Recurrent chemical reactivations of EBV promotes genome instability and enhances tumor progression of nasopharyngeal carcinoma cells



Nasopharyngeal carcinoma (NPC) is an endemic malignancy prevalent in South East Asia. Epidemiological studies have associated this disease closely with Epstein-Barr virus (EBV) infection. Previous studies also showed that EBV reactivation is implicated in the progression of NPC. Thus, we proposed that recurrent reactivations of EBV may be important for its pathogenic role. In this study, NPC cell lines latently infected with EBV, NA and HA, and the corresponding EBV-negative NPC cell lines, NPC-TW01 (TW01) and HONE-1, were treated with 12-O-tetradecanoylphorbol-13-acetate (TPA) and sodium n-butyrate (SB) for lytic cycle induction. A single treatment with TPA/SB revealed that DNA double-strand breaks and formation of micronuclei (a marker for genome instability) were associated with EBV reactivation in NA and HA cells. Examination of EBV early genes had identified several lytic proteins, particularly EBV DNase, as potent activators that induced DNA double-strand breaks and contribute to genome instability. Recurrent reactivations of EBV in NA and HA cells resulted in a marked increase of genome instability. In addition, the degree of chromosomal aberrations, as shown by chromosome structural variants and DNA copy-number alterations, is proportional to the frequency of TPA/SB-induced EBV reactivation. Whereas these DNA abnormalities were limited in EBV-negative TW01 cells with mock or TPA/SB treatment, and were few in mock-treated NA cells. The invasiveness and tumorigenesis assays also revealed a profound increase in both characteristics of the repeatedly reactivated NA cells. These results suggest that recurrent EBV reactivations may result in accumulation of genome instability and promote the tumor progression of NPC. © 2008 Wiley-Liss, Inc.

Epstein-Barr virus (EBV) is the etiological agent of infectious mononucleosis and is implicated in the development of several human malignancies including Burkitt's lymphoma (BL) and nasopharyngeal carcinoma (NPC). The life cycle of EBV includes latent and lytic stages. Infection with EBV usually results in latency in B lymphocytes.1 Nevertheless, reactivations of EBV have been observed in oropharyngeal epithelium2, 3 that may result from physiological changes4 or chemical induction by various compounds including phorbol esters (i.e., 12-O-tetradecanoylphorbol-13-acetate, TPA),5n-butyrates (sodium n-butyrate, SB).6 Upon reactivation to the lytic cycle, the immediate early genes Zta and Rta are expressed first, followed by the early genes (EAD, DNase, DNA polymerase, thymidine kinase, …etc.) and finally the late genes (VCA and MA, etc.).7

In addition to the importance of EBV latent infection on carcinogenesis,8, 9 several retrospective and prospective investigations suggest that the EBV reactivation may also play a substantial role in NPC.10–13 Studies carried out in China revealed regions with a high annual incidence of NPC were colocalized with the consumption of herbal drugs containing phorbol esters which effectively induce reactivation of EBV.14 The presence in NPC biopsies of lytic viral genome15 and various lytic gene products, such as BZLF1, BMLF1, BRLF1, gp220 and lytic LMP1,16–18 EAD, MA and VCA,19 as well as BALF120 further supports the notion that EBV replications do occur in some cases. Another line of evidence comes from serological studies on NPC patients. The elevated antibody titers against various EBV lytic gene products21 and increasing viral DNA load22 have been detected in plasma/serum of NPC patients, indicating that reactivation of lytic cycle occurs. The association of EBV reactivation and NPC was further strengthened by some clinical follow-up studies that showed antibody titers against EBV temporally correlate with the stages of cancer.23, 24 Concomitant with remission after therapy, the levels of EBV-specific antibodies and plasma EBV DNA decreased in NPC patients.22, 23 However, these two parameters of EBV reactivation rise again prior to relapse and metastasis.10, 11, 23 Recently, it has also been suggested that malaria antigens may induce EBV reactivation and increase the risk in BL development.25

Genome instability seems to be a hallmark of cancers.26 It has been defined as an increased frequency of genetic changes including subtle sequence changes and chromosomal alterations27 and is considered to be either as the cause or the result of carcinogenesis.28 On the basis of these conspicuous relationships between EBV reactivation and NPC described above, we postulated that recurrent EBV reactivations might induce genome instability and enhance the tumorigenecity of NPC. To test this prediction, we have analyzed the genome instability, chromosomal aberrations and cancerous properties of EBV-positive NPC cells with various incidences of TPA/SB-induced EBV reactivation. We have found that the degrees of genome instability and chromosomal aberrations in NPC cells were proportional to the frequencies of EBV reactivation. The invasiveness as well as tumorigenicity were also enhanced in NPC cells after recurrent TPA/SB-induced EBV reactivations. By examining a panel of EBV early genes, we have found that expression of EBV DNase, uracil DNA-glycosylase and EBV major DNA-binding protein can induce DNA double-strand breaks and enhance genome instability of the host cell. These results suggest that expression of lytic genes during EBV reactivation may result in genome instability and, over recurrent reactivations, lead to the accumulation of chromosomal abnormalities with a resultant enhancement of tumor progression in NPC.


