EBV Encoded miRNA BART8-3p Drives Radioresistance-Associated Metastasis in Nasopharyngeal Carcinoma

Background: Radiotherapy plays an important role in the treatment of nasopharyngeal carcinoma (NPC), however, 20 % of patients with NPC exhibit unusual radioresistance. Patients with radioresistance are at risk of recurrence, so it is imperative to explore the mechanism of resistance to radiotherapy. In the past, studies on the mechanism of radioresistance have been restricted to DNA damage and related cell cycle remodeling or apoptosis. So far, no studies have explored the relationship between radioresistance and metastasis. Methods: We analyzed the metastasis rate of patients with recurrent NPC and that of patients with primary NPC. Constructing an acquired radioresistant NPC cell line and detect their metastatic ability in vivo and in vitro. RNA-deep sequencing was performed to predict the targeted host genes of EBV-miR-BART8-3p. Western blotting, real-time PCR and immunochemistry were conducted to investigate the relationship of clinicopathologic features and EBV-miR-BART8-3p or PAG1. Results:Through the analysis of clinical samples, we observed that the metastasis rate of recurrent NPC was much higher than that of primary patients. In vitro and in vivo experiments showed that NPC cells with acquired radioresistance exhibited a stronger ability for invasion and metastasis. Mechanistically, we found that the Epstein–Barr virus (EBV)-encoded miRNA BART8-3p was increased in patients with NPC and its expression was positively correlated with adverse prognostic factors, such as radioresistance. Besides, miR-BART8-3p promoted the epithelial-mesenchymal transition (EMT), invasion, and metastasis of radioresistant NPC cells by targeting and inhibiting their PAG1 host gene. Conclusion: These results demonstrated a correlation between radioresistance and metastasis in NPC, which depended on the elevated levels of the EBV-encoded miRNA BART8-3p and the inhibition of the PAG1 host gene. These ndings suggested a novel role for EBV-miR-BART8-3p and PAG1 in recurrence


Background
Nasopharyngeal carcinoma (NPC) is remarkable for its distinct racial and geographic distributions, being a common malignant tumor in Southeast Asia, North Africa, Alaska, and the Mediterranean basin [1].
Although radiotherapy is the preferred treatment for NPC, radioresistance is an important problem in the treatment of NPC. Recurrence and metastasis of NPC due to radioresistance are known to occur in 20% of the total number of patients [2]. Both recurrence and metastasis are important signs of poor prognosis.
In particular, it has been reported that 20% to 30% of relapsed patients have distant metastatic tumors [2,3]. Compared with patients with better local control, patients with NPC with local recurrence are known to be more likely to have secondary distant metastasis [4,5]. Therefore, a better understanding of the relationship underlying radioresistance and metastasis of NPC is likely to improve survival and facilitate the design of novel therapeutic strategies for NPC.
Postradiotherapy plasma Epstein-Barr virus (EBV) DNA is known to play an important role in the risk strati cation of NPC [6], suggesting that EBV viral encoding products might play an important role in the metastasis of NPC. Noted, EBV is the rst human virus found to encode microRNAs (miRNAs). To date, a total of 25 EBV-miR precursors containing 48 mature miRNAs have been identi ed within 2 regions of the EBV genome. Analysis of the results comparing the miRNA microarray of NPC with that of normal nasopharyngeal (NP) tissues revealed that EBV-miR-BARTs were superior among all differentially expressed miRNAs [7]. Respectively, the mechanism of EBV-miR-BARTs in the radioresistance and metastasis of nasopharyngeal carcinoma has been reported, whereas the bridge function of EBV-miR-BARTs in the radioresistance associated metastasis of NPC remain elusive. In this study, we were interested in the role of viral and cellular miRNAs in the tumor radioresistance-associated metastasis and their clinical signi cance in NPC, in order to identify potential targets for improving the prognosis of recurrent NPC (rNPC).
In this study, using in vivo and in vitro experiments, we con rmed that NPC radioresistant cell lines were prone to metastasis. We further found that EBV-miR-BART8-3p was highly expressed in radioresistant NPC and closely associated with metastatic features of NPC. The underlying molecular mechanisms by which EBV-miR-BART8 caused tumor metastasis were revealed to involve the direct targeting of the major tumor suppressor phosphoprotein membrane anchor with glycosphingolipid microdomains 1 (PAG1), and promotion of its combination with vimentin, which induced epithelial-mesenchymal transition (EMT).
Our ndings provided new insights into the mechanisms of the EBV-regulated radioresistance in NPC and advocate for the development of clinical intervention strategies for NPC.

