Human HSC were isolated from healthy peripheral blood mononuclear cells (PBMC) cells using EasySep™ Human Progenitor Cell Enrichment kit with Platelet Depletion (cat. no. 19356), while human NK cells were isolated from healthy PBMC cells using EasySep™ Human NK Cell Isolation kit (Stemcell Technologies, Inc.) according to the manufacturers' instructions. A number of PBMCs, NK cells or isolated healthy HSC were conditionally immortalized using hTERT lentivirus vector with an extended life span to achieve a higher transfection efficiency and experimental stability (25,26). Tumor cell lines, including B95-8, Namalwa, HANK1, SNT8 and SNK6, were purchased from American Type Culture Collection. The B95-8 cell line was used for EBV viral production and subsequent concentration for the infection of HSC, PBMCs or NK cells according to a previously described protocol (27). The EBV LMP1 adenovirus for LMP1 transient infection and related empty adenovirus control were kindly provided by Dr Haimou Zhang (Hubei University, China). Antibodies against CdxA (Cdx1; cat. no. sc-515146), Ets1 (c-Ets; cat. no. sc-55581), forkhead box D3 (HFH2; cat. no. sc-517206), GA-binding protein α (GABPα; cat. no. sc-28312), GATA1 (cat. no. sc-265), GATA2 (cat. no. sc-267), Ki-67 (cat. no. sc-101861), nuclear respiratory factor 1 (NRF1; cat. no. sc-101102), NF-κB p65 (cat. no. sc-398442) and organic cation transporter 1 (Oct1; cat. no. sc-8024) were obtained from Santa Cruz Biotechnology, Inc. Antibodies against DRP1 (cat. no. ab140494), optic atrophy 1 (OPA1; cat. no. ab157457), mitofusin 2 (MFN2; cat. no. ab56889) and PGC1β (cat. no. ab240188) were obtained from Abcam. Antibodies against EBV LMP1 (cat. no. NBP1-79009) and 8-oxo-deoxyguanosine (dG) (cat. no. 4354-MC-050) were obtained from Novus Biologicals, Ltd. The small interfering RNA (siRNA) for GABPα and NRF1, and scrambled control siRNA were purchased from Ambion (Thermo Fisher Scientific, Inc.), and the target sequences of siRNA were provided in Table SI. The expanded section of Materials and methods is available in Appendix S1.

The DRP1 gene promoter (2 kb upstream) from Ensembl gene ID: DNM1L-201 ENST00000266481.10 (for DRP1) (with the following link: http://uswest.ensembl.org/Homo_sapiens/Transcript/Summary?db=core;g=ENSG00000087470;r=12:32679303-32744350;t=ENST00000266481), was amplified by PCR from SNK6 genomic DNA, and then subcloned into the pGL3-basic vector (cat. no. E1751; Promega Corporation) by using the following primers with restriction sites for MluI and XhoI, respectively: DRP1 forward, 5′-gcgc-acgcgt-agt tgg ggc cac agg tat gca-3′ (MluI) and reverse, 5′-atcg-ctcgag-aca gtt cgc ctc ctt cct cct-3′ (XhoI).

The shRNA lentivirus particles for human PGC1β, human NFκB-p65, EBV LMP1 and non-target control were designed and purchased from MilliporeSigma, and all the shRNA target sequences were provided in Table SI. These lentiviruses were used for infection of tumor cell lines (e.g. SNK6). The treated cells were cultured in the presence of 10 µg/ml puromycin, and the gene knockdown efficiency was evaluated by reverse transcription-quantitative PCR (RT-qPCR) based on an mRNA decrease of >65% compared with that in the control group. The primer sequences are shown in Table SII (23,24).

Treated SNK6 cells were transferred to coated cover slips, and then fixed with 3.7% formaldehyde, incubated with 1% BSA and 0.2% Triton X-100 for 1 h, and blotted with 40 µg/ml primary antibody against PGC1β, Ki-67 or 8-oxo-dG for 4 h. In triple-staining experiments, the live cells were stained with MitoTracker™ Green FM before fixation. After thorough washing, the cells were incubated with either FITC or Texas Red-labeled secondary antibody (1:1,000) for 1 h, and the cell nuclei were further stained with 300 nM DAPI (cat. no. D9542; Sigma-Aldrich; Merck KGaA). The cells were visualized under a confocal microscope and the staining was quantitated by ImageJ 1.52v software (National Institutes of Health).