BL, Burkitt's lymphoma; CA, chromosomal aberrations; CGH, comparative genomic hybridization; CNAs, copy-number alterations; DSB, double strand breaks; EBV, Epstein-Barr virus; mDBP, major DNA-binding protein; MN, micronucleus; NPC, nasopharyngeal carcinoma; SB, sodium n-butyrate; TK, thymidine kinase; TPA, 12-O-tetradecanoylphorbol-13-acetate; UDG, uracil DNA-glycosylase.

Material and methods

Cell lines

The EBV-positive NPC cell lines, NA and HA and the parental EBV-negative NPC cell lines, NPC-TW01 (TW01) and HONE-1 cells,29, 30 respectively, were prepared and maintained as described.31 The virus remains in latency II in NA and HA cells which mimic the tumor cells in NPC patients. EBV reactivation in these cells can be induced by treatment with 12-O-tetradecanoylphorbol-13-acetate (TPA) and sodium n-butyrate (SB) (Sigma-Aldrich, ST Louis, MO).31

Plasmids and transfection

The EBV Zta (BZLF1), uracil DNA-glycosylase (UDG, BKRF3), thymidine kinase (TK, BXLF1), major DNA-binding protein (mDBP, BALF2) and DNase (BGLF5) expression plasmids were constructed by cloning the respective coding sequences32 into the pEGFP-CPO-IRES-puro vector, which was derived from pEGFP-C1 (Clontech, Mountain View, CA), with insert of an IRES-puro cassette, by CPO site ligation. Plasmid DNA was transfected by Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.

Analysis of micronucleus formation

Detection of micronucleus (MN) formation was performed as described previously.33 The fixed cells were stained with Hoechst 33258 (0.2 μg/mL) (Sigma-Aldrich). Micronuclei were viewed using a fluorescence microscope and at least 1,000 cells were counted for each experiment.

Double immunofluorescence staining

Cells were fixed and permeabilized by exposure to ice-cold 100% methanol for 15 min. After rehydration and blocking, cells were first proceeded to indirect immunofluorescence staining, using mouse monoclonal antibody (mAb) to EAD as the primary antibodies34 and Rhodamine-labeled goat anti-mouse IgG (Jackson Laboratory, Bar Harbor, MA) as the secondary antibodies. Following a thorough wash, cells were stained subsequently with FITC-conjugated mouse Ab to γH2AX (Upstate, Charlottesville, VA).

siRNA-mediated inhibition of EBV reactivation

Experiments were performed according to the published procedures.35 For siRNA-mediated inhibition of EBV reactivation, NA cells were first transfected with 5 μg siRNA producing plasmids (siGFP and SiZta)35 for 48 hr and then either treated with TPA/SB or mock-treated for 24 hr. Noninsert pSuper vector was used as vector control. After treatment, the cells were harvested and assayed by western blot and anti-γH2AX flow cytometry as described below.

Analysis of γH2AX-presenting cells by flow cytometry

After fixation and immunofluorescence staining with antibody to γH2AX (Upstate), cells were subjected to flow cytometric analysis for the γH2AX-presenting cells, as described previously.36

Western blotting analysis

The cell extracts were separated by SDS-PAGE and transferred to nitrocellulose membrane. After blocking, the blot was incubated with indicated antibodies for 1 hr at room temperature. Antibodies used in this study include anti-EAD,34 anti-EBV Zta, anti-DNase37 and anti-β-actin (Sigma-Aldrich) antibodies. After washing, the blot was incubated with horseradish peroxidase–labeled goat anti-mouse IgG (Jackson Laboratory). After incubation, the blot was washed and developed with substrate.

EBV DNase activity assay

NA and TW01 cells were TPA/SB-treated or mock-treated. The cell lysates of indicated time were prepared, as described previously,38, 39 for DNase activity assay. Cell lysates were incubated with 1 μg/mL salmon DNA for 1 hr at 37°C. Picogreen dye (100 μL; Invitrogen) was then added to the samples and the fluorescence emission was detected at 520 nm. The DNase enzyme activity is represented as the percentage of DNA digested by EBV DNase.

Repeated inductions of EBV reactivation

Recurrent EBV reactivations were achieved by incubation with TPA/SB periodically. The cell cultures at the start of this experiment were defined as passage 0 (P0) cells. After 24 hr incubation, cells were mock-treated or TPA/SB-treated for another 24 hr. After treatment, the cells were recovered by replacing with fresh (treatment-free) medium and incubation for a further 24 hr. The resulting cells were defined as passage 1 (P1). Cells from P1 were trypsinized and transferred and the process repeated (passage 2). “NA(Rn)” represents TPA/SB-treated NA cells, where n stands for the passage number of the cells. “NA(Pn)” represents mock-treated NA cells, where n stands for the passage number of the cells. TW01 cells with mock-treatment or TPA/SB-treatment were named likewise.

