HDAC Inhibition Induces MicroRNA-182, which Targets RAD51 and Impairs HR Repair to Sensitize Cells to Sapacitabine in Acute Myelogenous Leukemia

Sapacitabine is an orally available nucleoside analogue prodrug that is structurally related to cytarabine. It is active against acute myelogenous leukemia (AML) both in elderly patients and in relapsed, refractory disease (1, 2). Sapacitabine is metabolized into its active form 2′-C-Cyano-2′-deoxy-1-β-D-arabino-pentofuranosylcytosine (CNDAC) by amidases. Following intracellular phosphorylation, the resulting triphosphate is incorporated into replicating DNA. In contrast to other nucleoside analogues, this does not terminate replication. Rather, after additional nascent DNA strand elongation, a β-elimination reaction occurs, which causes the analogue to rearrange to a 2′, 3′-dideoxyconfiguration. This terminates the 3′-end resulting in a single-stranded nick, which becomes converted to a double-strand break (DSB) upon a subsequent round of DNA replication (3, 4). The DNA DSBs generated by CNDAC (5) or ionizing radiation (6) are partly repaired by the homologous recombination repair (HR) pathway. HR is governed by multiple proteins such as ATM, the MRE11–RAD50–NBS1 complex (7), BRCA1/2 and RAD51, a key protein that directs the homology search, and DNA strand exchange that is critical for successful HR repair (8). High levels of RAD51 are linked to resistance to DNA-damaging therapies, relapsing disease, and poor survival (9, 10). Conversely, cells lacking RAD51 or its paralog RAD51D have an increased sensitivity to DNA-damaging agents (11) including CNDAC (5). Unlike breast and ovarian tumors that often harbor intrinsic defects in HR, AML cells usually exhibit a functional HR pathway (12), which may allow them to limit the toxicity of DNA DSB-inducing agents.

The histone deacetylases (HDACs) are a class of chromatin-modulating proteins that become recruited to target promoters to compact chromatin and silence gene expression. Conversely, inhibitors of HDACs (HDACi) function by promoting the acetylation of proteins, including histones at gene promoters to induce a transcriptionally active chromatin configuration and reverse HDAC-mediated gene expression (13, 14). HDACis such as vorinostat (SAHA) and belinostat are approved for the treatment of cutaneous T-cell leukemia (15), whereas panobinostat is approved for the treatment of refractory multiple myeloma (16, 17). In AML, HDACi's showed synergy with CNDAC in xenograft models (18) and were effective in phase I/II trials when combined with demethylating (19) or other chemotherapeutic agents (20). Mechanistically, panobinostat exposure led to decreases in the levels of ATM, BRCA1, and RAD51 to sensitize cells to cytarabine, daunorubicin, or ionizing radiation in solid tumor cell lines (21, 22). HDACi treatment also leads to the depression an important class of regulatory genes, the miRNAs (23–25). miRNAs bind to complementary sequences in target RNA to either destabilize it or prevent its transcription in a cell- and context-specific manner (26, 27). Specific miRNA such as miR-182 are overexpressed in breast cancer and intrinsically compromise HR in those cells (28). Other miRNAs, such as miR-103, miR-107, miR-96, and miR-182 target RAD51 and other components of the HR pathway in solid tumors (28–31); however, it is not known whether these miRNA target RAD51 in AML. As RAD51 is well expressed in AML, an investigation of the expression levels of miRNAs that potentially target RAD51 and the mechanisms that regulate their expression becomes important.

In this study, we investigated whether HDAC inhibition upregulated the expression of miRNA genes that targeted RAD51, and whether the resultant impairment in HR sensitized cells to sapacitabine, a clinically relevant DSB-inducing drug in AML.

Panobinostat (PS) was provided by Novartis. Vorinostat (SAHA) was purchased from Cayman Chemical), and decitabine (DAC) was from Sigma. All reagents were dissolved in 100% DMSO (Burdick & Jackson) to a stock concentration of 10⁻² mol/L and stored at −80 °C. CNDAC was supplied by Dr. A. Matsuda (Hokkaido University, Sapporo, Japan).

