Leukemias expressing the constitutively activated tyrosine kinases (TK) BCR-ABL1 and FLT3/ITD activate signaling pathways that increase genomic instability through generation of reactive oxygen species (ROS), DNA double-strand breaks (DSB), and error-prone repair. The nonhomologous end-joining (NHEJ) pathway is a major pathway for DSB repair and is highly aberrant in TK-activated leukemias; an alternative form of NHEJ (ALT-NHEJ) predominates, evidenced by increased expression of DNA ligase IIIα (LIG3) and PARP1, increased frequency of large genomic deletions, and repair using DNA sequence microhomologies. This study, for the first time, demonstrates that the TK target c-MYC plays a role in transcriptional activation and subsequent expression of LIG3 and PARP1 and contributes to the increased error-prone repair observed in TK-activated leukemias. c-MYC negatively regulates microRNAs miR-150 and miR-22, which demonstrate an inverse correlation with LIG3 and PARP1 expression in primary and cultured leukemia cells and chronic myelogenous leukemia human patient samples. Notably, inhibition of c-MYC and overexpression of miR-150 and -22 decreases ALT-NHEJ activity. Thus, BCR-ABL1 or FLT3/ITD induces c-MYC expression, leading to genomic instability via augmented expression of ALT-NHEJ repair factors that generate repair errors.
Implications: In the context of TK-activated leukemias, c-MYC contributes to aberrant DNA repair through downstream targets LIG3 and PARP1, which represent viable and attractive therapeutic targets. Mol Cancer Res; 13(4); 699–712. ©2015 AACR.
Constitutively activated tyrosine kinases (TK) BCR-ABL1 and FLT3/ITD generate increased levels of reactive oxygen species (ROS), DNA damage including double-strand breaks (DSB), and abnormal repair that is highly error-prone. Together, these features result in acquisition of genomic alterations which have the potential to drive disease progression and resistance to therapy (1, 2). DSBs are repaired by 2 pathways, homologous recombination (HR) and nonhomologous end-joining (NHEJ; refs. 3–5). The classic NHEJ (c-NHEJ) pathway frequently causes small alterations in DNA sequences around the break site but rarely joins previously unlinked DNA ends (6). However, there is an alternative (ALT) version of NHEJ that results in larger deletions and chromosomal translocations where DNA ends are joined at regions of DNA sequence microhomology (6, 7). Notably, ALT-NHEJ activity is increased when the c-NHEJ pathway is downregulated (6, 7). Expression of both BCR-ABL1 and FLT3/ITD significantly alters the expression of key DSB repair components at both the mRNA and protein levels(8) and increases the frequency of repair errors (9–12). These cells also demonstrate increased in ALT-NHEJ activity characterized by large DNA deletions and repair using DNA sequence microhomologies (8, 13, 14). Notably, treatment of BCR-ABL1- and FLT3/ITD-positive cells with TK inhibitors (TKI) leads to decreased expression of ALT-NHEJ components LIG3 and PARP1 (8, 13, 14) and decreased DSB repair errors, indicating that altered DNA repair is due to one or more effectors of these signaling pathways (12). However, the mechanisms responsible for this repair dysregulation were unknown.
There are several routes through which mammalian cells regulate gene expression, with transcriptional regulation being the most common mechanism. Both FLT3/ITD- and BCR-ABL1–activated leukemias exhibit induction of the oncogenic transcription factor c-MYC (15, 16). In turn, many downstream targets of c-MYC are transcriptionally activated, and this increased signaling may result in enhanced survival and genomic instability (17). In fact, c-MYC interacts with and regulates the promoters of DSB repair genes in several cancer cell lines (18). While the functional significance of this regulation remains unclear, these studies suggest that c-MYC might contribute to the DNA repair abnormalities observed in acute myeloid leukemia (AML) and chronic myelogenous leukemia (CML) cells with BCR-ABL1 and FLT3/ITD mutations. In addition, c-MYC has been shown to negatively regulate specific miRNAs, with c-MYC induction leading to repression of multiple miRNAs and miRNA clusters (19).
In this study, we investigated the role of c-MYC in error-prone repair through increased expression of ALT-NHEJ factors LIG3 and PARP1 in cell lines and primary myeloid malignancies and demonstrated that both c-MYC and c-MYC–regulated miRNAs, miR-150 and miR-22, contribute to increased levels of LIG3 and PARP1 expression, and consequently, ALT-NHEJ activity.
Materials and Methods
Frozen mononuclear cells (MNC) from bone marrow and peripheral blood samples were obtained from patients with CML [chronic phase (CP), n = 15; accelerated phase (AP), n = 1; blast crisis (BC), n = 14] at the Marlene and Stewart Greenebaum Cancer Center and at the University of Milano-Bicocca on protocols approved by the Institutional Review Board. Bone marrow MNC from healthy donors (HD; Lonza) were used as controls for the patient samples.
