Epstein-Barr virus-encoded EBNA1 inhibits the canonical NF-κB pathway in carcinoma cells by inhibiting IKK phosphorylation

Ad/AH (a human adenocarcinoma cell line derived from the nasopharynx), Hone1 (an EBV-negative NPC cell line), AGS (a human gastric-carcinoma derived cell line) and derivatives stably expressing EBNA1 at levels comparable to those found in EBV infection, Ad/AH cells stably infected with a recombinant EBV and C666-1 (an EBV-positive cell line derived from an undifferentiated EBV-positive NPC) were cultured as previously described [13, 14]. Neither Ad/AH cells stably infected with EBV or the C666-1 cell line express the EBV-encoded latent membrane protein 1 (LMP1). TNFα and IL-1β (Peprotech, London, UK) were reconstituted in serum-free growth medium to a concentration of 100 ng/ml, and stimulations carried out for 1 hour prior to harvesting.

Dual luciferase reporter assays were performed according to the manufacturer's instructions (Promega, Southampton, UK, cat. no. E1980) with cells cultured as previously described [13]. Cells were transfected with the following plasmids using Lipofectamine (Invitrogen, Renfrew, UK) following the manufacturer's instructions: pSG5-EBNA1 [20], pGL3-basic (Promega), 3 enhancer-ConA (an NF-κB-dependent luciferase reporter construct in which transcription of the firefly luciferase gene is driven by three NF-κB binding sites) [21], dnEBNA1 (M15 EBNA1 dominant-negative mutant [22]), and a control Renilla luciferase plasmid (pRL-TK; Promega). All assays were carried out in biological and technical triplicate and are represented as the mean of three independent experiments.

DNA encoding wild-type EBNA1 was excised from pSG5-EBNA1 [20] using EcoRI and BsaHI and end-filled using Klenow DNA polymerase. Adenine nucleotides were added using Taq DNA polymerase and the resulting DNA was ligated into pCR8 (Invitrogen, cat. no. K2500-20), following the manufacturer's instructions. DNA encoding the dGA, dnEBNA1 (d395-450), d8-67, d41-376, d61-83 and d325-376 EBNA1 mutants [23] was amplified by PCR using the following primers specific to the flanking vector sequences; 5'GCCGGATCCCCCACTGCTTACTGGCTTAT-3' and 5'-GCCGTCGACGGCAAACAACAGATGGCTGGCAA-3'. These were inserted into pCR8 following the manufacturer's instructions. An empty vector control (Vector) was generated by self ligation of pCR8. The vectors derived above were recombined with pLenti6/R4R2/V5-DEST (Invitrogen, cat. no. K591-10 and K5910-00) and pENTR5 containing the human metallothionein II promoter, following the manufacturer's instructions. In the transient transfection experiments presented here zinc stimulation was not required as the metallothionein II promoter was found to exhibit inherent leakiness (data not shown).

Nuclear extracts were prepared according to the manufacturer's instructions (Pierce Biotechnology, Illionois, USA, cat. no. 78833) and EMSA analysis was carried out on 5 μg of nuclear protein according to the manufacturer's instructions (LI-COR Biosciences, Cambridge, UK, doc. 982-07487) using an NF-κB consensus probe (sense oligonucleotide 5'-AGTTGAGGGGACTTTCCCAGGC-3') which was either 5' IRDye700 labelled or unlabelled for cold competition. EMSA gels were analysed and images were captured and quantified using the LI-COR Odyssey infrared laser imaging system. EMSAs were repeated for three independent biological replicates.

Nuclear protein extracts were isolated according to manufacturer's instructions (Active Motif, Rixensart, Belgium, cat. no. 40010). The NF-κB subunits p65 and p50 present in active dimers were measured using the ELISA based TransAM NF-κB family kit (Active Motif, cat. no. 43296) according to the manufacturer's instruction. Data are presented relative to the supplied internal Raji cell lysate control and are represented as the mean of three independent experiments.