OCI-AML3 and MV4-11 AML cells were obtained from ATCC and maintained in RPMI supplemented with 5% FCS and 1% glutamine in a 37°C incubator containing 5% CO₂. OCI-AML3 cells were authenticated at the Characterized Cell Line Core facility at The University of Texas MD Anderson Cancer Center (Houston, TX). MV4-11 and HeLa-DR-13 cell lines were authenticated by sequencing at the Ohio State University (OSU; Columbus, OH). The HeLa-DR-13 (HeLa-DR) and HEK cells were a gift from Dr. Parvin (Ohio State University, Columbus, OH) and maintained as described previously (32). Primary AML cells and nucleic acids from normal bone marrow were collected under an Institutional review board–approved protocol and obtained from the Leukemia tissue bank at OSU. Primary AML blasts were maintained in RPMI with 20% FCS and 1X Stem Span CC100 (Stem Cell Technologies).

OCI-AML3 and MV4-11 cells were exposed to 2 μmol/L CNDAC and increasing concentrations of panobinostat or SAHA for 48 hours following which Annexin assays were conducted by incubating cells with Annexin V–fluorescein isothiocyanate and propidium iodide (BD Biosciences) for 15 minutes and accumulating the fluorescence of at least 10,000 cells on a BD Biosciences FACSCalibur flow cytometer to determine the percentage of apoptotic and viable cells. Drug interactions were analyzed using the CalcuSyn (BioSoft). A combination index (CI) value of 1 indicated an additive drug interaction, whereas a CI value >1 suggested antagonism and a value <1 denoted synergism (33).

miRNA microarray analysis was carried out by LC sciences (http://www.lcsciences.com/) as described in Supplementary methods.

Immunoblotting was performed with antibodies against RAD51 (Millipore), H3K9Ac (Millipore), PARP (Cell Signaling Technology), cleaved caspase-3 (Cell Signaling Technology), γ-H2AX and H2AX (Cell Signaling Technology), or GAPDH (Cell Signaling Technology).

Equal amounts of cross-linked chromatin from OCI-AML3 or primary AML cells were used to carry out chromatin immunoprecipitation (ChIP) directed against HDAC1, HDAC2 (Millipore), H3K4me3, and H3K9AC. Rabbit IgG was used as a nonspecific control (Jackson Laboratories; ref. 34). DNA eluted from immunoprecipitates were purified and analyzed by PCR probes specific for the miR-182 promoter. For immunoprecipitation, HDAC1, HDAC2, and IgG immune complexes were prepared by incubating extracts from OCI-AML3 cells with antiserum for 1 hour, followed by 45-minute precipitation with protein A agarose beads, washed, and assayed for the presence of HDAC1 or HDAC2.

Total RNA was extracted from cells using the mir-Vana RNA extraction kit (Applied Biosystems). The expression of miR-182 was determined using 5 ng of total RNA, the step loop reverse primer, and the mir-Vana Reverse transcription kit followed by qPCR using primer probes from Applied Biosystems. Expression levels were quantitated by the comparative C_t method using snRNA RNU48 levels for normalization. Similarly, the levels of RAD51 were measured and normalized to GAPDH using the one-step, real-time (RT)-PCR procedure probes from Applied Biosystems.

Pre-miR-182, anti-miR-182, and Pre-miR–negative control #1 were obtained from Life Technologies. Oligonucleotides were transfected into OCI-AML3 or HeLa-DR cells at 150 nmol/L using the Nucleofection system, (Solution T, X-001; ref. 35; Amaxa) following manufacturer's instructions for 72 to 96 hours, after which the cells were harvested for RNA and protein analysis.

OCI-AML3 were exposed to drugs for 24 hours, washed with PBS (37°C), and plated in methylcellulose medium in triplicate. Clonogenicity was determined after 7 days. OCI-AML3 cells were exposed to scrambled (scr), miR-182, or anti-miR-182 for 48 hours, and then exposed to CNDAC for an additional 24 hours before being assayed for the ability to form colonies after 7 days.

