In Vivo Anti-Tumor Effect of PARP Inhibition in IDH1/2 Mutant MDS/AML Resistant to Targeted Inhibitors of Mutant IDH1/2

Abstract

Treatment options for patients with relapsed/ refractory acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) are scarce. Recurring mutations, such as mutations in isocitrate dehydrogenase-1 and −2 (IDH1/2) are found in subsets of AML and MDS, are therapeutically targeted by mutant enzyme-specific small molecule inhibitors (IDH^mi). IDH mutations induce diverse metabolic and epigenetic changes that drive malignant transformation. IDH^mi alone are not curative and resistance commonly develops, underscoring the importance of alternate therapeutic options. We were first to report that IDH1/2 mutations induce a homologous recombination (HR) defect which confers sensitivity to poly (ADP)-ribose polymerase inhibitors (PARPi). Here, we show that the PARPi olaparib is effective against primary patient-derived IDH1/2-mutant AML/ MDS xeno-grafts (PDXs). Olaparib efficiently reduced overall engraftment and leukemia-initiating cell frequency as evident in serial transplantation assays in IDH1/2-mutant but not -wildtype AML/MDS PDXs. Importantly, we show that olaparib is effective in both IDH^mi-naïve and -resistant AML PDXs, critical given the high relapse and refractoriness rates to IDH^mi. Our pre-clinical studies provide a strong rationale for the translation of PARP inhibition to patients with IDH1/2-mutant AML/ MDS, providing an additional line of therapy for patients who do not respond to or relapse after targeted mutant IDH inhibition.

Introduction

The isocitrate dehydrogenase-1 and −2 (IDH1/2) genes encode key metabolic enzymes in the Krebs cycle that catalyze the reversible conversion of isocitrate to alpha-ketoglutarate (αKG) ¹. IDH1/2 mutations are found in approximately 20% of AML and in 4 to 12% of MDS ^{2, 3}. Somatic missense mutations in IDH1/2 occur within the catalytic sites at arginine residues R132 for IDH1 (the cytoplasmic isoform) and R140 or R172 for IDH2 (the mitochondrial isoform) ^2–4. IDH1/2 enzymes function as homodimers, and when mutated, acquire the neomorphic function as heterodimers to catalyze the conversion of αKG to 2hydroxyglutarate (2HG) ¹. 2HG is an oncometabolite that has been shown to promote leukemogenesis via the competitive inhibition of αKG-dependent dioxygenases involved in the regulation of DNA and histone methylation and other functions ^5–8. Given their important role in tumor initiation and progression, mutant IDH1/2 enzymes represent an attractive therapeutic target leading to the development of selective small molecule inhibitors ⁹. Ivosidenib and enasidenib, inhibitors of mutant IDH1 and IDH2, respectively, have been approved in the treatment of adults with AML and MDS ^{10, 11}, however, when given as single agents are not curative, with complete response (CR) rates and median overall survival (OS) ranging between 20–30% and ~8–12 months, respectively ¹². IDH^mi induce cell differentiation ^{2, 13} with persistence of the mutant IDH1/2 clones even in patients who achieve CR ¹⁴. Resistance to IDH^mi is notable for either persistent or restored 2HG production ^15–18.

We previously reported that mutant IDH1/2-induced 2HG accumulation induces a HR defect that confers sensitivity to PARPi ¹⁹. Mechanistically, 2HG-induced inhibition of the lysine demethylase KDM4B results in aberrant hypermethylation of histone 3 lysine 9 (H3K9) at loci surrounding DNA breaks, masking the local H3K9 trimethylation signal essential for DNA double-strand break (DSB) repair via HR ²⁰. Tip60 and ATM fail to localize to DSBs, which results in impaired end-resection and recruitment of downstream HR factors. We identified similar oncogenic mechanisms secondary to TCA cycle gene mutations in fumarate hydratase and succinate dehydrogenase deficient cancers, highlighting the tumor agnostic nature of this “oncometabolite-induced BRCAness” and its potential as an exploitable vulnerability across different cancers ²¹ including leukemia and MDS. 2HG-induced HR deficiency renders cancer cells unable to resolve DSBs and sensitive to additional inhibition of single strand break (SSB) repair by PARPi ^{9, 19}. In vivo pre-clinical studies are necessary to bring novel concepts to the clinic and prove the validity of exploiting the 2HG-induced BRCAness defect with PARPi ^{19, 22, 23}.

Here we show that mutant IDH1 and IDH2 confer sensitivity to PARPi in vivo not only in IDH^mi naïve but also in resistant MDS/AML. Using genetically matched patient-derived IDH1/2 mutant and wildtype MDS/AML PDX models in cytokine humanized immunodeficient mice that recapitulate the mutational heterogeneity of AML/MDS ²⁴, we demonstrate that the PARP inhibitor olaparib significantly reduces MDS/AML engraftment and leukemia-initiating cell (LIC) frequency in IDH1/2 mutant but not wildtype MDS/AML. This sensitivity is maintained in IDH^mi resistant AML irrespective of co-occurring mutations. Our findings provide the rationale for the incorporation of PARPi into the treatment armamentarium for MDS/AML and the basis upon which to design combination therapies for relapsed/refractory and possibly treatment for naïve IDH1/2 mutant MDS/AML.