Cell lines and culture
The murine myeloid precursor cell lines 32Dcl3 and 32D-FLT3/ITD (kindly provided by Dr. Donald Small, Johns Hopkins University, Baltimore, MD), the human megakaryocytic cell lines MO7e and MO7e-BCR-ABL1 (kindly provided by Dr. Richard van Etten, Tufts University, Boston, MA), the human AML cell line MOLM14 (expressing FLT3/ITD), the human leukemia cell line REH (expressing wild-type FLT3), and the human CML cell line K562 (established from a patient in blast crisis; ref. 20) were grown as previously described (11, 14). 293T human embryonic kidney cells were grown in DMEM supplemented with 10% FBS.
The c-MYC expression vectors pBABEpuro-myc-ER (plasmid 19128) and pcDNA3-cmyc (plasmid 16011) were obtained from Addgene and described in Ricci and colleagues (21). The luciferase expression vector pGL4.10 was obtained from Promega. To construct the LIG3 promoter-reporter construct (pGL4.10-LIG3p), we cloned the LIG3 promoter from BAC clone 3143J8 (Invitrogen) using forward-AACCCTAACACCTCCTCTTCCTCT and reverse-TGATCAAGGCTCCCTGAGTCCCA primers (IDT Technologies). The cloned promoter was digested with NheI and EcoRV restriction enzymes (New England Biolabs) and inserted into a digested pGL4.10 vector using a T4 ligase kit (Promega) according to the manufacturer's instructions. The PARP1 promoter-reporter construct was obtained from Switch Gear Genomics. The carbamoyl-phosphate synthetase 2 (CAD) promoter was cloned from gDNA isolated from normal human bone marrow (Lonza) using forward-TGGGAGCCACCACTCTAT and reverse-CGCATCACAGAGTGGGATAA primers and then digested with SacI restriction enzyme and cloned into pGL4.10 using T4 ligase kit. For site-directed mutagenesis of the c-MYC binding sites within the constructs above, we used the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) according to the manufacturer's instructions. All promoter-reporter reporter constructs were sequenced to confirm correct orientation and sequence of promoters, as well as success of site-directed mutagenesis. A GFP-expressing plasmid, pmaxGFP (Lonza), was used throughout the study to control for transfection efficiency. pEF1alpha-DsRed-Express 2 vector (Clontech) was used as a transfection efficiency control for experiments using pEGFP-Pem1-Ad2 vector (kindly provided by Dr. Vera Gorbunova, University of Rochester, NY).
Western blot analysis was performed as previously described (14). c-MYC and Ku70 (Santa Cruz Biotechnology) antibodies were used at 1:500 and 1:1,000, respectively. Anti-LIG3 (clone 7, BD Biosciences) and anti-PARP1 (Cell Signaling Technologies) were used at 1:3000. Anti-LIG4 (GeneTex) was used at 1:1,000. Anti-GRB2 (Cell Signaling Technology) was used at 1:1,000. Anti-actin antibody (Sigma) was used as a loading control at 1:10,000. Secondary antibodies conjugated to horseradish peroxidase (HRP; KPL) were added at a 1:10,000 dilution. Signal was detected using ECL (GE Healthcare). Bands were quantified with ImageQuant software (Bio-Rad).
RNA isolation for real-time PCR of c-MYC, LIG3, and PARP1 transcripts was performed using the RNAspin mini-kit (GE Healthcare). Quantitect Primers (Qiagen) were used for each target, with GAPDH as housekeeping gene. Power SYBR Green RT-PCR Mastermix (Applied Biosystems) was used for reactions carried out in triplicates using a Mastercycler ep (Eppendorf). Relative quantification was determined using Eppendorf realplex software according to the ΔΔCT method. For ChIP-qPCR, we used HotStart-IT SYBR Green qPCR 2X Mastermix (Affymetrix) per the manufacturer's instructions.
Nanosting microRNA analysis
Paired CML-CP (n = 5), CML-BC (n = 5), and from HD (n = 3) MNCs were processed using Ficoll-Paque plus (GE Healthcare Life Sciences). Total RNA was extracted with TRIzol (Invitrogen) per the manufacturer's instructions, and 100 ng per sample was used to assess differences in miRNA expression using multiplexed NanoString enzyme-independent probe-based quantification system that allows digital counting of individual miRNA molecules (nCounter; NanoString Technologies), as previously described (22). NanoString assays were performed at the Ohio State University Comprehensive Cancer Center Microarray Facility (Columbus, OH). Heatmap analysis was performed using z-score.
Cells (2 × 106, 0.5 × 106 per mL) were washed in Opti-MEM medium (Invitrogen) and transfected with SMARTpool siRNA (Thermo Scientific Dharmacon), or single siRNA (Cell Signaling Technologies) using the Amaxa Nucleofection System (Lonza) as previously described (14).
Total RNA was isolated using miRNeasy kit (Qiagen) per the manufacturer's instructions. Levels of miRNAs were determined by qPCR using a TaqMan qRT-PCR kit (Life Technologies), with U18 serving as the housekeeping miRNA. miRNA expression and microarray analyses were carried out as previously described (23). For miRNA overexpression experiments, K562, MO7e-BCR/ABL1, and MOLM-14 cells were transfected with 50 nmol/L miR-34a, -22, -150, or a combination of miR-22 and -150 (Thermo Scientific-Dharmacon) and harvested for Western blotting and NHEJ repair analyses at 24, 48, or 72 hours posttransfection.