Patients and specimens
A cohort containing 62 NPC specimens with TNM staging was collected for the association analysis of the expression of EBV-miR-BART8 with pathological and clinical data (Supplementary Table 1-2). All clinical samples used for immunohistochemistry (IHC) analysis of the expression of PAG1 were collected from the Nanfang Hospital of Southern Medical University, Guangzhou, China (Supplementary Table  3).The biopsy specimens of the validation set were obtained from a clinical trial conducted by our study group (Clinical Trials.gov no.NCT01171235). Tumor stage was scored according to the American Joint Committee on Cancer staging system (7th edition).

Establishment of radioresistant cells
C666-1 cells were exposed to 6 Gy X-rays once every 2 wk for a total of 5 times (cumulative dose: 30 Gy), yielding C666-1R cells. After the nal exposure, cells were cultured under normal conditions for 4 wk.
Colony formation assay for radiosensitivity and irradiation Colony formation assays were performed to assess the radiosensitivity of cells after IR. Suspensions containing 200, 400, 800, 1600, and 3200 cells were seeded into 6-well plates and exposed to 0, 2, 4, 6, or 8 Gy (2 Gy perfraction), respectively, using a 6 MV X-ray beam from an Elekta linear accelerator (Precise 1120; Elekta Instrument AB, Stockholm, Sweden) at a dose rate of 220 cGy/min. Cells were incubated for 14 d until colony appearance. Colonies were xed for 15 min with carbinol (Huada, Guangdong, China) and stained for 30 min with 0.1 % Giemsa (AppliChem, Darmstadt, Germany). Colonies containing >50 cells were counted.

Transwell migration and Boyden chamber invasion assays
For the transwell migration assay, 10 5 cells in 100 mL serum-free RPMI-1640 media were triplicate seeded in each bronectin-coated polycarbonate membrane insert in a transwell apparatus (Corning, Shanghai, China). Consecutively, 600 mL of RPMI-1640 supplemented with 10 % NCS was added to the bottom chamber. Both C666-1 and C666-1R cells were incubated at 37 °C with 5 % CO 2 for 8 h. Concomitantly, 5-8F, HONE1, and HNE1-EBV cells were incubated for 8, 12, and 8 h, respectively. Cells adhered on the lower surface were xed with 100 % methanol (Huada) at 25 °C for 15 min and stained with hematoxylin (Macklin, Shanghai, China) for 15 min. Cell numbers in 6 predetermined elds in each replicate were counted under the microscope (NiKon ECLIPSE 80i system,NiKon, Shanghai, China ). All assays were independently repeated at least 3 times. Cell invasion assays were performed similar to the migration assay except that the transwell membrane was precoated with 24 mg/mL Matrigel (R&D Systems, Minn, USA).