Cytogenetic analysis

The analysis of metaphase chromosomes followed the established procedure.40 Chromosomal aberrations were scored in at least 200 metaphase plates for each cell line.

Array-based comparative genomic hybridization analysis

Procedures for array-based comparative genomic hybridization (array CGH) was performed as described previously.41 Briefly, the commercial genome-wide oligo array (Agilent Technologies, New Castle, DE) with an average spatial resolution of ∼35 kbp was used for our experiments. Genomic DNA from low-passage–number NA cells without TPA/SB treatment was used as controls. Experimental samples were derived from TPA/SB-treated NA cell. Genomic DNA fragmentation, labeling and array hybridization were performed according to the standard array-CGH protocol provided by Agilent Technologies.41 The data extraction was performed using Agilent Feature Extraction Software, version 8.1. Another custom analytical tool, Agilent CGH Analytics Software (version 3.4) was used for subsequent data mining. The chromosome aberration regions were calculated by Z-score statistical algorithm with moving average window 5 Mb. The Z-score threshold was set at 2.5 to make an amplification or deletion call.41

In vitro invasiveness assay

In vitro invasion assays were performed as described previously42 using HTS FluoroBlok inserts (Falcon, Cambridge, MA). The transwell membranes were first coated with Matrigel (Becton Dickinson, Franklin Lakes, NJ). Cells (1 × 105) were seeded onto the Matrigel-coated membranes and the inserts were incubated in 24-well plate for 24 hr. After incubation, the membranes were removed from the inserts, fixed with methanol and stained with 50 μg/mL propidium iodide (Sigma-Aldrich). The cells that had invaded and transmigrated to the lower surface of the polycarbonate membrane were photographed under a fluorescence microscope and the cell number was calculated using AIS software (Imagine Research Inc., Ontario, Canada).

In vivo tumorigenesis assay

Six-week-old SCID (severe combined immunodeficiency) mice were purchased from the animal center in Tzu Chi University, Taiwan, for this study. Mice were kept under sterile conditions and were used under protocols approved by the Institutional Animal Care and Use Committee. Tested cells (2 × 106 cells in each case) were resuspended in serum-free DMEM and injected subcutaneously into the dorsal flanks of SCID mice. Injected mice were examined weekly and tumor volumes were estimated from their length (l) and width (w), as measured by calipers, using the formula tumor volume = l w2 × 0.52.


Micronucleus formation and cellular DNA damage are accompanied with EBV reactivation

To determine whether the genome instability is a result of EBV reactivation in EBV-positive NPC cells by TPA and SB treatment, we evaluated the occurrence of micronuclei (MN), which arise from acentric chromatids and have been recognized as a marker for genome instability.43 We found that MN formation was low in EBV-negative TW01 cells (∼2.0%) and HONE-1 cells (∼1.1%), regardless of whether mock-treated or TPA/SB-treated (Fig. 1a). MN formation was slightly higher in the mock-treated EBV-positive NA cells (∼3.0%) and the HA cells (∼1.5%). It seemed that latent EBV infection of NA and HA cells did to some extent affect genome instability. However, an increase in the formation of MN was observed in NA cells (4.5%) and HA cells (2.3%) following TPA/SB treatment (p < 0.05, compared with the mock-treated cells) (Fig. 1a). Since treatment with TPA/SB interrupts EBV latency and triggers the lytic pathway in EBV-positive cells (the “EBV reactivation”),5, 6 these results indicate that genome instability is enhanced by EBV reactivation.

Figure 1.

Reactivation of EBV promotes genome instability and induces cellular DNA damage. (a) MN was examined in NPC cells after 24 hr of TPA/SB treatment. Occurrence of MN formation (MN%) is presented here as the percentage of cells presenting micronuclei, derived from the analysis of 1,000 cells. Bars represent mean value in MN% (triplicates ± SD). *p = 0.021, compared with the mock-treated NA; #p = 0.035, compared with the mock-treated HA. (b) Double immunofluorescence staining of NPC cell lines was performed with antibodies specific for EAD and γH2AX. NPC cell lines TW01, HONE-1, NA and HA cells were mock, TPA/SB, or bleomycin-treated for 24 hr. Bleomycin-treated (500 ng/mL) cells were used as positive controls for DNA breaks. Cells were first stained with anti-EBV EAD antibody (Rhodamine, red fluorescence) and then anti-γH2AX antibody (FITC, green fluorescence). The locations of cell nuclei in the same fields were revealed by staining with Hoechst 33258. Scale bar = 10 μm. (c) DNA damage is attenuated when EBV reactivation is inhibited by siZta. Left panel: Western blot analysis of TPA/SB-treated or mock-treated NA cells by using anti-EBV Zta, EAD and DNase. Transfections of siGFP and vector pSuper were used as negative control. Right panel: Anti-γH2AX flow cytometry of transfected NA cells after 24 hr TPA/SB treatment. Numbers indicated the percentage of γH2AX-presenting cells. Each flow cytometry assay was performed in duplicate with 10,000 cells.

