miR-21-mediated Radioresistance Occurs via Promoting Repair of DNA Double Strand Breaks*

miR-21, as an oncogene that overexpresses in most human tumors, is involved in radioresistance; however, the mechanism remains unclear. Here, we demonstrate that miR-21-mediated radioresistance occurs through promoting repair of DNA double strand breaks, which includes facilitating both non-homologous end-joining (NHEJ) and homologous recombination repair (HRR). The miR-21-promoted NHEJ occurs through targeting GSK3B (a novel target of miR-21), which affects the CRY2/PP5 pathway and in turn increases DNA-PKcs activity. The miR-21-promoted HRR occurs through targeting both GSK3B and CDC25A (a known target of miR-21), which neutralizes the effects of targeting GSK3B-induced CDC25A increase because GSK3B promotes degradation of both CDC25A and cyclin D1, but CDC25A and cyclin D1 have an opposite effect on HRR. A negative correlation of expression levels between miR-21 and GSK3β exists in a subset of human tumors. Our results not only elucidate miR-21-mediated radioresistance, but also provide potential new targets for improving radiotherapy.

MicroRNA-21 (miR-21) overexpresses in almost all types of human tumors (1,2). Previously, we reported that up-regulating miR-21 in human non-tumorigenic cells could promote the development of the cells into tumorigenic cells (3), and other groups have used mouse models to show that up-regulating miR-21 promotes tumorigenesis (4 -6). These results demonstrate that miR-21 is an onco-miR. In addition, miR-21 was involved in tumor cell resistance to ionizing radiation (IR) 2 (7,8); however, the underlying mechanism remains unclear. IRinduced cell killing occurs mainly by generating DNA double strand breaks (DSB). Non-homologous end-joining (NHEJ) and homologous recombination repair (HRR) are the two main DNA DSB repair pathways in mammalian cells (9,10). In general, any gene-mediated radioresistance should directly or indirectly promote NHEJ, HRR, or both. Therefore, the purpose of this study is to examine whether miR-21-mediated cell radioresistance occurs through promoting repair of DNA DSB, and if so, how miR-21 stimulates DNA DSB repair.

miR-21-mediated Radioresistance Occurs through Promoting
DNA DSB Repair-To confirm miR-21-mediated radioresistance, we examined the survival sensitivity of miR-21 knock-in mice (generated in our lab (29)) and miR-21 knock-out mice (obtained from Dr. Olson's lab (30)) to IR. The miR-21 knockout mice or mouse embryo fibroblast (MEF) cells derived from the mice did not show a detectable miR-21 level. The miR-21 levels in the miR-21 knock-in mice were shown to be 3-12-fold higher in different organs versus levels in the wild type counterparts. The miR-21 levels in the MEF derived from the miR-21 knock-in mice were shown to be 6 -8-fold higher than that from the wild type mice. The survival results showed that when compared with the wild type counterparts, miR-21 knock-in mice or MEF cells were much more radioresistant and miR-* This work was supported by National Institutes of Health Grants CA186129 and CA185882 (to Y. W.) and P30CA138292 (to Winship Institute). This work was also supported by China National High Technology Research and Development Program Grant 2014AA020604 (to X. Y.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 1  21 Ϫ/Ϫ mice or their MEF cells were much more radiosensitive (Fig. 1, A and B). These results demonstrate that miR-21 upregulation contributes to radioresistance in mice. To study whether miR-21-mediated radioresistance is involved in promoting DNA DSB repair, we examined the kinetics of ␥-H2AX foci (DNA DSB marker) following IR exposure as we described previously (31) among wild type, miR-21 knock-in, or miR-21 Ϫ/Ϫ MEF cells at different times following IR (2 Gy) (Fig. 1C). The results showed that there was no significant difference in the foci number (per cell) at 0.5 h after IR (Fig. 1D), indicating that miR-21 does not affect the IR-induced yield of DSB. However, the foci number among these cell lines had significant differences at 4 h (p Ͻ 0.05) and 8 h after IR (p Ͻ 0.01) (Fig. 1D), supporting that miR-21-mediated radioresistance occurs through promoting repair of DNA DSB.