Animal experiments
Animal experiments were approved by the Ethical Committee for Animal Research of Southern Medical University (protocol number: NFYY-2018-76) and conducted based on the state guidelines from the Ministry of Science and Technology of China. All nude mice (4-6 wk old, male) were purchased from the Central Animal Facility of Southern Medical University. The in vivo metastasis model was established by tail vein injection. Brie y, 1 × 10 6 NPC cells were suspended in 100 μL serum-free 1640 medium and injected into the tail vein of nude mice (10 mice in each group). Accordingly, 6 wk later, whole bodies, and freshly dissected internal organs, including lungs and livers were collected for uorescence imaging with the LT-9MACIMSYSPLUS whole-body imaging system (Encinitas, CA, USA). Organs were xed in 4 % paraformaldehyde (Macklin) for 48 h and transferred to gradient ethanol (Huada). Then, organs were embedded in para n (Shitai, Jiangsu, China), sectioned using a Leica RM2235 microtome (Leica Biosystems, Weizler, Germany) and processed for histological examinations.
Immunohistochemistry staining Para n sections prepared from patients were applied to IHC staining for the detection of the levels of PAG1 protein, using the indirect streptavidin-peroxidase method. All antibodies used for IHC are listed in Supplementary Table 4. The intensity of immunostaining was scored as weak (1), medium (2), and strong (3). The extent of staining, de ned as the percentage of positive staining cells, was scored as 1 (≤25 %), 2 (26-50 %), 3 (51-75 %), and 4 (>75 %). An overall expression score, ranging from 0 to 7, was obtained by adding the score of the intensity and that of the extent of staining. The nal staining score was given as low expression (overall score of 1-4), or high expression (overall score of 4-7).
Wound scratch assay Cells (5 × 10 5 ) were seeded in a 6-well culture dish and grown to 90 % con uence. A single wound was made in the center of the cell monolayer and cell debris was removed by washing twice with PBS (Invitrogen). Complete medium was added and 5-8F or HONE1 cells were allowed to migrate into the clearing area for 24 or 12 h, respectively. Wound closure areas were visualized under an inverted microscope with a 100× magni cation, and the migrated areas were counted.  Table 4) at 4 °C overnight. Consecutively, cells were incubated with Alexa Fluor 488 goat anti-rabbit IgG (1:500, Proteintech, Rosemount, IL, USA) and Alexa Fluor 647 goat anti-mouse IgG (1:500, Proteintech) at 37 °C for 1 h. Coverslips were mounted on slides using anti-fade mounting medium with DAPI (Invitrogen). Accordingly, IF images were acquired on an OLYMPUS confocal micrograph system, and analyzed using the FV10-ASW1.7 viewer software (Olympus, Japan).

RNA oligos and cell transfection
The control mimic, miR-BART8-3p mimic, control inhibitor, and miR-BART8-3p inhibitor were synthesized by Integrated DNA Technology (GenePharma, Suzhou, China) (Supplementary Table 5). The culture medium was changed to fresh RPMI-1640 with 10 % FBS 24 h before transfection. The mimics and inhibitors were transfected to cells using Lipofectamine 3000 (Invitrogen, Waltham, MA, USA) at a nal concentration of 50 nM. The medium was changed again to fresh medium 6 h after transfection.
qRT-PCR analysis Total RNA was extracted using the TRIzol reagent (Invitrogen), and complementary DNA (cDNA) was synthesized with the PrimeScript RT reagent Kit (TaKaRa, Dalian, China). Accordingly, qRT-PCR analysis was performed in triplicate using the SYBR Premix ExTaq (TaKaRa). The primers used for ampli cation of genes of interest are listed in Supplementary Table 6. Quanti cation of EBV-miR-BART8 was conducted with TaqMan microRNA assays (ABI, Shanghai, China). Mature miRNAs were reverse transcribed, and qRT-PCR was performed using the All-in-One miRNA qRT-PCR Detection Kit (GeneCopoeia, Guangdong, China) following the manufacturer's protocol. RPU6B and β-actin were used for normalizing the expression of miRNA and mRNA, respectively. Fold changes were calculated using the 2 -ΔΔCq method.

Plasmid preparation and cell transfection
The GV230 expression vector (http://www.genechem.com.cn) containing the whole coding sequence of PAG1, and HA-vimentin, as well as the GV170 control vector were purchased from GeneChem (Shanghai, China). Plasmid DNAs were puri ed using the TIANprep Mini Plasmid Kit (TIANGEN, Beijing, China) and transduced into NPC cells following the manufacturer's instructions.