HeLa-DR cells were transfected with 300 or 500 nmol/L miR-182 or scr for 48 hours before being transfected with the pCBASceI plasmid. For the experiments with miR-182 and scr, we performed a second round of transfection along with the PCBASce1 plasmid. After 96 hours, cells expressing green fluorescence were measured using a Gallios FACSCalibur instrument as described previously (32). For experiments testing HDAC inhibitors, cells were exposed to 10 or 15 nmol/L panobinostat for 24 hours before transfection with PCBASce1 and assayed for HDR as described previously (32).

The pmiRTarget vector with RAD51 WT insert (full-length 3′-UTR) or mutated RAD51 3′-UTR (containing a deletion of the miR-182 target site) were cotransfected into HeLa or HEK cells with miR-182 or scr in HeLa cells. Cells were lysed 48 hours after transfection and luciferase and red fluoresence protein intensity (used as an transfection control) were measured for 36 to 48 hours and data expressed as the mean ± SEM from triplicate determinations from two to four independent transfections.

Quantitative data are shown as the mean ± SEM for atleast three independent experiments. Comparative statistics used the paired Student t test or one-way ANOVA with Tukey post hoc tests to evaluate differences between experimental groups. P < 0.05 and lower was considered statistically significant.

The pan-HDAC inhibitor, vorinostat (SAHA), has been shown to synergize with CNDAC, the active metabolite of sapacitabine, in killing MV4-11 AML cells both in vitro and in a xenograft mouse model (18). To determine whether panobinostat, a related HDACi, would synergize with CNDAC in AML cells, we exposed OCI-AML3 and MV4-11 cells to either 2 μmol/L CNDAC, 0.1, 0.2, or 0.5 nmol/L panobinostat, or to 0.1, 0.2, or 0.5 μmol/L of SAHA and observed minimal cytotoxicity (Fig. 1A and B and Supplementary Figs. S1A and S1B). However, when these cells were exposed to the combination of 2 μmol/L CNDAC with increasing concentrations of panobinostat up to 0.5 nmol/L (Fig. 1A and Supplementary Fig. S1A) or SAHA up to 0.5 μmol/L (Fig. 1B and Supplementary Fig. S1B) there was a greater than additive loss in cell survival (P value between P < 0.01 and P < 0.05). Drug interactions were analyzed by the Chou–Talalay method; the combination index (CI) values in OCI-AML3 cells for the interaction of 2 μmol/L CNDAC with 0.1, 0.2, or 0.5 nmol/L panobinostat and 0.1, 0.2, or 0.5 μmol/L SAHA were below 0.68 ± 0.24 (Supplementary Table S1). Similarly, the CI value in MV4-11 cells for the interaction of 2 μmol/L CNDAC with 0.1, 0.2, or 0.5 nmol/L panobinostat and 0.1, 0.2, or 0.5 μmol/L SAHA were below 0.72 ± 0.02 indicating synergy (Supplementary Table S1). We then chose the highest concentrations of panobinostat (0.5 nmol/L) and SAHA (0.5 μmol/L) that were synergistic with CNDAC in the apoptotic assays and determined whether they could synergize with CNDAC in colony forming assays. Concentrations of 0.5 μmol/L CNDAC were chosen for the cloning assay because exposure to 1 μmol/L or higher CNDAC inhibited colony growth in the OCI-AML3 cell line (5) attesting to the higher sensitivity of the clonogenic assays. OCI-AML3 cells were exposed to 0.5 μmol/L CNDAC, 0.5 nmol/L panobinostat, 0.5 μmol/L SAHA, or to combinations of CNDAC and panobinostat or CNDAC and SAHA for 24 hours. Our results indicate that the combination of 0.5 nmol/L panobinostat as well as 0.5 μmol/L SAHA was synergistic with 0.5 μmol/L CNDAC in inhibiting colony growth (P < 0.0001 and P < 0.001, respectively; Fig. 1C and D). Taken together, our data indicate that HDACis synergize with CNDAC to limit the survival of AML cell lines.