Chromatin immunoprecipitation (ChIP) was performed according to Luoto and colleagues (18), with the following modifications. Lysates were sonicated on ice 10 times using a Branson 450 Sonifier for 10 seconds each time, at 40% output control and 2.5 duty cycle, with 30-second refractory periods between sonications. ChIP products were purified using a Qiaquick kit (Qiagen) followed by qPCR using the following primer sets:
CAD(forward) 5′- TAGCCACGTGGACCGACT-3′
CAD(reverse) 5′- TACGGAGAAGCGGGAAGGA-3′
Ch22(forward) 5′- GGATGACAGGCATGAGGAATTA-3′
Ch22(reverse) 5′- TGCTGCTTACTTGGGATATGAG-3′
32D-FLT3/ITD and MO7e-BCR-ABL1 cells were transfected with control or c-MYC–targeting siRNA as described above. Transfected cells were incubated for 48 hours and then harvested for transfection with luciferase constructs pGL4.10 (promoterless control, Promega) or pGL4.10-LIG3p and pGL4.10-PARP1p (luciferase vectors containing LIG3 and PARP1 promoters, respectively). Transfection efficiency was determined by cotransfection with pRL-TK vector (Promega) containing HSV thymidine kinase promoter. Measurement of luciferase activity was performed with Dual Luciferase Reporter Assay System (Promega) using an HT Synergy plate reader (Bio-TEK).
In vivo NHEJ assay
The in vivo NHEJ assay was performed as previously described (14). Briefly, 0.2 μg of linearized pUC18 plasmid was transfected into 2 × 106 cells 48 hours posttransfection with siRNA against c-MYC or miR-22/150 combination, respectively. Repaired plasmid clones were sequenced at the repair junction, and sequences were analyzed using BioEdit sequence alignment editor (http://www.mbio.ncsu.edu/bioedit/bioedit.html). Three independent c-MYC knockdown or miRNA overexpression experiments, followed by 3 independent in vivo NHEJ assays were performed to confirm results. To determine the relative end-joining efficiency after siRNA knockdown, we used the GFP-based end-joining assay described by Fattah and colleagues (Fattah, 2010, #11559). Briefly, cells were transfected with siRNA for 48 hours, followed by transfection with HindIII-linearized pEGFP-Pem1-Ad2. Twenty-four hours later, the cells were analyzed by flow cytometry for GFP. For this assay, a DsRed expression plasmid (pEF1alpha-DsRed-Express 2) was used to normalize for any variations in transfection efficiency.
Statistical analysis comparing experimental and control groups was performed using the Student t test and z-test (for differences in frequencies of microhomology-mediated repair and proportions of patient samples). The correlation coefficient (Pearson R) was used to determine positive or negative correlations.
c-MYC expression is elevated in TK-activated myeloid leukemias and correlates with LIG3 and PARP1 expression levels
Our previous studies showed that TK-activated leukemias had increased steady-state levels of PARP1 and LIG3 mRNA and protein (12), suggesting that they are transcriptionally regulated. Following database analysis (using Matinspector from Genomatix) and TF Search (Parallel Application TRC Laboratory) of promoter regions of these genes, c-MYC binding sites were revealed and thus c-MYC was investigated as a candidate in transcriptional regulation of these genes.
To determine whether c-MYC plays a role in regulating expression of LIG3 and PARP1, we initially examined microarray data for multiple myeloid leukemia cell lines and primary cells to correlate mRNA expression levels of c-MYC with either LIG3 and PARP1 mRNA expression levels. Analysis of AML cell lines (HL60, Kasumi1, KG1a, ML2, Mo7e, and U937), the BCR-ABL1+ CML cell line in blast crisis K562 and primary AML patient samples (n = 18; Supplementary Table S1) revealed that endogenous levels of c-MYC showed a significant positive correlation, with levels of LIG3 (Pearson R = 0.6492; Fig. 1A) and PARP1 (Pearson R = 0.8422; Fig. 1B).
Given the strong correlation of c-MYC mRNA levels with LIG3 and PARP1 mRNA levels in AML, we hypothesized that increased c-MYC expression may augment expression of LIG3 and PARP1 in TK-activated leukemias, in which c-MYC is known to be expressed at high levels (15, 16). We first assessed protein levels of c-MYC, LIG3, and PARP1 in cell extracts from hematopoietic cell lines that stably express BCR-ABL1 (MO7e-BCR-ABL1) or FLT3/ITD (32D-FLT3/ITD) and their respective parental controls (MO7e and 32Dcl3). In addition, leukemia cell lines endogenous protein levels were compared in FLT3/ITD-positive MOLM14 versus wild-type FLT3 REH and BCR-ABL1–positive K562 versus BCR-ABL1–negative MO7e leukemia cell lines. Figure 1C shows that cells expressing TKs have an approximately 2-fold increase in steady-state protein levels of c-MYC, LIG3, and PARP1 compared with controls (P < 0.05). In addition, c-MYC–positive control Cyclin A (CCNA2) shows an approximately 2-fold increase in TK-activated cells, whereas GRB2, which has been reported not to be controlled by c-MYC (24), is not significantly altered (Fig. 1C, bottom).