RNA Sequencing
RNA-deep sequencing was performed and analyzed in Aksomics, Inc, Shanghai, China. In brief, mRNAs were isolated from DNase-treated total RNA using the Dynabeads mRNA Puri cation Kit (Invitrogen) . According to the manufacturer's instructions, mRNAs were fragmented with divalent cations and converted to single-stranded cDNA using random hexamer primers and Superscript II reverse transcriptase (Invitrogen). The second strand of cDNA was generated by RNase H (Enzymatics,Beverly, MA,USA ) and DNA polymerase (Enzymatics). Subsequently, cDNA products were puri ed using Ampure beads XP (Beckman, Indianapolis, IN, USA). After converting the overhangs into blunt ends using the T4 and Klenow DNA polymerases (Enzymatics), an extra "A" base was added to the 3′-end of cDNA by the Klenow enzyme. Sequencing adapters were then ligated to the end of cDNA by T4 DNA Ligase (Enzymatics). Fragments of 200 bp were selected using Ampure beads XP (Beckman) and enriched through 12 cycles of PCR. PCR products were loaded into a owcell to generate clusters and then sequenced using the Hiseq 2000 system (Illumina). Selected results are shown in Supplementary Table 7.

Co-immunoprecipitation and mass spectrometry assays
To determine potential PAG1-binding proteins, HONE1 and 5-8F cells transfected with the empty vector, as well as ag-PAG1-expressing HONE1 and 5-8F cells were used for coIP employing anti-ag beads. The PAG1 protein complex was eluted using a 0.1 M blycine solution (Fude), separated on an SDS gel, visualized by silver staining using the silver staining kit (Invitrogen), and analyzed by MS at Genepharma (Supplementary Table 9). CoIP assays were performed using 1 mg cell lysates in NP-40 buffer (Fude), with anti-HA and anti-Flag antibodies being employed to pull down the PAG1 and vimentin protein, respectively.

Ethical statement
This study was reviewed and approved by the Ethics Committee of Nanfang Hospital, Southern Medical University (Guangzhou, Guangdong, China) and was conducted in accordance with the Declaration of Helsinki.

Statistical analysis
All experiments were performed in triplicate. Data shown are mean ± s.e.m. (unless otherwise speci ed), from at least 3 independent experiments. The SPSS 16.0 software (IBM SPSS Statistics, Chicago, IL, USA ) was used for statistical analyses. Differences were considered to be statistically signi cant at values of P < 0.05 by Student's t-test or χ 2 test (categorical variables) for 2 groups, or by one-way ANOVA (analysis of variance) analysis for multiple groups. Correlations were analyzed using the two-tailed Spearman's correlation analysis. Single, double, and triple asterisks indicate statistical signi cance *P < 0.05, **P < 0.01, and ***P < 0.001.

Epstein-Barr virus (EBV)-positive radioresistant nasopharyngeal carcinoma cells promoted metastasis
Recurrence of NPC has often been associated with radiotherapy resistance. Metastasis is a signi cant characteristic of the recurrence of NPC. Based on preliminary analysis of clinical samples, we found that the metastasis rate of patients with recurrent NPC was much higher than that of patients with primary NPC (Supplementary Figure 1). To explore the correlation between metastasis and radioresistance, we generated an acquired radioresistant NPC cell line, termed C666-1R. The radioresistance of this cell line was veri ed by colony formation assay ( Fig. 1A and 1B). Moreover, we used transwell and Boyden assays to detect the migration and invasion ability of C666-1R cells. These assays showed that C666-1R cells had a more visible migration and invasion ability than C666-1 cells, which was con rmed by quantitative analysis (Fig. 1C). In addition, we generated a tumor model in nude mice through Caudal vein injection of C666-1R and C666-1 cells. On day 14, live-animal imaging technology and observation of major organs (lungs and liver) revealed the occurrence of more obvious metastasis in the C666-1R group. The results of the histologic evaluation and statistical results from different groups are shown in Figure 1D-G. The above in vitro and in vivo experiments con rmed that radioresistant NPC cells exhibited stronger invasion and metastasis ability.