In colon cancer cells, HDAC inhibition was linked to decreases in RAD51 and an inability to complete repair by HR (21). To determine the action of HDAC inhibition on RAD51 in AML, we exposed OCI-AML3 cells to increasing concentrations of panobinostat, SAHA, or combinations of panobinostat or SAHA with 2 μmol/L CNDAC for 48 hours. HDAC inhibition resulted in the hyperacetylation of histones H3 and H4 at lysine residues (K9/14). This was accompanied by dose-dependent decreases in the levels of the RAD51 protein in cells exposed to either panobinostat, SAHA, and to combinations of panobinostat or SAHA with 2 μmol/L CNDAC (Fig. 1E), whereas cells exposed to CNDAC alone did not significantly decrease RAD51 protein. Similar results were obtained when MV4-11 cells were exposed to increasing concentrations of panobinostat either alone or in combination with 2 μmol/L CNDAC (Supplementary Fig. S1C). We then conducted time course experiments where OCI-AML3 and MV4-11 cells were exposed to 2 μmol/L CNDAC or 2 μmol/L CNDAC in combination with 0.5 nmol/L panobinostat for up to 48 hours or to 0.5 nmol/L panobinostat alone for 48 hours. Our results demonstrate that cells exposed to CNDAC minimally decreased the levels of RAD51 in both cell lines. However, those exposed to either a combination of CNDAC with panobinostat or panobinostat alone significantly decreased RAD51 levels (Supplementary Fig. S2A and S2B) indicating that it was HDAC inhibition that was responsible for the observed losses in RAD51 protein.

Multiple mechanisms such as downregulation of mRNA transcript (36), caspase-3–mediated cleavage (37) and miRNA-mediated targeting regulate the abundance of any given protein (36). To identify the mechanisms by which HDAC inhibition decreased RAD51 protein, we exposed OCI-AML3 to 0.2 or 0.5 nmol/L panobinostat for varying times. Exposure of 0.2 nmol/L panobinostat or 0.5 nmol/L panobinostat for up to 48 hours led to minimal decreases in the levels of RAD51 mRNA (Supplementary Fig. S3A). The levels of RAD51 protein, on the other hand, decreased by 20% by 48 hours in response to 0.2 nmol/L panobinostat and by over 60% within 24 hours in response to 0.5 nmol/L panobinostat (Fig. 1F and Supplementary Fig. S3B). Loss of RAD51 was accompanied by the appearance of cleaved caspase-3 and some PARP cleavage in cells exposed to 0.5 nmol/L but not 0.2 nmol/L panobinostat (Fig. 1F). In contrast to RAD51, the levels of the closely related HR repair proteins RAD50, ATM, and NBS1 remained constant for up to 48 hours (Supplementary Figs. S3B and S4) suggesting that within the HR pathway, RAD51 was selectively targeted by HDAC inhibition. Because RAD51 is a known substrate of caspase-3 (37) and cells exposed to 0.5 nmol/L panobinostat showed processing of caspase-3, we exposed OCI-AML3 cells to 0.5 nmol/L panobinostat with or without the presence of the pan-caspase inhibitor, z-vad-fmk. The results indicated that levels of RAD51 declined to the same extent regardless of inhibition in caspase-3 processing, indicating that caspase-3–mediated cleavage was not a major mechanism for the observed declines in RAD51 in response to HDAC inhibition (Supplementary Fig. S3C and S3D).

Antagonizing the action of the HDACs leads to a reversal of silencing of a large number of genes, including those of miRNA (23, 24). To determine whether HDAC inhibition could induce the expression of epigenetically silenced miRNA that were predicted to target RAD51, we exposed OCI-AML3 cells to 0.5 nmol/L panobinostat for 6 hours. The expression of a total of 163 miRNA was altered in response to HDAC inhibition; the miRNA significantly up- or downregulated (P < 0.01) are shown in Fig. 2A. Many of the miRNA induced in response to panobinostat putatively targeted multiple genes of the HR repair pathway, such as RAD51, its paralogs (RAD51D, RAD51B, XRCC2/3), BRCA1 and NBS1 (Supplementary Table S2); miRNA predicted to target RAD51 in silico by a minimum of two algorithms (Target Scan and miRanda) and are highlighted in bold in Supplementary Table S2. To validate the results of the microarray, we chose three miRNAs, miR-182, miR-211, and miR-320, that were induced after HDAC inhibition as well as predicted to target RAD51 for further evaluation. Following exposure of OCI-AML3 cells for varying times with 0.5 nmol/L panobinostat, the expression of these miRNA was induced between four- and sixfold in response to HDAC inhibition (Fig. 2B–D).