RNA was extracted using EZ-RNA total RNA isolation kit (Geneflow, Staffordshire, UK) and was reverse transcribed for RT-PCR with Superscript III (Invitrogen), following the manufacturer's instruction. RT-PCR was carried out using standard procedures with the primers listed in Table 1. Standard immunoblotting procedures [14] were used to detect proteins using the antibodies listed in Table 2. All assays were carried out in biological and technical triplicate and are represented as the mean of three independent experiments. Tissue arrays were constructed and assayed at the Department of Pathology, Cancer Centre, Sun Yat-Sen University (Guangzhou, Guangdong, China) as follows; formalin-fixed paraffin-embedded blocks were obtained from the archives of the Sun Yat-Sen University pathology department. The matching H&E-stained slides were reviewed and screened, and samples containing both NPC tumour and adjacent nasopharyngeal mucosae were chosen for tissue array construction. Each case was represented by a mean of 4 cores with 2 tumours and 2 normal nasopharyngeal mucosa using a 0.6 mm punch. Immunohistochemistry was performed using the agitated low temperature epitope retrieval (ALTER) method [24]. Immunohistochemical staining was carried out for p65 and sections were counterstained with haematoxylin.

Where appropriate, statistical significance was calculated by performing a Sudent's t-test having first determined equal or unequal variance by using an F-test.

To assess whether EBNA1 influenced NF-κB activity we initially performed luciferase reporter assays in a range of carcinoma cell lines using a synthetic NF-κB reporter and found that NF-κB activity in Ad/AH, AGS and Hone1 cells stably expressing levels of EBNA1 comparable to those found in EBV infection was inhibited by 8, 5 and 2.6 fold, respectively (Fig. 1). Transient expression of EBNA1 in Ad/AH cells achieved by transfection using a range of concentrations of EBNA1 plasmid DNA resulted in a dose-dependent decrease in NF-κB reporter activity with a 2-fold reduction seen at the highest concentration of input DNA (Fig. 2A). To assess whether inhibition of NF-κB required a fully functional EBNA1, a dominant-negative EBNA1 (dnEBNA1) was titrated against wild-type EBNA1 in Ad/AH cells. Increasing doses of dnEBNA1, which dimerises with wild-type EBNA1 impairing its function, resulted in almost complete abrogation of the ability of wild-type EBNA1 to inhibit NF-κB activity (Fig. 2B). In addition, the dnEBNA1 alone had no effect on NF-κB reporter activity (Fig. 2B). Next we performed electromobility shift assays (EMSA) using a consensus NF-κB probe to assess whether the reduction in reporter activity was due to a reduction in NF-κB DNA binding. EMSAs performed on Ad/AH cells stably expressing EBNA1 indicated a 2-fold basal reduction in band intensity indicating a reduction in nuclear protein binding to the NF-κB probe, when compared with the neo control cells, which was consistent with our observed reduction in reporter activity (Fig. 3). Furthermore, the ability of two different pro-inflammatory cytokines, TNFα and IL-1β, to enhance NF-κB DNA binding was ablated in those cells stably expressing EBNA1 (Fig. 3 and data not shown). Similar results were observed in Ad/AH cells stably infected with a recombinant EBV (rEBV) carrying the neomycin drug selectable marker (Fig. 4A). In addition, the ability of TNFα, a potent activator of the canonical NF-κB pathway, to enhance binding with the NF-κB probe in C666-1 cells was considerably lower than in Ad/AH parental cells (Fig. 4B). EMSAs performed on Ad/AH cells stimulated with TNFα incubated with both the labeled NF-κB probe and a 100-fold excess of unlabeled probe (cold competition) resulted in complete abrogation of probe binding (Fig. 3, middle panel), thus demonstrating the high degree of specificity in the EMSA assays.