Epstein-Barr virus (EBV)-encoded miR-BART8-3p impelled the migration, invasion, and epithelialmesenchymal transition of nasopharyngeal carcinoma cells
In order to explore the mechanism of radiotherapy resistance promoting metastasis of NPC, we detected the expression level of miRNAs in C666-1R cells and discovered the increased expression of miR-BART8-3p (Supplementary Figure 2). In previous studies, we found that the EBV-encoded miRNA-BART8-3p was closely related to the radioresistance of NPC. Next, we explored whether miRNA-BART8-3p might also play a role in invasion and migration. Respectively, we generated miR-BART8-3p transfected cell lines using the 5-8F and HONE1 EBV-negative NPC cell lines (EBV products, including miRNA-BARTs are not encoded) (Supplementary Figure 3). Wound-healing and transwell assays demonstrated that upregulation of miR-BART8-3p dramatically increased the ability of either 5-8F-BART8 (6 h, p<0.001, 12h, p<0.001) or HONE1-BART8 (12 h, p<0.05, 24h, p<0.001) cells for migration compared with the relative control ( Fig. 2A). Boyden assays revealed that the upregulation of miR-BART8-3p significantly increased the ability for cell invasion (Fig. 2B-C). To verify these results, we downregulated the expression of miR-BART8-3p in HNE1-EBV and 5-8F-BART8-3p cells through transfection with BART8-inhibitory oligonucleotides. Consistent with the upregulation results, downregulation of miR-BART8-3p by in-BART8-3p was observed to dramatically decrease the migration and invasion abilities of cells ( Supplementary Figure 4-7). In addition, we noted a change in cell morphology under microscopic observation, following the transfection of 5-8F and HONE1 cells with miR-BART8-3p; the key characteristics of EMT are known to include a morphological change from a cobblestone-like epithelial appearance to an elongated, spindle-like broblastic shape (Fig. 2D). This phenomenon led us to further explore the effect of miR-BART8-3p on EMT of NPC cells. Besides the morphological change, other important features of EMT are known to include cytoskeletal reorganization, cadherin switching involving downregulation of epithelial E-cadherin and upregulation of mesenchymal N-cadherin, enhanced resistance to cell death, and acquisition of a migratory phenotype. Accordingly, we detected the cadherin switching and our results further showed that miR-BART8-3p increased the RNA and protein levels of stromal markers (N-cadherin, vimentin) and decreased those of the epithelial marker (E-cadherin) (Fig. 2E-F). We observed stained NPC cells using confocal microscopy and clearly observed the position of E-cadherin and vimentin (Fig. 2G). These results suggested that miR-BART8-3p regulates the migration, invasion, and EMT of NPC cells.
Epstein-Barr virus (EBV)-encoded miR-BART8-3p directly targeted PAG1 in nasopharyngeal carcinoma cells To further reveal the mechanism of the miR-BART8-3p regulation of NPC metastasis, we compared the gene expression profiling of HONE1-BART8-3p versus HONE1-NC cells employing RNA-sequencing analysis. Our results identified 692 downregulated genes, from which many metastasis-associated candidates could be retrieved (Fig. 3A). Previous studies reported that the VHOT algorithm could predict 112 genes directly targeted by miR-BART8-3p (Supplementary Table 8). Combined with the above 2 methods, we screened out the potential target gene PAG1 (Fig. 3B). To clarify whether PAG1 was a direct cellular target for miR-BART8-3p, we performed luciferase reporter assays by cotransfection of a wild-type (WT) or mutant (MUT) PAG1 3′-UTR-containing luciferase reporter vector with a miR-BART8-3p mimic. The luciferase activity of the WT PAG1 3′-UTR, but not that of the MUT 3′-UTR was shown to be significantly reduced by the BART8-3p mimic (Fig. 3C-D).Concomitantly, we found that the expression level of PAG1 was decreased in NPC cells and radioresistant cells (Fig. 3E-F). Further evaluation of the regulatory effect of miR-BART8-3p on PAG1 showed that upregulation of miR-BART8-3p resulted in signi cant downregulation of PAG1 at both the RNA and protein levels ( Fig. 3G-H). The above results suggested that miR-BART8-3p could regulate PAG1 in both structure and composition. To verify the plausibility of this conclusion, we reexamined the expression levels of PAG1 in patient tissues using IHC and observed a signi cant decrease in the expression of PAG1 in NPC tissues compared with NP tissues (Fig. 3I).

