The mTOR inhibitor everolimus in combination with azacitidine in patients with relapsed/refractory acute myeloid leukemia: a phase Ib/II study

Therapeutic options are limited in relapsed/refractory acute myeloid leukemia (AML). We evaluated the maximum tolerated dose (MTD) and preliminary efficacy of mammalian target of rapamycin (mTOR) inhibitor, everolimus (days 5–21) in combination with azacitidine 75 mg/m2 subcutaneously (days 1–5 and 8–9 every 28 days) in 40 patients with relapsed (n = 27), primary refractory (n = 11) or elderly patients unfit for intensive chemotherapy (n = 2). MTD was not reached following everolimus dose escalation (2.5, 5 or 10 mg; n = 19) to the 10 mg dose level which was expanded (n = 21). Major adverse events (grade > 2) were mostly disease-related: neutropenia (73%), thrombocytopenia (67%), mucositis (24%) and febrile neutropenia (19%). Overall survival (OS) of the entire cohort was 8.5 months, and overall response rate (ORR; including CR/CRi/PR/MLFS) was 22.5%. Furthermore, a landmark analysis beyond cycle 1 revealed superior OS and ORR in patients receiving 2.5 mg everolimus with azoles, compared to those without azoles (median OS 12.8 vs. 6.0 months, P = 0.049, and ORR 50% vs. 16%, P = 0.056), potentially due to achievement of higher everolimus blood levels. This study demonstrates that everolimus in combination with azacitidine is tolerable, with promising clinical activity in advanced AML.


INTRODUCTION
Mammalian target of rapamycin (mTOR) is a serine/threonine protein kinase regulating cell growth, proliferation, survival, autophagy, gene transcription and protein synthesis. Activation of the PI3K/AKT/mTOR signaling pathway is prevalent in acute myeloid leukemia (AML) [1,2], and constitutive activation of AKT is associated with inferior survival in AML [3]. mTOR phosphoactivates downstream 4EBP1 and p70S6K. Its inhibition mediates cell cycle arrest through 4EBP1 dephophosphorylation, increased p27 and reduced cyclin D1 activity. Rapamycin-induced targeting of mTOR complex 1 (mTORC1) may result in paradoxical feedback hyperactivation of AKT [4,5]. Although mTOR complex 2 (mTORC2) is thought to be rapamycin insensitive, it may be inhibited by higher doses of rapamycin, leading to AKT and ERK suppression [6]. Rapamycin derivatives, temsirolimus and everolimus, have been shown to inhibit both mTORC1 and mTORC2, leading to inhibition of AKT in AML [7].
Small molecule targeting of mTOR has been shown to selectively inhibit the clonogenic function of leukemic,
To date, only one clinical study has combined an mTOR inhibitor and a hypomethylating agent in relapsed/ refractory AML. Decitabine 20 mg/m 2 was given on days 1-5 followed by first generation mTOR inhibitor rapamycin on days 6-25. Among 12 evaluable patients, the median survival was 4 months and only one CR was achieved. [29]. To determine if everolimus, a more potent mTOR inhibitor, would be tolerable and effective at augmenting the activity of hypomethylating agents in AML, we herein report a dose-finding study of azacitidine in combination with the rapamycin derivative everolimus in relapsed and refractory AML.

Toxicity and maximum tolerated dose
Three patients withdrew early due to disease progression, leaving 37 patients evaluable for toxicity. Two patients developed a first cycle dose-limiting toxicity (DLT) during the phase I dose escalation phase. At the 2.5 mg everolimus dose level, one patient developed grade 3 mucositis, whereas at the 5 mg dose level another patient developed grade 3 septic arthritis. No DLTs were experienced at the 10 mg dose level and thus the maximum tolerated dose (MTD) was not defined. The 10 mg everolimus dose level was the recommended phase 2 dose.
The majority of patients exiting the study did so due to disease progression or relapse (n = 29, 76.3%), including two early deaths from AML progression. Other reasons for study discontinuation included allogeneic hematopoietic stem cell transplantation (n = 5, 13.2%), deranged liver function tests due to graft versus host disease (n = 1) and metastatic melanoma (n = 1).
Three patients initially in the 2.5 mg cohort and 9 patients initially in the 10 mg cohort subsequently received azole antifungal prophylaxis from cycle 2 onwards, with the everolimus dose capped at 2.5 mg due to the higher drug levels expected from the interaction with CYP3A4 inhibitory azoles. A landmark analysis was performed on patients from cycle 2 day 1 (resulting in exclusion of 9 patients). Interestingly, ORR and OS were higher among patients who received everolimus and concurrent azoles, compared to those not receiving concurrent azoles: ORR 6/12 (50%) vs. 3/19 (16%, P = 0.056) and median OS 12.8 vs. 6.0 months (P = 0.049, Figure 1C), respectively.
Median OS for patients with baseline intermediate cytogenetic risk was 8.6 months versus 3.8 months for patients with adverse risk (P = 0.06). Median survival outcome was comparable for patients with relapsed and primary refractory AML ( Figure 1D).

