Poly (ADP-Ribose) Polymerase Inhibitor Hypersensitivity in Aggressive Myeloproliferative Neoplasms

Myeloproliferative neoplasms (MPN) represent a heterogeneous group of clonal diseases with a common propensity to progress to acute leukemia (1–4). Essential thrombocythemia, polycythemia vera, primary myelofibrosis (PMF), and mixed myelodysplastic/myeloproliferative neoplasms, such as chronic myelomonocytic leukemia (CMMoL), are all clonal neoplasms derived from aberrant early hematopoietic precursors but have varied clinical manifestations. Upon progression, they are uniformly refractory to standard acute leukemia therapies, with median survival less than 6 months.

One common mutation in MPNs, an activating V617F point mutation in the tyrosine kinase JAK2, is found in more than 80% of cases of polycythemia vera (5), 40% of cases of essential thrombocythemia, and 30% of cases of MF (6). JAK2 mutation and overexpression have been associated with increased homologous recombination (HR) and genomic instability (7–9). Other alterations conferring an MPN-like phenotype, such as BCR/ABL translocations in chronic myelogenous leukemia (CML), FIP1L1–PDGFR rearrangements in eosinophilic leukemias, and FLT3 mutations in acute myeloid leukemia (AML), have also been associated with changes in the DNA repair pathways, leading to increased genomic instability and drug resistance (9, 10). For example, even though early studies indicated that RAD51, a critical component of the HR pathway, is upregulated in BCR/ABL–positive CML cells (11), subsequent studies demonstrated that repair in these cells is error prone and leads to mutations and large deletions or insertions (12). Further analysis traced this genomic instability to several changes, including (i) enhanced tyrosine phosphorylation of RAD51, leading to its aberrant function (13); (ii) downregulation of BRCA1 (14); (iii) stimulation of single-strand annealing, an error-prone DNA repair pathway (15); and (iv) other changes in the Fanconi anemia/BRCA pathway that can be reverted by ectopic BRCA1 expression (16).

Repair defects in BCR/ABL–negative MPNs are not as well characterized. Gross chromosomal lesions are common in MF and accelerating MPN; and SNP array karyotyping identified additional subcytogenetic abnormalities (17, 18). These types of changes are reminiscent of chromosomal aberrations observed in HR-deficient solid tumors, such as BRCA1- or BRCA2-mutant breast and ovarian cancer (19). Although the upstream source and pathologic consequences of these extensive genomic rearrangements are not as well understood in MPN, previous studies have reported increased error-prone double-strand break repair in MPNs as well as CML (15).

PARP inhibitors are a class of antineoplastic agents being widely tested in solid tumors (20–25). These agents target PARP1, PARP2, and PARP3, three enzymes that contribute to various aspects of DNA repair (21, 23, 24, 26–28). PARP inhibition not only diminishes base excision repair but also impairs alternative end-joining (29) and accelerates nonhomologous end-joining (30). In cells with diminished HR, these changes lead to error-prone DNA repair and cell death (21, 23, 31). In addition, PARP inhibition leads to trapping of PARP1 on DNA, preventing access of downstream repair proteins to sites of DNA damage (32–34) and providing a potential mechanism for PARP inhibitor–induced killing in HR-proficient cells that contain high levels of PARP1 protein. Importantly, chromosome 1q (including the PARP1 locus at 1q42) is amplified in a subset of chronic phase MPNs and even more commonly in transformed MPNs (17, 35), providing a potential opportunity for PARP1 trapping even in MPNs without HR defects.

Consistent with the BCR/ABL–induced DNA repair abnormalities described above, PARP inhibitor hypersensitivity has been reported in CML (36). In view of the repair defects observed in other MPNs, as well as the copy number increases of the PARP1 locus in these disorders (17, 35), we have performed a survey comparing PARP inhibitor sensitivity of various MPNs, including CML, with that of normal hematopoietic progenitors. For this study, we have examined two PARP inhibitors that are undergoing extensive clinical testing. Veliparib, which is in NCI-sponsored trials in a variety of neoplasms, including hematologic neoplasms (www.clinicaltrials.gov), has recently been shown to exhibit favorable properties for combining with DNA-damaging agents (34). Olaparib, which is roughly 10-fold more potent in vitro, has attained regulatory approval for the treatment of ovarian cancer (24, 25).