PAG1 regulated radioresistance and metastasis of nasopharyngeal carcinoma cells
To study the correlation between PAG1 and the radioresistance or metastasis of NPC, we transfected a PAG1 expression vector into HOEN1 and 5-8F cells. Increased expression of PAG1 was shown to significantly increase the radiosensitivity of the cell lines (Fig. 4A). Besides, we transfected a PAG1 small interfering RNA plasmid vector into HOEN1 and 5-8F cells (supplementary gure 8), which was observed to lead to the increased migration and invasion of both HOEN1-siPAG1 and 5-8F-siPAG1 cells compared with those induced by the relative siPAG1 plasmid control (Fig. 4B-C). Likewise, EMT related detection revealed that siPAG1 increased the levels of stromal markers, whereas decreased those of epithelial markers (Fig. 4D-F). These results suggested that PAG1 regulates the radioresistance and metastasis of NPC cells.

PAG1 interacted with vimentin
To further investigate the underlying mechanism by which PAG1 promoted NPC migration and metastasis, we performed an immunoprecipitation assay followed by mass spectrometry to identify PAG1-interacting proteins (Fig. 5A). Respectively, we found vimentin as one of the potential PAG1interacting proteins from a list of identi ed proteins, and our coIP assay veri ed the correlation between PAG1 and vimentin (Fig. 5B). Immunoprecipitation of endogenous PAG1 resulted in the detection of the presence of vimentin, with the reciprocal performed coIP also con rming the correlation between PAG1 and vimentin (Fig. 5B). In addition, we clearly observed the colocalization of PAG1 with HA-vimentin, but not HA alone, in HA-vimentin transfected cells, under the same conditions (Fig. 5C), which further con rmed the combination of PAG1 and vimentin in NPC cells.