Having demonstrated that general NF-κB activity and NF-κB DNA binding in EBNA1 expressing cells was reduced, we sought to determine the relative abundance of the canonical NF-κB subunits p65 and p50 in active nuclear complexes bound to target DNA. We chose to study p65 and p50 because heterodimers of these NF-κB subunits are the most abundant form of NF-κB and exhibit the most powerful transcriptional activation potential [25]. Furthermore, the canonical NF-κB pathway is most commonly associated with innate immunity in general and with cellular differentiation in epithelial cells, both of which impact upon the pathogenesis of NPC. TransAM analysis in Ad/AH cells demonstrated a 4-fold enrichment of p65-containing dimers in neo control cells stimulated with TNFα (relative to un-stimulated neo cells), whereas the abundance of p50 did not change significantly (Fig. 5). In contrast, there was a 2-fold reduction in the basal amount of p65-containing dimers in Ad/AH cells stably expressing EBNA1, when compared with the un-stimulated neo control, and these cells were refractory to stimulation with TNFα. However, the amount of p50-containing dimers in Ad/AH cells stably expressing EBNA1 both pre and post TNFα stimulation was not significantly different from the levels observed in the neo control cells. Results similar to those found in Ad/AH cells stably expressing EBNA1 were also observed in Ad/AH cells stably infected with rEBV and in C666-1 cells. Interestingly, the amount of p50-containing dimers was marginally increased in Ad/AH cells stably infected with rEBV and in the C666-1 cells and this was unaffected by stimulation with TNFα.

We next sought to determine whether the EBNA1-induced reduction in NF-κB activity and reduced levels of p65 in active NF-κB dimers was as a result of alterations in the expression or phosphorylation status of p65. In its transcriptionally inactive form unphosphorylated p65 NF-κB is retained in the cytoplasm via interactions with specific inhibitors, the IκBs. Upon stimulation of the canonical NF-κB pathway the inhibitory IκBα protein becomes phosphorylated, ubiquitinated and degraded by the 26S proteasome. NF-κB dimers subsequently translocate to the nucleus where phosphorylated p65-containing dimers can modulate the expression of target genes [26]. Specifically, phosphorylation of p65 at Ser 536 by IKKα and IKKβ has been implicated in p65 nuclear translocation and transcriptional activity [26]. Immunoblot analysis confirmed enhanced serine 536 phosphorylation of p65 following TNFα stimulation of Ad/AH and Hone1 neo control cells as expected (Fig. 6A and 6B, respectively). In contrast, there was almost complete inhibition of p65 phosphorylation in stable EBNA1-expressing Ad/AH and Hone1 cells basally and following TNFα stimulation. Furthermore, the reduction in p65 phosphorylation in EBNA1-expressing cells was not due to a reduction in total levels of p65 protein (Fig. 6A and 6B, respectively). EBNA1-expressing Hone1 cells were also refractory to stimulation with another potent activator of the canonical NF-κB pathway, IL-1β, exhibiting reduced p65 phosphorylation in EBNA1-expressing cells which was in contrast to enhanced levels of phospho-p65 observed in the neo control cells (Fig. 7A). Immunofluorescence staining in Ad/AH cells also revealed that the ability of TNFα to stimulate translocation of p65 from the cytoplasm to the nucleus, as seen in the neo control cells, was almost completely inhibited in cells stably expressing EBNA1 (Fig. 7B). These observations were, therefore, in agreement with the above reporter assays, EMSA and TransAM data.

The activation of the canonical NF-κB pathway is tightly regulated by signals such as pro-inflammatory cytokines, that stimulate the IκB kinase complex (IKK) to phosphorylate the inhibitory IκBs which marks them for ubiquitin-mediated degradation. This allows free NF-κB to translocate to the nucleus where it activates target genes. The IKK complex is composed of two catalytic subunits, IKKα and IKKβ, which contain N- terminal serine/threonine kinase domains and a regulatory subunit, IKKγ (NEMO), that does not exhibit kinase activity but is essential for IKK phosphorylation and activation of upstream kinases [27]. Whilst both IKKα and IKKβ cooperate for IκB phosphorylation, IKKβ is indispensable for signalling via the canonical NF-κB pathway [28].