Treated cells were fixed with 2.5% glutaraldehyde at room temperature for 15 min, and then dehydrated by using a series of increasing concentrations of ethanol, and embedded in epoxy resin and propylene oxide overnight. The 70 nm-thick sample sections were then stained with lead citrate, and mitochondria morphology was detected by double-blinded technicians using a transmission electron microscope (28).

DNA methylation on the human PGC1β promoter was evaluated by methylation-specific PCR according to a previously described method with minor modifications (29–31). Genomic DNA was extracted and purified from treated cells, and then treated by bisulfite modification with EpiJET Bisulfite Conversion kit (cat. no. K1461; Thermo Fisher Scientific, Inc.), and the DNA was then amplified by using the following primers: Methylated primers: Forward 5′-ttt tta aag tgt tgg gat tat agg c-3′ and reverse 5′-acg tta cgt taa cgc taa acg a-3′; and unmethylated primers: Forward 5′-ttt aaa gtg ttg gga tta tag gtg t-3′ and reverse 5′-tca cat tac att aac act aaa caa a-3′. The product sizes were as follows: 142 bp (methylated) with melting temperature (Tm) 66.4°C and 142 bp (unmethylated) with Tm 65.8°C; CpG island size: 197 bp. The methylation results were calculated by normalization of the unmethylated PCR results (32).

Tumor cells were treated, and then 2×10⁶ viable cells in 0.1 ml PBS were injected into the lateral tail vein of mice. The null mice receiving tail vein injections were separated into the following 4 groups (n=9): i) Group 1 [control (CTL)], conditionally immortalized HSC at passage #6 after vehicle lentivirus infection; ii) group 2 (EBV), conditionally immortalized HSC at passage #6 after EBV infection; iii) group 3 (LMP1↑), conditionally immortalized HSC at passage #6 after LMP1 adenovirus infection; and iv) group 4 (EBV/shPGC1β), conditionally immortalized HSC at passage #6 after EBV infection and shPGC1β lentivirus infection in the presence of 10 µg/ml puromycin. Mouse survival was monitored and calculated; the tumor growth was calculated; the lungs were isolated and stained with hematoxylin and eosin (H&E); and the gene expression and superoxide anion (O₂^−.) release from tumor tissues were evaluated (7,23,33).

The mean level and standard deviation (SD) were calculated and presented, and each experiment was performed for at least 4 independent times (n=4) unless otherwise mentioned. The data was analyzed as normal distribution using Shapiro-Wilk test to evaluate the normality of the data in SPSS 22 software (34), and the one-way ANOVA (analysis of variance) together with Tukey-Kramer test were employed to determine significant difference of different groups. The mouse survival curve was established by Kaplan-Meier survival analysis followed by the log-rank test through SPSS 22 software, and P<0.05 was considered to indicate a statistically significant difference (33).

The effect of LMP1 on PGC1β expression was determined. SNK6 cells were treated with CTL, shLMP1, shNF-κB-p65 or shPGC1β lentivirus before being harvested for biological assays. It was found that knockdown of either NF-κB-p65 (shNF-κB-p65) or PGC1β (shPGC1β) significantly reduced PGC1β mRNA levels in SNK6 cells compared with the findings in the CTL group, while LMP1 knockdown (shLMP1) had no effect (Fig. 1A). Next, the levels of proteins corresponding to the LMP1, NF-κB-p65 and PGC1β genes were measured, and it was revealed that the protein levels were similar to the mRNA levels (Figs. 1B and C, and S1A). Next, PGC1β reporter activity was evaluated in the cells, and it was identified that shNF-κB-p65 treatment significantly reduced PGC1β reporter activity, while shLMP1 and shPGC1β showed no effect (Fig. 1D). Gene expression was also determined in other EBV-positive cell lines, including HANK1 (Fig. S2A) and SNT8 (Fig. S2B) cells, and it was found that treatment with either shNF-κB-p65 or shPGC1β significantly reduced the PGC1β mRNA levels, while shLMP1 showed no effect.