To verify the effects of miR-21 on radioresistance in human cells, we initially compared the miR-21 levels in 30 pairs of human lung cancer tissue with their adjacent non-tumor tissue in wax blocks (obtained from the Department of Pathology and Laboratory Medicine at the Emory University School of Medicine) as well as in 12 human non-small cell lung cancer (NSCLC) cell lines (compared with two human bronchial epithelial cell lines, 3KT (obtained from Dr. Mina's lab (32)), and NL20 (purchased from the American Type Culture Collection (ATCC)) using NanoString technology (33) (Fig. 2A). The results showed that the miR-21 levels were up-regulated in 24 tumors and 12 cell lines (A549, H226B, H358, H460, H522, H1299, H2792, H1957, Calu1, SKMES-1, 95C, and 95D) among the 30 pairs of samples and 12 human NSCLC lines when compared with their non-tumor counterparts (Fig. 2B). Only H157 and A2780 cells demonstrated lower miR-21 levels than in 3KT and NL20 control cells (Fig. 2B). We then examined the effects of regulating the miR-21 level in human cell lines on cell radiosensitivity. The NL20 cell lines, miR-21-1 and miR-21-2, which are stably up-regulated with miR-21 (generated in our lab (34)), and four human NSCLC cell lines, A549, H460, H358, and H2792, which are down-regulated with miR-21 using a miR-21 inhibitor (antisense RNA) to block miR-21 function, were used for the radiosensitivity experiments. The results showed that when miR-21 was up-regulated in the normal cells (6-fold), the cells became more resistant to IR (Fig. 2C), and when miR-21 was blocked by its inhibitor, the radioresistance was abolished in the human tumor cells (Fig.  2D). These results further demonstrate that miR-21 is associated with radioresistance in human cells.
miR-21 Promotes Both NHEJ and HRR-To investigate which DNA DSB repair pathway, NHEJ or HRR, was promoted by miR-21, we used the reporter assay. Interestingly, up-regulating miR-21 in miR-21 knock-in MEF cells (Fig. 3A) or using miR-21 mimics in human cells (Fig. 3E, top panel) promoted NHEJ and HRR efficiency in both MEF (Fig. 3, B and C) and human cells (Fig. 3E, middle and bottom panels). To identify which NHEJ or HRR relative factors were affected by miR-21, we examined the major NHEJ factors including DNA-PKcs, Ku70, ligase IV, and XRCC4, and the major HRR factors including RAD51, RPA34, RPA70, XRCC2, and XRCC3, as well as cyclin D1. We did not find any change in the protein level for the tested NHEJ factors in MEF (Fig. 3D, left panel) or human cells (Fig. 3F, top panel), but found that DNA-PKcs autophosphorylation signal increased after up-regulating miR-21 in MEF (Fig. 3D, left panel) or human cells (Fig. 3F, top panel). Because DNA-PKcs autophosphorylation is essential for promoting NHEJ efficiency (35,36), these results indicate that miR-21 promotion of NHEJ occurs through stimulating DNA-PKcs phos-FIGURE 1. miR-21 mediates radioresistance and promotes DNA DSB repair. A, mice (male, 6 weeks old, 14 mice/group) were exposed to X-rays (8 Gy). p Ͻ 0.001 between groups. B, the different types of MEF cells were exposed to different doses of X-ray. The miR-21 level in knock-out cells was not detectable and was ϳ6 -8-fold higher in knock-in cells than in wild type cells. The sensitivity of the cells to IR-induced killing was determined using a clonogenic assay. The data are the mean Ϯ S.D. and were obtained from three separate experiments. C, the image shows ␥-H2AX foci in non-irradiated or irradiated cells, following a similar protocol as described in our previous publication (31). D, cells were irradiated with 2 Gy, incubated at 37°C at different times, and then processed for ␥-H2AX foci staining (50 cells were counted for each slide). Data shown are the mean Ϯ S.D. from three independent experiments.

