The PRIME trial: An investigator‐initiated, multicentre, phase II study of the poly(ADP‐ribose) polymerase inhibitor olaparib in isocitrate dehydrogenase (IDH)‐mutated relapsed/refractory acute myeloid leukaemia and myelodysplastic syndromes

Isocitrate dehydrogenase (IDH)1/2 mutations in acute myeloid leukaemia (AML) and myelodysplastic syndromes (MDS) confer a neomorphic activity on the encoded enzymes, such that they catalyse the conversion of isocitrate to the oncometabolite (R)‐2‐hydroxyglutarate (2‐HG), instead of α‐ketoglutarate (αKG), with pleiotropic detrimental effects including inhibition of αKG‐dependent enzymes that are important for normal deoxyribonucleic acid (DNA) and histone methylation and cellular differentiation. ¹ Many patients with IDH1/2‐mutated MDS/AML will receive IDH1/2 inhibitors, which are approved in the relapsed/refractory (R/R) setting, but confer modest benefit and simply block the neomorphic function of the mutant proteins, without evidence of impact on the mutant IDH1/2 clone. ² , ³ Collectively, these data suggest that enzymatic inhibition, while clinically active, may be insufficient for sustained disease control, underscoring the need to better define the biology of IDH1/2‐mutated disease to enable therapies that more directly target the pathogenic mutant clone.

We previously made the novel discovery that IDH1/2 mutations induce a homologous recombination (HR) defect which renders tumour cells exquisitely sensitive to poly(adenosine diphosphate‐ribose) polymerase (PARP) inhibitors. ⁴ , ⁵ As 2‐HG accumulates in IDH1/2‐mutant cells, HR disruption causes cells to rely on alternate low fidelity DNA repair pathways, thus hastening genomic instability and cell death. Remarkably, this Breast Cancer susceptibility genes (‘BRCAness’) phenotype can be reversed by treatment with small‐molecule inhibitors of mutant IDH1 and, conversely, it can be entirely recapitulated by treatment with 2‐HG alone in cells with wild‐type IDH1/2. ⁴ IDH mutation‐dependent PARP inhibitor sensitivity was demonstrated in a range of clinically relevant models, including patient derived AML cells and genetically matched tumour xenografts in vivo; furthermore, 2‐HG production is restored at the time of failure of IDH1/2 inhibitor exposure, thus also ‘restoring’ a HR defect and PARP sensitivity. ⁴ , ⁶ , ⁷ Similar synthetic lethal interaction between IDH1/2 mutations and PARP inhibitors were also reported by other laboratories, which further strengthened the validity of this discovery. ⁷ Collectively, these findings provided an alternative to the current therapeutic strategy to block IDH1/2‐mutant function, and offered a novel approach to treat IDH1/2‐mutated tumours including MDS/AML by exploiting IDH1/2‐mutant DNA damage repair deficiency.

We thus conducted an investigator‐initiated, National Cancer Institute‐Cancer Therapy Evaluation Program (NCI‐CTEP)‐sponsored, multicentre, phase 2 trial of the PARP inhibitor olaparib in IDH1/2‐mutated R/R AML/MDS or the PARP Inhibition in IDH Mutant Effectiveness (PRIME) trial (NCI 10264, NCT03953898).