Veliparib (ABT-888, Enzo Life Sciences) and olaparib (ChemieTek) were dissolved in DMSO at stock concentrations of 10 mmol/L. Stocks were aliquoted in 10 μL volumes, stored at −80°C, and thawed once immediately before use. All samples in any given experiment contained identical concentrations of DMSO (0.1% v/v).

After patients provided informed consent, specimens were obtained by purifying mononuclear cells from the peripheral blood of patients with MPN on Ficoll–Hypaque gradients. The MPN cohort included fresh samples from cases of BCR/ABL–positive CML; the classical BCR/ABL–negative MPNs essential thrombocythemia, polycythemia vera, primary and secondary MF and mixed myelodysplastic syndrome (MDS)/MPN diagnoses (CMMoL, aCML, MDS/MPN-Unclassifiable; Table 1).

Ten million Ficoll/Hypaque–purified mononuclear cells (MNC) from normal controls and MPN patients were exposed to 10 Gy ionizing radiation from a Rad Source RS200 X-ray irradiator, then allowed to recover for 6 hours in a humidified 37°C tissue culture incubator equilibrated at 5% (v/v) CO₂. Leukocytes were pelleted by centrifugation at 200 × g for 10 minutes and fixed in 2% (w/v) paraformaldehyde in Dulbecco's calcium- and magnesium-free PBS for 10 minutes at 20°C to 22°C. Leukocytes were repelleted as above, washed with PBS, and stored in PBS at 4°C. For analysis, 2.5 × 10⁴ leukocytes were deposited onto glass coverslips by cytocentrifugation and processed as described previously (37). Briefly, coverslips were washed 3 times with PBS, permeabilized in PBS + 0.25 % (v/v) Triton X-100 for 10 minutes, washed an additional 3 times with PBS, and then incubated for 1 hour in blocking buffer [PBS, 1 % (v/v) glycerol, 0.1% (w/v) gelatin from cold-water fish, 0.1% (w/v) BSA, 5% (v/v) goat serum, and 0.4 % (w/v) sodium azide] for 1 hour at room temperature. Coverslips were incubated with RAD51 rabbit polyclonal (Active Motif) and phospho-Ser¹³⁹-H2AX mouse monoclonal (Millipore) antibodies diluted 1:500 in blocking buffer overnight at 4°C. Coverslips were then washed 3 times with PBS, followed by incubation for 1 hour in secondary Alexa Fluor 488–conjugated goat anti-mouse IgG and Alexa Fluor 568–tagged goat anti-rabbit IgG (Invitrogen) diluted 1:1,000 in blocking buffer. Coverslips were further washed 3 times with PBS, counterstained with 1 μg/mL Hoechst 33258 in PBS, and mounted using ProLong Antifade Reagent (Invitrogen). PEO1 and PEO4, cell lines with a truncating BRCA2 mutation and a reversion mutation, respectively, were utilized as negative and positive controls for radiation-induced RAD51 foci. Confocal images were captured on a Zeiss LSM 710 scanning confocal microscope with a 100×/1.4 N.A. oil immersion objective. For quantitation, 100 cells per sample from more than 5 fields were manually scored for RAD51 and phospho-H2AX foci by an investigator blinded to clonogenic assay results. Cells with >5 foci were graded as positive. Quantitation and image processing were performed with the Zeiss Zen software package and Adobe Photoshop CS3.