Having demonstrated that stable EBNA1 expression results in reduced levels of phospho-p65 and its translocation to the nucleus we sought to determine whether EBNA1 achieved this by affecting the expression and/or phosphorylation status of the NF-κB inhibitory subunit IκBα. Immunoblotting demonstrated that stable EBNA1 expression in Ad/AH and Hone1 cells resulted in a decrease in phosphorylation of IκBα at ser32, relative to the neo control cells (Fig. 8A). In contrast, stable EBNA1 expression did not alter the expression of total IκBα in the Ad/AH cells. In the Hone1 cells stable EBNA1 expression resulted in a marginal reduction in the expression of total IκBα, which is itself an NF-κB regulated gene.

We then performed immunoblotting to determine whether EBNA1 achieved a reduction in IκBα and p65 phosphorylation by affecting components of the IKK complex. There was a marked reduction in the levels of serine phosphorylation within the activation loops of both IKKα/β, a prerequisite for activation, in Ad/AH and Hone1 cells stably expressing EBNA1, when compared with the neo control cell lines (Fig. 8B). In contrast, there was no appreciable difference in the total levels of IKKα, IKKβ or IKKγ (Fig. 8B) which demonstrated that the reduction in phospho-IKKα/β in EBNA1 expressing Ad/AH and Hone1 cells was not due to a reduction in the expression of these catalytic subunits.

Having demonstrated that EBNA1 inhibited the phosphorylation of IKKα/β we asked whether specific domains of EBNA1 were responsible for this phenomenon. Therefore, we cloned wild-type EBNA1 and a selection of EBNA1 domain mutants [23] into a lentiviral expression vector. Amounts of DNA found to yield equal levels of protein expression (data not shown) were transiently transfected into Ad/AH cells and NF-κB luciferase reporter assays were performed (Fig. 9). Transfection of wild-type EBNA1 (wtEBNA1) and EBNA1 lacking the gly-ala repeat region (dGA), from which all subsequent mutants were derived [23], resulted in a significant reduction in NF-κB activity, relative to the empty vector control (Vector), in agreement with our data presented in Fig. 2A. Transfection of a dominant-negative EBNA1 (dnEBNA1) carrying the same DNA sequence as that used in Fig. 2B did not result in a significant reduction in NF-κB reporter activity. Transfection of EBNA1 mutants d8-67, d41-376, d61-83 and d325-376 did not result in a significant reduction in NF-κB reporter activity. Thus deletion of domains of EBNA1 essential to its ability to transactivate viral gene expression ablated its ability to inhibit NF-κB activity. In contrast, transfection of an EBNA1 mutant deleted for the binding site of the deubiquitinylating enzyme USP7 [4] (d395-450) resulted in significant inhibition of NF-κB reporter activity. Therefore deletion of the USP7 binding domain had not altered the ability of EBNA1 to inhibit NF-κB activity. These data therefore suggested that the mechanism by which EBNA1 was able to inhibit phosphorylation of IKKα/β was likely to be through EBNA1s ability to modulate the expression of cellular genes. Therefore we sought to determine whether EBNA1 influenced the expression of a selection of genes reported to regulate IKK activity and the canonical NF-κB pathway in general.

As IL-1 and TNF activate the canonical NF-κB pathway by binding with their cognate receptors we next examined whether the EBNA1-induced reduction in IKKα/β phosphorylation was as a consequence of EBNA1 modulating the expression of IL-1 and/or TNF receptors. RT-PCR analysis revealed that EBNA1 did not alter the natural levels of expression of the IL-1 receptors 1 and 2 (IL1R1 and IL1R2) or TNF receptors 1 or 2 (TNFR1 and TNFR2) in Ad/AH and Hone1 cells (Fig. 10A). Therefore it is unlikely that EBNA1 modulates NF-κB by influencing the ability of cells to respond to these pro-inflammatory ligands.