Because genome instability likely is a consequence of chromosomal damage and defective genome surveillance mechanism(s),44 we made an assumption that DNA damage may be detectable with EBV reactivation-related genome instability. To examine this possibility, double immunofluorescence staining was performed by using monoclonal antibody against EAD (an EBV lytic antigen) and γH2AX (a marker for DNA double-strand breaks, DBS)45 at 24 hr post-TPA/SB treatment (Fig. 1b). Interestingly, DNA DSB were observed predominantly in TPA/SB-treated NA and HA cells with EAD expression, but neither detected in TPA/SB-treated HA and NA without EAD expression, nor in TPA/SB-treated TW01, HONE-1 and mock-treated cells (Fig. 1b). These results indicate the coincidence of DNA damage (evaluated here as DNA DSB) and EBV reactivation and that the DNA lesions are not caused by TPA/SB treatment.

To provide additional evidence in support of EBV reactivation, rather than the cytotoxicity of TPA/SB, being the main cause of the DNA damage observed in TPA/SB-treated NA and HA cells, we used interference EBV Zta siRNA to inhibit EBV reactivation under TPA/SB treatment. Transfection of siZta prior to TPA/SB treatment can efficiently inhibit EBV reactivation in NA cells (Fig. 1c). The anti-γH2AX flow cytometry indicated that the percentage of γH2AX-presenting cells was markedly reduced in siZta-transfected NA cells, whereas the percentage of γH2AX-presenting cells was no change when TPA/SB-treated NA cells were transfected with green fluorescent protein (GFP) RNA interference (siGFP) and vector pSuper (Fig. 1c). To further examine whether γH2AX arised from cellular DNA DSB or from viral replicating genome, we carried out γH2AX immunofluorescence staining in the presence of phosphonoacetic acid (PAA), an inhibitor of EBV reactivation.46 There were still substantial γH2AX detected in TPA/SB-treated NA cells in the presence of PAA (Supp. Info. Fig. 1), indicating the observed γH2AX in TPA/SB-treated NA cells did not arise from replicating viral DNA. Altogether, these results suggest that the accelerated DNA lesions in NA and HA cells are mainly a consequence of EBV reactivation.

Expression of EBV lytic genes induce the formation of micronuclei and DNA double strand breaks

Since reactivation of EBV induces DNA DSB and genome instability of host cells, we sought to identify which effectors of EBV reactivation may contribute to these activities. Expression of immediate early and/or early viral gene products has been proved to be required for the induction of chromosomal damage.47 While the induction of DNA DSB can be detected as early as 24 hr post-TPA/SB treatment, we postulate that expression of EBV early genes may contribute to these activities. To address this question, several EBV lytic genes, including EBV Zta, thymidine kinase (TK), major DNA-binding protein (mDBP), uracil DNA-glycosylase (UDG) and DNase, were cloned and transfected into TW01 cells. The formation of MN and phosphorylation of H2AX were examined in the transfected cells expressing respective EBV lytic protein. Comparing with the vector control (2.3%), the MN formation were significantly elevated in EBV UDG and DNase-expressing TW01 cells (3.9% and 6.3%, respectively, Fig. 2a). DNA DSB, as evaluated by the percentage of γH2AX presenting cells, was also found to increase in cells expressing mDBP (15.9%), UDG (23.2%) and EBV DNase (26.0%), when compared with the pEGFP control (9.4%) (Fig. 2b). Among these lytic genes, the EBV DNase showed the very noteworthy effects on both MN and DNA DSB formation of TW01 cells. To confirm that EBV DNase do exert its effects during the reactivation of EBV, the lysates of NA cells were assayed at 12 hr intervals after TPA/SB treatment. The expression of Zta, as well as EAD, was significantly increased after 12 hr of induction, indicating that EBV had been reactivated (Fig. 2c). EBV DNase can be detected at high levels 24 hr after induction (Fig. 2c). The enzyme activity of EBV DNase also increased continuously after induction (Fig. 2d). These data indicate that, following reactivation, the expression of EBV lytic genes, particularly DNase, can induce DNA damage and enhance genome instability of host cells.

Figure 2.