Colony-forming assays were performed as described previously (38). Ficoll/Hypaque–purified MNCs were washed and resuspended in RPMI1640 medium (Gibco). Aliquots (0.5 mL) were combined with 4.5 mL of MethoCult medium (STEMCELL Technologies #H4435) containing increasing concentrations of veliparib (0, 0.5, 1, 2.5, 5, 10, and 20 μmol/L) or olaparib (0, 0.25, 0.5, 1, 2, and 5 μmol/L). Two or four 1-mL aliquots of each mixture were plated onto replicate 35-mm gridded culture plates to give final MNC concentrations of 1–5 × 10⁵ cells per plate. After incubation at 37°C in 5% CO₂, most samples were scored for total colony number after 10 to 14 days; however, some CMMoL samples were scored earlier (5–7 days) due to hypersensitivity to growth factors. Colonies were defined as clusters of >40 cells. Drug concentrations at which total colony counts were inhibited to 50% of diluent treated controls (IC₅₀) were determined using linear regression analysis of the dose–response curves after linear transformation using an exponential model (CalcuSyn software, Biosoft, Inc).

DNA was extracted from Ficoll/Hypaque–purified MNCs using a DNA Midi Kit (Qiagen). To search for somatic mutations in HR pathway proteins, aliquots of DNA (1 μg) were subjected to targeted capture and massively parallel sequencing using BROCA as described previously (39) for a panel of genes involved in HR (ATM, ATR, BABAM1, BARD1, BAP1, BLM, BRCA1, BRCA2, BRE, BRIP1, BRCC3, CHEK1, CHEK2, FAM175A, FANCC, FANCP, MRE11A, NBN, PALB2, RAD50, RAD51, RAD51B, RAD51C, RAD51D, RBBP8, TOPBP1, UIMC1, XRCC2, and XRCC3), regulation of HR (CDK12, C11orf30, ID4, TP53BP1, and USP28), or NHEJ (LIG4, XRCC4, XRCC5, XRCC6, PRKDC, and DCLRE1C) and TP53.

Genomic DNA was bisulfite treated with the EZ DNA Methylation Kit (Zymo Research). The initial methylation-specific PCR (MSP) screen was performed as described previously (40). Quantitative MSP (qMSP) for BRCA1 promoter methylation, normalized to β-actin levels, was carried out using the iTaq SYBR Green mix and 300 nmol/L of each primer. qMSP, MSP primer sequences, and annealing temperatures are listed in Supplementary Table S1. BRCA1 qMSP primers were adapted from Estellar and colleagues and cover the region −175 to +9 relative to the TSS (41). Cycling conditions are 95°C for 5 minutes, followed by 40 cycles of 95°C for 5 seconds and 64°C for 60 seconds. Melt curve readings were recorded at 0.5°C increments from 65°C to 95°C. The controls for unmethylated and methylated templates were bisulfite-treated samples from normal lymphocyte DNA and CpG Methylated Jurkat Genomic DNA (New England Biolabs), respectively. Methylated samples are defined as amplicons with melting temperatures matching that of the methylated control.

To assess whether PARP inhibitors might have any promise in non-CML MPNs, this study examined primary samples from 41 patients with a variety of MPNs (Table 1). Following up on the observation that MPNs are associated with multiple genetic aberrations (17, 18), we examined a subset of 14 samples for integrity of the HR pathway. Upon exposure to ionizing radiation, cells with normal HR repair form foci of DNA repair complexes that can be detected by immunofluorescence after staining for RAD51, the recombinase responsible for pairing homologous sequences. Cells with impaired HR typically will not form RAD51 foci when exposed to ionizing radiation. Of the 14 MPN samples examined using this assay, 6 (43%) displayed markedly diminished RAD51 foci formation (Fig. 1; Table 1). In contrast, 0 of 14 controls performed with these assays exhibited diminished RAD51 foci formation (P = 0.016, Fisher exact test).

To search for potential explanations for this HR deficiency, we sequenced a panel of repair genes found to be mutated in a variety of malignancies, including BRCA1, BRCA2, RAD51, RAD51 paralogs and several Fanconi anemia pathway genes, in MPN samples from 24 of these patients. Results of this analysis are summarized in Table 1. Mutations observed included a well-established BRCA1 stop mutation (BRCA1 c.5382insC) and CHEK2 frameshift mutation (CHEK2 c.1100delG), as well as a deleterious mutation in the RAD50 gene (RAD50 c.3476delA). Although a variety of rare normal alleles and conservative single-nucleotide variants were also detected in other genes (Table 1), no deleterious homozygous or hemizygous mutations were observed that could, by themselves, account for the defect in RAD51 protein recruitment to sites of DNA damage. We also assessed BRCA1, BRCA2, FANCC, FANCF, and FANCL promoter CpG island methylation in these MPN samples. No hypermethylation was identified in the BRCA2, FANCC, FANCF, or FANCL promoters. However, 6 of 27 samples (22.2%) analyzed in this study demonstrated hypermethylation of the BRCA1 promoter (Table 1).