c-MYC inhibition results in decreased LIG3 and PARP1 levels in FLT3/ITD- and BCR-ABL1–positive cells
Next, to determine whether depletion of endogenous c-MYC results in decreased levels of LIG3 and PARP1 transcripts and proteins, siRNA-mediated knockdown of c-MYC was performed in MO7e-BCR-ABL1 and 32D-FLT3/ITD, followed by qPCR and Western blotting of mRNA and protein levels, respectively. To ensure that observed effects of siRNA-mediated c-MYC knockdown were not due to cell death, we performed experiments over a 48-hour period and confirmed cell viability by trypan blue dye exclusion as well as by MTS-based viability assays (Supplementary Fig. S1A and S1B). siRNA-mediated knockdown in above cell lines, using pooled oligonucleotides of c-MYC (>50%), led to a significant reduction (P < 0.01) in both mRNA and protein levels of LIG3 and PARP1 (Fig. 1D and E). Similar results were observed upon c-MYC knockdown in the FLT3/ITD-positive cell line MOLM14 (Supplementary Fig. S1C). siRNA-mediated c-MYC knockdown, using single oligonucleotides, yielded similar results to experiments using pooled oligonucleotides (Supplementary Fig. S1D). Furthermore, chemical inhibition of c-MYC [10058-F4, 60 μmol/L (selleckchem.com)] which inhibits MYC-MAX binding and prevents transcriptional activation of c-MYC targets (25) resulted in a significant reduction of LIG3 and PARP1 transcripts (Supplementary Fig. S1E).
c-MYC induces the transcription of LIG3 and PARP1 in FLT3/ITD- and BCR-ABL1–positive cells
Database searches using Matinspector (Genomatix) and TF Search (Parallel Application TRC Laboratory) revealed several predicted c-MYC binding sites (E-box motifs) within the defined PARP1 and LIG3 promoter regions (Fig. 2A). In addition, ChIP-seq data from ENCODE (UCSC Genome Browser) revealed the increased binding of c-MYC to several DSB repair gene promoters, including PARP1 and LIG3 in the BCR-ABL1–positive cell line K562, similar to well-established c-MYC targets, CAD and CCNA2 (Fig. 2B). In addition, genes reported not to be controlled by c-MYC, such as GRB2 (24) or immediate early gene FOS (26), had relatively little c-MYC binding (Fig. 2B). To determine the significance of c-MYC binding to the promoters of ALT-NHEJ genes, the promoter regions of LIG3 and PARP1 (containing the putative c-MYC binding sites) were cloned into pGL4.10 luciferase expression plasmids. Plasmid constructs were transfected into 32D/ITD and MO7e-BCR-ABL1 cells, and induction of luciferase expression via endogenous c-MYC was determined by luciferase assay (Materials and Methods). Cells transfected with a plasmid containing either the LIG3 or PARP1 promoter, compared with a promoterless control vector, induced luciferase expression 4- to 25-fold (Supplementary Fig. S2A). Figure 2C shows that baseline LIG3 and PARP1 promoter induction of luciferase expression in MO7e-BCR-ABL1 and 32D/ITD cells was approximately 3-fold higher (P < 0.01) compared with that in parental control cells. Importantly, c-MYC knockdown in 32D/ITD and MO7e-BCR-ABL1 cells resulted in a significant reduction (P < 0.05) in promoter activity as shown in Fig. 2C.
To further confirm that c-MYC induces the transcription of the LIG3 and PARP1 genes, we used 2 c-MYC expression vectors described by Ricci and colleagues (21). First, pBABE-puro-mycER was transfected into MO7e-BCR-ABL1 cells which were subsequently treated with either vehicle (ethanol) or 300 nmol/L 4-hydroxytamoxifen (4-OHT). 4-OHT induces translocation of mycER from the cytoplasm to the nucleus and activation of c-MYC–mediated transcription. Induction of c-MYC resulted in approximately a 1.7-fold (P = 0.014) increase in LIG3 mRNA and a 2.5-fold (P = 0.005) increase in PARP1 mRNA levels at 12 hours postinduction with 4-OHT. Cyclin D2 (CCND2) mRNA, which was induced approximately 2-fold (P = 0.01) by mycER, was used as a positive control, as Cyclin D2 was previously shown to be a downstream target of mycER-induced transcription (27). MycER protein expression was also verified by Western blot analysis (Fig. 2D). We also used a well-established c-MYC overexpression model (21) in which the c-MYC expression vector, pcDNA3-cmyc, is cotransfected into 293T cells along with PARP1 and LIG3 promoter luciferase constructs and luciferase activity is measured in comparison to empty vector pcDNA3 controls. c-MYC overexpression (Fig. 2E) resulted in 2.5-fold (P = 0.013) and 1.8-fold (P = 0.005) increase in LIG3 and PARP1 promoter activities, respectively. Notably, when compared with parental controls, LIG3 and PARP1 luciferase activity levels were similar to those in TK-positive cells (2- to 3-fold increase, Fig. 2C). Furthermore, site-directed mutagenesis of the highest probability c-MYC binding sites (as determined by Genomatix software) within the LIG3 and PARP1 promoter–reporter constructs (Fig. 2A), resulted in a significant decrease in LIG3 (40%; P = 0.02) and PARP1 (30%; P = 0.004) promoter activities, compared with cells transfected with wild-type promoter-reporter vectors (Fig. 2E). Importantly, cells transfected with the mutant promoter–reporter constructs showed significantly decreased luciferase activity, compared with controls (Fig. 2E). We obtained similar results as above using a promoter–reporter vector of carbamoyl-phosphate synthase 2 (CAD), a well-established target of c-MYC–induced transcription (28), as a positive control (Supplementary Fig. S2B).