Materials and Methods

Primary patient samples

All patients’ PB and BM samples were obtained with donor’s written consent and accessed from the Yale hematology tissue bank. All human studies were approved by the Yale University Human Investigation Committee. Mononuclear cells were isolated using Ficoll-Paque separation (GE Healthcare, Cat#17–1440-03) and cryopreserved in fetal bovine serum (Gemini-Bioproducts, Cat#100–106) with 10% dimethyl sulfoxide.

Animal studies

All animal experiments were approved by and carried out in compliance with the Yale Institutional Animal Care and Use Committee protocols.

Primary murine AML model:

FLT3^ITD/WT IDH2^R140Q/WT mice were a gift from Dr. Ross Levine’s lab and have been previously described ²⁵. Mx1Cre⁺ Flt3^ITD/WT IDH2^R140Q/WT mice were treated every other day with Polyinosinic: polycytidylic acid [poly (I:C)] (Calbiochem, Cat#528906) via intra-peritoneal (IP) injection at a dose of 20μg/g starting 4–6 weeks after birth. 6–9 months later, donor mice were euthanized and BM cells harvested from femurs and tibias, and lineage depleted using anti-mouse lineage cell detection cocktail antibody and BD IMag Streptavidin Particles Plus (BD Biosciences, Cat# 557812, RRID: AB_10050580).

MISTRG patient derived xenotransplantation (PDX) studies:

MIS^h/hTRG mice (CSF1^h/h IL3/CSF2^h/h SIRPA^h/h THPO^h/h Rag2^−/− Il2rg^−/−) were bred to MIS^m/mTRG mice (CSF1^h/h IL3/CSF2^h/h Sirpa^m/m THPO^h/h Rag2^−/− Il2rg^−/−) to generate human cytokine homozygous and hSIRPA heterozygous mice (MIS^h/mTRG, labeled MISTRG throughout the study) ^{24, 26}. Mice were maintained on continuous treatment with enrofloxacin in the drinking water (0.27 mg/mL, Baytril, Bayer Healthcare). MISTRG mice have been deposited at the Jackson laboratory. IDH1/2 mutant and wild type AML patient samples were transplanted into newborn (3–5 days) MISTRG mice after irradiation twice with 150 cGy (XRAD 320, PXI X-Ray Systems) 4 hours apart. AML patient PB or BM mononuclear cells were incubated with anti-human CD3 antibody (clone Okt3, BioXCell, Cat# BE0001–2, RRID: AB_1107632) at 5 μg/100 μl for 30min prior to injection. Cells were injected intra-hepatically in a volume of 20 μL with a 22-gauge Hamilton needle and Hamilton syringe (Hamilton, Reno). Mice were analyzed at time points indicated via flowcytometry as previously described ²⁴. For secondary transplantation, cryopreserved BM cells from primary MISTRG recipient were thawed and depleted of murine cells via negative selection of murine CD45⁺ and Ter119⁺ cells by magnetic labeling with biotin-anti-mouse CD45 (clone 30-F11, Biolegend, Cat#103103, RRID: AB_312968) and biotin anti-mouse TER119 (clone TER-119, Biolegend, Cat#116204, RRID: AB_313705) and BD IMag Streptavidin Particles Plus (BD Biosciences, Cat# 557812, RRID: AB_10050580). Purified cells were incubated with anti-human CD3 antibody and injected intra-hepatically into MISTRG mice as described above. 8 to 12 weeks (AML) and 12 to 16 weeks (MDS) post transplantation, BM was aspirated from transplanted mice to quantitate human CD45⁺ engraftment levels by flow cytometry and to assign mice to treatment groups based on equal engraftment levels. For tertiary transplantation to determine LIC frequency, equal numbers of huCD45⁺ cells obtained from vehicle or drug treated mice were transplanted into newborn recipients, and analysis was performed at 16 weeks post transplantation.

Drug treatment

Olaparib (AZD2281) (Selleckchem, Cat#S1060) was dissolved in PBS containing 10% (w/v) 2-hydroxy-propyl-beta-cyclodextrin (Sigma) at a concentration of 20mg/ml. For in vivo studies, olaparib was administered IP at 100 mg/kg once daily. Enasidenib was purchased from LeadGen Labs, USA and dissolved in 0.5% methylcellulose and 0.2% Tween 80 in PBS at a concentration of 4 mg/ml and administered via oral gavage at 40 mg/kg once daily. 2-HG was measured in plasma before and after treatment with vehicle, enasidenib, or olaparib in triplicates with the D-2-Hydroxyglutarate (D2HG) assay kit (Biovision, Cat#K970–100, Sigma, Cat# MAK320–1KT) according to manufacturer’s protocol.