With the observation that a substantial fraction of MPN samples have a deficient DNA repair response, as manifested by the lack of formation of radiation-induced RAD51 foci, we performed colony-forming assays with continuous exposure to the PARP inhibitor veliparib using MNCs from 41 MPN patients and 13 normal controls. As shown in Supplementary Fig. S1, the sensitivities of granulocyte and erythroid colonies generally paralleled each other in both normal and MPN samples. Accordingly, for the remainder of this work, we tabulated various colony types with each assay but combined them at the time of graphing. Assay reproducibility was established by comparing results of assays performed on samples from several patients at two points in time in the absence of changes in treatment (Supplementary Fig. S2).

As shown in Fig. 2A and B, the drug concentrations that inhibited colony formation by 50% (IC₅₀) were an average of 4-fold lower in the MPN samples than in normal controls (2.5 vs 9.6 μmol/L, P < 0.0001). Focusing on the subset of samples that were subjected to assays for both RAD51 foci and colony formation (Fig. 2C), PARP inhibitor sensitivity was greater (IC₅₀ on average 3-fold lower) in MPN samples with impaired RAD51 foci formation compared with those without (mean 9.3 vs. 2.8 μmol/L, P = 0.028). In contrast, BRCA1 promoter methylation status did not track with veliparib sensitivity (Supplementary Fig. S3).

To determine whether PARP inhibitor sensitivity was linked to clinical characteristics, we analyzed colony-forming assay results based on clinical descriptions of disease at the time of sample acquisition. This analysis suggested broad sensitivity across clinical subtypes, with samples from MF patients demonstrating the greatest heterogeneity (Fig. 3A). Although the number of samples in any MPN category was somewhat limited, CML, CMMoL, and MDS/MPN-U samples were most sensitive, with IC₅₀ values averaging 3, 3, and 1.5 μmol/L, respectively.

We also examined whether JAK2 mutation status, which was suggested as a surrogate marker for impaired DNA repair (15), correlated with PARP inhibitor sensitivity. In our sample set (Fig. 3B), JAK2-nonmutated samples were significantly more sensitive to PARP inhibition than JAK2V617F-mutated samples (mean IC₅₀ 2.8 vs. 5.7 μmol/L, P = 0.04). Notably, however, 5 of 16 individual JAK2-mutated samples demonstrated veliparib IC₅₀s below 2 μmol/L despite the relative insensitivity of this subset as a whole.

While this work was in progress, it was reported that the PARP inhibitors PJ-34 (a preclinical tool compound) and veliparib might also inhibit certain kinases when assayed at high concentrations under cell-free conditions (42, 43). Importantly, the structurally dissimilar PARP inhibitor olaparib lacks this off-target kinase inhibitory activity (43). To provide assurance that the veliparib hypersensitivity observed in MPN samples was due to PARP inhibition, olaparib sensitivity was also examined using colony-forming assays (e.g., Fig. 4A) in a subset of samples. For samples assayed using both veliparib and olaparib, there was a strong correlation (R² = 0.61, P = 0.003) between sensitivity to the two agents (Fig. 4B). Accordingly, as was the case with veliparib, a substantial fraction of these MPN samples was also hypersensitive to olaparib when compared with normal controls (Fig. 4C).

Because genomic instability is a hallmark of advanced MPNs, we examined whether clinical MPN specimens exhibit HR defects and whether they are hypersensitive to PARP inhibitors, which are known to be particularly effective in neoplasms with HR deficiency (20–24). Here, we demonstrate PARP inhibitor hypersensitivity for the first time in a large subset of MPN samples.