EBV lytic proteins induce the formation of MN and DNA strand breaks. (a) TW01 cells were transfected with pEGFP-CPO-IRES-puro (vector), Zta, thymidine kinase (TK), major DNA-binding protein (mDBP), uracil DNA-glycosylase (UDG) and DNase plasmids. Cells were harvested and were stained by Hoechst for MN counting, 24 hr posttransfection. The values are a mean ± SD from 3 individual experiments (*p < 0.01). (b) TW01 cells were transfected with plasmids the same as described above. Cells were fixed and incubated with γH2AX antibody for cytometry analysis, 24 hr posttransfection. Numbers indicated the percentage of γH2AX-presenting cells. (c) NA cells were treated with TPA/SB. The cell lysates were collected 12, 24 and 36 hr after treatment for western blotting with Zta, EAD, DNase and β-actin antibodies. (d) NA and TW01 cells were mock or TPA/SB-treated. The cell lysates of indicated time point were collected and assayed for DNase enzyme activity. Enzyme activity is represented as percentage of DNA digested by EBV DNase.

Recurrent EBV reactivations by TPA/SB increase genome instability in NA and HA cells

Because recurrent reactivations of EBV may occur before the relapse and metastasis of NPC, we tried to mimic the situation in vivo and carried out recurrent reactivations of EBV in NA and HA cells. To assess the impact of recurrent EBV reactivations on genome instability, a culture system was established to carry out a longitudinal study by repeated incubation of NA and HA cells in medium containing TPA/SB (described in Material and methods). The formation of MN was used to determine how genome instability was affected by recurrent EBV reactivations. As shown in Figure 3, formation of MN was low in TW01 cells (∼2.1%) and HONE-1 cells (∼1.9%) throughout 15 passages, regardless of whether mock or TPA/SB-treated. A similar pattern was observed in the mock-treated NA cells (∼3.0%) and the HA cells (∼2.8%), in which the formation of MN remained at a low-level and did not rise significantly as the passages progressed (p > 0.05). It seemed that the intrinsic genome instability in TW01, HONE-1, NA and HA cells was not increased markedly throughout the 15 passages. However, a profound increase in the formation of MN was observed in NA cells (4.5%–9.1%) and HA cells (2.3%–6.4%), as proportional to the frequency of TPA/SB treatment. These results indicate that genome instability is significantly enhanced by recurrent EBV reactivations in NPC cells.

Figure 3.

Recurrent EBV reactivations concomitant with progressive genome instability in NPC cells. Recurrent EBV reactivations increase MN formation in NA and HA cells. Cells were stained with Hoechst 33258 and MN was examined using a fluorescence microscope. Occurrence of MN formation (MN%) was presented here as the percentage of the number of micronuclei-presenting cells per 1,000 analyzed cells and plotted as function of frequency of EBV reactivation. At least 1,000 cells were counted for each experiment. Data indicate mean value in MN% (triplicates ± SD). *p < 0.001, compared with the NA, TW01 and TW01-TPA/SB. #p < 0.001, compared with the HA, HONE-1 and HONE-1-TPA/SB.

Since the formation of DSB and increased MN observed in NA and HA cells are equivalent, we assume that these 2 cells have similar responses to EBV reactivation. Therefore, the laborious and expensive experiments, karyotype analysis, array CGH, invasiveness assay and SCID mouse transplantation were carried out only in NA cells and their counterpart TW01 cells.

Recurrent EBV reactivations induced by TPA/SB increase extensive chromosomal aberrations and DNA copy-number alterations in NA cells

Since the genome instability is significantly enhanced by recurrent EBV reactivations in NPC cells, the recurrent reactivations may induce an accumulation of genetic alterations and lead to chromosomal aberrations. Chromosomal instability has been associated with carcinogenesis and most human cancers.27, 28 Therefore, karyotype analysis was carried out to elucidate the effect of recurrent reactivations on chromosome instability. Chromosomal aberrations (CA), including dicentric chromosomes, chromosome fragments, chromatid gaps, and chromosome rings were scored in 200 metaphase plates of each mock or TPA/SB-treated TW01 and NA cell lines (Table I). CA were low in TW01 cells at passage 1 and passage 15, regardless of whether mock or TPA/SB-treated. For NA cells, single TPA/SB treatment had slightly increased the frequencies of chromosome fragments and chromatid gaps [NA(R1) vs. NA(P1)]. Serial rounds of passages did show some effects on the CA of the EBV-positive NA cells. By comparing the passage 15 with passage 1 of the NA cells [NA(P15) vs. NA(P1)], the frequencies of dicentric chromosomes, chromosome fragments and chromatid gaps were increased, while chromosome rings were seen rarely. However, the most noteworthy CA were detected in the passage 15 of recurrent EBV reactivated NA(R15) cells, with dicentric chromosomes, chromosome fragments and chromatid gaps increased by several folds compared with the mock-treated NA(P15) cells and with the NA(R1) cells with single treatment (Table I). These data indicated that CA are prevalent among the EBV-positive NPC cells. Moreover, the genome of NA cells becomes more aberrant as the number of EBV reactivation increases.

Table I. Recurrent EBV Reactivations Induce Structural Chromosomal Aberrations
  1. Karyotype analysis of NA and TW01 cells with different times of TPA/SB treatment. Numbers indicated the percentages of aberration patterns observed within 200 metaphase genomes analyzed for each experiment.