Everolimus trough level
Several patients switched from the allocated everolimus dose to the modified 2.5 mg dose when azoles were added. To gain insight into the unexpectedly superior outcomes observed in the everolimus/azole cohort, trough everolimus levels were examined. This revealed that everolimus levels were highest in the everolimus/azole cohort, even higher than patients administered everolimus 10 mg without azoles (Figure 2A). Substantial intra-patient increases in everolimus levels were demonstrated following the addition of antifungal azoles ( Figure 2B). Furthermore, the cumulative everolimus dose was similar between the everolimus/azole cohort (median 394 mg) and those not receiving concurrent azoles (median 525 mg, P = 0.40).

Plasma inhibitory activity (PIA) assays
To assess the pharmacodynamic activity of everolimus, a PIA assay was used to assess patient plasma collected on day 5 and 19 (prior to and after 2 weeks of everolimus therapy), to examine the potential for plasma to inhibit P-p70S6K in the OCI-AML3 reporter cell line (Supplementary Figure S1). PIA responses correlated with increasing everolimus levels recorded at the same time in a subset of patient samples that were available for analysis ( Figure 2C). PIA responses also correlated with bone marrow blast reductions to a greater extent than everolimus drug levels ( Figure 2D), though neither achieved statistical significance.

DISCUSSION
The combination of everolimus with azacitidine was tolerable and deliverable to a group of predominantly  This cohort comprised 3 patients initially in the 2.5 mg "without-azole" cohort and 9 patients initially in the 10 mg expansion cohort with everolimus dose reduced to 2.5 mg from cycle 2 onwards in combination with an antifungal azole. b 3 patients in the 5 mg cohort did not have an evaluable bone marrow examination but had persistent leukemia in peripheral blood. c 2 patients in the 10 mg cohort died from disease progression before completion of therapy. 3 patients in the 5 mg cohort withdrew due to toxicities before completion of the initial therapy. Abbreviations: CR, complete remission; CRi, complete remission with incomplete blood count recovery; MLFS, morphologic leukemia-free state; PR, partial remission In an attempt to facilitate the use of antifungal azoles in this patient population, everolimus was combined with voriconazole or posaconazole, with the concurrent everolimus dose reduced to 2.5 mg daily, due to the known pharmacokinetic interactions between these two drugs. Trough everolimus blood concentrations were noticeably higher among patients taking concomitant azoles, exemplified by the increased intra-patient concentrations (Figure 2), consistent with the known pharmacokinetic interaction with CYP3A4 [30][31][32]. Interestingly, landmark analysis beyond cycle 1 revealed that the everolimus/ azole cohort had a higher ORR (50% vs. 16%, P = 0.056) and longer median OS (12.8 vs. 6.0 months, P = 0.049), when compared to patients not receiving concurrent azole, despite comparable everolimus doses delivered to each cohort. Fungal chest infection developed in 0/12 in the everolimus/azole cohort, and 2/19 in those without concurrent azole (P = 0.51). The correlation between superior response and drug exposure is consistent with another phase 1b study by Park et al. where everolimus was combined with conventional 7+3 chemotherapy in patients < 65 years of age at first relapse [15].
In our study, azacitidine and everolimus were given sequentially, similar to the sequential decitabine and rapamycin approach reported by Liesveld et al. [29], but in contrast to pulsed everolimus (up to 70 mg) on days 1 and 7 in combination with 7+3 chemotherapy [15]. Hypomethylating agents result in delayed demethylation, peaking 1-2 weeks after therapy. Sequential administration over 3 weeks was designed to maximise exposure of blasts to mTOR inhibition during the period of peak demethylation agent response. The schedule was also aimed at avoiding potential cell cycle inhibitory effects of mTOR inhibitors on azacitidine activity. mTORC1 plays an important role in mRNA translation and autophagy regulation, whereas mTORC2 regulates the phosphorylation of AKT on Ser 473 (pAKT-S473) [33]. mTORC1 targeting by rapalogs could lead to unintended feedback amplification of the PI3K/ AKT/mTOR pathway in AML, potentially subverting the clinical effect of targeting mTORC1 alone [4,5]. Several studies have demonstrated improved anti-leukemic effects