Combination romidepsin and azacitidine therapy is well tolerated and clinically active in adults with high‐risk acute myeloid leukaemia ineligible for intensive chemotherapy

Azacitidine (AZA) is important in the management of patients with acute myeloid leukaemia (AML) who are ineligible for intensive chemotherapy. Romidepsin (ROM) is a histone deacetylase inhibitor which synergises with AZA in vitro. The ROMAZA trial established the maximum tolerated dose (MTD) of combined ROM/AZA therapy in patients with AML, as ROM 12 mg/m2 on Days 8 and 15, with AZA 75 mg/m2 administered for 7/28 day cycle. Nine of the 38 (23·7%) patients treated at the MTD were classified as responders by Cycle 6 (best response: complete remission [CR]/incomplete CR n = 7, partial response n = 2). Correlative next‐generation sequencing studies demonstrated important insights into therapy resistance.


Introduction
Treatment options for patients with acute myeloid leukaemia (AML) who are ineligible for intensive chemotherapy are limited and, as a consequence, patient outcomes remain poor. 1 A significant advance has been the use of epigenetic therapies in patients with AML and high risk myelodysplastic syndromes (MDS), and azacitidine remains the backbone of regimens involving novel agents, including venetoclax. 2,3 However, in patients with relapsed/refractory AML, the activity of azacitidine monotherapy is reduced 4 and strategies which increase the activity of azacitidine are required. In vitro studies demonstrated evidence of synergistic antileukaemic activity between azacitidine and histone deacetylase inhibitors (HDACi). 5 Emerging clinical data suggest that azacitidine used with the HDACi, romidepsin, may have significant anti-tumour activity and synergy. 6 Romidepsin belongs to the cyclic peptide family 7 and inhibits different pathways to previously investigated HDACi. 8 Romidepsin has been shown to be clinically potent and well-tolerated as monotherapy in patients with peripheral T-cell lymphoma, 9 but has not been investigated in AML. Therefore, in this phase I/II ROMAZA trial, we examined the safety and efficacy of romidepsin and azacitidine in patients with newly diagnosed, relapsed or refractory AML who are ineligible for conventional chemotherapy.

Study conduct
Review boards of participating institutions approved the study protocol, which was conducted according to the Declaration of Helsinki and Good Clinical Practice Guidelines of the International Conference on Harmonization (EudraACT No: 2011-005023-40 and ISCRTN:69211255).

Patients and investigations
Patients with newly diagnosed, relapsed or refractory AML as defined by the World Health Organisation classification and deemed ineligible for intensive chemotherapy on the grounds of age or co-morbidities, were eligible for trial inclusion. Patients with prior treatment with demethylating agents were ineligible.
Treatment. The maximum tolerated dose of romidepsin in combination with azacitidine was determined using an escalating/de-escalating 3 + 3 cohort design. Patients were recruited in planned cohort sizes of three up to a sample size of 18, with an additional 35 patients recruited to an expansion cohort. ( Figure S1, Data S1). Patients initially received up to six cycles of combination therapy and continued if benefiting clinically.
Adverse event reporting. Tolerability and safety were assessed according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. A dose-limiting toxicity (DLT) was defined as a clinically significant grade 3 or 4 non-haematological toxicity, excluding nausea (manageable with anti-emetics) and fatigue (persisting for more than seven days), which occurred within the DLT monitoring period and was related to the trial medication.
Endpoints. The primary endpoint for the escalation cohort was to determine the maximum tolerated dose (MTD) of the romidepsin/azacitidine combination ( Figure S3). The objective for the expansion cohort was to provide preliminary efficacy and safety data of the combination at the MTD with response defined as acquisition of complete remission (CR), incomplete CR (CRi) or partial response (PR) as the primary endpoint. This was assessed at the end of three and six cycles of treatment according to modified Cheson criteria. 10 Secondary endpoints were tolerability and safety of the combination and overall survival for patients treated at the MTD. Further details of statistical and mutational analysis using nextgeneration sequencing (NGS) are available in Data S1.

Baseline characteristics
Forty-eight patients were recruited from October 2013 to October 2017 ( Figure S1). Thirteen patients were recruited to the dose-finding cohorts, which included four replacements for those who did not complete the DLT monitoring period. A subsequent 35 patients were recruited to the expansion cohort of patients treated at the MTD (cohort 2) ( Figure S2). Baseline patient characteristics are listed in Table I, 36/48 (75%) had relapsed/refractory disease.

Treatment administration
Patients received a median of 2Á5 cycles of treatment across the whole trial. Twenty out of 38 patients treated at the MTD discontinued their treatment before the third cycle of treatment and were not assessed for their response. Only 4/ 20 discontinued treatment due to treatment toxicity.

MTD assessment
In the absence of DLTs in the first cohort (10 mg/m 2 romidepsin on Day 8 and 15, and 75 mg/m 2 azacitidine on Day 1-9 (in a 5-2-2 combination)) and second cohort (12 mg/m 2 romidepsin on Day 8 and 15, and 75 mg/m 2 azacitidine on Day 1-9) ( Figure S1), a third cohort of four patients was opened at 12 mg/m 2 romidepsin on Day 8, 15 and 22 and 75 mg/m 2 azacitidine on Day 1-9. A total of four patients were recruited as one patient was unevaluable. All patients experienced at least one serious adverse event (SAE). Whilst there were no reported DLTs at this dose level, a proportion of patients suffered from nausea, vomiting and associated weight loss. As a result, the dose was de-escalated to that of Cohort 2 (12 mg/m 2 romidepsin on Day 8 and 15, and 75 mg/m 2 azacitidine on Day 1-9 (in a 5-2-2 combination)) which was used as the MTD.