To determine whether c-MYC actually binds to the promoters of the LIG3 and PARP1 genes and to confirm the ChIP-seq data (from ENCODE) observed in K562 cells (Fig. 2B), we performed ChIP in extracts from MO7e-BCR-ABL1–positive (Fig. 2F) and FLT3/ITD-positive MOLM14 cells (data not shown), followed by PCR amplification of LIG3 and PARP1 promoter regions spanning the putative c-MYC binding sites. qPCR analysis of ChIP products confirmed c-MYC binding to the LIG3 and PARP1 promoters, compared with positive (CAD), negative (Ch22), and IgG controls (Fig. 2F).
c-MYC regulation of LIG3 and PARP1 expression is dependent on expression of FLT3/ITD and BCR-ABL1
Given that c-MYC is a downstream target of the FLT3/ITD and BCR-ABL1 TKs (15, 29), we sought to investigate whether c-MYC regulation of LIG3 and PARP1 is dependent on the activities of these TKs. Therefore, FLT3/ITD-positive MOLM14 and MO7e-BCR-ABL1 cells were treated with TKIs (50 nmol/L of the FLT3 inhibitor CEP701 or 5 μmol/L of the BCR-ABL1 inhibitor imatinib, respectively) or DMSO. Thereafter, qPCR analysis of mRNA and Western blot analysis of protein levels were performed. FLT3/ITD inhibition resulted in greater than 75% (P < 0.005) decrease in c-MYC, LIG3, and PARP1 mRNA levels (Fig. 3A and B). Imatinib treatment of MO7e-BCR-ABL1 cells resulted in smaller (20%–30%) yet significant decrease (P < 0.04) in mRNA and protein levels of c-MYC, LIG3, and PARP1 (Fig. 3A and B). As seen in Fig. 3C, TKI treatment of MOLM14 and MO7e-BCR-ABL1 resulted in similar effects on the protein levels of LIG3 and PARP1 as was observed for mRNA levels (Fig. 3A and B). Of note, we did not observe the presence of cleaved PARP1 (Fig. 3C), suggesting that the downregulation observed was not a result of the cells undergoing apoptosis. To ensure that PARP1 can undergo cleavage at this time point, we treated cells with 50 nmol/L (MOLM14) or 1 μmol/L (MO7e-BCR-ABL1) etoposide and demonstrated PARP1 cleavage (Fig. 3D). Given that c-MYC phosphorylation at Serine 62 (Ser62) is an indication of its signaling activity (30), we performed Western blotting for this c-MYC phosphorylation site following treatment with the TKIs described above and showed that phospho-Ser62 is decreased significantly following treatment in both cell lines (Fig. 3E).
c-MYC negatively regulated miRNAs decrease LIG3 and PARP1 expression in FLT3/ITD- and BCR-ABL1–positive cells
c-MYC is known to specifically regulate expression of multiple miRNAs, the majority of which undergo negative regulation by this transcription factor (19). To determine whether c-MYC–mediated decrease in miRNA levels in leukemia cells may be an additional mechanism of LIG3 and PARP1 regulation, we examined the 3′-untranslated region (UTR) and coding sequences of LIG3 and PARP1 for potential c-MYC–regulated miRNA binding sites (Supplementary Table S2). LIG3 and PARP1 are predicted targets of multiple c-MYC–regulated miRNAs, including miR-22, miR-27a, miR-34a, and miR-150. In our initial studies, we used microarray analysis of mRNA expression and mature miRNA levels from multiple AML cell lines and primary samples (Supplementary Table S1) shown in Fig. 1A and B, to look for a correlative relationship between LIG3 or PARP1 and the above miRNAs. There was a significant inverse correlation between LIG3 and miR-22 and -150 (Pearson r < −0.3, Fig. 4A); similarly, there is a significant inverse correlation between PARP1 and miR-22 and -150 (Pearson r < −0.3, Fig. 4B). Not surprisingly, there was a significant inverse correlation between c-MYC and miR-22 and -150 (Pearson r < −0.3, Fig. 4C). In contrast, miR-34a showed neither a positive nor inverse correlation with either LIG3 or PARP1 (Supplementary Fig. S3A). While miR-27a demonstrated an inverse correlation with PARP1, it did not significantly correlate with LIG3 (Supplementary Fig. S3B). Accordingly, we focused on miR-22 and -150 for the remainder of this study.