During this longitudinal study, chromosome numbers of NA cells on metaphase were also calculated. Interestingly, we found that the chromosome numbers has a tendency to increase after recurrent TPA/SB treatments and EBV reactivations, as compared with the initial NA cells [average chromosome number per cell increased from 88.2 of NA(P1) to 91.85 of NA(R15), as calculated from metaphase plate]. This suggests that recurrent EBV reactivations may be able to induce aneuploidy by increasing chromosome numbers.

Array-based comparative genomic hybridization (array CGH) is an established high resolution method to study amplification or deletion of chromosomal segments, i.e., copy-number aberrations (CNAs). To assess the impact of repeated EBV reactivations on chromosome integrity, array CGH was applied for NA(P15), NA(R1), NA(R15) and NA(P1) to examine the effect of the iterative EBV reactivations on the host DNA copy-number. We observed significant differences in the overall level of CNAs with respect to the frequency of TPA/SB-induced EBV reactivation. When compared with the NA(P1) common reference genome, the regions of CNAs in NA(R15) increased dramatically (69 amplified, 6 deleted), whereas those detected in NA(R1) were limited (4 amplified, 5 deleted) and those in NA(P15) were relatively few (8 amplified, 28 deleted) (Supp. Info. Table I). As shown in Figure 4, the alteration patterns of NA(R15) had a tendency of gain rather than loss and most of the altered loci were restricted on chromosomes 3, 7, 8, 9 and X. These results indicate large scale of CNAs occurred under the conditions that highly recurrent EBV reactivations took place, but single accidence of EBV reactivation and long-term culture did not cause severe CNAs. Another set of array CGH analysis was also performed with TPA/SB-treated TW01 cell, relatively less CNAs were observed in TW01(R15) (data not shown), supporting the influence of recurrent EBV reactivations on alterations of DNA copy-number. All the array CGH results were verified by dye-swap experiments and the minimum information about a microarray experiment (MIAME) is provided at the GEO site: The raw data are available for FTP download by GEO accession number GSE6472 at the FTP site:

Figure 4.

Highly recurrent EBV reactivations result in genomic copy-number alterations (CNAs) in NA cells. Array CGH analysis of TPA/SB-treated or mock-treated NA cells. Genome obtained from NA (P1) was used as common reference. Representative CNAs in NA(P15), NA(R1) and NA(R15) cells are shown for chromosomes 3, 7, 8, 9 and X as the probability heat-map on the right. Chromosomal regions of loss (green), gain (red), and no change (white) are color-coded using aberration definition as described in Material and methods. Graphics were produced by software CGH Analytics version 3.4.

These results, coupled with the observed increased MN formation of NA(R15) (Fig. 3) and aberrant chromosome structures (Table I), suggested that recurrent EBV reactivations in NA cells can be a critical factor that lead to aggravating genome instability.

Recurrent EBV reactivations aggravate malignant phenotypes of NPC cells

By database searching on the Internet at the Babelomics (, we found that hot regions revealed by array CGH in NA(R15) contained several genes involved in processes responsible for genome instability including chromosome segregation, nuclear division, as well as mitotic cell cycle (Supp. Info. Table II). This result is consistent with increasing chromosomal aberrations and MN formation in NA(R15). Furthermore, there are also numerous cancer-related genes mapped to the alteration regions (Supp. Info. Table III). Based on these results, we proposed that some malignant phenotypes may be aggravated in NA(R15) cells. To examine whether the recurrent TPA/SB-induced EBV reactivations enhance cancerous properties of NA cells, invasiveness assay in vitro and tumorigenicity assay in vivo were performed. As shown in Figure 5, we observed that there is no obvious correlation between invasiveness and passage numbers of TW01 cells, regardless mock or TPA/SB-treated, indicating long-term in vitro cultivation and recurrent TPA/SB treatments have no significant effect on the invasiveness of TW01. For EBV-infected NA cells, we observed a steady increase in invasiveness with respect to passage number under both mock and TPA/SB-treated conditions. These results suggested that EBV infection may enhance the invasiveness of NA cell, no matter in viral latent or lytic state. Furthermore, the results also showed the numbers of invading cells in NA(P15) and NA(R15) multiple to approximately 2-fold and 4-fold as compared with NA(P1) and NA(R1), respectively. These results indicate NA cells with highly recurrent EBV reactivations are more invasive than those with EBV latent infection, emphasizing the importance of repeated EBV reactivations on the invasiveness of NPC cells.

Figure 5.

Recurrent EBV reactivations promote the invasiveness of NA cells. EBV reactivation increases invasiveness, dependent on the number of rounds of reactivation. Mock or TPA/SB-treated TW01 and NA cells were assayed for their ability to invade a Matrigel-coated membrane. Data indicate the average number of invaded cells in 3 independent experiments (mean ± SD). *p < 0.01, compared with the other untreated and repeatedly TPA/SB-treated NA cells.