miR-21 Promotes DNA Repair
phorylation. On the other hand, we did not find any changes in levels of RAD51, RPA34, RPA70, XRCC2, or XRCC3 between miR-21 up-regulated MEF or human cells and their control counterparts (Fig. 3D, right panel, and Fig. 3F, bottom panel); however, we found that the levels of cyclin D1 increased after up-regulating miR-21 in MEF (Fig. 3D, right panel) and human cells (Fig. 3F, bottom panel). Because it is known that cyclin D1 promotes RAD51 to recruit to damaged DNA sites and thus facilitates HRR (20), these results suggest that miR-21 promoting HRR is linked to up-regulating cyclin D1. It is known that CDC25A is a miR-21 target; therefore, miR-21 targeting CDC25A might also contribute to miR-21-increased HRR and radioresistance.
Identifying GSK3B as a New Target of miR-21 That Can Stimulate DNA-PKcs Activity-To search for key factors that affect the cyclin D1 level or determine whether DNA-PKcs autophosphorylation could be targeted by miR-21, we focused on GSK3B because GSK3B is involved in decreasing both DNA-PKcs activity and cyclin D1 levels. To examine whether GSK3B is a real miR-21 target, we used the mature miR-21 sequence during a match search at the 3Ј-UTR of mouse or human GSK3B. We found that two conservative sequences at 3Ј-UTR of mouse or human GSK3B match miR21-5p and miR-21-3p, respectively (Fig. 4A). A luciferase reporter assay showed that a wild type sequence dramatically reduced the luciferase activity but mutation at the key sites could not (Fig. 4, A and B), indicating that miR-21 could bind to such sequences to inhibit a GSK3B expression level and confirming that GSK3B is a target of miR-21 in both mice and humans. Up-regulating miR-21 decreased GSK3B levels in both MEF (Fig. 4C) and human cells (Fig. 4D), which provides additional evidence to confirm GSK3B as a target of miR-21. Although it is known that inhibition of GSK3B stimulates DNA-PK activity and protects mouse hippocampal neurons from irradiated-induced damage (37,38), the underlying mechanism remains unclear. To address this question, we compared the CRY2 level after up-regulating miR-21 because CRY2 is also a target for GSK3B phosphorylation-induced degradation (15) and CRY2 could interact with PP5 and inhibit PP5 phosphate activity (18), which is important for dephosphorylating DNA-PKcs (19). Up-regulating miR-21 resulted in increased CRY2 levels in both MEF (Fig. 4C) and human cells (Fig. 4D), which demonstrates that inhibition of GSK3B-stimulated DNA-PK activity and NHEJ efficiency (37) occurs through the CRY2/PP5 pathway. Cell survival data from MEF (Fig. 4E) and human cells (Fig. 4F) provide strong evidence to support that GSK3B as a novel target of miR-21 is involved in miR-21-mediated radioresistance.
miR-21-mediated Radioresistance Occurs through Targeting Both GSK3B and CDC25A-To examine whether miR-21-mediated radioresistance is involved in targeting both GSK3B and CDC25A, we examined the protein levels of CDC25A and cyclin D1 as well as the radiosensitivities of GSK3B Ϫ/Ϫ cells or cyclin D1 Ϫ/Ϫ cells with or without knocking down cyclin D1 or CDC25A. The results showed that both cyclin D1 and CDC25A increased in the GSK3B Ϫ/Ϫ cells but GSK3B Ϫ/Ϫ cells were only mildly resistant to IR when compared with their wild type counterpart cells (Fig. 5A). Knocking down cyclin D1 sensitized GSK3B Ϫ/Ϫ cells to IR, but knocking down CDC25A made the cells more radioresistant when compared with the wild type cells treated with the control RNA (Fig. 5A). Because CDC25A is a known miR-21 target (22), up-regulating miR-21 reduces CDC25A levels, which decreases the CDC25A accumulation induced by targeting GSK3B (Figs. 4C and 5A). Next, we examined the DNA DSB repair activities in these cells in which some proteins were manipulated as described for MEF cells (Figs. 4C and 5A). GSK3B Ϫ/Ϫ MEF cells demonstrated increased NHEJ efficiency but no significant changes in HRR efficiency, and knocking down CDC25A promoted HRR in the cells (Fig. 5, A and