, and the median OS was 2.4 months . The outcomes in our study were almost identical to these historical expectations, suggesting a lack of benefit over standard of care treatments. Furthermore, the response rate observed was similar to that achieved with trametinib monotherapy (overall response rate of 20% in one study), also suggesting that the triplet combination may not provide additional benefit over trametinib monotherapy . While 2 of 3 patients without prior venetoclax exposure responded, the number of venetoclax-naive patients was very small in our study, and therefore, it is not possible to make any conclusions about the efficacy of this regimen in this population – or its benefit over HMA plus venetoclax alone, where CR/CRi/MLFS rates as high as 62% have been reported . It is also notable that 2 out of 5 patients (40%) with both TET2 and NRAS mutations responded to the combination regimen. Interestingly, the presence of a concomitant NRAS mutation and TET2 knockdown has been suggested to increase sensitivity to MEK inhibition in a preclinical mouse model . While no definitive conclusions can be drawn from this observation, it raises the question whether certain co-mutation patterns may sensitize AML to RAS pathway-targeted therapies.
Tolerability was also a concern with this regimen. Adverse events relating to this combination occurred in 81% of patients, and 50% of patients required at least dose adjustment of trametinib. Mucositis and diarrhea in particular made this regimen difficult to tolerate. While the overall adverse event profile was generally similar to previous trials in solid tumors [15‒18], the added toxicity in addition to myelosuppressive therapy made this regimen particularly difficult to tolerate in this AML population. Additionally, compared with a previous study of trametinib monotherapy in relapsed/refractory AML , this combination was associated with a higher rate of grade 3–4 adverse events (88% and 32%, respectively) and dose reductions (31% and 5%, respectively), suggesting possible potentiation of trametinib-related toxicity with the addition of azacitidine and venetoclax.
There are several possible mechanisms that could explain the general lack of efficacy in our population. Proteins upstream of MEK interact with a variety of pathways, including the PI3K/AKT, the JAK/STAT, and the NFĸB pathways, promoting survival despite MEK inhibition [19, 20]. Additionally, upregulation of upstream receptor tyrosine kinases (c-Kit, EGFR, etc.) allows for an adaptive kinome response despite MEK inhibition [19, 21]. One potential solution to overcome this resistance is targeting proteins upstream of MEK such as RAF or RAS. Unfortunately, RAS proteins are ubiquitous, and the binding pocket shares structural similarities to many other proteins, making it notoriously difficult to target [22, 23]. Sotorasib, a novel KRAS G12C inhibitor, has shown promising results in non-small-cell lung cancer . Unfortunately, this mutation is extremely rare in AML. Targeting ERK is also an option that is being investigated at this time [25, 26]. Combination of multiple small-molecular inhibitors of different kinases in the RAS pathway can potentially overcome this resistance. Cytarabine-based chemotherapy regimens have been shown to be particularly efficacious in patients with RAS mutations, and therefore, cytarabine-based regimens in addition to RAS pathway inhibitors might lead to better results [1, 27].
In conclusion, the combination of trametinib, azacitidine, and venetoclax resulted in a modest overall response rate of 25% and a median OS of 2.4 months in this heavily pretreated population of patients with relapsed/refractory AML harboring a RAS pathway mutation, outcomes that are similar to historical expectations in this population. The regimen was also associated with significant toxicity, requiring frequent dose adjustments of trametinib. As RAS pathway mutations remain a major driver of resistance in AML, novel agents that target the MAPK pathway with a more tolerable toxicity profile need to be conducted.
Statement of Ethics
This study was approved by the University of Texas MD Anderson Cancer Center Institutional Review Board and cleared by the Ethics Committee under protocol 2020-0506 and was registered at ClinicalTrials.gov (NCT04487106). All patients provided informed consent according to institutional guidelines and the Declaration of Helsinki. Written consent was obtained from all participants prior to enrollment.
Conflict of Interest Statement
Nicholas J. Short has served as consultant for Takeda Oncology and AstraZeneca, reported receiving research grants from Takeda Oncology and Astellas Pharma Inc., and has received honoraria from Amgen. The authors report no other relevant disclosures.
This research was supported by an MD Anderson Cancer Center Support Grant (CA016672) and SPORE. Nicholas J. Short is supported by the K12 Paul Calabresi Clinical Oncology Scholar Award and the American Society of Hematology Junior Faculty Scholar Award in clinical research.
Conception and design: Nicholas J. Short and Farhad Ravandi; administrative support: Nicholas J. Short and Farhad Ravandi; provision of study materials or patients: Nicholas J. Short, Farhad Ravandi, Naveen Pemmaraju, Marina Konopleva, Gautam Borthakur, Elias J. Jabbour, Naval Daver, Nitin Jain, Kelly S. Chien, Abhishek Maiti, Guillermo Montalban-Bravo, and Tapan M. Kadia; collection and assembly of data: Monica Kwari, Ricardo DeLumpa, Sanam Loghavi, and Nicholas J. Short; data analysis and interpretation: Nicholas J. Short, Sai Prasad Desikan, and Walid Macaron; manuscript writing: Sai Prasad Desikan and Nicholas J. Short; and critical revision for important intellectual content: all the authors. All the authors reviewed and approved the final version of the manuscript.
Data Availability Statement
All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.