Flow cytometric analysis of human cell engraftment in MISTRG mice

Engraftment of human CD45⁺ cells and their subsets were determined by flow cytometry. In brief, cells were isolated from engrafted mice, blocked with human BD Fc block antibody (BD Biosciences, Cat#564219, RRID: AB_2728082) and mouse BD Fc block (2.4G2, BD Pharmingen, Cat#553141, RRID: AB_394656), and stained with combinations of antibodies purchased from Biolegend: Hematopoietic Stem and Progenitor Cell (HSPC) panel: APC/Cy7 mCD45 (30-F11, 1:300, Cat#103116, RRID: AB_312981), APC/Cy7 mTer119 (Ter-119, 1:300, Cat#116223, RRID: AB_2137788), BV510 hCD45 (HI30, 1:100, Cat#304036, RRID: AB_2561940), BV421 huCD38 (HIT2, 1:100, Cat#303526, RRID: AB_10983072), PE huCD34 (561, 1:100, Cat#343606, RRID:AB_1732008), PE/Cy7 huCD10 (HI10a, 1:100, Cat#312214, RRID: AB_2146548). Human cell engraftment panel: APC/Cy7 mCD45 (30-F11, 1:300), APC/Cy7 mTer119 (Ter-119, 1:300), BV510 hCD45 (HI30, 1:100), FITC huCD3 (OKT3, 1:100, Cat#317306, RRID: AB_571907), PE/Cy7 huCD19 (HIB19, 1:100, Cat#302216, RRID: AB_314246), APC huCD33 (WM53, 1:100, Cat#983902, RRID: AB_2810824), PE huCD34 (561, 1:100). Data were acquired with FACSDiva on an LSR Fortessa (BD Biosciences) equipped with 5 lasers and analyzed with FlowJo V10 software.

Histologic analysis

Tissues were fixed in BBC Biochemical B-PLUS FIX Fixative solution (Thermo Fisher Scientific, Cat#NC9496646) and embedded in paraffin. Femurs were decalcified with Formic Acid Bone Decalcifier (Decal Chemical, NY, USA). Paraffin blocks were sectioned at 4 μm and stained with hematoxylin and eosin (H&E) and antigen-specific antibodies by the Yale Clinical Pathology and Yale Pathology Tissue Services. Antibodies: human CD45, Leucocyte Common Antigen, (clone PD7/26+2B11, Dako, Cat#GA75161–2). Images were acquired using a Nikon Eclipse 80i microscope.

DNA cloning and cell culture

Site-directed mutagenesis to introduce the IDH2^Q316E and IDH2^I319M mutations was performed using the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies, Cat#200521) with primers (IDH2^Q316E Forward: CTATGACGGAGATGTGGAGTCAGACATCCTGGC; IDH2^Q316E Reverse: GATACTGCCTCTACACCTCAGTCTGTAGGACCG; IDH2^I319M Forward: AGATGTGCAGTCAGACATGCTGGCCCAGG; IDH2^I319M Reverse: CCTGGGCCAGCATGTCTGACTGCACATCT). Wild type (WT) and mutant IDH2 were cloned into the LeGO-iC plasmid, a gift from Boris Fehse (Addgene plasmid, #27462) and verified by Sanger sequencing. Lentiviral particles were produced by co-transfection with transgene, VSVG and psPAX2 (gifts from Tannishtha Reya (Addgene plasmid # 14888 and #12260) in HEK293T cells (ATCC Cat# CRL-3216) using polyethylenimine transfection reagent (Polysciences Cat# 23966–2) followed by spin-concentration (Beckman, 27,000xg, 90min). Lineage depleted murine Flt3^ITDIDH2^R140Q BM cells were pre-cultured with 8 μM Cyclosporine H (Sigma, SML1575) for 4 hours and then infected with viral particles via spinfection (1000xg at 25C for 1 hour) in StemSpan™ medium (Stem Cell Technologies) with addition of 8 μg/ml polybrene (Sigma) and recombinant murine cytokines (20 ng/ml Interleukin-3 (mIL-3, Cat#300–324P), 100 ng/ml Stem Cell Factor (mSCF, Cat#300–348P), 50 ng/ml Thrombopoietin (mTPO, Cat# 300–351P), 100 ng/mL Flt-3 Ligand (mFLT-3L, Cat#300–306P)) and 10% Penicillin-Streptomycin (Sigma). All cytokines were purchased from Gemini-Bioproducts. Cells were sorted for mCherry fluorescence 72 hours after transduction on a FACS Aria (BD Biosciences). Sorted cells were plated in methocult (StemCell Technologies Cat# M3434) in the presence of DMSO, enasidenib (50nM), or olaparib (1.25μM) in 35mm dishes in triplicates at a density of 1.2 × 10⁴ cells per ml. Colonies were counted after 7 days and serially re-plated and re-counted.

Targeted next generation sequencing

Genomic DNA (gDNA) was extracted from BM (diagnosis, Y1597) or PB (relapse, Y2260) mononuclear cells using QIAGEN DNeasy Blood & Tissue kit according to manufacturer’s instructions. Concentration and purity of the extracted gDNA were assessed using NanoDrop 1000 spectrophotometer (Thermo Scientific). During library preparation, IDT’s xGen Acute Myeloid Leukemia Cancer Panel v1.0 (Integrated DNA Technologies) was used for the targeted enrichment of 264 genes associated with AML (11731 probes, 1.19 Mb of target regions from 6235 exons). Hybridization capture of DNA libraries and sequencing (Illumina HiSeq, paired-end 100bp) were performed at the Yale Center for Genome Analysis (YCGA). Raw FASTQ files were subjected to quality control through FastQC v0.11.5 ²⁷. Sequencing reads were then aligned to the human genome (GRCh38.p10) with Burrows-Wheeler Aligner (BWA-MEM, v0.7.15) using default parameters ²⁸. Duplicate reads were removed with Picard v2.9.0 (http://broadinstitute.github.io/picard/). The number of total aligned reads before and after duplicate removal was obtained using SAMtools ²⁹ v1.5. Median coverage at each position in each enriched targeted region was calculated using BEDTools v2.27.1 ³⁰. Single-nucleotide variants (SNVs) and indels were detected with Genome Analysis Toolkit (GATK) v4.0.6.0 (https://github.com/broadinstitute/gatk) running Mutect2 ³¹ in tumor-only mode after base quality recalibration. Variants were annotated using ANNOVAR (2019Oct24) ³² and high-confidence variants were filtered as follows: exclusion of false positives (technical artifacts, germline variants and sequencing errors) through GATK4 filtering; exclusion of variants with synonymous or unknown effect; selection of deleterious variants (FATHMM-MKL prediction score) ³³; selection of variants annotated in COSMIC v90 ³⁴. This last filter was not applied to IDH2 variants to deeply explore regions potentially affected by PARPi treatment. All reported variants were visually inspected with the Integrative Genomics Viewer (IGV) ³⁵.

