EBV–encoded miRNAs can sensitize nasopharyngeal carcinoma to chemotherapeutic drugs by targeting BRCA1

Abstract Nasopharyngeal carcinoma (NPC) is an Epstein‐Barr virus (EBV)‐associated epithelial malignancy. The high expression of BART‐miRNAs (miR‐BARTs) during latent EBV infection in NPC strongly supports their pathological importance in cancer progression. Recently, we found that several BART‐miRNAs work co‐operatively to modulate the DNA damage response (DDR) by reducing Ataxia‐telangiectasia‐mutated (ATM) activity. In this study, we further investigated the role of miR‐BARTs on DDR. The immunohistochemical study showed that the DNA repair gene, BRCA1, is consistently down‐regulated in primary NPCs. Using computer prediction programs and a series of reporter assays, we subsequently identified the negative regulatory role of BART2‐3p, BART12, BART17‐5p and BART19‐3p in BRCA1 expression. The ectopic expression of these four miR‐BARTs suppressed endogenous BRCA1 expression in EBV‐negative epithelial cell lines, whereas BRCA1 expression was enhanced by repressing endogenous miR‐BARTs activities in C666‐1 cells. More importantly, suppressing BRCA1 expression in nasopharyngeal epithelial cell lines using miR‐BART17‐5p and miR‐BART19‐3p mimics reduced the DNA repair capability and increased the cell sensitivity to the DNA‐damaging chemotherapeutic drugs, cisplatin and doxorubicin. Our findings suggest that miR‐BARTs play a novel role in DDR and may facilitate the development of effective NPC therapies.

reg/facts.html). However, NPC still poses serious socio-economic and healthcare problems in Hong Kong because the peak incidence of the disease is in the main workforce population with age between 3560 years (≥30 per 100 000 populations in males).
Hence, understanding the nature of NPC for the development of effective, target-specific therapies is still the main research focus in this field.
In NPC, the clonal Epstein-Barr virus (EBV) genome is consistently detected in both dysplastic lesions and invasive carcinoma, suggesting the crucial role of the virus in cancer progression. 1,2 EBV resides in NPC as a type II latent infection, in which only latent membrane proteins (LMPs) and EBV nuclear antigen 1 (EBNA1) are expressed. 3 The oncogenic properties of these viral proteins have been well characterized. [4][5][6][7][8] Because of their high immunogenic potential, LMPs are usually expressed at low levels in the infected cells to escape the host immune surveillance. In contrast, the non-immunogenic non-coding RNAs, such as EBERs and viral microRNAs, are abundantly expressed in NPC. EBV was the first virus reported to encode miRNAs, in 2004, 9 and subsequent work by other teams eventually identified a total of 44 mature EBV-miRNAs. 10,11 EBV-miRNAs are located in two viral genome regions and are named miR-BHRF1s and miR-BARTs. All four miR-BHRF1s are generated in the untranslated region of the early lytic gene, BHRF1 and are restricted expression in the EBV type III latent infection. However, the rest of the 40 miRNAs derived from the two clusters within the non-coding BamH1-A rightward transcripts, BARTs (miR-BART1 to miR-BART22), are abundantly expressed in all EBV-positive epithelial malignancies. 12 The miR-BARTs constitute 38% of the total miRNAs in NPC, 13 and their diverse functions in augmenting cancer development have been extensively reported; they include maintaining viral latency, 14,15 promoting survival, [16][17][18] invasiveness, 19 metastasis 20,21 and controlling the host cells' immunity. 10,22,23 Cellular DNA is constantly damaged by different sources of stimuli. Therefore, cells need to preserve genome integrity using the error-free homologous recombination (HR) pathway for DNA repair. Once DNA double-strand breaks (DSBs) occur, the ATM rapidly localizes to the damage site and phosphorylates H2AX, which, in turn, recruits a variety of proteins such as BRCA1 and MRE11-Rad50-NBS1 (MRN) complexes to form nuclear foci for repairing the damaged DNA. In the nuclear foci, BRCA1-MRN complexes activate the end resection of DSBs to produce 3' single-strand DNA (ssDNA). To ensure accurate DNA correction, RAD52, RPA and RAD51 sequentially bind to the ssDNA and stimulate DNA strand exchange events using the undamaged sister chromatid as a repair template. 24 Because the availability of sister chromatids is necessary, HR only occurs in the S-and G2/M phases of the cell cycle. Both BRCA1 and ATR contribute to HR by activating CHK1 through phosphorylation at S345. The active CHK1 subsequently represses CDK1 activity, resulting in arresting the G2/M checkpoint for cell fate decision; in other words, either HR or apoptosis occurs. Disrupting the activities of HR proteins not only contributes to genomic instability in tumour development but also sensitizes the cells to radio-and chemotherapies. We have previously demonstrated that miR-BARTs directly inhibit ATM expression in NPC cells and sensitize the cells to irradiation treatment. 13  and HeLa cell lines were used in this study. 13,[36][37][38][39] The clinically frozen specimens for Reverse transcription-quantitative PCR (RT-qPCR) (Table S1) and paraffin-embedded specimens for Immunohistochemistry (IHC) analysis (Table S2)