Safety and tolerability
The combination therapy was well tolerated: five nonhaematological treatment-related adverse events of Grade 3 or higher affecting at least 10% of all trial patients were reported in the study across the 38 patients treated at the MTD (Tables SI and SII).

Response and survival
Using an intention-to-treat (ITT) approach, nine of the 38 (23Á7%) patients treated at the MTD were classified as responders by Cycle 6 (best response: CR/CRi n = 7, PR n = 2) ( Figure S4) (a pre-specified success criteria by A'herns design required 10/38 responders). Of the 11 patients who were previously untreated, four (36Á4%) had a response. In the 27 patients with relapsed/refractory disease there were five responses (18Á5%). Of the 29 non-responders, 20 discontinued treatment prior to end of Cycle 3 and consequently did not undergo response assessment.
At the date of data cut-off (21st May 2020), one patient in Cohort 2 remained on treatment and there were 45/48 deaths. The median overall survival of the 38 patients treated at the MTD was 6Á21 months (95% confidence interval [CI]: 3Á71, 8Á87), whilst the median overall survival for the 9/38 patients who responded from Cohort 2 was 15Á3 months (95% CI: 4Á3, 29Á1).

Mutational analysis by NGS at baseline and sequentially post-treatment
We obtained diagnostic mutational status across 97 genes in 35 patients using NGS. The median number of variants per patient was six ( Figure S5). The most common mutations were in RUNX1 and FLT3, reflecting predominance of adverse risk genetic mutations, 11,12 consistent with our population disease history.

Serial clonal structures in responding and nonresponding patients
We next examined the pattern of clonal mutational architecture at different sequential time points in response to azacitidine and romidepsin, and related this to their clinical responses. In two patients achieving a CR (Fig 1A), the dominant clone present at commencement of therapy was suppressed, suggesting clonal selection in responding patients. For example, in TNO 7 mutations, pre-treatment include STAG2, NRAS, RUNX1 and SRSF2. However, NGS analysis of the sample after Cycle 12, at a time when the patient remained in CR, a subdominant clonepresent at diagnosis consisting of ATM/KIT/CUX1 mutations becomes the dominant clone.
In contrast, recurrent myeloid mutations tended to persist, at similar VAFs, in patients who achieved a CRi (n = 3,   Figure S6B). This could be consistent with the persistence of pre-leukaemic or leukaemic clones. If these clones are leukaemic then it would suggest that the mechanism of action of romidepsin and azacitidine is to promote differentiation.
In patients who responded and subsequently have a documented relapse (Fig 1C), complex patterns of clonal evolution were seen. In TNO22, a FLT3-ITD clone decreased in size when the patient was in PR and remained low at relapse. In contrast, cells with RUNX1/CBL mutations were selected for at PR and the VAF of these mutations increased at relapseconsistent with resistance to azacitidine and romidepsin. In contrast, in patient TNO32 the VAF of mutations in TET2, U2AF1 and NRAS were reduced at CR, only to increase at relapse. This is consistent with leukaemic cells showing initial therapy sensitivity followed by therapy resistance, possibly through epigenetic mechanisms. Finally, we observed that in patients with resistant disease (n = 5, Fig 1D; Figure S6A) mutant clones persist or expand.

Discussion
Responses in the relapsed/refractory setting are challenging to obtain and this study exemplifies the difficulty in treating this cohort of less fit patients with rapid kinetic disease. 13 Of particular note, 17 (35%) had previously received an allogeneic stem cell transplant.
The tolerability of the ROM/AZA regimen is notable because previous experiences of HDACi were associated with increased toxicity. Romidepsin belongs to a different class of HDACi, as previously investigated in AML, 7,14 and we demonstrate that this combination can be safely delivered in the outpatient setting.
In order to obtain insights into the mechanism of this novel combination we monitored the clonal architecture of the disease through the serial NGS detection of mutations. The prevalence of TP53 mutations appear to reflect the heavily pre-treated nature of this cohort of patients, and in patients who subsequently have progressive disease, these mutant clones expand, in keeping with previous results of patients treated with decitabine/panobinostat. 15 In summary, this study established a MTD for combined ROM/AZA therapy that is safe and clinically active within adults with relapsed AML. Further studies will be required to compare the clinical activity of ROM/AZA directly to a comparator treatment arm.
(MC_UU_12009/11). The Centre for Haematology at Imperial College London receive funding from the NIHR BRC.

Author contributions
JL, MM, RB, AJ, PV and CC analysed the data. LH, RB, AJ and SF were involved in the management of the trial. ET, JL, AP, MD, PV and CC recruited patients to the study. CC designed the trial. All authors had access to the primary data and contributed to manuscript development. All authors assumed responsibility for executing the study according to the protocol and statistical analysis plan, completeness and integrity of the data, and the decision to submit the manuscript.

Disclosures
PV is on the advisory boards for BMS, Daiichi Sankyo, Astellas, AbbVie, Pfizer, Jazz; speaker bureau for AbbVie, Novartis, Jazz, Daiichi Sankyo, BMS, the Takeda scientific advisory board for Auron, Oxford Biomedica and receives research support from CD47 Inc, BMS, Novartis, GSK. CC has received honoraria from Celgene, Daichi-Sankyo, Novartis and Pfizer as well as research funding from Celgene. JL has received travel funding from Novartis and Daichi-Sankyo, and honoraria from Pfizer, Janssen and Amgen. JP has received travel funding and honoraria from Daichi-Sankyo and Jazz.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Data S1. Supplementary Methods. Table SI. Treatment-related adverse events. Table SII. Treatment-related serious adverse events.