We next sought to determine whether EBNA1 inhibited NF-κB activity by down-regulating the expression of IL-1 and TNFα. RT-PCR demonstrated that whilst TNFα and IL-1β expression could not be detected we surprisingly found IL-1α to be up-regulated by EBNA1 in both Ad/AH and Hone1 cells (Fig. 10B and microarray data not shown). This suggested that reduced stimulation of the NF-κB pathway is not responsible for the low levels of NF-κB activity observed in EBNA1 expressing epithelial cells. Intriguingly, however, elevated IL-1α is a marked feature of NPC suggesting that EBNA1 may contribute to this phenotype [29, 30].

The transcription factor C/EBPβ has been implicated in NF-κB inhibition by preventing p65 phosphorylation [31]. RT-PCR revealed that EBNA1 expression in the nasopharyngeal cell lines was associated with C/EBPβ down-regulation and was therefore unlikely to be involved in the mechanism by which EBNA1 inhibited NF-κB activity in epithelial cells (Fig. 10B).

The microarray data used for the in silico promoter analysis that implicated EBNA1 as having a role in NF-κB modulation reported that expression of A20, which inhibits NF-κB machinery up-stream of the IKK complex, was up-regulated 2.5-fold. RT-PCR analysis, however, indicated that A20 was down-regulated in Ad/AH and Hone1 cells stably expressing EBNA1 and was therefore unlikely to be involved in the mechanism by which EBNA1 inhibited NF-κB activity in epithelial cells (Fig. 10B)

Previous reports have demonstrated in EBV-positive NPC xenografts and EBV-positive NPC biopsies that p65 is localised to the cytoplasm in tumour cells [32, 33]. Having demonstrated that there was a reduction in active nuclear p65 in carcinoma cells stably expressing EBNA1, infected with rEBV and in C666-1 cells we sought to confirm the p65 status in NPC biopsies. Immunohistochemical staining for p65 was carried out on tissue arrays containing 11 NPC biopsies with matched normal nasopharyngeal control tissue. This demonstrated that the level of p65 staining was elevated in the tumour cells of 6 out of 11 NPC biopsies, relative to the matched normal tissue sections, with the p65 staining in the remaining 5 tumour samples being indistinguishable from the controls. However, the p65 staining in all cases was exclusively cytoplasmic with no detectable nuclear staining despite the pro-inflammatory environment characteristic of NPC. The p65 staining in the cellular infiltrate of all NPC biopsies was negative. Two examples are presented in Fig. 11.

NF-κB signalling regulates a variety of major cellular processes, including cell growth, differentiation and apoptosis, and it is therefore not surprising that aberrant NF-κB signalling has been implicated and documented in the pathogenesis of a wide range of cancers. In addition, NF-κB signalling impacts upon both adaptive and innate immunity, the latter being crucial to the ability of host cells to mount effective defences against oncogenic viruses such as EBV [34]. In this study we have demonstrated for the first time that NF-κB activity is repressed by both transient and stable expression of EBNA1 in a number of carcinoma cell lines; confirming that the phenomenon is not merely cell line specific or due to clonal variation. This study therefore reveals that, like its homologues from KSHV and MuHV-4 (LANA and ORF73, respectively) [18, 19], EBNA1 also has a role in inhibiting the NF-κB pathway and that this is therefore most likely a conserved function amongst these gammaherpesvirus nuclear proteins.