| Reverse transcription-quantitative PCR
The total RNAs were extracted using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and reverse-transcribed with miScript II RT Kit (Qiagen, Hilden, Germany). The BRCA1 RT-qPCR product was amplified using the SYBR Green PCR Master Mix Kit (Applied Biosystems, Foster City, CA, USA). The relative gene expressions were normalized with actin, and the fold-change was calculated using the 2 (∆∆-Ct) method. The method for miR-BART expression has been previously described. 12 The qPCR primer sequences for BRCA1 and cellular BRCA1-responsive miRNAs are listed in Table S3.

| Immunohistochemistry
The BRCA1-IHC staining was performed using the Polymer Refined Detection Kit on Leica Bond-Max, fully automated staining system.
The primary antibody for BRCA1 (1:100 dilution, clone MS110; Millipore, Quincy, MA, USA) was used. The expression level of BRCA1 was determined using a scoring system that considered both the staining intensity and prevalence of intensities as described previously. 40 The specimens with moderate to strong BRCA1 signal were considered IHC positive.

| Prediction of microRNA targets
The sequences of miRNAs and the BRCA1 transcript (NM_007294.2) were extracted from miRBase 41 and the NCBI, respectively. The putative binding site of miR-BARTs on BRCA1 was predicted with mi-Randa and RNAhybrid programs as described previously. 10,[42][43][44] The cut-off point for the selection was MEF <−16kcal/mol.

| The miRNA mimics, inhibitors, expression vectors and transfection
The function of miRNAs was investigated using synthetic, chemically modified, small RNAs that either mimicked (miRNA mimics) or inhibited (miRNA inhibitors) the activity of the specific miRNA in vitro.

| Luciferase reporter assay
The construction of the luciferase reporter plasmids and the procedure of the reporter assay has been previously described. 10 The cell-cycle analysis, RAD51 staining, comet assays and clonogenic survival assays were described previously with some modifications. 13,46 In brief, the cells were transfected with the desired siRNAs or miR-BART mimics for six hours and then evenly seeded into the new cultureware for subsequent analysis. The cells were treated with the specific drug for 24 hours and then stained with propidium iodide for cell-cycle analysis. For RAD51 staining assays, the cells were seeded on glass coverslips, treated with the drug for 16 hours, returned to growth in the normal medium for three hours, and subsequently fixed and stained with the RAD51 antibody for counting. In the comet assays, the cells were incubated with the specific drug for four hours and then returned to culture for three hours before single-cell gel electrophoresis was performed using an OxiSelect Comet Assay Kit (Cell Biolabs, San Diego, CA, USA) following the manufacturer's instructions.
For the clonogenic survival assay, 500 or 1000 transfected cells were evenly seeded into the 6-well plate and exposed with either mock medium or drugs for 24 hours. After 18 days cultured with the growth medium, the colonies were fixed with methanol and then stained with 0.5% crystal violet for visualization. The colonies containing more than 30 cells were counted. The plating efficiency (PE) of the average of the three colonies counted in each treatment was calculated as follows: Then the surviving fraction was determined with the following equation and compared with the mock treatment control (100%):

| Statistical analysis
Two-sided Student's t-test was used to compare the differences between the two groups unless otherwise specified. The analysis of each experiment was performed in triplicate, and the results are expressed as mean + SD. All analyses were performed using GraphPad Prism 5.