In addition, the potential effects of the treatment on epigenetic changes on the PGC1β promoter were measured, and it was observed that none of the treatments had a significant effect on histone 3 methylation (Fig. S3A), DNA methylation (Fig. S3B), histone 3 acetylation (Fig. S3C) or histone 4 methylation (Fig. S3D). Finally, the effect of LMP1/PGC1β expression on tumor cell growth was explored, and it was demonstrated that shLMP1 treatment had no effect, while treatment with either shNF-κB-p65 or shPGC1β significantly reduced cellular proliferation, as evaluated by thymidine incorporation (Fig. 1E), glucose uptake (Fig. 1F), colony formation assay (Fig. 1G) and Ki-67-positive cell rate (Fig. 1H and I). It was concluded that LMP1 knockdown did not affect PGC1β expression or its subsequent tumorigenesis in EBV-positive cells.

The current study determined the mechanism behind PGC1β-mediated DRP1 regulation in immortalized HSC. Different progressive 5′-promoter deletion constructs for DRP1 were generated and transfected into HSC together with full-length DRP1 reporter plasmids. The cells were then infected with either PGC1β or empty lentivirus (CTL) for luciferase activity assay, and the results showed that PGC1β-induced DRP1 reporter activity did not change among the −2,000, −1,600, −1,200, −800, −700, −600, −500 and −400 deletion constructs (numbered according to the Ensembl gene ID: DNM1L-201 ENST00000266481.10), while PGC1β-induced reporter activation was partly decreased at −300 deletion and completely disappeared at the −200-deletion construct. This indicates that the PGC1β-responsive binding element is located in the range of −400 to −200 on the DRP1 promoter (Fig. 2A).

All the potential transcription factor binding sites in the range of −400 to −200 were analyzed from the database, and the following motifs were identified: HFH2, GABPα (marked in red), CdxA, c-Ets, Oct1, CCAAT-enhancer-binding protein, GATA1, NRF1 (marked in red) and GATA2 (Fig. 2B). All the single mutations were then generated on the full length of the DRP1 reporter construct (p-DRP1-2000) for these binding motifs for the luciferase reporter assay, and the results showed that only single mutants for GABPα (from ggaa to ttcc, in green at −383) and NRF1 (from cctg to aagt, in green at −232) showed a significant reduction in reporter activity, indicating that GABPα and NRF1 may be responsible for PGC1β-induced DRP1 activation (Fig. 2C). Next, GABPα or NRF1 single or double mutants (M-383GABPα/-232NRF1) were transfected for the reporter assay, and it was found that both single mutants partly reduced, while the double mutants completely reduced, the PGC1β-induced DRP1 reporter activation (Fig. 2D).

The binding abilities were then determined through chromatin immunoprecipitation analysis on the DRP1 promoter. The results revealed that both GABPα and NRF1 increased the binding ability after PGC1β infection in HSC compared with that of the CTL group, while other transcriptional factors had no effect (Fig. 2E). Additionally, both GABPα and NRF1 showed reduced binding ability after shPGC1β lentivirus infection in SNK6 cells compared with that in the CTL group, while other transcriptional factors had no effect (Fig. 2F).

Finally, the potential effect of GABPα and NRF1 on DRP1 expression in SNK6 cells was determined by siRNA techniques, and it was observed that GABPα and NRF1 were knocked down successfully by siGABP and siNRF1 respectively. Additionally, the DRP1 mRNA level was significantly decreased by single knockdown of either GABPα or NRF1 compared with that of the CTL group, and the DRP1 mRNA level was further decreased by double knockdown (siGABPα/NRF1) compared with that of the single siGABPα group (Fig. 2G). The DRP1 reporter activity was also determined, and it was found to be similar to the DRP1 mRNA levels (Fig. 2H). It was concluded that DRP1 expression was regulated by PGC1β through GABPα and NRF1 binding sites on the DRP1 promoter.