Statistical analysis

Data was analyzed using Prism 8 (GraphPad Software, La Jolla, CA) with the use of paired Ratio t-test and unpaired Mann-Whitney U test as indicated. P value was considered significant at values less than 0.05, n.s. not significant, * statistically significant (p<0.05), ** statistically significant (p<0.01), *** statistically significant (p<0.001), **** statistically significant (p<0.0001).

Data availability

All relevant data are available from the authors upon reasonable request.

Results

Enhanced PARPi sensitivity in IDH2-mutant AML in vivo

Our earlier work on the oncometabolite 2HG suggested that the HR defect and resultant PARPi sensitivity would be tumor agnostic ^19–21.

To determine whether PARPi would be efficacious in primary human MDS/AML within the context of their clonal complexity and variable co-mutational patterns, we generated a panel of IDH1/2-mutant and matched wildtype MDS and serially transplantable AML PDXs in cytokine humanized immunodeficient MISTRG mice (Table 1 and Supplementary Fig. S1A). MISTRG mice express physiologic levels of human cytokines (M-CSF^h/h, IL-3/GM-CSF^h/h and hTPO^h/h) on the RAG2^−/− γc^−/− background with additional humanization of the CD47 receptor hSIRPα ^{26, 36}; they support the engraftment, maintenance, and differentiation of AML and other myeloid malignancies ^{24, 37}. 12–16 weeks post transplantation we confirmed primary engraftment levels of human CD45⁺ (huCD45⁺) AML in MISTRG mice (Supplementary Fig. S1B, C) and generated secondary recipients for drug administration as described previously ²⁴. Eight to twelve weeks after secondary transplantation, MISTRG mice engrafted with newly diagnosed (Y566, Y577) or relapsed/refractory (Y747) IDH2^R140Q human AML were assigned to either vehicle or olaparib based on similar engraftment levels determined by BM aspiration and dosed intraperitoneally with vehicle or olaparib (100mg/kg) 6 days a week for 3 weeks. While in vehicle treated mice leukemia burden (Fig. 1A–C) and 2HG levels in plasma (Fig. 1D) rose sharply, especially for Y577 and Y747, engraftment and 2HG levels in olaparib treated mice remained stable or fell compared to pre-treatment levels (Fig. 1A–D and Supplementary figure S1D–F). Representative histologic assessment of huCD45⁺ engraftment confirmed the olaparib effect on BM leukemic burden (Fig. 1E). Increased γH2AX⁺ huCD45 cells in bone marrow confirmed that olaparib induced DNA damage in IDH mutant cells (Fig. 1F). To assure that olaparib toxicity against human AML was specific to IDH mutant AML, we engrafted MISTRG mice with IDH wildtype AML and treated mice as before with vehicle or olaparib (100mg/kg) 6 days a week for 3 weeks. In MISTRG mice engrafted with IDH wildtype AML (Y689, Y406, and Y456) olaparib administration showed no benefit when compared to vehicle regardless of mutational composition, with significantly increased PB and BM tumor burden compared to pre-treatment bone marrow aspiration levels (Supplementary Fig. S2A–C).

Table 1. Clinical characterization of the selected IDH2R140Q and IDHWT AML/MDS patients.

Patients 1–7 were IDH2^R140QAML; patients 8–10 were IDH1/2^WT; patients 11–12 were IDH1^R132K, and patients 13–14 were IDH2^R140QMDS. N/A, not applicable.