Epstein-Barr virus (EBV)-miR-BART8-3p promoted metastasis of nasopharyngeal carcinoma cells in vivo.
In order to evaluate the necessity of PAG1 in the regulatory mechanism of miR-BART8-3p, we restituted the expression of PAG1, and found that it significantly reduced the migration, invasion, and EMT phenotypic transformation of 5-8F-BART8-3p cells compared with those induced by the relative PAG1negative plasmid control ( Fig. 6A-D). Considering that overexpression of miR-BART8-3p in EBV-negative NPC cells did not accurately simulate the pathogenesis, we further used EBV-positive NPC cells for validation ( Fig. 6A-D). So, we came to the conclusion that restored PAG1 could rescue the phenotypes produced by miR-BART8-3p. Then, we xenografted 5-8F-BART8-3p cells into the caudal vein of nude mice. Upregulation of miR-BART8-3p was demonstrated to significantly increase the metastasis in liver and lungs compared with the 5-8F-NC (Fig. 6E-H), in which miR-BART8-3p was shown to promote the metastasis of NPC.
Epstein-Barr virus (EBV)-miR-BART8-3p and PAG1 were independent prognostic factors for the clinical outcome in patients with nasopharyngeal carcinoma, respectively To investigate whether the expression of miR-BART8-3p might be associated with clinicopathologic features of NPC, we determined the expression of miR-BART8-3p in a cohort of 82 NPC samples with known TNM-stage (Fig. 7A). The expression of miR-BART8-3p was shown to be dramatically increased in N2-3 as compared with N0-1 stages. Similarly, a substantial higher level of miR-BART8-3p was observed in advanced clinical stage III-IV compared with early clinical stage I-II (Fig. 7A). Thus, these results suggested that EBV-encoded miR-BART8 might contribute to the metastasis of NPC. Using IHC, we further determined the expression of PAG1 in another cohort of 111 NPC samples with known TNM-stage. The expression of PAG1 was observed to be decreased in M1 as compared with M0 stages. Similarly, a substantial lower level of PAG1 was observed in advanced clinical stage III-IV relative to early clinical stage I-II. Besides, the expression level of PAG1 was shown to be lower in advanced than primary NPC (Fig. 7B-C). Among the 110 patients, the 5-y overall (OS), progression-free (PFS), local relapse-free (LRFS), and distant metastasis-free (DMFS) survival were observed to be signi cantly improved in the PAG1 high expression group compared with the PAG1 low expression group (Fig. 7B-E). In summary, PAG1 was demonstrated to be closely related to the clinical prognosis in NPC.