The present study evaluated the potential effect of LMP1 expression on epigenetic changes on the PGC1β promoter. Conditionally immortalized HSC were treated with CTL, transient EBV infection, transient LMP1 infection by LMP1 adenovirus (LMP↑), or EBV infection together with permanent PGC1β knockdown through shPGC1β lentivirus in the presence of 10 µg/ml puromycin (EBV/shPGC1β). The cells were continuously cultured from passage #1 to #6 before being harvested for analysis on different passages. The properties of cells at passage #1 were first evaluated, and the results revealed that EBV infection, including EBV and EBV/shPGC1β treatments, significantly increased the quantity of EBV genome copies compared with those of the CTL group (Fig. S4A). Additionally, treatments with EBV, LMP1↑ and EBV/shPGC1β significantly increased LMP1 mRNA levels. Moreover, treatments with EBV and LMP1↑ significantly increased, while EBV/shPGC1β treatment significantly decreased, the PGC1β mRNA levels. None of treatments had a significant effect on NF-κB-p65 expression compared with that of the CTL group (Fig. S4B). The effect on epigenetic changes on the PGC1β promoter at passage #1 was also determined, and it was found that treatments with EBV, LMP1↑ and EBV/shPGC1β significantly decreased the epigenetic modifications on H3K9me2 and H3K9me3 (Fig. S4C) in addition to DNA methylation (Fig. S4D). These treatments had no effect on histone 4 methylation (Fig. S4E) or histone 3 acetylation (Fig. S4F) compared with the CTL group.

Next, the properties of cells at passage #6 were determined. The results showed that the EBV genome was almost non-detectable upon treatment with EBV, LMP1↑ or EBV/shPGC1β after continuous culturing for 6 passages (Fig. 3A). Additionally, LMP1 mRNA was not detectable, while the NF-κB-p65 mRNA levels showed no changes. On the other hand, PGC1β mRNA remained increased after EBV and LMP1↑ treatment, while it remained decreased upon EBV/shPGC1β treatment compared with that in the CTL group (Fig. 3B). The protein levels were also measured, and the protein expression was similar to the mRNA levels (Figs. 3C and D, and S1B). Epigenetic changes on the PGC1β promoter were then determined, and it was found that treatments with EBV, LMP1↑ and EBV/shPGC1β significantly decreased DNA methylation (Fig. 3E) in addition to epigenetic modifications on H3K9me2 and H3K9me3 (Fig. 3F), while having no effect on histone 3 acetylation (Fig. S5A) or histone 4 methylation (Fig. S5B) compared with the findings in the CTL group, where the epigenetic modifications were similar to those noted in cells at passage #1.

Finally, the PGC1β levels in cells from passage #1 to #6 were measured, and it was found that treatments with EBV and LMP1↑ significantly increased, while EBV/shPGC1β decreased, PGC1β mRNA. This remained persistent from passage #1 to #6 (Fig. 3G). The present study also measured DNA methylation on the PGC1β promoter, and found that treatments with EBV, LMP1↑ or EBV/shPGC1β significantly decreased DNA methylation, which remained persistent from passage #1 to #6 (Fig. 3H). The potential effect of LMP1 expression on PGC1β expression in other cells was also determined, and it was revealed that treatments with EBV and LMP1↑ significantly increased, while EBV/shPGC1β treatment significantly decreased, PGC1β expression in both PBMCs (Fig. S6A) and NK cells (Fig. S6B); this remained persistent from passage #1 to #6. Thus, it was concluded that transient EBV/LMP1 exposure caused persistent epigenetic modifications on the PGC1β promoter, which subsequently resulted in persistent PGC1β upregulation even in the absence of EBV/LMP1.