DNA hypermethylation acts in conjunction with the histone deacetylases to epigenetically silence gene expression (38). To determine the contribution of DNA hypermethylation to the silencing of miR-182 in AML, we exposed MV4-11 AML cells to 0.5 μmol/L decitabine (DAC, a DNA methylation inhibitor), 10 nmol/L panobinostat, or a combination of panobinostat and DAC for up to 24 hours. We exposed MV4-11 cells to 10 nmol/L panobinostat based on the concentrations utilized to demonstrate synergy with other DNA-damaging agents in this cell line (39). The results show that exposure to DAC alone for up to 48 hours induced miR-182 by less than twofold (Fig. 2E and Supplementary Fig. S5). However, exposure to panobinostat alone or in combination with DAC induced miR-182 levels by sevenfold within 24 hours, indicating that HDACs and not DNA promoter hypermethylation played a dominant role in the silencing of miR-182 in AML (Fig. 2E). Correspondingly, the panobinostat -induced increases in miR-182 in cells exposed to panobinostat alone or in combination with DAC led to reciprocal decreases in the levels of RAD51 (Fig. 2F).

The ability of a miRNA to target a given mRNA is highly context and tissue specific (26, 27) making it important to validate the interaction between miRNA and its targets in the relevant tissue of interest. Therefore, we focused on 2 miRNAs, miR-320 and miR-182 that had highly conserved binding sites clustered within the central portion of the 3′-UTR of RAD51. The RAD51 gene 3′- UTR has two binding sites for miR320a/c and a single binding site for miR-182 (Fig. 3A). Mimetics for miR-182 and miR-320 were transfected into OCI-AML3 cells along with the nontargeting controls (scr) for 48 hours. Of these, miR-320 did not reduce RAD51, whereas miR-182 consistently reduced the levels of RAD51 by 40% to 50% after 48 hours compared with cells expressing scr oligonucleotides (Fig. 3B–E; P < 0.01, paired t test, n = 6). Time course experiments then confirmed that cells expressing miR-182 targeted RAD51 protein even up to 72 hours (Fig. 3C). Next, to definitively determine the role of miR-182 in targeting RAD51, we overexpressed antagomirs to miR-182 for 48 hours. Our data indicate that in the presence of anti-miR-182, the levels of RAD51 were preserved, whereas expression of miR-182 caused the expected loss in RAD51 levels (Fig. 3D and E). Finally, to test whether miR-182 directly targeted RAD51, HeLa-DR as well as HEK cells were transfected with the pmiRTarget luciferase reporter containing either the full-length RAD51 3′-UTR or one with the miR-182-binding site deleted with miR-182 (Fig. 3F, underlined) or nontargeting controls at 300 nmol/L. Ectopic expression of miR-182 caused a 30% decrease in luciferase activity (P < 0.05, paired t test, n = 4) in HeLa cells and a 70% decrease in luciferase activity (P < 0.05, paired t test, n = 2) in HEK cells in comparison with the scr, whereas deletion of the 3′-UTR binding site for miR-182 restored the luciferase activity in both cell types (Fig. 3G and H). This indicates that miR-182 directly targets RAD51.

We compared the expression of miR-182 in normal bone marrow and AML (Supplementary Table S3) and found that the expression of miR-182 was 10- to 1,000-fold lower in AML blasts (Fig. 4A). As we had determined that HDAC inhibition led to an upregulation of miR-182 in AML, we evaluated whether HDAC1 and HDAC2 were recruited to the miR-182 promoter to silence its expression. miR-182 is an intergenic miRNA transcribed as part of a cluster (40) and contains a major (S1) recruitment site for HDAC1 and HDAC2 (Fig. 4B), the major class I HDACs. HDAC1 and HDAC2 are found together in repressive transcriptional complexes which suggest a high degree of functional redundancy between the two proteins (41). Reciprocal immunoprecipitations from nuclear extracts using antibodies against HDAC1, HDAC2, or IgG were resolved on gels and immunoblotted for the presence of HDAC1 or HDAC2. The results demonstrated that HDAC1 and HDAC2 form a complex in nuclei (Fig. 4C). ChIP assays demonstrated that HDAC1 and HDAC2 were recruited specifically over IgG at the miR-182 promoter in OCI-AML3 cells (Fig. 4D). Conversely, HDAC inhibition led to the accumulation of transcriptionally permissive chromatin modifications such as the trimethylation of histone H3 (H3K4me3) over nonspecific IgG at the miR-182 promoter (Fig. 4E). This was accompanied by a six- to eightfold increase in the expression of miR-182 by 18 hours (Fig. 2D).