To evaluate the effect of recurrent TPA/SB-induced EBV reactivations on tumor growth, a tumorigenicity assay was performed using SCID mice injected with various cells. A dramatically increased tumor sizes were observed in mice inoculated with NA(R15), as compared with tumors obtained from mice inoculated with TW01(R1), TW01(R15), NA(P1) and NA(P15) and NA(R1) (Fig. 6a). The tumor nodules taken from the mice at day 40 also showed that NA(R15) cells were found to have more accelerating tumorigenicity (Fig. 6b). These results demonstrated that NA cells with highly recurrent EBV reactivations acquired enhancing malignancy.

Figure 6.

Recurrent EBV reactivations increase the tumorigenicity of NA cells in vivo. (a) The tumorigenicity of cells after recurrent EBV reactivations was evaluated in vivo by injecting cells subcutaneously into SCID mice. The size of the tumors at the inoculation site was measured using a standard procedure at 7-day intervals postinjection. Data indicate mean tumor size (n = 6) ± SD. *p < 0.001, compared with the NA(R15), NA(P1) and NA(R1) cells. (b) Representative photographs of tumor nodules taken from the mice at day 40.

These results suggest that the recurrent EBV reactivations, but not a single reactivation event, may play an essential role in the tumor progression of NPC.


Reactivation of EBV was proposed to contribute to the development of NPC.14 It was demonstrated that combined treatment of P3HR-1 and Raji cells with n-butyrate and croton oil (which is a source of phorbol) increased the lytic cycle of EBV in these cells dramatically. However, since both P3HR-1 and Raji cells are lymphocytes origin, the results may not be applied to epithelial cells. Because of the difficulty in establishing a latently-infected epithelial system, the question whether EBV reactivation induce genome instability in human epithelial cells is difficult to be explored at that moment. In this study, based on the pair-wise EBV-positive and EBV-negative NPC cell in vitro system, we demonstrate that recurrent TPA/SB-induced EBV reactivations have profound influences on genome instability, chromosomal aberrations and malignant phenotypes of NPC cells. The choice of TPA and SB for EBV reactivation is a reflection of the previous study indicating that people living in regions with high annual incidence of NPC were colocalized with the consumption of the herbal drugs containing phorbol esters.14 Although the TPA itself is a potent carcinogen, however, it is considered as nongenotoxic.48 Butyrate, a histone deacetylase inhibitor, is known to inhibit proliferation, induce apoptosis and differentiation. Recent studies have indicated that butyrates can lead to an enhanced chemoprotection and reduce the impact of endogenous genotoxicity of colon cancer cells.49, 50 In this study, their effects on the EBV-negative TW01 and HONE-1 cells were moderate during the 15 recurrent cycles (Fig. 3). Latent EBV infection may also contribute to certain extent of genome instability; however, recurrent EBV reactivations induced by TPA/SB promote the most noteworthy genome instability in NPC cells (Fig. 3). Furthermore, Matrigel invasiveness and SCID mice tumorigenesis assays have demonstrated the effect of recurrent EBV reactivations on enhancing tumorigenesis of NPC cells (Figs. 5 and 6). Thus, we conclude that recurrent chemical reactivations of EBV enhance genome instability and promote the tumorigenicity of NPC cells. These data, combined with the observations that EBV reactivation take place prior to relapse and metastasis,10, 11, 23 may indicate that recurrent EBV reactivations can enhance the tumor progression in NPC patients.

A clue supporting the association between recurrent EBV reactivations and NPC comes from cytogenetic studies. Array CGH analysis performed in our laboratory with NA(R1), NA(R5), NA(R10) and NA(R15) show the regional copy-number alterations (CNAs) were nonrandom, specifically involving gains at chromosome 3, 8 and 9 (Fig. 4).41 Moreover, these chromosome-specific CNAs are progressively increased in direct proportion to the frequency of TPA/SB-induced EBV reactivation. Whereas the CNAs detected in TW01 cells with mock or TPA/SB treatment, and in the mock-treated NA counterparts were all randomly distributed and independent of the times of TPA/SB treatment. It is noteworthy that several previous karyotyping analysis and CGH studies using NPC neoplasias as sample sources identified extensive nonrandom numerical abnormalities with respect to gains of 3p, 3q, 8p and 8q.51–53 Furthermore, in a recent meta-analysis of all available CGH data for a total of 188 primary NPCs, Li et al.,51 reconfirmed these chromosome-specific aberrations. Interestingly, they also investigated the chromosomal abnormalities of 30 NPC tumors with evidence of EBV. The results were largely similar to those observed in whole population analysis (i.e., the 188 primary NPCs) and showed that genetic imbalances involving chromosomes 3 and 8 are strongly specific for this malignancy.51 These consistent aberrations observed between our in vitro system receiving recurrent TPA/SB-induced EBV reactivations and NPC biopsies, along with the observations that EBV reactivation takes place in NPC patients,11–13 suggest that the causative repeated viral reactivations may occur in vivo, and thus contribute to the chromosomal aberrations of NPC tumors.