in AML by targeting both mTORC1 and mTORC2 simultaneously using active site or catalytic inhibitors [34]. Interestingly, mTORC2 may be inhibited by high doses of rapamycin, leading to inactivation of AKT and ERK [6]. Everolimus has also been reported to suppress mTORC2 and the AKT and ERK pathways in AML at doses approaching 20 nM [7], which is the concentration range achieved by the everolimus/azole combination in  (Figure 2). It is therefore possible that the superior outcome observed with the everolimus/azole combination is due to the inhibition of mTORC2, resulting from the higher blood concentrations of everolimus consequent to CYP3A4 inhibition by antifungal azoles. Although limitation of available patient samples prevented an analysis of pAKT-S473 in AML blasts among patients receiving everolimus/azole combination, this should be the objective of future validation studies.
Reliable molecular biomarkers of response to mTOR inhibitors remain to be determined. Ecotropic viral integration site 1 (EVI1) has been shown to directly repress PTEN transcription, resulting in increased sensitivity to rapamycin in EVI1-related mouse leukemia [35]. Poor prognosis karyotypes including monosomy 7 and t(11q23) are characterized by EVI1 overexpression [36,37]. Although 2/2 patients with MLL-rearranged AML in the current study had a > 50% reduction in bone marrow blasts, further studies will be required to determine the clinical significance of this association. Three patients with refractory FLT3-ITD AML had > 80% bone marrow blast reductions with everolimus/azacitidine therapy. Two of these patients went on to have allogeneic stem cell transplantation but subsequently relapsed and died.
FLT3-ITD in AML is frequently found in combination with genomic lesions affecting epigenetic regulators [38], which may result in synergistic hypermethylation of genes, such as GATA2 [39]. Future studies exploring azacitidine in combination with rapalogs should therefore include analysis of FLT3-ITD and mutations affecting epigenetic modifiers. Interestingly, IDH1/2 mutation was considerably prevalent (15/35; 43%) in our predominantly relapsed and refractory study population, in contrast to the 16-21% observed in recently published large series of newly diagnosed AML [38,40].
Although the MTD of everolimus (in combination with azacitidine) was not defined in this study, three patients developed delayed and unexpected cardiac dysfunction that only became apparent following allogeneic stem cell transplantation. All three patients received high-dose cyclophosphamide and high-dose total body irradiation (TBI) conditioning. Two other patients in the study allografted with either busulphan/cyclophosphamide or lowdose TBI conditioning regimen did not develop subsequent cardiac compromise. The pre-transplant cardiac exposure to mTOR inhibitors followed by TBI is uncommon, so these three cardiac events should promote caution for future studies among patients likely to receive potent Oncotarget 52276 www.impactjournals.com/oncotarget mTOR inhibitors in combination with azacitidine prior to irradiation therapy involving the heart. Disruption of cardiac mTORC1 is associated with a significant reduction in cardiomyocytes and cardiac dilation in mice, suggesting an important role of mTOR in cardiomyocyte survival during development [41,42]. mTOR inhibition may decrease the apoptotic threshold and enhance autophagy, leading to hypersensitivity of tissues to radiotherapy in patients treated with a mTOR inhibitor [43]. In our study, this scenario may have been further exacerbated by the frequent exposure of our patient population to prior anthracyclines.
In summary, the combination of everolimus and azacitidine is a safe and well-tolerated treatment for relapsed and refractory AML. Future studies should explore the potentially beneficial pharmacokinetic interaction between everolimus and antifungal azoles in delivering higher everolimus concentrations in combination with azacitidine in vivo, and to examine the impact this might have on the regulation of mTORC1 and mTORC2-mediated pathways in AML blasts. Further studies are also warranted to define the potential for FLT3-ITD, MLL rearrangement and other genomic markers to predict response to this therapy. In addition, our results also support the clinical investigation of hypomethylating agents in combination with newer generation dual mTORC1/mTORC2 inhibitors [44,45] or compound PI3K/mTOR inhibitors [46,47] in patients with AML.