| Down-regulation of BRCA1 in NPC
We have previously reported that several miR-BARTs contribute to the disruption of the DNA damage repair by suppressing the ATM signalling pathway. 13 Although ATM expression is consistently down-regulated in NPCs, the expression level of its downstream target, BRCA1, is variable in a panel of NPC samples ( Figure 1A).  (Table S5). In our IHC analysis, we detected the predominant positive BRCA1 expression in 83% of NP cases, whereas only 42% of NPC cases scored positive ( Figure 1D and Table S5). Since BRCA1 protein expression is not directly correlated to its mRNA level in NPC cell lines and primary tumours, we postulated that BRCA1 expression in NPCs may be regulated in the post-transcriptional level, likely due to the miRNAs derived from EBV.

F I G U R E 2
The BRCA1 is the potential target of miR-BARTs. (A) The relative luciferase activity of the reporter plasmids harbouring a full length of BRCA1-3'UTR (sFL-3'UTR) or a full length of BRCA1-3'UTR in reversed orientation (asFL-3'UTR) was co-transfected together with the indicated miRNAs. The luciferase signal with the co-transfection of negative miRNA mimic control (miR-NEG) was set at 1 for comparison. (B) The direct interaction between the putative binding sites on BRCA1 and miR-BARTs were demonstrated in the reporter assays. The firefly luciferase reporter activity was normalized to the Renilla luciferase control. The data shown is the mean + SD from three independent experiments. The result with the co-transfection of miR-NEG and pMIR-CTL was set at 1. pMIR-CTL = pMIR-REPORTTM vectors containing unrelated sequences; pMIR-B = pMIR-REPORTTM vector harbouring the predicted miR-BART binding site, pMIR-CDS = predicted binding site on CDS (Table S5). B2-3p = BART2-3p; B12 = BART12; B17-5p = BART17-5p; B19-3p = BART19-3p. *P < 0.05, **P < 0.001

| BRCA1 is a direct target of miR-BARTs
To further dissect the involvement of EBV in BRCA1 regulation, we screened the suppressive effect of miR-BARTs on the BRCA1-3'UTR using luciferase reporter assay. As a well-known BRCA1 modulator, hsa-miR-182-5p was included as a positive control. 48 When compared with the co-transfection of miRNA control mimics (miR-NEG), miR-182-5p, BART2-3p, BART12 and BART17-5p exerted strong suppressive effects on the full length of BRCA1-3'UTR, whereas the inhibitory effects disappeared when the BRCA1-3'UTR was cloned into the reporter in reverse orientation (Figure 2A). Therefore, we predicted the putative binding sites of miR-BARTs by using the default settings of two publicly available computer programs, miRanda and RNAhybrid. In these in silico analyses, a number of putative binding sites of miR-BART2-3p, miR-BART12, miR-BART17-5p and miR-BART19-3p on the BRCA1 transcript were suggested (Table S6).
We subsequently cloned each putative binding site into the 3'UTR