B). Similar results were observed in human cells (Fig. 5, C and D). These results explain why GSK3B Ϫ/Ϫ cells did not promote HRR, which might be due to two reasons: (i) increased CDC25A in GSK3B Ϫ/Ϫ cells might inhibit the cyclin D1 activity because it is known that CDC25A has an inhibition effect on the cyclin D1 activity (21); and (ii) increased CDC25A in GSK3B Ϫ/Ϫ cells reduces the checkpoint response and thus reduces HRR efficiency, which neutralizes the increased cyclin D1-promoted HRR. Overexpressing GSK3B in miR-21 up-regulated cells abolished the increased NHEJ efficiencies but had stably overexpressed NL20 cell lines, miR-21-1 and miR-21-2) or without miR-21 (vector transfection control), to IR were measured using a clonogenic assay. The data are the mean Ϯ S.D. and were obtained from three separate experiments. D, the NSCLC cell lines (1, A549; 2, H460; 3, H358; 4, H2792) were treated with either control RNA or anti-miR-21 RNA. At 48 h after transfection, the cells were exposed to 4 Gy and then the radiosensitivity was measured using a clonogenic assay. The data are the mean Ϯ S.D. and were obtained from three separate experiments. FEBRUARY 24, 2017 • VOLUME 292 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 3533
less effect on HRR (Fig. 5, B and D); overexpressing CDC25A in miR-21 up-regulated cells only abolished the increased HRR efficiency and had no effect on NHEJ (Fig. 5, B and D).
There Is a Correlation between High miR-21 Levels and Low GSK3B Levels in Some Human Cancers-We discovered in this study that GSK3B as an important target of miR-21 requires miR-21-mediated radioresistance. Because most human tumors have a high level of miR-21 (1, 2), we wanted to see whether our discovery has any link to human tumor data. For this purpose, we searched The Cancer Genome Atlas (TCGA) database and found that miR-21 up-regulation in human tumors has a negative correlation with GSK3B down-regulation in different human tumors including pheochromocytoma/ paraganglioma (Fig. 6, A and B), kidney tumor (Fig. 6C), and testicular germ cell tumors (Fig. 6D). These results indicate that our data pertaining to miR-21 targeting GSK3B have an important translational potential, and these data provide useful information for developing strategies to improve radiotherapy.
Our results reveal that miR-21-mediated radioresistance occurs through promoting NHEJ and HRR of DNA DSB. That cells were examined by a Western blotting assay at 1 h after the cells were exposed to IR (4 Gy). Similar results were obtained from two independent experiments. p-DNA-PKcs, phospho-DNA-PKcs; Lig IV, ligase IV; KD, knockdown. E, top panel, the miR-21 levels were measured from human 293FT cells at 48 h using quantitative PCR after the cells were treated with miR-21 mimics or control RNA. Middle panel, NHEJ efficiency in 293FT cells was determined as described in our previous publication (31). The quantity of NHEJ efficiency was determined based on the fluorescence signals. Data shown are the mean Ϯ S.D. from three independent experiments, *, p Ͻ 0.05. Bottom panel, HRR efficiency in 293FT cells was determined as described under "Experimental Procedures." The left panel shows the GFP signals measured by a flow cytometer, and the right panel shows the quantity of NHEJ efficiency based on the GFP signals. Data shown are the mean Ϯ S.D. from three independent experiments, *, p Ͻ 0.05. F, the levels of HRR-or NHEJ-related factors in human cells treated with miR-21 mimics were examined by a Western blotting assay at 1 h after cells were exposed to IR (4 Gy). Similar results were obtained from two independent experiments.

miR-21 Promotes DNA Repair
is, promoting NHEJ occurs because targeting GSK3B increases DNA-PKcs activity through the CRY/PP5 pathway; promoting HRR occurs through targeting GSK3B, thus increasing the cyclin D1 level, and targeting CDC25A neutralizes the effects of targeting GSK3B-induced accumulated CDC25A, which thus increases the checkpoint response (Fig. 7).