Patients
Eligible patients were relapsed or refractory AML who had received up to two prior lines of intensive therapy, and untreated AML in patients aged ≥ 60 years unfit for intensive chemotherapy. A full list of eligibility criteria is detailed in Supplementary Table S1.

Study design and treatment
This open-label, phase Ib/II study was conducted at the Alfred Hospital, approved by an independent Human Research and Ethics Committee and registered with the Australian and New Zealand Clinical Trials Registry (ACTRN12610001031055). Azacitidine 75 mg/m 2 was injected subcutaneously on days 1-5 and 8-9 [48]. The everolimus dose was escalated in a standard 3 × 3 study design to determine the MTD. The cohort doses of everolimus examined were 2.5, 5 and 10 mg with no intra-patient dose escalation permitted; the dosage was capped at 10 mg as this was the established dose used in the treatment of solid cancers [49]. After establishment of the MTD, a dose expansion group was included to better estimate the tolerability of this dose and to provide preliminary assessment of efficacy. Due to the CYP3A4 interaction between everolimus and antifungal azoles, azole administration was excluded from the first cycle of therapy, but permitted (voriconazole or posaconazole) in subsequent cycles with the dose of everolimus limited to 2.5 mg. Everolimus was administered orally daily on days 5-21 of each 28-day cycle. Hydroxyurea was permitted up to 48 hours preceding commencement of study medication, but no other anti-leukemic therapies were permitted. Patients were permitted to receive transfusion support in accordance with established guidelines along with granulocyte colony stimulating factor (G-CSF) and prophylactic antimicrobials at physician discretion. Patients were continued on therapy until AML relapse or progression, unacceptable toxicity or completion of study related procedures.

Objectives and endpoints
The main objective was to examine the safety and tolerability of everolimus in combination with azacitidine in relapsed/refractory AML. Secondary objectives were to provide preliminary data regarding clinical outcomes and the correlation between clinical responses and exploratory biomarkers of outcome.

Assessments Toxicity
Adverse events were graded according to the National Cancer Institute CTCAE version 4. DLT was defined as treatment-related Grade 3-4 non-hematologic toxicity attributable to either study drug, or persistent grade 4 neutropenia or thrombocytopenia by day 42 of first cycle with bone marrow cellularity < 10% not related to persistent leukemia (bone marrow blasts < 5%). Patients were evaluable for toxicity if ≥ 85% of cycle 1 therapy was completed, or if the early withdrawal was toxicityrelated. Patients with early withdrawal due to disease progression were excluded from the safety analysis, but included for the efficacy analysis.

Efficiancy
Response criteria were graded as per the IWG criteria [50]. OS was calculated from first day of study therapy to date of death. RFS was calculated from date of confirmed response, inclusive of CR, CRi, MLFS and PR, to the date of confirmed relapse or loss of PR. Time-toevent analyses were censored for patients who underwent transplant or lost to follow-up.

Correlative assessments
Everolimus trough blood levels were measured by liquid chromatography mass spectrometry ~day 19, two weeks after the commencement of the mTOR inhibitor. Pharmacodynamic assessment of everolimus activity was Oncotarget 52277 www.impactjournals.com/oncotarget by PIA of P-p70S6K levels by Western blot using the OCI-AML3 cell line, with plasma taken on days 5 and 19. Briefly, frozen plasma samples were thawed, clarified by centrifugation, and washed twice with ice-cold phosphatebuffered saline prior to cell lysis. After immunoblotting for phosphorylated p70S6K, densitometric analysis was performed on the bands, and the PIA for each plasma sample was calculated by expressing the density of its corresponding band as a percentage of the density of the baseline band (which was arbitrarily set at 100%).

Statistical analysis
Statistical analysis was performed using Graphpad Prism version 6 (GraphPad Software, La Jolla, California, USA) and the R statistical software version 3.2.3 (R foundation for statistical computing, Vienna, Austria). Survival curves were calculated by the Kaplan-Meier method and the log-rank test was used to compare patient groups. A χ 2 test and Fisher's exact test was used to compare correlative response outcomes.