First, we sought to confirm c-MYC repression of miR-22 and -150 in our cell lines. We performed siRNA-mediated knockdown of c-MYC in 3 different cell lines (MOLM14, K562, and MO7e-BCR-ABL1) followed by qRT-PCR analysis of miR-22 and -150 levels. There was a significant increase in levels of miR-22 and/or -150 in cells transfected with siRNA targeting c-MYC, compared with cells transfected with control siRNA (Supplementary Fig. S3C). To confirm that miR-22 and -150 are involved in regulating the expression of LIG3 and PARP1 in TK-activated leukemias, we transiently transfected a combination of miR-22 and miR-150 mimics (50 nmol/L each) into MOLM14 and MO7e-BCR/ABL cells and assessed LIG3 and PARP1 protein levels at 72 hours posttransfection. We verified the transfection efficiency and confirmed high levels of miRs by qPCR in several cell lines (Supplementary Fig. S3D). Transfection of a combination of miR-22 and -150 produced an approximately 35% to 40% decrease in LIG3 (P < 0.05) and PARP1 (P < 0.05) protein levels, respectively, compared with MOLM14, and MO7e-BCR-ABL1 cells transfected with a nonspecific control (NSC) miRNA (Fig. 4D and E).
Depletion of c-MYC and overexpression of c-MYC–regulated miR-150 and miR-22 decrease ALT-NHEJ activity in FLT3/ITD- and BCR-ABL1–positive cells
To determine whether c-MYC regulation of key ALT-NHEJ components in TK-activated leukemias also plays a role in ALT-NHEJ repair activity, we performed an established in vivo plasmid-based end-joining assay (11, 12, 31, 32) following chemical and siRNA inhibition of c-MYC. In this assay (Fig. 5A), pUC18 plasmid is linearized by restriction enzyme digestion to simulate a DSB and then transfected into cells. The repaired plasmids are isolated from cells and transformed into DH5α bacteria, which are then plated on agar plates. Colonies (representing clones of one repaired plasmid) are picked for plasmid isolation and sequencing of repair junctions. c-MYC was inhibited in MOLM14 and MO7e-BCR/ABL1 using chemical inhibition (10058-F4) or siRNA technology, followed by the above end-joining assays. c-MYC knockdown and consequent decrease in LIG3 and PARP1 levels were confirmed by Western blot analysis (Supplementary Fig. S4A and S4B). We also verified that the transfection efficiency of the linearized pUC18 plasmid was not affected by c-MYC knockdown by cotransfection with a GFP-expressing vector (Supplementary Fig. S4C). Approximately 30 colonies from 3 independent experiments were analyzed by colony PCR and sequencing of the region that encompasses the repaired DSB. MOLM14 cells depleted of c-MYC by chemical inhibition had a general reduction in the size of DNA deletions when compared with DMSO-treated cells (Fig. 5B, P < 0.015). Similar results were obtained following siRNA-mediated knockdown of c-MYC in MO7e-BCR-ABL1 cells (Fig. 5C, P < 0.05). Importantly, repair assays performed with linearized pUC18 DNA controls showed very few colonies (Supplementary Fig. S4D). In addition, we analyzed the data for the proportion of large deletions (>20 bp) versus small deletions (<20 bp) that are characteristic of c-NHEJ (Supplementary Fig. S4E). Approximately 50% of DSBs are repaired in TK-activated cell lines MOLM14 and MO7e-BCR-ABL1 with small deletions occurring at the repair junctions. Following siRNA knockdown of c-MYC, the fraction of small deletions actually increases significantly (Supplementary Fig. S4E). In fact, siRNA knockdown of c-MYC results in a slight but insignificant increase in NHEJ efficiency using a different NHEJ reporter (pEGFP-Pem1-Ad2) assay (Fattah, 2010, #11559). In contrast, knockdown of control c-NHEJ factor, LIG4, reduces NHEJ efficiency significantly (Supplementary Fig. S4F and S4G). This suggests that c-MYC may not contribute significantly to c-NHEJ activity. Moreover, c-MYC depletion decreased the frequency of microhomology-mediated repair (Fig. 5D, P < 0.05). Notably, overexpression of miR-150 and miR-22 in MO7e-BCR-ABL1 cells produced the same results seen with c-MYC inhibition (Fig. 5E and F). The reduction in LIG3 and PARP1 protein levels after miR overexpression was confirmed by Western blot analysis (Supplementary Fig. S4H).
c-MYC mRNA levels correlate with LIG3 and PARP1 mRNA and inversely correlate with miR-22 and -150 in primary samples from patients with CML
To determine whether primary CML samples demonstrated increased expression levels of c-MYC and decreased levels of miR-150 and -22, we first examined mRNA samples from 21 patients with CML (Table 1) and compared transcript levels of c-MYC, LIG3, and PARP1 in peripheral blood and bone marrow bone marrow mononuclear cells (BMMNC) samples with those of normal bone marrow (NBM) controls. As seen in Fig. 6A and B, there was a strong positive correlation between c-MYC mRNA levels and LIG3 (Pearson r = 0.88, P < 0.001) and PARP1 (Pearson r = 0.84, P < 0.001) mRNA levels. c-MYC expression was increased in 12 of 21 samples (57%), compared with normal controls, Of the 12 CML samples with increased c-MYC levels, 9 (75%) also had increased levels of LIG3 and PARP1 (Table 1). In these samples with increased levels of c-MYC, PARP1, and LIG3, there was no bias toward a particular disease phase or response to TKIs. Interestingly, however, 2 of 3 TKI-resistant patients with CML (Table 1, PT #8 and #16) with T315I-BCR-ABL1 mutations (34, 35) had increased levels of c-MYC, LIG3, and PARP1.
|Patient .||Phase .||Mutations .||IM status .||LIG3 up .||PARP1 up .||c-MYC up .|
|Patient .||Phase .||Mutations .||IM status .||LIG3 up .||PARP1 up .||c-MYC up .|
NOTE: Increased LIG3, PARP1 and c-MYC expression relative to normal control bone marrow MNC and based on P < 0.05 by the t test.