Discussion
In this study, we observed that radioresistant NPC cells exhibited a greater tendency for metastasis than radiosensitive NPC cells. In order to nd the crux factor for this problem, we tested the expression of EBVencoded miRNAs in radioresistant or radiosensitive C666-1 NPC cells and found that the expression of miR-BART8-3p was signi cantly increased in these cells. More speci cally, miR-BART8-3p could promote the invasion, metastasis, and EMT of NPC cells by targeting and inhibiting the PAG1 host gene, which combined with vimentin targets the activity of skeleton proteins and might thus affect their regulation and function. Moreover, miR-BART8-3p and PAG1 were closely related to TNM stages and survival, suggesting that they are important prognostic factors in NPC.
The principal obstacle to long-term survival after NPC radiotherapy is radioresistance. It is noteworthy that the rate of metastasis in patients with recurrent NPC is signi cantly higher than that in patients with primary NPC. The pathogenesis of this process remains uncertain. We conducted analysis of clinical samples and in vivo and in vitro functional experiments and demonstrated the phenomenon that radioresistant NPC has a greater tendency to EMT and metastasis. To explore the plausible molecular mechanisms, and whether changes in genes induced by radiotherapy might lead to the improvement of the ability of cells for invasion and metastasis, we screened the EBV-encoded miRNA BART8-3p identi ed from the table of differences between radiosensitive and radioresistant NPC cells. The miR-BART8-3p is known to target the PAG1 host gene in NPC and regulate the mutual binding of PAG1 and vimentin, partly explaining the enhanced ability for invasion and metastasis caused by radioresistance. However, there are often 2 sides to an issue. An increasing number of studies have suggested that EMT or EMT-factors represent a critical process affecting the DNA damage response (DDR)-mediated sensitivity of cancer cells to radiotherapy and chemotherapy [8][9][10]. For instance, the Snail, Slug, and ZEB1 EMT-factors were determined to be DNA repair regulators of radiotherapy and chemoresistance in colorectal, bladder cancer, and so on [11][12][13][14] Further studies are needed to discover various other mechanisms connecting EMT and radioresistance.
Infection by EBV is the most important characteristic in NPC. EBV-encoded miRNA BARTs have been reported to regulate several functions by targeting and regulating host genes. Our team has been focusing on the functional and clinical signi cance of EBV-encoded BARTs, but so far has focused on the single aspect of the treatment of resistance or metastasis. Based on these reports, both EBV-miR-BART1, and EBV-miR-BART7-3p have been shown to induce tumor metastasis by regulating various kinds of pathways in NPC [7,15]. We further discovered that EBV-miR-BART7-3p increased the chemoresistance of NPC [16]. With the development of the research of BARTs, we began to explore the bridge between the treatment of resistance and metastasis. Studies have shown that 52 % of NPC demonstrated high level expression of EBV-miR-BART8-3p and could induce EMT, promoting the metastasis of cells [17], which was consistent with our results. In our previously reported study, we found that the EBV-encoded miRNA BART8-3p promoted radioresistance in nasopharyngeal carcinoma by regulating the ATM/ATR signaling pathway [18]. In this study, we combined radioresistance with metastasis and found that radioresistant NPC cells were characterized by the overexpression of miR-BART8-3p , promoting PAG1-dependent EMT and metastasis. These primary study results suggested that radioresistant NPC cells have a stronger ability for metastasis due to a gene advantage. In addition to the positive effect of radioresistance on metastasis, we also preliminary observed the reverse trend of the effect of metastasis to radioresistance.
In our evaluation of the EMT status of NPC cells, we noticed that changes in ZEB1 were associated with EBV-miR-BART8-3p and PAG1 and further detected that the expression of ZEB1 was exhibiting a positive correlation with radioresistance or overexpression of miR-BART8-3p, but a negative correlation with PAG1 (Supplementary Figure 9); however, the mechanism needs to be further explored.
The protein encoded by PAG1 is a type III transmembrane adaptor protein that binds to the tyrosine kinase family of proteins. First, it is thought to be involved in the regulation of the activation of T-cells and mast cells. It has been also reported to regulate the formation of immunological synapses and cell adhesion signaling by preventing the dynamic arrangement of lipid raft proteins. Then, a study found that PAG1 might exhibit ambivalent functions in several carcinomas by interacting with the Src family kinase (SFK), regulating downstream effector pathways [19]. Especially important, upregulation of PAG1 has been shown to contribute to the promotion of tumor progression and chemoresistance of adipose-derived mesenchymal stem cells in breast cancer [20]. In addition, we also found that the expression of PAG1 exhibited obvious organizational heterogeneity. Although, high levels of expression of PAG1 has been shown in cancers, such as breast cancer and liver hepatocellular carcinoma, PAG1 has been reported to exhibit low expression in lung squamous cell carcinoma and rectum adenocarcinoma (Supplementary Figure 10). A recent research study suggested that PAG1 might be associated with radioresistance, promoting the inherent radioresistance of laryngeal cancer cells via activation of STAT3 or interaction with integrin β1 [21,22]. At present, the expression and function of PAG1 in NPC remains unclear. In this study, we found that the expression of PAG1 was negative correlated with the status of patients with NPC, especially the recurrence of NPC. In terms of its function and mechanism, we found that PAG1 binds to the vimentin protein, which is an EMT-related cytoskeletal protein, and inhibits the invasion, migration, and EMT of NPC cells. Besides, overexpression of PAG1 promoted the radiosensitivity of NPC cells.
Based on the analysis of clinical data, we found the OS, PFS, LRFS, and DMFS were signi cantly improved in the low PAG1-expressing group compared with the high PAG1-expressing group. So, we will continue to explore the role of PAG1 in the DNA repair process (intrinsic radioresistance) and metastasisassociated radioresistance in NPC.

Conclusions
In summary, we demonstrated a correlation between radioresistance and metastasis in NPC, which depended on the elevated levels of the EBV-encoded miRNA BART8-3p. Additionally, we also found that miR-BART8-3p promoted EMT and metastasis of NPC by targeting the inhibition of the PAG1 host gene.
Overall, our ndings indicated that high expression of miR-BART8-3p and low expression of PAG1 predicted a poor clinical outcome, and thus could be exploited as critical targets for the development of new therapeutic strategies for the treatment of NPC.      miR-BART8-3p increased tumor metastasis in the lungs, and liver. N = 10. Differences were evaluated with the χ2-test. ***P < 0.001.