Patient ID	Sample type	Age	Gender	Diagnosis	Blast%	Karyotype	IDH mutation status	IDH mutation VAF	Co-occuring mutations
Y566	PB	61	M	AML arising from chronic myelomonocytic leukemia	PB:7, BM:35	47,XY,add(5)(q34),add(6)(q23),+8,−17, der(22)t(1;22)(q21;p11.2)[cp7]/46, XY[8]	IDH2R140Q	42.3	NRAS, IDH2, JAK2, SRSF2
Y577	PB	26	F	New diagnosis AML with no prior chemotherapy	PB:68, BM:90	NK	IDH2R140Q	38.4	NPM1
Y747	BM	62	M	Relapsed refractory AML (progressed from MDS)	PB: 28, BM:59	46,XY,der(15)t(1;15)(q12;p12)[7]/46,idem, del(20)(q11.2q13.3)[8]	IDH2R140Q	47	SRSF2, TET2
Y1597	BM	77	M	New diagnosis AML	PB:15, BM:48	46,XY,del(8)(q23.2q24.13)[8]/46, XY[7]	IDH2R140Q	50.1	SRSF2, JAK2, ASXL1
Y2260	PB	78	M	Relapsed AML after azacitidine+enasidenib	PB:13	46,XY,del(8)(q23.2q24.13)[8]/46, XY[7]	IDH2R140Q	NA	SRSF2
Y1550	BM	75	M	Relapsed AML	PB:3, BM:23	NK	IDH2R140Q	NA	SRSF2, RUNX1
Y1840	PB	75	M	Relapsed AML after enasidenib	PB:25	NK	IDH2R140Q	NA	SRSF2, RUNX1
Y689	BM	63	M	New diagnosis AML, progressed from MDS after failed azacitidine therapy	PB:12, BM:13	NK	WT		SRSF2, ASXL1, CBL, NRAS, KRAS
Y406	PB	59	M	New diagnosis AML	PB:15, BM:40	46,XY,inv(3)(q21q26),del(5)(q14q34), del(13)(q13q21)[15]	WT		RUNX1
Y456	BM	46	F	AML	PB:5, BM:36	NK	WT		NPM1, WT1, FLT3ITD, FLT3D835
Y276	BM	39	F	New diagnosis AML on Hydrea	PB:92, BM:77	NK	IDH1R132K	NA	NPM1, FLT3D835
Y493	PB	59	F	Relapsed/Refractory AML from MDS- on clinical trial with IDH1 inhibitor with new anemia	PB:6, BM:30	NK	IDH1R132K	NA
Y1952	BM	86	F	MDS-EB1	BM:8	NK	IDH2R140Q	47.3	SRSF2, ASXL1, STAG2
Y1365	BM	58	M	MDS-EB2	PB:3, BM:9	NK	IDH2R140Q	39	DNMT3A

N/A, not applicable

Fig 1. Olaparib targets IDH2-mutant AML blasts and leukemia initiating cells in vivo.

A-E, Secondary MISTRG recipients engrafted with BM cells collected from primary MISTRG mice, transplanted with three different IDH2^R140Q AML patient samples. Secondary mice were treated with either vehicle or olaparib (100mg/Kg) 6 days a week for 21 days. A-C, Pre- and post-treatment comparison of human CD45⁺ engraftment levels in BM of secondary treated mice. A, Y566 (vehicle, n=4, olaparib, n= 4); B, Y577 (vehicle n= 12, olaparib n= 8); C, Y747 (vehicle n= 4, olaparib n= 4). Individual mice are represented by matching pre-post symbols; P values were determined by ratio paired t test for pre-post treatment within groups and by Mann-Whitney test for comparison between treatment groups; n.s. not significant, *p < 0.05, ****p < 0.0001). D, Plasma concentrations of D-2-HG in pre-versus post administration of vehicle or olaparib. Individual mice are represented by symbols with mean ± S.E.M.; P values were determined by Mann–Whitney test; n.s. not significant, *p < 0.05, **p < 0.01. E, Immunohistochemistry (IHC) staining for hCD45 of BM of vehicle (top panel) and olaparib (lower panel) treated mice (scale bars 100 μm, original magnification 10x). F, Comparison of γH2AX⁺ huCD45 cells in bone marrow of secondary mice of Y577 treated with vehicle and olaparib (vehicle n= 8, olaparib n= 6). G-I, Secondary (2°) and tertiary (3°) transplantation of IDH2^R140Q AML (Y577). G, Comparison of huCD45⁺lin⁻huCD34⁺CD38⁻ population in BM of vehicle (n=8) vs olaparib (n=6) treated mice. Quantification of huCD45⁺ engraftment levels (H) and huCD45⁺lin⁻huCD34⁺CD38⁻ population (I) in the BM of tertiary mice engrafted with equal numbers of huCD45⁺ BM cells (5×10⁵) of vehicle (n=9) vs. olaparib (n=8) treated mice. J-L, Secondary and tertiary transplantation of IDH2^R140Q AML (Y747). J, Comparison of huCD45⁺lin⁻huCD34⁺CD38⁻ population in the BM of vehicle (n=4) vs olaparib (n=4) treated mice. Quantification of huCD45⁺ engraftment levels (K) and huCD45⁺lin⁻huCD34⁺CD38⁻ population (L) in the BM of tertiary mice engrafted with equal numbers of huCD45⁺ BM cells (5×10⁵) of vehicle (n=6) vs. olaparib (n=7) post-treatment mice. Individual mice are represented by symbols with means ± S.E.M.; symbols for corresponding 2° and 3° recipient mice are color coded; statistics represent Mann–Whitney test; n.s. not significant, *p < 0.05, **p < 0.01.