Abbreviations: IM, imatinib; N, no; Y, yes.
aPatient samples used for miRNA analysis.
CML patient samples were next evaluated for expression of c-MYC–regulated miRs. We first examined mRNA for miR-150 in a subset of the above-described patient samples (n = 8; Table 1, PT #6, #8, #13, #17, #18, #20, #21) for which mRNA was available and 2 additionally acquired patients with CML (Supplementary Table S3; #22 in chronic phase and #23 in blast crisis) by miRNA qPCR analysis and compared results with NBM (n = 2). There was a significant decrease (P < 0.05) in c-MYC–repressed miR-150 levels in all CML samples, compared with NBMC (Fig. 6C). We next examined miR-22 expression levels in CML patient samples in chronic phase (n = 4; Supplementary Table S3; PT #13, #24, #25, and #26) versus those in blast crisis (n = 4; Supplementary Table S3; #8, #18, #20, and #23). In contrast to results with miR-150, miR-22 expression levels appear to be decreased significantly (P < 0.01) in patients with CML-BC when compared with patients with CML-CP (Fig. 6D). To confirm these data, we analyzed NanoString mIR expression arrays from a separate cohort of CML patient mRNAs from paired MNC samples of patients with CML-CP (n = 5) and -BC (n = 5) healthy donor (HD) controls (n = 3). As expected from the earlier correlation studies (Supplementary Fig. S3B), miR-27a showed no significant differences in expression between CML patient samples and healthy donor controls (Fig. 6E). In contrast, and as seen in the first cohort of CML patient samples (Fig. 6C), levels of miR-150 were significantly decreased in both chronic phase and blast crisis compared with healthy donors, and there were no significant expression differences between chronic phase and blast crisis in the 5 paired patient samples examined (Fig. 6E). Similar to the above results (Fig. 6D), compared with healthy donor controls, miR-22 expression was decreased significantly only in blast crisis, compared with healthy donor controls (Fig. 6E).
Leukemia cells expressing constitutively activated BCR-ABL1 or FLT3/ITD TKs are characterized by a highly error-prone and alternative form of NHEJ (ALT-NHEJ) involving increased expression of LIG3 and PARP1, that is likely responsible for much of the error-prone end-joining repair (12, 14). To date, the mechanism through which TKs activate ALT-NHEJ factors had not been elucidated. Our studies have, for the first time, linked c-MYC, a key downstream target of both BCR-ABL1 and FLT3/ITD to increased expression of the ALT-NHEJ factors LIG3 and PARP1 and increased ALT-NHEJ repair activity.
While elevated expression of c-MYC occurs frequently in human cancers and is associated with tumor progression and poor clinical outcome (36), the effect of high levels of c-MYC on global gene regulation is poorly understood. Recent studies suggest that high levels of c-MYC accumulate in the promoter regions of active genes and cause transcriptional amplification, producing increased levels of transcripts within the cell's gene expression program (37). Our studies demonstrate that c-MYC binds to the promoter regions of both LIG3 and PARP1, thereby increasing transcriptional activity, leading to increased ALT-NHEJ in TK-activated cells. Thus, ALT-NHEJ activity, which is present at low levels in normal cells (11), is increased through transcriptional activity of the key pathway components LIG3 and PARP1, likely leading to acquisition of genomic alterations. Given the role of c-MYC in cell-cycle progression, the possibility exists that the expression of PARP1 and LIG3 and therefore ALT-NHEJ may be influenced by the cell cycle. Studies performed in cells enriched for particular phases of the cell cycle should clarify this question. Given that c-MYC is known to bind several DSB repair genes (18), increased expression of LIG3 and PARP1 may be part of a dysregulated expression program involving abnormal DSB repair that leads to genomic instability. Therefore, similar studies are merited for other DSB repair genes. Thus, while increased ALT-NHEJ activity may be a plausible mechanism for genomic instability, and studies that measure actual chromosomal instability are warranted (38).
The most extensively used point of gene regulation is at the level of transcription, whereas miRNAs are thought to be involved in “fine-tuning” gene regulation. miRNAs are involved in the regulation of many cancer-specific signaling pathways in hematopoiesis and are aberrantly expressed in hematologic malignancies (39). c-MYC is known to negatively regulate a group of miRNAs, which is thought to contribute to tumorigenesis (19). In the context of LIG3 and PARP1 regulation, we propose in addition to its transcriptional regulation of these genes, c-MYC also represses miRNAs (miR-22 and miR-150) that target LIG3 and PARP1 transcripts. Indeed, this dual regulation by c-MYC was previously reported; c-MYC was shown to regulate the transcription factor E2F1 through direct transcription and activation of miRNAs that target E2F1 (40). Another group showed that expression of EZH2, a protein involved in regulation of gene expression, was amplified as a result of c-MYC–mediated repression of miR-26a, which targets EZH2 (41).