Whilst the mechanism for ORF73-mediated NF-κB inhibition has been determined to be via poly-ubiquitination and subsequent proteasomal-dependent nuclear degradation of p65, which is dependent upon a SOCs box motif present in ORF73, the mechanism for LANA-mediated NF-κB repression remains unknown, although it may be similar as LANA also contains a SOCs box [19, 35]. However, examination of EBNA1 reveals no obvious SOCs motif. In contrast, the findings presented here indicate that EBNA1 inhibits the canonical NF-κB pathway in carcinoma lines by inhibiting phosphorylation of the IKK complex upon which several pro-inflammatory signalling cascades converge, enabling EBNA1 to block NF-κB activation in response to a broad range of stimuli.

Whilst we have not fully elucidated the mechanism by which EBNA1 inhibits IKK phosphorylation our data do indicate that deletion of domains of EBNA1 reported to be essential to its ability to transactivate EBV encoded genes [23] abrogated the ability of EBNA1 to inhibit NF-κB activity, whereas deletion of the domain of EBNA1 known to bind with the cellular deubiquitinylating enzyme USP7 [4] had no effect. These data therefore suggest that the ability of EBNA1 to inhibit IKK phosphorylation is most likely via EBNA1 regulating the expression of a cellular gene(s) involved in this process. Subsequent RT-PCR analysis determined that a reduction in IKKα/β phosphorylation was not likely to be as a result of inhibition of the expression of IL-1 or TNFα receptors or their ligands, or by up-regulation of the NF-κB inhibitors C/EBPβ or A20. Interestingly, RT-PCR analysis indicated that A20 was down regulated in Ad/AH and Hone1 cells stably expressing physiological levels of EBNA1, adding credence to our observations that EBNA1 inhibits NF-κB activity as A20 expression is positively regulated by NF-κB [36]. In addition, preliminary observations suggest that EBNA1 does not bind with or relocalise IKKβ to the nucleus and so it is unlikely that this is the mechanism by which nuclear EBNA1 inhibits the phosphorylation and activity of the IKK complex (Figure S1, Additional file 1). The precise mechanism by which EBNA1 inhibits IKKα/β phosphorylation and NF-κB activity is therefore currently under investigation.

Chronic activation of NF-κB is associated with the development of a number of malignancies. Therefore on face value our observation that EBNA1 inhibits canonical NF-κB would appear counter intuitive with regard to the pathogenesis of NPC. However, within the context of epithelial cells it has been reported that NF-κB activation is growth inhibitory. For example, Seitz et al. [37] found that the expression of constitutively active p50 and p65 canonical NF-κB subunits in normal epithelial cells resulted in irreversible cell cycle arrest, whereas Gapuzan et al. [38] reported that p65 knockout fibroblasts have a transformed phenotype. In addition, expression of a dominant-negative IκBα "super repressor" in murine and human epidermis led to hyperplasia and the development of squamous cell carcinoma (SCC) [39]. Furthermore, Dajee et al. [40] showed in SCC biopsies that p65 staining was predominantly cytosolic.

The ability of EBV-encoded LMP1 to activate both the canonical and non-canonical NF-κB pathway has been the subject of many, mostly in vitro, studies. However it is becoming increasingly evident that the ability of LMP1 to activate the NF-κB cascade in vivo where other EBV latent genes, including EBNA1, are expressed is not so well defined. Recent studies have gone some way to addressing this question by demonstrating in EBV-positive NPC biopsies and xenografts that p65 is located almost exclusively in the cytoplasm and that this is independent of LMP1 expression [32, 33]. In agreement with and expanding upon these data, we have demonstrated that whilst the level of p65 was elevated in 6 out of 11 NPC tumours examined, relative to matched normal control tissue from the same patient, in all cases p65 staining was cytoplasmic. Thornburg et al. [33] propose that inhibition of p65 in NPC may protect against growth arrest whilst p50/p50 and p50/BCL3 NF-κB could still maintain the tumourigenic effects of NF-κB. It is therefore interesting to speculate that EBNA1 may modulate the ability of LMP1 to activate specific aspects of NF-κB signaling and that this in turn may impact upon the pathogenesis of NPC and other EBV-related tumours. This clearly warrants further investigation.