An impaired DNA damage response, as evidenced by the lack of development of RAD51 foci in response to ionizing radiation, was observed in approximately 40% of MPN samples assayed (Fig. 1; Table 1). This observation suggests that the HR pathway is impaired in those samples. These results are consistent with prior observations of extensive chromosomal and subchromosomal copy number changes in MPN (17, 18), which are another hallmark of HR deficiency (19). Because formation of RAD51 foci only provides an assessment of HR pathway integrity upstream of RAD51 and is unaffected by changes downstream, for example, loss of RAD51C (44), it is important to emphasize that the present observations might underestimate the true frequency of HR repair defects in MPNs.

The causes of this HR pathway dysfunction in MPNs are incompletely understood at present. Some of the samples with impaired RAD51 foci formation also exhibited diminished formation of H2AX foci (Fig. 1B; Table 1). Because H2AX phosphorylation after double-strand breaks typically reflects the activation of ATM and phosphorylation of its substrate MDC1 (45), diminished formation of both phospho-H2AX and RAD51 foci might reflect a defect in ATM or its activation as described in other neoplasms. Other samples formed phospho-H2AX foci but, nonetheless, failed to form RAD51 foci, suggesting one or more defects between these two events in the DNA damage response. Consistent with this heterogeneity, we have previously examined 144 cases of MPN via SNP arrays and found that 26% have heterozygous deletions in genes encoding one or more DNA repair pathway proteins such as BRCA2, ATM, FANCC, or FANCL (46). Because the other copy remains intact, however, it is unclear whether these heterozygous deletions cause sufficient changes in protein expression to impact the HR pathway. Accordingly, we have examined a subset of MPN samples for methylation changes in FANC proteins, ATM, and BRCA2 but did not observe frequent changes in methylation that would confer increased genomic instability. BRCA1 promoter methylation was found in 6 of 27 samples analyzed (Table 1) but was not associated with increased sensitivity to PARP inhibition ex vivo (Supplementary Fig. S3). Given the results of drug sensitivity testing in this study, further investigation of repair defects in MPNs appears warranted.

PARP inhibitors are currently undergoing extensive testing in other neoplasms with HR deficiencies (20–25). Here, we found that MPN samples with impaired RAD51 foci formation exhibited increased sensitivity to the PARP inhibitor veliparib, suggesting that these cells depend on alternative repair pathways containing one or more veliparib-sensitive PARPs for their survival (Fig. 2C). Further examination of MPN samples as a group revealed broad sensitivity to PARP inhibition across many subtypes of MPN (Fig. 3A). This hypersensitivity relative to normal myeloid progenitors might reflect multiple alterations in MPNs. In addition to the sensitizing effects of HR deficiency, the presence of high concentrations of PARP1, for example, as a consequence of chromosome 1q gains in copy number seen in a subset of MPNs (see Introduction), might also sensitize cells to the PARP trapping that accompanies treatment with PARP inhibitors (32, 37). Accordingly, there are multiple potential explanations for the observed hypersensitivity of MPNs relative to normal myeloid progenitors, just as there are in solid tumors (20–25), and further investigation is required to understand this hypersensitivity on a case-by-case basis in both instances.

While this work was in progress, it was reported that veliparib can inhibit purified kinases when applied under cell-free conditions at concentrations 100- to 1,000-fold higher than those required to inhibit purified PARP1 and PARP2 (43). Importantly, the PARP inhibitor olaparib was shown to lack these off-target kinase inhibitor effects (43). In our study, there was a strong correlation between olaparib and veliparib sensitivity (Fig. 4B), suggesting that the PARP inhibitor hypersensitivity of MPNs (Figs. 2B and 4C) reflects PARP inhibition rather than an off-target effect.

To our knowledge, this study provides the first examination of PARP inhibitors in non-CML MPN. Because hypersensitivity of some MPN samples was apparent within the first few samples, we surveyed a broad range of MPN samples in an attempt to determine the types of diseases lumped under the categories of MPN and mixed MDS/MPN that might display this hypersensitivity. Further studies examining additional samples with each type of MPN or MDS/MPN syndrome at various stages of disease (e.g., initial diagnosis vs. advanced disease requiring treatment vs. recurrence after therapy or stem cell transplant) are required to assess whether PARP inhibitor hypersensitivity persists after current treatments, as appears to be the case in a subset of ovarian cancers, for example (23).