Persistent infection is a common phenomenon in virus-associated human malignancies, such as human T-cell leukemia virus Type 1 (HTLV-1) and adult T-cell leukemia/lymphoma,54 human papilloma virus (HPV) and cervical cancer55, 56 and Kaposi's sarcoma-associated herpesvirus (KSHV) and Kaposi's sarcoma.57 For each virus, a very long period of persistent latent infection may be accompanied by occasional reactivations of the virus.57, 58 It is accepted widely that activation of oncogenes, repression of tumor suppressor genes and insertional mutagenesis may contribute to viral carcinogenesis.59, 60 During latency, the HTLV-1 Tax, HPV E6 and E7 and KSHV LANA have been demonstrated to contribute to the induction of genomic instability.61–63 After the discovery of Epstein-Barr virus for 40 years, it was proposed that EBV latent infection may participate in the carcinogenesis of NPC.8 It was also demonstrated that latently infected EBV may promote the growth of NPC.64 We have demonstrated previously that EBV LMP1 is able to enhance MN formation through inhibition of damaged DNA repair in human epithelial cells by host cell reactivation assay.33, 65 The host cell reactivation assay has been used to detect nucleotide-excision repair-associated instability (NIN) which reveals subtle changes in DNA sequence.66 This may also contribute to the mild increase in chromosomal aberrations and invasiveness of NA cells after 15 time passages without TPA/SB treatment (Table I and Fig. 5). In this study, we demonstrate by γH2AX assay that chromosomal instability (CIN) is associated closely with the induction of DSB (Figs. 1a and 1b). We expressed EBV LMP1 in human epithelial cells to see whether this protein would induce γH2AX. However, no remarkable changes were revealed (unpublished observations). The result indicates that LMP1 may exert its effects predominantly in the latently infected NPC cells. Recently, it was demonstrated that LMP1 can disrupts microtubule structures and induces chromosomal aberrations in epithelial cells.67 The result obtained in our study showing that the elevation of chromosomal aberrations in the EBV-positive cells when compared with the EBV-negative counterparts and increasing aberrations during the long-term passage of the mock-treated EBV-positive cell (Table I) may support this notion.

Because, in this study, EBV reactivation is markedly associated with genome instability, we examined further the effects of several EBV early genes on γH2AX and MN formation in TW01 cells. A striking effect was found to be contributed by the EBV DNase, an early lytic gene in EBV replication (Fig. 2). The effect of DNase to host cell DNA damage was in a dose-dependent manner and not apparent in the DNase-activity-null mutant (unpublished observations). We also found previously that EBV DNase has a Type II restriction enzyme-like catalytic center.68 It was reported that the expression of restriction endonucleases is able to lead to cell killing, chromosomal aberration, oncogenic transformation and genetic mutation.69–71 The expression of EBV DNase has been detected in NPC tissues.72 These studies suggest that the nuclease activity of EBV DNase might damage the host DNA and lead to the DNA DSB and consequent genome instability. We are currently carrying out more extensive studies for EBV DNase to elucidate its contribution on the genome instability and carcinogenesis of NPC.

Studies in another herpesvirus, the herpes simplex virus Type 1 (HSV-1), have provided evidence that viral genes may induce cytogenetic changes to the host cells. Infection of human cells by HSV-1 resulted in chromosomal breaks, gaps and pulverization as well as cytogenetic changes and DNA strand breaks.47, 73 Temperature-sensitive mutants were employed to elucidate which genes were responsible for these effects. The results showed that the HSV-1 transcriptional regulatory protein ICP 4, DNA polymerase and major DNA-binding protein ICP 8, as well as HSV-2 alkaline nuclease, might contribute to the chromosomal damage.47 In this study, the EBV DNase, UDG and mDBP were found to induce MN formation or DNA DSB in transfected cells (Figs. 2a and 2b). In addition, we have found that EBV BHRF1, a homologue of the BCL-2 proto-oncogene, can increase MN formation when expressed in epithelial cells (unpublished observations). We have also demonstrated previously that EBV BGLF4 induced premature chromosome condensation,74 which may also contribute to genome instability of the host cell. These studies indicate that, in addition to the well-accepted concept of latent EBV may contribute to carcinogenesis of NPC, expression of lytic genes during EBV reactivation can induce DNA damage and enhance genome instability and subsequently contribute to the carcinogeneisis of NPC cells. However, more studies are required to elucidate how many EBV lytic genes are involved and the underline mechanisms in this viral contribution to host cell genome instability.


We are indebted to Dr. Tim J Harrison at the Royal Free and University College Medical School, University College London for critical reading of the manuscript. We are grateful to Dr. Allan Hildesheim at Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda for comments and discussions. We thank Drs. Ko-Jiunn Liu and Shine-Gwo Shiah of National Institute of Cancer Research, National Health Research Institutes, Taiwan, for much assistance in carrying out experiments.