Discussion
In this study, we demonstrate that miR-21-mediated radioresistance occurs through promoting repair of DNA DSB, which is involved in targeting both GSK3B and CDC25A. Because whether radiation kills cells depends mainly on the generation of DNA DSB, any factor that affects cell radiation sensitivity should either increase the yield of DNA DSB or increase the cell ability to repair DNA DSB. miR-21-mediated cell resistance to IR-induced killing is not due to increasing the yield of DNA DSB, but is due to promoting DNA DSB repair. Previously, it was reported that miR-21 prevents cell apoptosis via targeting PTEN (39), which, as a phosphatase, is an apoptosis promoter and a tumor suppressor, suggesting that miR-21-mediated B, the effects of the potential miR-21 binding sites at the 3Ј-UTR of GSK3B on luciferase activity were measured as described under "Experimental Procedures." 293FT cells were transfected with a firefly luciferase reporter plasmid containing a partial 3Ј-UTR of GSK3B with the putative miR-21 binding site (WT) or deleted the key binding site (DM). Luciferase activity was assayed 24 h after transfection with miR-21 mimic (miR-21) or control (Ct) RNA and was normalized to the activity of the firefly luciferase expressed from the same pGL3-control vector, **, p Ͻ 0.01. C, wild type or miR-21 knock-in MEF cells were transfected with the GSK3B vector or CDC25A vector, or co-transfected with GSK3B/CDC25A vectors for 48 h. The cells were either exposed to IR (4 Gy for DNA-PKcs detection) or given no IR exposure, and then the cells were collected at 1 h after IR for the standard Western blotting assay. Similar results were obtained from two independent experiments. p-DNA-PKcs, phospho-DNA-PKcs. D, human 293FT cells were transfected with miR-21, GSK3B vector, or CDC25A vector or cotransfected with GSK3B/CDC25A vectors for 48 h. The cells were either exposed to IR (4 Gy for DNA-PKcs detection) or given no IR exposure, and then the cells were collected at 1 h after IR for the standard Western blotting assay. Similar results were obtained from two independent experiments. E, at 48 h after transfection as described in C, the cells were exposed to IR (4 Gy) and collected for sensitivity measurements using a clonogenic assay as described in the legend for Fig. 1. Data shown are the mean Ϯ S.D. from triple sets of two independent experiments; ND, no significant difference. F, at 48 h after transfection as described in D, the cells were exposed to IR (4 Gy) and collected for sensitivity measurement using a clonogenic assay as described in the legend for Fig. 2. Data shown are the mean Ϯ S.D. from triple sets of two independent experiments; ND, no significant difference. FEBRUARY 24, 2017 • VOLUME 292 • NUMBER 8 radioresistance may involve reducing apoptosis via targeting PTEN. However, knocking PTEN out does not affect cell sensitivity to radiation (40), which might be due to the fact that apoptosis is not a considerable factor affecting cell radiosensitivity (41). Therefore, targeting PTEN may not contribute to miR-21-mediated cell radioresistance. miR-21 expression is stimulated by the IR-activated EGFR/ STAT3 pathway (3) or by the IR-activated ATM (ataxia telangiectasia-mutated) pathway through phosphorylating KSRP (KH-type splicing regulatory protein) to stimulate global pri-miRNA biogenesis (42). After ATM is activated by IR, both ATM and its downstream target, CHK2, could phosphorylate CDC25A on serine 123, which promotes CDC25A degradation and results in an S phase checkpoint (43) to facilitate HRR. Thus, miR-21-mediated radioresistance not only depends on its endogenous expression level as well as its targets but also partially involves an IR-activated DNA damage response.