| Regulation of endogenous BRCA1 expression by miRNAs
To dissect the miR-BARTs expression levels in the 20 available NPC biopsies, in which the BRCA1 protein expression status was studied in IHC, we revealed that the total expressions of BART2-3p, BART12, BART17-5p and BART19-3p in BRCA1-positive NPCs (n = 7) were significantly lower than in BRCA1-negative NPCs (n = 13) (P = 0.039) ( Figure 1E). In the cell lines study, the normal nasopharyngeal epithelial cells (NP69) had higher BRCA1 expression than the NPC cell lines in the immunoblotting analysis. Among the NPC cells, the two newly derived EBV-positive cells (NPC43 and C17 cells) clearly had lower BRCA1 protein levels than the EBV-negative cells (HK1 and NPC53) ( Figure 3A). It is noteworthy that the BRCA1 protein was only barely detected in C17 cells, even though they had similar BRCA1 mRNA levels to NPC43 ( Figure 1B). As the expression levels of most previously reported BRCA1-repressive miRNAs, except miR-146a-5p, were only slightly different (<4 folds) in NPC cell lines ( Figure S5), the relatively high expression levels of total BART2-3p, BART12, BART17-5p and BART19-3p in C17 cells may be heavily involved in modulating BRCA1 expression in the post-transcriptional level. The high miR-146a-5p level in C17 cells may also contribute to the BRCA1 suppression ( Figure 3B). The significant difference in miR-146a-5p expression between HK1 and NPC53 may also result in similar BRCA1 protein expression but considerably different BRCA1 transcript levels between these two EBV-negative cells ( Figure 3A and B).
To directly prove the regulatory effect of miR-BARTs on BRCA1 expression, we introduced the miR-BART mimics into HeLa, NP69, HK1 and NPC53 for analysis. The endogenous BRCA1 protein levels in these four EBV-negative epithelial cells decreased after the transfection of individual BART2-3p, BART12, BART17-5p and BART19-3p mimics for 24 hours ( Figure 3C). Moreover, BRCA1 was readily suppressed in the HK1 cells that had been stably transfected with either the BART17 or BART19 expression vector ( Figure S6). On the contrary, BRCA1 expression in EBV-positive

C666-1 cells was increased by suppressing the endogenous miR-
BARTs activity with specific miRNA inhibitors ( Figure 3D). Overall, the findings provide convincing evidence to support the regulatory role of BART2-3p, BART12, BART17-5p and BART19-3p on BRCA1 expression.

| Down-regulation of BRCA1 in NP69 and HK1 cells sensitizes the cells to chemo-drug treatment
We had previously established a comprehensive transcriptional profile of miR-BARTs in NPC PDXs and realized that BART2-3p (<0.05% of total miR-BARTs) and BART12 (~0.3%) were expressed at extremely low levels in NPC. In contrast, both BART17-5p (2.5%) and BART19-3p (4.9%) were highly expressed. 13  and NP69 transfected with a si-BRCA1, BART17-5p or BART19-3p mimic remarkably induced cell-cycle arrest at S phase or G 2 /M phase after exposure to CDDP (7.5 µmol/L for HK1 and 2 µmol/L for NP69) and DOX (0.25 µmol/L for HK1 and 0.1 µmol/L for NP69) for 24 hours ( Figure 4B). Consistent with the cell-cycle analysis, the elevation of the phosphor-Chk1 level was detected in the BRCA1

F I G U R E 4
The CDDP and DOX sensitivity in HK1 and NP69 cells. Approximately 500 or 1000 transfected cells were seeded into the 6-well plate and treated with CDDP or DOX for 24 h. The cells were cultured for 14-18 d in normal medium before staining, and colonies containing more than 30 cells were counted. The number of colonies generated from the mock treatment was compared (set as 100%). All the experiments were performed in triplicate and the Student's t-test was conducted, compared with the control transfected cells. *P < 0.05; **P < 0.01 knockdown cells after treatments ( Figure 4C). As a control, the transfected cells alone did not affect the normal cell-cycle progression ( Figure 4B and Figure S7). This data suggests that both CDDP- When compared with the control, the si-BRCA1 transfected HK1 and NP69 cells remarkably reduced the colony formation rates by more than 50% after incubation with CDDP (7.5 µmol/L for HK1 and 2 µmol/L for NP69) or DOX (0.25 µmol/L for HK1 and 0.1 µmol/L for NP69) for 24 hours. Similarly, the increased BART17-5p or BART19-3p level in the cells suppressed colony formation, although the magnitude of the effect was clearly reduced ( Figure 4E and Figure S8).
To investigate whether the increased cytotoxic effect of the transfected cells was caused by defects in the DNA repair system, we extended our study to examine the DNA damage recovery ability of the cells with a RAD51 foci formation assay ( Figure 5). Since  In this study, we identified that BRCA1, a tumour suppressor gene responsible for DNA double-strand break-repair, is the cellular target of BART2-5p, BART12, BART17-5p and BART19-3p, in which BART17-5p and BART19-3p constitute ~8.6% of the total viral miR-NAs in NPC. 13 The direct interaction of miR-BARTs on six in silico predicted binding sites on the BRCA1 3'UTR was confirmed using reporter assays ( Figure 2B). Interestingly, the miRNA response element (MRE) of BART19-3p is located on the CDS of BRCA1. Although miRNAs commonly suppress target gene translation by binding with their 3'UTRs, their MRE can mechanistically occur in any position on the target mRNA. 53 Moreover, the MRE located on the CDS can be validated by luciferase assays, with a reporter vector containing the specific MRE in the 3'UTR. 54 In line with our findings, one of the BART17-5p (BRCA1-1038) and BART19-3p (BRCA1-CDS-2694)