The IDH^m inhibitors enasidenib and ivosidenib are known to induce differentiation of leukemic blasts, and we have previously been able to recapitulate the differentiation effect in IDH2 mutant MDS in MISTRG mice ²⁴. Unlike IDH^mi, olaparib exhibits direct toxicity on IDH mutant cells secondary to the 2HG-induced BRCAness and dependence on PARP-mediated DNA repair^{9, 38, 39}. Consistent with this mechanism, olaparib did not alter lineage representation (Supplementary Fig. S3A) but significantly reduced the population of huCD45⁺Lin⁻CD34⁺CD38⁻ cells, previously shown to be enriched for LICs ^{40, 41} (Fig. 1G and Supplementary Fig. S3B). To validate that functional LIC were targeted by PARPi, we transplanted equal numbers of huCD45⁺ BM cells from vehicle or olaparib treated mice into tertiary recipient mice (Supplementary Fig. S3C). 16 weeks after transplantation of olaparib versus vehicle treated BM, we quantitated engraftment levels in recipient mice as readout of olaparib toxicity against LIC in treated donor mice. Overall huCD45⁺ AML engraftment levels and LICs were significantly lower in the recipients of olaparib- compared to vehicle-treated BM cells (Fig. 1H and Supplementary Fig. S3D, E and Fig. 1I, respectively), suggesting that olaparib exerted direct toxicity on LICs. These findings held true in mice engrafted with a IDH2^R140Q-mutant AML (Y747) relapsed after chemotherapy. Similarly, huCD45⁺Lin⁻CD34⁺CD38⁻ LICs were significantly reduced in BM of olaparib compared to vehicle treated mice (Fig. 1J) again translating into reduced leukemia burden in tertiary recipients (Fig. 1K, L and Supplementary Fig. S3F–H).

In summary, our data confirm the effectiveness of the PARPi olaparib in vivo specifically against IDH2^R140Q, but not IDH^WT AML, regardless of co-mutational pattern with toxicity against LICs.

Olaparib-induced toxicity is preserved in IDH^mi-resistant IDH2-mutant AML

IDH^mi rarely effect cure and ultimately patients relapse with restoration of 2HG levels suggesting an oncometabolite dependence of IDH mutant AML ^15–18. We therefore sought to test whether PARPi will be effective in enasidenib-resistant IDH2-mutant AML. We first resorted to a syngeneic mouse model of AML relying on co-mutation of Flt3 with an internal tandem duplication (ITD) (Flt3^ITD) and of IDH2^R140Q. Polyinosinic: polycytidylic acid [poly (I:C)] induced activation of Mx1 promoter-driven Cre recombinase (M) results in heterozygous expression of IDH2^R140Q (I2) in the background of constitutive Flt3^ITD (F) expression and mice develop a transplantable (MFI2) leukemia as previously described ²⁵. To test olaparib in vitro in colony forming assays we established that olaparib at 1.25μM was non-toxic to IDH wildtype cells while effectively reducing colony formation in IDH2^R140Q mutant cells (Supplementary Fig. S4A). We generated enasidenib-resistant murine AML by transduction of MFI2 AML with lentiviral vectors expressing IDH2 carrying second site mutations Q316E or I319M, previously reported to confer enasidenib resistance in patients ¹⁸. Lentivirally transduced MFI2 AML expressing IDH2^I319M, IDH2^Q316E or IDH2^WT were then subjected to serial colony formation in vitro in the presence of vehicle, enasidenib (50nM), or olaparib (1.25μM). Enasidenib inhibited serial replating capacity of cells expressing IDH2^R140/WT at the third replating (CFU3) as previously shown, while the Q316E and I319M mutations conferred resistance to enasidenib but not to olaparib (Fig. 2A). These data suggest that restoration of 2HG may be sufficient to confer sensitivity to PARPi. While second site mutations have been identified only in a minority of patients ¹⁸, restoration of aberrant 2HG production is found in a significant subset of IDH^mi resistant patients, suggesting that PARP inhibition may be more widely applicable to mutant IDH inhibitor-resistant AML ^{14, 16}.

Fig 2. Enasidenib resistant AML is sensitive to olaparib administration both in vitro and in vivo.

A, Colony forming unit assay (CFU) re-plating of hematopoietic stem and progenitor cells (HSPCs) from FLT3^ITDIDH2^R140Q linage depleted primary murine cells expressing IDH2 WT or Q316E or I319M in trans and cultured in methocult containing vehicle, AG-221 at 50nM, or olaparib at 1.25μM. Data are mean ± SEM. for triplicate (CFU1/2) culture. P values were determined by Mann-Whitney; *p < 0.05, **p < 0.01, ***p<0.001. This is a representation of 2 independent experiments. B-C, Clinical and laboratory features for a AML patient at the time of IDH2^R140Q AML diagnosis (Y1597) (B) and after relapse on azacytidine + enasidenib treatment (Y2260) (C), including blood absolute neutrophil count (ANC) and white blood cell count (WBC), and blast percentage in PB and BM. D, Human CD45⁺ engraftment levels in the BM of secondary mice of (Y1597) treated with vehicle (n=4), enasidenib (n=4), or olaparib (n=5); enasidenib (40mg/kg) was administered via oral gavage and olaparib (100mg/kg) intraperitoneally for 6 days per week for 3 weeks. E, Human CD45⁺ engraftment levels in the BM of secondary mice of (Y2260) treated with vehicle (n=5), enasidenib (n=5), or olaparib (n=5). Individual mice are represented by matching pre-post symbols; P values were determined by ratio paired t test for pre-post treatment within groups and by Mann-Whitney test for comparison between different treatment groups; n.s. not significant, *p < 0.05, ***p<0.001. F-G, Comparison of huCD45⁺lin⁻huCD34⁺CD38⁻ population in BM of treated mice of Y1597 (F) and of Y2260 (G).