The current study evaluated the effect of EBV/LMP expression on oxidative stress in treated cells at passage #6, and found that both EBV and LMP1↑ treatments significantly decreased reactive oxygen species (ROS) (Fig. 4A) and 3-nitrotyrosine formation (Fig. 4B) formation compared with those of the CTL group, while PGC1β knockdown (EBV/shPGC1β) completely reversed this effect. The oxidative stress-mediated DNA damage was also determined, and the results showed that both EBV and LMP1↑ treatments significantly decreased γH2AX (Figs. 4C and D, and S1C), 8-OHdG (Fig. 4E) and 8-oxo-dG (Fig. 4F and G) formation, while PGC1β knockdown completely reversed this effect. Thus, it was concluded that EBV/LMP1 exposure-mediated PGC1β expression ameliorated oxidative stress.

The present study evaluated the effect of EBV/LMP1 expression on mitochondrial function, and found that both EBV and LMP1↑ treatments significantly increased mitochondrial DNA copies (Fig. 5A), intracellular ATP level (Fig. 5B) and mitochondrial membrane potential (Fig. 5C), in addition to decreasing the apoptotic rate (Figs. 4D and 5E) compared with the findings in the CTL group. Notably, PGC1β knockdown completely reversed this effect.

Mitochondrial mass and gene expression were then evaluated, and the results revealed that both EBV and LMP1↑ treatments significantly increased PGC1β expression and mitochondrial mass by immunostaining (Fig. 5F and G) compared with those of the CTL group. Both EBV and LMP1↑ treatments significantly increased DRP1 expression and decreased MFN2 expression, but had no effect on OPA1 expression (Figs. 5H-J and S1D). Finally, changes in mitochondrial morphology were evaluated by using an electron microscope, and it was found that both EBV and LMP1↑ treatments significantly increased mitochondrial fission compared with that of the CTL group, while PGC1β knockdown completely reversed this effect (Fig. 5K). Therefore, it was concluded that EBV/LMP1 exposure-mediated PGC1β expression potentiated mitochondrial function by upregulation of mitochondrial fission.

The present study evaluated the effect of EBV/LMP1 exposure on tumor cell proliferation, and found that both EBV and LMP1↑ treatments significantly increased [³H]-2-deoxyglucose uptake (Fig. 6A), cellular proliferation by thymidine incorporation (Fig. 6B), colony formation (Fig. 6C and D) and Ki-67 positive cells (Fig. 6E and F), compared with those in the CTL group. Of note, PGC1β knockdown completely reversed this effect. Thus, it was concluded that EBV/LMP1 exposure-mediated PGC1β expression potentiated tumor cell proliferation.

The current study determined the potential effect of EBV/LMP1 expression on in vivo tumor growth. Treated cells at passage #6 were harvested for tail vein injection to monitor the process of tumor growth. Gene expression was first determined in the tumor tissues, and it was found that cells that received treatment with either EBV or LMP1↑ showed significantly increased mRNA levels of PGC1β and DRP1 but decreased mRNA levels of MFN2 compared with those of the CTL group. PGC1β knockdown completely reversed this effect (Fig. 7A).

Next, the protein levels were determined, and they were identified to be similar to the mRNA levels (Figs. 7B and C, and S1E). Furthermore, both EBV and LMP1↑ treatments significantly decreased O₂⁻. release (Fig. 7D) and increased the number of lung metastatic nodules (Fig. 7E) and lung tumor spots (Fig. 7F and G) compared with the CTL group. In addition, both EBV and LMP1↑ treatments significantly decreased mouse survival compared with that of the CTL group (Fig. 7H). PGC1β knockdown completely reversed this effect. It was concluded that EBV/LMP1 exposure-mediated PGC1β expression significantly potentiated tumor growth in vivo and shortened mouse survival.

The current study established a schematic model for EBV/LMP1-mediated tumorigenesis through persistent epigenetic changes on the PGC1β promoter and subsequent upregulation of PGC1β and DRP1: Briefly, transient EBV/LMP1 exposure triggers epigenetic changes, including decreased histone 3 methylation and DNA methylation on the PGC1β promoter, resulting in persistent PGC1β upregulation. Increased PGC1β expression upregulates DRP1 expression through the transcription factors GABPα and NRF1, resulting in mitochondrial fission with potentiated glycolysis, minimized oxidative stress and DNA damage. Taken together, transient EBV/LMP1 exposure triggers persistent PGC1β upregulation and tumorigenesis (Fig. 8).