Similarly, HDAC1 and HDAC2 were specifically recruited over IgG to the promoter for miR-182 in primary AML blasts (Fig. 4F). The recruitment of HDAC1 and HDAC2 at the miR-182 promoter did not change significantly before or after exposure to panobinostat in AML cell lines or primary samples (Supplementary Fig. S6). The IC₅₀ of panobinostat against AML blasts is less than 20 nmol/L (39). On the basis of this finding and other published results (42), we exposed AML blasts to 10 nmol/L panobinostat which led increases in the levels of the activating chromatin mark, AcH3K9, in the primary AML samples (Fig. 4G), a 10- to 50-fold increase in the levels of miR-182 in primary AML samples (Fig. 4H) as well as reciprocal and sustained losses in the levels of RAD51 in all primary AML cells (Fig. 4I).

Because the loss of RAD51 was associated with an attenuation of DNA repair in solid tumor cells (21), we evaluated the consequences of miR-182 overexpression or of HDAC inhibition on the HR repair process using the HDR assay. We utilized HeLa-DR cells that had the p-DR-GFP (with 2 inactive copies of GFP) integrated into its genome (43). Transfection with the I-SceI endonuclease caused a single DSB, which, if repaired using homologous sequences from the donor GFP gene, resulted in a functional GFP (32). HeLa-DR cells were transfected with scrambled (scr) or miR-182 oligonucleotides at 300 nmol/L or 500 nmol/L for 48 hours before being transfected with the I-SceI endonuclease. Overexpression of miR-182 but not scr resulted in a >50% decrease in RAD51 in HeLa-DR cells (Fig. 5A). Correspondingly, transfection of the scr oligos together with I-SceI endonuclease resulted in 20% GFP-positive cells indicative of successful HR repair (Fig. 5B, C, and E). In contrast, transfection of miR-182 in conjunction with the I-SceI endonuclease led to a >50% inhibition of HR as measured by a reduction in the percentage of GFP-positive cells (Fig. 5B, D, and F; P < 0.001). Inhibition of HR would be expected to lead to higher levels of unrepaired damaged DNA, which can be assayed by measuring the levels of γ-H2AX that persist after DNA damage (44). Therefore, OCI-AML3 cells were exposed to either 0.5 μmol/L CNDAC, or to miR-182 or scr oligonucleotides for 48 hours before being challenged with 0.5 μmol/L CNDAC for 24 hours. Cells exposed to CNDAC alone or those expressing scr did not decrease RAD51, but responded with some increases in γ-H2AX after exposure to CNDAC for 24 hours (Fig. 5G). In contrast, cells transfected with miR-182 showed the expected decrease in RAD51 and exhibited higher levels of γ-H2AX when exposed to CNDAC for 24 hours (Fig. 5G). We then tested the consequence of the miR-182–mediated decrease in RAD51 and inhibition of DNA repair on the survival of cells exposed to CNDAC. OCI-AML3 cells were left untreated, exposed to 0.5 μmol/L CNDAC for 24 hours, or transfected with scr, miR-182, or anti-miR-182 oligonucleotides after which a portion was exposed to 0.5 μmol/L CNDAC for 24 hours before being assayed for colony growth. Untreated cells or those expressing scrambled showed similar numbers of colonies (280 ± 41 and 282 ± 14; Fig. 5H). Expression of miR-182 decreased colony formation by 23% to 218 ± 7 in comparison with cells expressing scr (P < 0.05, two-tailed p test), whereas expression of anti-miR-182 increased the number of colonies to 243 ± 6 in comparison with those expressing miR-182 (P < 0.05, two-tailed p test). Untreated cells or those expressing scr showed a 40% decrease in colony formation to 60 ± 9 and 62 ± 7 after exposure to CNDAC (P < 0.001, two-tailed p test). However, exposure of cells expressing miR-182 CNDAC caused a greater than 65% inhibition in colony formation to 38 ± 9 in comparison with cells expressing scr that were exposed to CNDAC (P < 0.001, two-tailed p test). Finally, exposure of cells expressing anti-miR-182 to CNDAC supported 185 ± 6 colonies (decrease of 24% vs. 65% in cells exposed to miR-182+CNDAC; P < 0.01, two-tailed p test; Fig. 5H). Finally, to determine whether the miR-182–mediated loss in RAD51 augmented the toxicity of panobinostat in combination with CNDAC, we exposed OCI-AML3 cells to scrambled oligonucleotides, miR-182 or anti-miR-182 for 48 hours before challenging cells to a combination of 0.5 nmol/L panobinostat with 0.5 μmol/L CNDAC. Transfection with scrambled oligonucleotides elicited 17% ± 2% Annexin-positive cells. In comparison, cells transfected with miR-182 showed an increase in Annexin-positive cells to 28% ± 7%, whereas those transfected with anti-miR-182 showed decreases in cell death to 13% ± 11%. Addition of panobinostat and to cells transfected with scrambled, miR-182, and anti-miR-182 increased Annexin positivity to 52% ± 2%, 66% ± 12%, and 32% ± 22%, respectively (P = ns, Supplementary Fig. S7).