We therefore investigated the olaparib response in enasidenib-resistant IDH2^R140Q AML from 2 patients treated at our institution. The first patient presented with newly diagnosed, intermediate-risk AML with inversion 8 (inv (8)), with an elevated WBC of 42×10³/μl with 48% blasts in BM and 10% blasts in PB (Fig. 2B, C). Targeted next generation sequencing (NGS) of bone marrow cells at time of diagnosis revealed an IDH2^R140Q mutation along with mutations in SRSF2, JAK2, and ASXL1 (Table 1). The patient received azacitidine and enasidenib, which induced a significant reduction in WBC count and blast percentage (Fig. 2B, C). After almost 12 months of continuous treatment with enasidenib, the patient relapsed with rapid rise in the WBC with leukemic blast predominance (Fig. 2B, C), which preceded the patient’s death. Resequencing of the mutant AML did not identify the mechanism of acquired resistance neither in the clinical targeted panel (Table 1) nor using a comprehensive AML targeted panel (Supplementary Table S1). We transplanted MISTRG mice with BM and PB mononuclear cells banked at the time of diagnosis (Y1597) and at time of relapse on enasidenib (Y2260) (Table 1 and Fig. 2B, C). Both diagnosis and relapsed AML engrafted MISTRG mice had elevated 2HG levels (Supplementary Fig. S4B, C). Secondary recipient mice were treated with vehicle, olaparib or enasidenib (40mg/kg via oral gavage) for 6 days per week for 3 weeks. While both enasidenib and olaparib effectively reduced engraftment and plasma 2HG levels in mice engrafted with the new AML diagnosis sample (Y1597), enasidenib failed to limit proliferation of the relapsed AML (Y2260) with significant rise in disease burden during the treatment course (Fig. 2D, E and Supplementary Fig. S4B–H). Olaparib on the other hand significantly inhibited leukemic progression compared to enasidenib (Fig. 2D, E and Supplementary Fig. S4C–H). Moreover, olaparib significantly reduced Lin⁻CD45⁺CD34⁺CD38⁻ LICs in both diagnosis (enasidenib-naïve) and enasidenib-resistant AML compared to vehicle and enasidenib, confirming effectiveness of olaparib against LICs (Fig. 2F, G).

Persistent sensitivity to olaparib despite acquisition of resistance to IDH^mi was also evident in a second patient with relapsed AML (Fig. 3A, B) with IDH2^R140Q and RUNX1 mutations (Table 1). While the patient’s AML was sensitive to both enasidenib and olaparib prior to treatment with enasidenib (Y1550, Fig. 3C), after progression of disease on enasidenib, olaparib but not enasidenib effectively limited leukemic cell expansion in MISTRG (Y1840, Fig. 3D). Both enasidenib and olaparib significantly inhibited Lin⁻CD45⁺CD34⁺CD38⁻ LICs in the enasidenib-naïve AML engrafted mice (Y1550, Fig. 3E) while only olaparib significantly reduced LICs in the enasidenib-resistant AML engrafted mice (Y1840, Fig. 3F).

Fig 3. Olaparib remains effective in a second enasidenib-resistant IDH2-mutant AML PDX.

A-B, Clinical and laboratory features for a relapsed IDH2^R140Q AML patient prior to treatment with enasidenib (Y1550) (A) and after relapse on enasidenib treatment (Y1840) (B), including blood absolute neutrophil count (ANC) and white blood cell count (WBC), and blast percentage in PB and BM. C, Human CD45⁺ engraftment levels in the PB and BM of secondary mice of (Y1550) treated with vehicle (n=3), enasidenib (n=4), or olaparib (n=3). D, Human CD45⁺ engraftment levels in the PB and BM of secondary mice of (Y1840) treated with vehicle (n=3), enasidenib (n=3), or olaparib (n=3). E-F, Comparison of huCD45⁺lin⁻huCD34⁺CD38⁻ population in BM of treated mice of Y1550 (E) and of Y1840 (F). Individual mice are represented by matching pre-post symbols; P values were determined by ratio paired t test for pre-post treatment within groups and by Mann-Whitney test for comparison between different treatment groups; n.s. not significant, *p < 0.05, ***p<0.001.

Collectively, these data show that PARP inhibition is effective in IDH^mi resistant AML and may offer a novel treatment approach for patients with relapsed AML.

PARP sensitivity is IDH isoform- and disease-agnostic

AML and MDS carry mutations at two hotspot sites in IDH2 and in IDH1and different mutations may have in part distinct oncogenic mechanisms ⁴². We therefore tested whether susceptibility to PARP inhibition was indeed independent of the mutant IDH isoform and whether PARPi could be effective in patients with MDS. We transplanted MISTRG mice with one newly diagnosed (Y276) and one relapsed/refractory secondary (Y493) IDH1^R132K -mutant AML sample (Table 1 and Supplementary Fig. S5A). Olaparib effectively treated both IDH1^R132K -mutant AMLs compared to vehicle (Fig. 4A, B) and despite prior exposure of Y493 to an IDH1 inhibitor (Fig. 4C, D). Olaparib administration reduced the huCD45⁺lin⁻CD34⁺CD38⁻ LIC population compared to vehicle treatment in both AMLs (Supplementary Fig. S5B, C) as shown for IDH2 mutant AML. Overall, these results demonstrated that olaparib monotherapy was also active against IDH1-mutant AML and thus isoform- agnostic.

Fig 4. PARP inhibition is also effective in IDH1-mutant AML and IDH1/2-mutant MDS.