We previously reported that treatment of BCR-ABL1- and FLT3/ITD-positive cells with TK inhibitors leads to decreased expression of the ALT-NHEJ components LIG3 and PARP1 (8, 13, 14) and decreased DSB repair errors, indicating that the altered DNA repair is due to one or more of the effectors of these signaling pathways (12). Our studies here show that the downstream TK target c-MYC mediates expression of the ALT-NHEJ factors PARP1 and LIG3 that is dependent on BCR-ABL1 or FLT3/ITD. Furthermore, our studies demonstrate that the TKI CEP-701 strongly downregulates expression of c-MYC, LIG3, and PARP1 in FLT3/ITD-positive cells, compared with a weaker effect of imatinib in BCR-ABL1–positive cells. One explanation is that the mechanisms of drug action of the specific inhibitors may differ. Whereas imatinib targets BCR-ABL1 activity, studies have shown that CEP701 inhibits FLT3 and Jak2 (42), both of which are important for c-MYC activity (15, 29). Another explanation may be that the mechanisms of c-MYC activation may differ in the 2 models of TK-activated leukemia. In Bcr-Abl1–positive leukemias, Jak2 mediates the increase c-MYC RNA expression but also interferes with proteasome-dependent degradation of c-Myc protein (29, 43). In contrast to studies in BCR-ABL1–positive cells, while gene expression induced by FLT3/ITD or constitutively activated wild-type FLT3 signaling induces high expression levels of c-MYC (15), few mechanistic studies have been reported to elucidate how c-MYC is regulated by FLT3/ITD.
Previous studies demonstrate that c-MYC expression increases as CML progresses to blast crisis, and this overexpression induces aberrant DNA synthesis, suggesting a role for c-MYC in disease progression through promoting genomic instability (44). Our studies shows that in the majority (75%) of CML patient samples (irrespective of disease stage) with increased c-MYC expression, LIG3 and PARP1 expression is also increased. This suggests that increased ALT-NHEJ is promoted throughout the course of CML disease. However, in a minority (25%) of primary CML samples, c-MYC expression is not correlated with LIG3 and PARP1 expression, suggesting that additional molecular mechanisms may be involved in regulating LIG3 and PARP1 and ALT-NHEJ. Expression of miR-150 which is negatively regulated by c-MYC was significantly decreased (compared with normal bone marrow) in all CML patient samples tested, regardless of disease stage, suggesting that miR-150 is repressed in a BCR-ABL1–dependent manner (45). In contrast, miR-22 appears to be preferentially decreased in patients with CML-BC. One study (46) demonstrated a role for miR-22 in inducing cellular senescence, thereby preventing cancer progression. Considering that one of the hallmarks of blast crisis is increased genomic instability, it is tempting to speculate that miR-22 suppression in blast crisis may override DNA damage–induced senescence, leading to promotion of genomic instability. Studies into the exact function of miR-22 in blast crisis with respect to genomic instability would be of particular interest.
Published studies also demonstrate higher c-MYC expression in samples from patients not responding to imatinib therapy, compared with those responding to therapy (44). Interestingly, 2 of 3 bone marrow samples with T315I mutations, connoting the greatest resistance to TKIs showed elevated steady-state levels of c-MYC, LIG3, and PARP1. Therefore, increased ALT-NHEJ may exist in a cellular milieu of increased c-MYC activity in which multiple aberrant repair pathways are active, leading to acquisition of TKI-resistant BCR-ABL1 mutations.
From a therapeutic standpoint, targeting c-MYC in leukemia may not be a viable option, considering the myriad of roles c-MYC plays in normal cells. However, the identification of other “drugable” downstream targets of c-MYC may be considered. With respect to our current study, PARP1 represents an attractive target for therapy, as PARP1 inhibitors are currently used in the clinic, and new ones are being investigated in clinical trials. In fact, a recent study by our collaborators demonstrated that PARP1 inhibition is synthetically lethal in TK-driven leukemias that exhibit defects in the HR pathway (47). Therefore, given that DSBs are lethal if not repaired, the c-MYC downstream targets LIG3 and PARP1 and other DSB repair factors may be attractive targets for therapy (48).
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: N. Muvarak, C. Gambacorti-Passerini, C. Civin, F. Rassool
Development of methodology: K. Scheibner
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kelley, M.R. Baer, D. Perrotti
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Muvarak, S. Kelley, C. Robert, K. Scheibner, F. Rassool
Writing, review, and/or revision of the manuscript: N. Muvarak, M.R. Baer, C. Civin, F. Rassool
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Kelley
Study supervision: N. Muvarak, C. Gambacorti-Passerini, F. Rassool
The authors thank Jason G. Harb, Paolo Neviani, and Justin Ellis (Ohio State University, Columbus, OH) for scientific and technical assistance with the NanoString array preparation and data analysis.
This work was supported by Cigarette restitution funds of Maryland (F. Rassool), NIH grants CA095512 and CA163800 (D. Perrotti), The CARIPLO Foundation Biomedical Scientific Research (C. Gambacorti-Passerini and D. Perrotti), MSCRF (K. Scheibner and S. Kelley), and by NCI Cancer Center Grant P30 CA134274.
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.