Adult patients with IDH1/2‐mutated R/R AML/MDS were enrolled into one of four cohorts depending on disease state (i.e. AML vs. MDS) and prior exposure to an IDH1/2 inhibitor (i.e. naïve vs. exposed) with latter to account for the possibility of IDH inhibitor exposure and consequent impact on 2‐HG levels affecting response. Patients received olaparib 300 mg twice a day (the maximum tolerated dose per the Investigator's Brochure), continuously on 28‐day cycles until progression, unacceptable toxicity or consent withdrawal. The primary objective was to determine the rate of composite complete response (CR) + CR with incomplete haematological response (CRi) + CR with partial haematological response (CRh) (CR + CRi + CRh) to olaparib, after 6 cycles of treatment, according to the AML International Working Group (IWG) 2003 criteria for patients with AML and according to the MDS IWG 2006 criteria for patients with MDS. ⁸ , ⁹ Only patients completing ≥1 cycle of therapy were considered evaluable for response. A total of 94 patients (n = 39 with IDH inhibitor naïve AML, n = 39 with IDH inhibitor naïve MDS, n = 14 with IDH inhibitor exposed AML, n = 12 with IDH inhibitor exposed MDS; all of which with a potential 15% drop‐out rate) were anticipated to be enrolled over 2 years; a futility analysis was planned to be performed using a Simon 2‐stage design assuming a 30% and 5% null hypothesis CR rate and 50% and 35% alternative CR rate for IDH inhibitor naïve and exposed patients respectively. Toxicities were defined according to the Common Terminology Criteria for Adverse Events Version 5 (CTCAE v5). Haematological toxicities were defined by the absence of the recovery of a cytopenia to baseline levels within 42 days of a cycle start, in the absence of signs of active disease in blood and bone marrow. Correlative studies included the evaluation of a relationship between treatment response and plasma or bone marrow 2‐HG levels and IDH1/2 variant allele frequency (VAF), biomarkers identified via whole exome sequencing (WES), single‐cell DNA sequencing (sc‐DNAseq), the response to therapy based on the different IDH1/2‐mutant genotypes and other genomic alterations by which treatment may be assigned, and to evaluate drug response and potential drug combinations in xenotransplantation models. Eligibility criteria, secondary objectives and correlative methods are further described in the Supporting Information.

This study was largely conducted during the coronavirus disease (COVID)‐19 pandemic and was administratively closed by sponsor prior to the futility analyses due to low accrual. A total of 14 patients were enrolled with 12 eventually starting study treatment (2 patients died prior to study treatment). Median patient age was 76 (range [R]: 42–86) years with most patients being white (n = 10, 71%) and having an Eastern Cooperative Oncology Group (ECOG) Performance Status of 0–1 (n = 12, 86%). Half of patients were male (n = 7, 50%) and half female (n = 7, 50%). Most patients had a diagnosis of AML (n = 13, 93%). Median total white blood cell count was 2.7 (R: 0.5–25.1), median absolute neutrophil count was 0.8 (R: 0–10.2), median haemoglobin was 9.1 (R: 7.3–12.6) g/dL, and median platelet count was 54 (R: 8–146). Median peripheral blasts were 8% (R: 0%–81%) and 33% (R: 7–79.5) in blood and marrow respectively. IDH1, IDH2 and dual IDH1/2 mutations were detected in 3 (21%), 10 (71%) and 1 (7%) patient respectively. Other detected co‐mutations included DNA Methyltransferase 3A (45%), runt‐related transcription factor (RUNX1) (41%), serine/arginine‐rich splicing factor (SRSF2) (38%), additional sex combs like (ASXL1) (28%) and fms‐Like tyrosine kinase3 (17%) (Figure 1A). Most patients (n = 8, 57%) had prior IDH inhibitor exposure. Baseline demographic and clinical characteristics of enrolled patients are demonstrated in Table S1.

Median duration of treatment was 1.79 months with a median number of 1 cycles received (R; 0–6). The most common adverse events, irrespective of attribution, were haematological in nature and serious adverse events occurred in 9 (75%) patients and those occurring in >10% of patients included febrile neutropenia (n = 3, 25%), sepsis (n = 4, 33%), lung infection (n = 4, 33%), hypoxia (n = 2, 17%) and respiratory failure (n = 2, 17%) (Table S2).