We then tested the consequence of HDAC inhibition on the ability of Hela-DR cells to carry out HR. Cells transfected with the I-SceI endonuclease alone showed 24% GFP-positive cells indicating ongoing HR, whereas those preexposed to 10 and 15 nmol/L panobinostat responded with a 75% to 80% inhibition in HR (Fig. 6A–E). We exposed HeLa-DR-13 cells to concentrations of 10 and 15 nmol/L panobinostat because those concentrations decreased RAD51 protein by >50% (Fig. 6F). We then tested whether HDAC-mediated inhibition of HR would affect the ability of cells to resolve DNA damage. When OCI-AML3 cells were exposed to IR, an agent that causes instantaneous DSB, there was a rapid increase in the levels of γ-H2AX which decreased to baseline within 4 to 6 hours indicating the successful repair of DNA damage. However, if these cells were pretreated with 0.5 nmol/L panobinostat before being exposed to IR, the γ-H2AX signal persisted indicating a delay in DNA repair (Fig. 6G). In parallel, when OCI-AML cells were exposed to CNDAC which causes DSB after 2 replication cycles, there was an increase in the γ-H2AX starting at 24 to 48 hours. However, pretreatment with panobinostat resulted in γ-H2AX accumulating earlier (16 hours) and to a greater level suggesting that the HDACi-mediated loss in RAD51 led to a delay in DNA repair and higher levels of residual DNA damage (Fig. 6H).

The HDAC inhibitors panobinostat and SAHA have been shown to synergize with cytarabine, daunorubicin (22), and CNDAC using cell viability assays (18). We utilized cell viability as well as colony forming assays to demonstrate that panobinostat and SAHA synergized with CNDAC in killing AML cells. Pharmacokinetic evaluations show that the maximum serum plasma concentrations of vorinostat are approximately 1.4 μmol/L, 1 to 3 hours postdosing in AML (45, 46) and approximately 19 nmol/L for panobinostat (46) indicating that the concentrations of vorinostat and panobinostat achieved in patients are more than sufficient to inhibit both the proliferative capacity as well as kill AML cells. Correlative studies during therapy would help validate the action of this combination on the proliferative index and survival of circulating AML blasts.