| D ISCUSS I ON
binding sites were previously identified in the primary effusion lymphoma cell lines using the highly sensitive PAR-CLIP method. 55 We further found that the inhibition of BRCA1 or overexpression of BART17-5p and BART19-3p in HK1 and NP69 cells can make the cells more susceptible to cisplatin and doxorubicin treatments ( Figure 4D and E).

F I G U R E 6
The overexpression of miR-BARTs increase DNA damage response in nasopharyngeal epithelial cells. The transfected HK1 and NP69 cells were treated with the indicated concentration of cisplatin (CDDP) for four hours and subsequently returned to the fresh culture medium for three hours before they were analysed by comet assays. The representative images for different transfected cells are shown in the upper panel. At least 70 cells were randomly selected for analysis using ImageJ software. The mean of the tail moment is shown in the bar chart: mean + SEM. Student's t-test was used to compare them with the miR-NEG transfected control (black bar). *P < 0.05; **P < 0.01; ***P < 0.001 This study had limitations as well. We have demonstrated that overexpressing miR-BARTs in EBV-negative cells leads to increased cell sensitivity to chemotherapeutic drugs. However, we do not have suitable EBV-positive NPC cell lines to prove the hypothesis in the miR-BART knockdown experiment. There are only three EBV-positive NPC cell lines available worldwide, but both NPC43 and C17 cells were established from recurrent NPC 37,39 and have characteristics different from those of primary NPC; for instance, they have no p53 protein expression ( Figure 4A). The C666-1 is the model commonly used for the study of viral-host interaction, and the positive correlation between endogenous BRCA1 expression and the chemoresistance of the drugs can be demonstrated in C666-1 ( Figure S9).
However, both BRCA1 protein and miR-BARTs were highly expressed in C666-1, indicating that other mechanisms in C666-1 may diminish the involvement of miR-BARTs in BRCA1 regulation, such as shortening the BRCA1 3'UTR. 56 Together with our previous findings, we have identified the roles of eight miR-BARTs in regulating two essential DNA homologous recombination factors, ATM and BRCA1. These eight miRNAs occupy about 20% of the total EBV-encoded miRNAs in NPCs. 13 The repression of either ATM or BRCA1 expression makes the cells hypersensitive to the treatment with olaparib, an FDA approved PARP inhibitor. 48,57,58 Hence, the relationship between the effect of olaparib, BRCA1 expression and miR-BARTs activity should be evaluated in NPC. Unfortunately, the NPC cell lines available in our laboratory were highly resistant to olaparib (concentration ≤ 300 µmol/L), albeit PARP1 proteins were abundantly expressed (Table 1 and Figure 1A).
Thus, the effect of olaparib on NPC was excluded from the study.
This study is the first report to demonstrate the interaction of four miR-BARTs on the BRCA1 transcript. Our findings at least partially support the hypothesis that DNA repair factors are tightly regulated by miR-BARTs during tumorigenesis in a group of EBVinfected NPCs. Since overexpression of miR-BARTs can potentiate the sensitivity to chemotherapeutic agents in some nasopharyngeal epithelial cells, our study may contribute to the development of effective therapies for NPC management.