The efficacy of olaparib in this study population was limited. Five patients did not complete a full cycle of therapy (two patients died after enrolment but before start of therapy, one patient withdrew consent 13 days into cycle 1 and opted for hospice, one patient died 13 days into cycle 1 due to acute hypoxic respiratory failure attributed to disease and one patient died 18 days into cycle 1 due to severe acute respiratory syndrome coronavirus 2 [SARS‐CoV2] infection). Among the nine patients who were evaluable for response, no patient achieved a protocol‐defined response and no decreases in marrow blasts were observed; four patients had stable disease and five patients had progressive disease. Median progression‐free survival was 2.3 (range: 0.4–5.6) months and median overall survival was 3.4 (range: 0.4–35.3) months.

Oncogenic/likely oncogenic VAFs were relatively static amongst the 10 patients with serial samples on therapy (Figure 1B); however, we observed VAF decreases of 25%–50% among three of the four patients with spliceosome‐mutated disease (i.e. SRSF2, SF3B1 = splicing factor 3b subunit 1 [SF3B1], ZRSR2 = zinc finger, RNA‐binding motif and serine/arginine‐rich 2 [ZRSR2]) (Figure 1C).

Total levels of 2‐HG as well as levels of L‐2‐HG and D‐2‐HG were measured in plasma from both peripheral blood and bone marrow. Measured levels of total 2‐HG were higher in bone marrow versus blood and were predominantly D‐2‐HG (Figure S1A). Total 2‐HG levels tended to increase at later time points in both blood and bone marrow, although statistical comparisons were hampered by sample availability to be tested at later time points (Figure S1B). Low 2‐HG levels raise the concern that a BRCAness phenotype may not have been present in those patients.

sc‐DNAseq results were generally consistent with what was seen in the WES data. Additionally, from these data we can infer the order of mutational acquisition. Figure 1D (also captured in Figure 1C) demonstrates two patients with IDH2‐mutated disease, which was also co‐mutated with RUNX1 and SRSF2. Based on these data, patients 4 and 7 acquired the IDH2 mutation first, followed by RUNX1 and then SRSF2, with only patient 7 showing a decrease in mutant clone burden by the end of treatment. The one caveat to this analysis is that for patients with SRSF2 mutations, only a fraction of the cells sequenced could be included in the analysis because of the inefficiency of probing/sequencing the SRSF2 locus because of the high guanine (G) and cytosine (C) (GC) content.

Particular mutational signatures associated with DNA repair were derived from WES and identified in certain patients. We identified patients with Catalogue Of Somatic Mutations In Cancer (COSMIC) v2 signature 3, which is associated with an increased number of large insertions and deletions with overlapping microhomology at breakpoint junctions, thus associated with failure of DNA double‐strand break‐repair by HR; it has also been associated with olaparib response in solid tumours. ¹⁰ , ¹¹ COSMIC v2 signatures 6 and 15, often co‐occurring, were identified and are associated with an increase of small insertions and deletions at mono/polynucleotide repeats, with consequent defective DNA mismatch repair, and are found in microsatellite unstable tumours. ¹¹ , ¹² We found that patients with COSMIC v2 signature 3 had a higher rate of stable disease when compared with progressive disease while on therapy (Figure 1A).

We report the first prospective clinical trial of PARP inhibition for patients with MDS/AML and a candidate biomarker predictive of response. Although we have demonstrated a preclinical susceptibility of IDH1/2‐mutated disease to the intervention, our study showed that olaparib, despite being reasonably tolerated given the study population, demonstrated no appreciable clinical benefit amongst evaluable patients. Low accrual prompting an administrative study closure and an appreciable rate of patients being too ill to embark on study treatment limit firm conclusions about clinical efficacy. Future attempts to leverage the consistent pre‐clinical data for the ‘BRCAness’ phenotype of IDH1/2‐mutated MDS/AML and resultant synthetic lethality by PARP inhibition may need to include further biomarker optimization such as evidence of homology‐directed repair (HDR) defect. Improved PARP inhibitors, namely with PARP1‐selective inhibition to enhance efficacy and moderate any drug‐related myelosuppression or combination strategies could be considered.