A-B, Assessment of human CD45⁺ chimerism levels in vehicle vs olaparib treated secondary MISTRG mice engrafted with BM cells collected from primary mice transplanted with IDH1^R132K AML. A, Y276 (vehicle, n=4, olaparib, n= 4); B, Y493 (vehicle n= 2, olaparib n= 3). Primary patient BM CD34⁺ HSPCs were pre-incubated with anti-CD3 antibody (OKT3) and injected intra-hepatically into newborn MISTRG mice conditioned twice with 150cGy. C-D, Clinical and laboratory data for a relapsed/refractory IDH1^R132K AML (Y493) while on ivosidenib trial treatment, including ANC, WBC, (C) and blast percentage (D) in PB. E-F, Chimerism detection of huCD45⁺ percentages in BM preceded treating the mice with either vehicle, enasidenib, or olaparib. Mice were analyzed 21 days after drug administration. E, Comparison of overall human CD45⁺ engraftment in BM of mice engrafted with IDH2^R140Q MDS-EB1 (Y1952) treated with vehicle (n=3), enasidenib (n=3), or olaparib (n=4). F, Comparison of overall human CD45⁺ engraftment in BM of mice engrafted with IDH2^R140Q MDS-EB2 (Y1365) treated with vehicle (n=4), enasidenib (n=5), or olaparib (n=6). Individual mice are represented by matching pre-post symbols; P values were determined by ratio paired t test for pre-post treatment within groups and by Mann-Whitney test for comparison between treatment groups; n.s. not significant, *p < 0.05, **p<0.01.

Lastly, to determine whether PARP inhibition is a relevant therapy for IDH-mutant MDS, we transplanted MISTRG mice with CD34⁺ cells from two patients with IDH2 mutant MDS, namely IDH2^R140Q-mutant MDS-excess blast (EB) 1 (Y1952) and EB2 (Y1365) (Table 1). Primary engrafted mice were assigned to treatment groups based on pre-treatment marrow engraftment and treated with vehicle, enasidenib or olaparib for 21 days. Enasidenib and Olaparib showed efficacy in both MDS samples with reduction in huCD45⁺ engraftment in both BM and PB (Fig. 4E–F and Supplementary Fig. S5D–E) and with significant reduction in the huCD45⁺Lin⁻CD38⁻CD34⁺ LIC populations (Supplementary Fig. S5F–G).

Collectively, our data showed the effectiveness of the synthetic lethality-mediated approach as a novel alternative therapeutic strategy in IDH2-mutant MDS.

Discussion

There is significant need for improvement for patients with relapsed/ refractory AML and MDS despite the availability of IDH inhibitors, which have modest and transient effectiveness. Both primary and acquired resistance mechanisms to IDH inhibitors have been identified ^15–18. Primary resistance has been highly associated with the presence of NRAS, KRAS and MAPK pathway mutations, and increased incidence of relapse has been linked to higher mutational burden ^{12, 14, 43}. In addition, adaptive resistance develops through the acquisition of second site mutations or via isoform switching or yet unknown mechanisms, but almost invariably aberrant 2HG levels are restored ¹⁸. Post-resistance treatments are thus critical to prolong patients’ lives.

Recent advances in the understanding of metabolic vulnerabilities induced by IDH mutations have led to the identification of an alternative strategy to leverage the concept of synthetic lethality associated with the impaired HR conferred by 2HG ⁴⁴. We here validated the efficacy of PARPi exploiting the 2HG-induced BRCAness defect in primary AML and MDS samples in vivo building upon our and others’ prior in vitro studies and our previously reported highly efficient MDS/AML PDX model ^{19, 22, 24}.

Cytokine humanized MISTRG mice express physiological levels of human cytokines instead of their murine counterparts and have overcome many challenges in modeling myeloid malignancies ^{36, 45}. Accordingly, our group previously reported MISTRG as an advanced and highly efficient model of MDS that faithfully replicates the intra- and inter-disease heterogeneity, the dysplastic morphology, and the mutational profile and clonal complexity ²⁴. We leveraged the MISTRG model to test whether PARP inhibition would be effective against a variety of IDH mutant AML, AML resistant to IDH^mi, and against IDH mutant MDS. Olaparib treatment demonstrated antileukemic efficacy irrespective of co-occuring mutations as long as cells carried IDH1/2 mutations, while olaparib was ineffective against IDH1/2 wildtype AML. In addition, olaparib demonstrated toxicity against phenotypic and functional LICs, a rare subpopulation of leukemic cells that possess stem cell properties, including leukemia initiation and self-renewal, as evident by engraftment in immunodeficient mice as shown here, and drug resistance ^{46, 47}.

Our data also demonstrates that PARP inhibition could represent a new choice for patients with primary refractoriness to IDH inhibitors or for those who relapse on IDH1/2 inhibitor treatment irrespective of tumor type ¹⁹. The PRIME clinical trial (NCI10264) is currently testing the overall response of IDH1/2-mutant relapsed/refractory AML and MDS to PARP inhibitor monotherapy, which further underlines the clinical relevance of the results presented here. Future studies combining PARP inhibitors with other DNA damaging agents or agents with complementary activity will be important to leverage this mechanism against MDS/AML as monotherapy rarely results in cure.

Collectively, our data provide critical insights which support the pursuit of PARP inhibition as a promising therapeutic tool for IDH mutated MDS/AML patients.