miR-21 Promotes DNA Repair
We show here that overexpression of miR-21 promotes both NHEJ and HRR, but the phenotype of cell radioresistance is not as significant as that of NHEJ-or HRR-deficient cells. We believe that such results occur mainly because of the following two reasons. (i) miRNA knockdown of a gene is not as efficient as siRNA knockdown (31) because miRNA occurs by partially matching the sequence in the 3Ј-UTR of the gene, and siRNA occurs by completely matching the sequence in the coding region of the gene. (ii) Most importantly, miR-21 targets so many factors (44) that some of the factors (both discovered and undiscovered) may indirectly neutralize the effects of miR-21 on promoting NHEJ or HRR, and thus, reduce the miR-21mediated cell radioresistance. miR-21 overexpression is found FIGURE 5. miR-21-mediated radioresistance occurs through targeting both GSK3B and CDC25A. A, upper panel, whole cell lysates were prepared from wild type, GSK3B Ϫ/Ϫ , and cyclin D1 Ϫ/Ϫ MEF cells that were treated with control RNA (CtRNA, lanes 1-3), cyclin D1 siRNA (lanes 4 -6), or CDC25A siRNA (lanes 7-9). The protein levels were measured using a standard Western blotting assay. Actin was used as an internal loading control. Lower panel, cell survival fraction from 4 Gy irradiated cells. The data were mean Ϯ S.D. from three independent experiments, **, p Ͻ 0.01. B, NHEJ or HRR efficiency was examined in MEF cells that were transfected with the reagents as described above for 24 h and then transfected with either the HRR or NHEJ reporter for an additional 24 h. The NHEJ or HRR efficiency assays were as described in the legend for Fig. 3. The data presented are the mean Ϯ S.D. from three independent experiments. C, the protein levels were examined in human 293FT cells at 48 h after the cells were transiently transfected with GSK3B siRNA, CCND1 (cyclin D1) siRNA, or CDC25A siRNA. Actin was used as an internal loading control. D, after 293FT cells were transfected with the reagent for 24 h, the cells were transfected with the NHEJ or HRR reporter for an additional 24 h, and then the cells were collected to detect NHEJ or HRR efficiency. Data shown are the mean Ϯ S.D. from triple seta of two independent experiments; ND, no significant difference.

miR-21 Promotes DNA Repair
in most types of human tumors; however, not all of these tumors demonstrated the same radioresistance. This is mainly due to the heterogenic features of human tumors; even different cells isolated from the same human tumor show dramatically different radiosensitivities due to different expressions of some DNA DSB repair factors (45).
Taken together, our results in this study reveal the mechanism underlying miR-21-mediated cell radioresistance, and can provide useful information for clinical consideration about how to treat miR-21-mediated resistance to radiotherapy in the near future.

Experimental Procedures
Mice, Cell Lines, and Irradiation-All animal experiments were conducted following an animal protocol (#2002753) approved by the Institutional Animal Care and Use Committee (IACUC) of Emory University. C57BL/6J mice were purchased from The Jackson Laboratory. miR-21 knock-in mice with a C57BL/6J genetic background were generated in our lab as described (29). miR-21 Ϫ/Ϫ mice were obtained from Dr. Olson's lab (46). All mice were male, 6 weeks old, and either irradiated (whole body) with 8 Gy or sham-irradiated at the same time as described previously (47). The related mouse embryo fibroblast cells derived from the mice were generated in our laboratory. GSK3B Ϫ/Ϫ cells were obtained from Dr. Woodgett's lab (12), and cyclin D1 Ϫ/Ϫ cells were obtained from Dr. Chenguang Wang's lab (48). Non-viral immortalized human bronchial epithelial cells (3KT) were obtained from Dr. Mina's lab (Ramirez et al. (32)). Human 293FT cells, immortalized human bronchial epithelial cells (NL20), and human lung tumor cell lines A549, H460, H358, H522, Calu1, and Skemes-1 were purchased from the ATCC. Human lung tumor cell lines 95C, 95D, H226B, H1597, and H2797 are as described previously (49). All cell lines used in this study were checked for mycoplasma contamination. These cell lines were grown in Dulbecco's modified Eagle's medium supplemented with 10% FBS. Irradiation of mice or cells was performed with an X-ray machine (X-RAD 320, North Branford, CT) at 320 kV, 10 mA, and the filtration was performed with 1.5-mm aluminum filter, 0.8-mm tin filter, and 0.25-mm copper filter for mice, and with 2-mm aluminum filter for cells in our laboratory. The dose rate was 1-2 Gy/min.