AUTHOR CONTRIBUTIONS

Namrata Chandhok: Conceptualization; investigation; writing – review and editing. Steven D. Gore: Writing – review and editing; project administration; supervision. Rory M. Shallis: Conceptualization; investigation; writing – original draft; formal analysis; methodology. Richard F. Little: Writing – review and editing; supervision; project administration. Jan Phillip Bewersdorf: Investigation; writing – review and editing. Diane Durecki: Project administration; writing – review and editing. Thomas Prebet: Conceptualization; investigation; funding acquisition; writing – review and editing; methodology. Jennifer VanOudenhove: Formal analysis; data curation; writing – review and editing; methodology. Deepa Jeyakumar: Investigation; writing – review and editing. Scott Boerner: Project administration; writing – review and editing. Justin Watts: Investigation; writing – review and editing. Jing Li: Methodology; formal analysis; writing – review and editing; data curation. Daniel R. Reed: Investigation; writing – review and editing. Ranjit S. Bindra: Conceptualization; funding acquisition; writing – review and editing; methodology. Xun Bao: Data curation; formal analysis; writing – review and editing. Sayak Ghatak: Data curation; formal analysis; writing – review and editing. Amer M. Zeidan: Investigation; writing – review and editing; supervision. Stephanie Halene: Conceptualization; investigation; funding acquisition; writing – review and editing; methodology; formal analysis. Li Chen: Data curation; formal analysis; writing – review and editing.

FUNDING INFORMATION

The study was supported by the National Cancer Institute of the National Institutes of Health under the NCI‐Cancer Therapy Evaluation Program (NCI‐CTEP) and UM1CA186689. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Mention of trade names, commercial products, or organizations does not imply endorsement by the United States government. SH received funding from National Institutes of Health, NCI grant 1R01CA266604, Blood Cancer United and the Frederick A. DeLuca Foundation. RSB received funding from National Institutes of Health, NCI grant 1R01CA266604.

CONFLICT OF INTEREST STATEMENT

RMS has served in a consulting/advisory role and/or received honoraria from Bristol Myers Squibb, Cardinal Health, Geron, Gilead Sciences Inc., Kura Oncology, Rigel, Servier, Syndax Pharmaceuticals and TScan Therapeutics, and has participated in a Steering Committee for Servier. AMZ participated in advisory boards, consulted, participated in clinical trial committees and/or received honoraria from AbbVie; Akesobio; Agios; Amgen; AstraZeneca/Alexion, Astellas; BioCryst; BioThera, BeiOne; Boehringer‐Ingelheim; Celgene/BMS; Chiesi/Cornerstone biopharma; Daiichi Sankyo; Dr. Reddy; Epizyme; Faron; Fibrogen/Kyntra Bio; GSK; Glycomimetics; Genentech/Roche; Gilead; Geron; Innocare, Janssen/J&J; Jasper; Karyopharm; Kyowa Kirin; Keros; Kura; Novartis; Notable; Orum; Otsuka; Oncoverity, Pfizer; Puretech/Gallop Oncology, Regeneron; Rigel; Seattle Genetics; Shattuck labs; Schrodinger; Sumitomo; Syros; Syndax; Servier; Takeda; Treadwell; Taiho; Vincerx; and Zentalis. SH receives research funding from Astra Zeneca and STORM Therapeutics. All other authors report no relevant conflicts of interest.

ETHICS STATEMENT

This study was performed in line with the principles of the Declaration of Helsinki. The study protocol was approved by the Ethics Committees at all participating sites.

PATIENT CONSENT STATEMENT

Informed consent was obtained from all individual participants included in the study.

CLINICAL TRIAL REGISTRATION

The study was registered on clinicaltrials.gov under the registration number NCT03953898.

Supporting information

DATA AVAILABILITY STATEMENT

Data are available upon request.