Our results indicate that MPN samples with the activating JAK2 V617F mutation are, on average, less sensitive than MPN samples without JAK2 mutations. Because the assays measured the sensitivity of proliferating cells, this difference is unlikely to reflect any difference in the rate of proliferation in JAK2 wild-type versus JAK2-mutant samples. Instead, it is possible that the difference reflects the ability of JAK2 V617F to activate STAT-mediated transcription and inhibit apoptosis (47).

On the other hand, the therapeutic window for PARP inhibition in MPN, for example, as measured by differences in mean IC₅₀ of RAD51 foci–deficient MPNs and normal progenitors (Fig. 2C), is somewhat narrower than the 100-fold difference in IC₅₀ observed when comparing BRCA1- or BRCA2-deficient murine embryonic stem cells and their HR proficient counterparts (48). However, it is important to point out that the difference in sensitivity between BRCA2-deficient human ovarian cancer cells and their isogenic BRCA2-restored counterparts is also only 5- to 10-fold (30), yet clinical activity of PARP inhibitors is observed in ovarian cancer (20–25). The limited, albeit easily observed, sensitivity difference between normal and MPN progenitors suggests that the activity of PARP inhibitors as single agents in MPN might merit further investigation, particularly in JAK2 wild-type MPNs, which sometimes lack therapeutic options (1–4). On the other hand, we have previously examined the impact of adding veliparib to topotecan and carboplatin (37), two agents that have exhibited some activity when combined to treat relapsed AML (49), and have observed synergistic cytotoxic effects with the topotecan/veliparib combination in AML lines in vitro (37). Moreover, because JAK2 mutations were associated with diminished PARP inhibitor sensitivity (Fig. 3B), we have begun examining the effect of combining PARP inhibitors with other agents used in the treatment of MPNs, looking for synergistic cytotoxic effects with PARP inhibitor/hydroxyurea and PARP inhibitor/ruxolitinib combinations. These observations will provide the impetus for further study of PARP inhibitors, alone and in combination, in MPNs.

M.A. McDevitt is an employee of AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.W. Pratz, A.G. Patel, J.E. Karp, M.A. McDevitt, S.H. Kaufmann

Development of methodology: K.W. Pratz, A.G. Patel, W. Poh, J.G. Herman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.W. Pratz, B.D. Koh, A.G. Patel, W. Poh, R. Dilley, M.I. Harrell, B.D. Smith, J.E. Karp, E.M. Swisher, M.A. McDevitt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.W. Pratz, B.D. Koh, A.G. Patel, K.S. Flatten, J.G. Herman, R. Dilley, M.I. Harrell, J.E. Karp, E.M. Swisher, M.A. McDevitt

Writing, review, and/or revision of the manuscript: K.W. Pratz, B.D. Koh, A.G. Patel, K.S. Flatten, W. Poh, J.G. Herman, R. Dilley, B.D. Smith, J.E. Karp, S.H. Kaufmann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.W. Pratz, K.S. Flatten, M.I. Harrell, M.A. McDevitt

Study supervision: K.W. Pratz, M.A. McDevitt, S.H. Kaufmann

Procurement of specimens by the Sidney Kimmel Cancer Center at Johns Hopkins Tumor and Cell Procurement Bank (supported by NIH Grant P30 CA006973) is gratefully acknowledged by all authors.

K.W. Pratz's effort on these studies was supported as a co-investigator for translational science team at Johns Hopkins (NIH Grant UM1 CA186691). These studies were also supported by educational funds from the Mayo Foundation, including the M.D.-Ph.D. Program (to A.G. Patel, NIH Grant T32 GM65841) and Clinical Pharmacology Training Program (to B.D. Koh, NIH Grant T32 GM008685) and Clinician Investigator Training Program (to B.D. Koh).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.