To elucidate the mechanism underlying the synergy between HDACi and CNDAC, we focused on RAD51, a key component of the homologous repair pathway in AML cell lines and primary AML lymphocytes. The one published study did not elucidate the mechanism underlying the observed synergy between SAHA and CNDAC (18), whereas two additional publications showed that panobinostat at concentrations of 10 nmol/L decreased RAD51, BRCA1, and ATM transcripts (22) and protein to inhibit HR (21). However, at the lower concentrations of panobinostat (0.5 nmol/L) used in our study, RAD51 transcript levels did not decrease. Rather, we identified an alternative mechanism whereby RAD51 protein levels decreased due to targeting by miR-182. miRNAs have been shown to regulate components of the HR pathway. In breast cancer cells, miR-182 was found to target BRCA1, P53BP1, as well as CHEK (28, 30). In addition, miR-182, miR-96, miR-103, and miR-106 exhibited high levels of expression in breast cancer cells to target RAD51 (29, 31) and intrinsically suppress HR (28, 31). AML differs from these solid tumors in that the miRNAs targeting RAD51 are underexpressed in AML leading to robust levels of expression of RAD51 as well as a functional HR pathway. In addition, miRNA target proteins in a context-dependent manner and miRNA usually express a reverse correlation with their cellular targets (29, 44). Therefore, although miR-103 and miR-106 targeted RAD51 in solid tumors (29), it was unlikely they would do the same in AML because they and their target, RAD51 were expressed at high levels in OCI-AML3 cells (Supplementary Table S2). This conclusion is supported by our findings with miR-320 which is expressed at high levels in OCI-AML3 cells (Supplementary Table S2) and failed to target RAD51 even though RAD51 was a predicted target.

Thus, we focused on miR-182 because it was expressed at low levels in AML, whereas its predicted target RAD51 had robust expression. Mechanistically, targeting the HDACs with panobinostat or SAHA rapidly reversed the silencing of miR-182, whereas targeting DNA hypermethylation did not indicating that while DNA promoter methylation has a key role in suppressing many miRNAs in AML (38), miR-182 is not regulated by this mechanism. miR-182 directly targeted RAD51 at the protein level and in luciferase assays. Targeting RAD51, either via HDACi exposure or ectopic expression of miR-182 inhibited HR and strongly sensitized AML cells to the cytotoxic action of CNDAC in colony forming assays highlighting the importance of RAD51 as a determinant of the sensitivity of AML cells to double-strand–damaging agents. Manipulating the levels of miR-182 with mimics and anti-miRs moderately increased or attenuated the cytotoxicity caused by the combination of panobinostat and CNDAC. This is not surprising given that the combination of panobinostat and CNDAC induces a very robust cell death response. However, this observation raises the possibility that miR-182 may have additional targets the loss of which may cooperate with panobinostat and CNDAC to impact AML survival.

In conclusion, our findings define the strategy of blocking HDACs to induce miR-182 and limit RAD51 expression. Furthermore, if miR-182 targets multiple components of the HR pathway in AML cells as shown for solid tumors (28, 30); epigenetic reactivation of miR-182 could be used to compromise HR at multiple levels. Because, HDAC inhibition induced miR-182 which targeted RAD51 in AML was sufficient to decrease HR by over 80% and sensitized cells to DNA strand breaking agents, it represents a novel pharmacologic strategy. HR is implicated in the repair of damage caused by diverse agents including topoisomerase I inhibitors (47), IR (6) DNA interstrand cross-linkers (48) as well as sapacitabine (5). This suggests that investigations in other malignancies could establish whether targeting RAD51 expression with HDAC inhibitors may also serve to augment the selective toxicity of DSB-inducing DNA-damaging agents across cancers. Thus, this mechanism-based strategy may have multiple applications for clinical evaluations.

No potential conflicts of interest were disclosed.

Conception and design: W. Plunkett, D. Sampath

Development of methodology: T.-H. Lai, B. Ewald, C. Liu, D. Sampath

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-H. Lai, B. Ewald, A. Zecevic, C. Liu, M. Sulda, D. Papaioannou, R. Garzon, D. Sampath

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.-H. Lai, B. Ewald, W. Plunkett, D. Sampath

Writing, review, and/or revision of the manuscript: W. Plunkett, D. Sampath

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.S. Blachly

Study supervision: W. Plunkett, D. Sampath

The authors thank the Leukemia Tissue Bank at Ohio State University and Dr. John Byrd for the primary AML samples and intellectual discussions.

This study was supported by the National Cancer Institute, Department of Health and Human Services (R01-CA28596), The MD Anderson Cancer Center, and OSU Cancer Center Support grants (NCI-CA016672 and NCI-CA06058).

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.