Plasmid Construction-The plasmids containing mouse GSK3B (pSP72 GSK3B) or the human GSK3B gene (HA-GSK3B wt pcDNA-3) were purchased from Addgene. The plasmid containing mouse CDC25A was purchased from OriGene Inc., and the human CDC25A was obtained from Dr. Jiri Bartek's lab (43). The primers used to generate an expression plasmid from   FEBRUARY 24, 2017 • VOLUME 292 • NUMBER 8 the cloning plasmid of pSP72 GSK3B are listed in Table 1. The mouse GSK3B gene was verified with enzyme digestion (KpnI and XbaI) and sequencing. The DNA encoding partial 3Ј-UTR of GSK3B was PCR-amplified from mouse or human genomic DNA; DNA fragments from partial 3Ј-UTR of GSK3B that host the predicted complementary sites of miR-21 or the mutated sites were cloned downstream of firefly luciferase reporter gene in pGL3-control plasmid (Promega Corp.) as described in the following section. The primer information is listed in Table 1. The DNA sequence encoding miR-21 together with surrounding precursor sequences (ϳ300 bp in total) were purchased from GeneCopoeia Inc., and then inserted into pcDNA3.1 (Invitrogen) at HindIII and XhoI sites to produce miRNA expression plasmids.

miR-21 Promotes DNA Repair
Mimic/Inhibitor of MiRNA, SiRNAs, and Transfection-NiR-21 mimic or inhibitor and control RNAs were purchased from Ambion Inc., and the control RNA and a pool of siRNA against GSK3B, CCND1 (cyclin D1), or CDC25A were purchased from Dharmacon Inc. Cells were transfected with the plasmid, mimic, or inhibitor siRNA for 24 -48 h, and then collected for further experiments. Plasmid and siRNA transfections were performed using Lipofectamine 3000 (Invitrogen), according to the manufacturer's protocol.
Cell Survival Assay-Cell sensitivity to radiation was evaluated for loss of colony-forming ability as described previously (28).
HRR or NHEJ Reporter Assay-The NHEJ assay reporter (pEGFP-pem1-Ad2) vector was obtained from Dr. Eric Hendrickson's lab (50), and the human 293FT cells expressed with an integrated NHEJ reporter were obtained from Dr. Pandita's lab (51). The HRR reporter was obtained from Dr. Jasin's lab (52,53). Wild type MEF or human 293FT cells were treated with control RNA, miR-21 mimic, or miR-21 inhibitor and GSK3B-overexpressing plasmid, GSK3B, or CDC25A siRNA for 24 h and then transfected with the reporter (NHEJ or HRR) and I-SceI for an additional 48 h. NHEJ or HRR efficiency was measured by quantifying fluorescent signals in cells after transfection with an I-SceI plasmid using flow cytometry as described previously (54).
Antibodies and Immunoblotting-The whole cell lyses were prepared as described previously (Wang et al. (28)). The antibodies against human and mouse DNA-PKcs, RPA70, Ku70, ligase IV, XRCC4, XRCC2, XRCC3, CDC25A, HA, ␤-actin, and CRY2 were purchased from Santa Cruz Biotechnology Inc. The antibody against human cyclin D1 was purchased from Cell Signaling Technology Inc. The antibody against phosphorylated DNA-PKcs was purchased from Abcam Inc. The antibodies against human and mouse GSK3B, Rad51, and ␥-H2AX were purchased from EMD Millipore Inc. The antibody against human CRY2 was purchased from Bethyl Laboratories Inc.
Negative Correlation between miR-21 and GSK3B in Human Tumors-TCGA data were downloaded from the Broad Institute FireBrowse Data Portal. TCGA mRNA-seq data (build 2015060100) and TCGA miRNA-seq data (build 2015110100) were downloaded from Broad Firehose and used for the analysis.
Statistical Analysis-The statistical significance of comparisons between two groups was determined using Student's t test. p values of less than 0.05 were considered statistically significant. Pearson's correlation analysis was performed with GraphPad Prism 5.
Author Contributions-Y. W. designed the study. B. H., X. W., S. H., X. Y., P. W., X. Z., J. W., and H. W. conceived and performed the experiments. B. H., X. W., and Y. W. analyzed data and prepared the manuscript. All authors reviewed and approved the manuscript.