Epstein-Barr virus-encoded LMP1 induces a hyperproliferative and inflammatory gene expression programme in cultured keratinocytes

Epstein–Barr virus (EBV) is a human gammaherpesvirus that is associated with a number of human malignancies of lymphoid and epithelial cell origin. Of the latter, nasopharyngeal carcinoma (NPC) is numerically of greatest importance. Although early-stage tumours respond well to chemotherapy (Qin et al., 1988), the aggressively metastatic character of NPC results in death within 1 year for 85 % of patients having metastatic disease (Teo et al., 1996). The tumour has a multifactorial aetiology involving virological (EBV), genetic and environmental components (Tao & Chan, 2007). A proportion of NPCs express the EBV-encoded latent membrane protein 1 (LMP1), which is considered to be the major cellular-transforming gene of the virus. LMP1 acts as a classical oncogene in rodent fibroblast transformation assays, is essential for EBV-induced B-cell transformation, induces the expression of cell survival genes, cytokines and oncogenes and, with potential significance for the development of NPC, can transform epithelial cells, increase their motility, block terminal differentiation (Dawson et al., 1990, 2000, 2003 and references therein) and downregulate the expression of adhesion molecules such as E-cadherin (Fåhraeus et al., 1992).

Several of the above LMP1-induced effects have been observed previously in SCC12F epithelial cells stably transfected with the LMP1 gene of the B95-8 strain of EBV (Dawson et al., 2000). The SCC12F cell line has proved to be a useful model system to study the impact of EBV gene expression on keratinocyte behaviour, given its unique ability to respond to signals that regulate normal growth and differentiation. These qualities provide significant benefits over the use of fully tumorigenic keratinocyte lines which, by definition, have acquired multiple defects in signalling pathways that regulate these processes.

Here, we sought to define the effect of B95-8 LMP1 expression on SCC12F cellular gene expression using a tetracycline-inducible system. The use of an inducible system overcomes the problems associated with clonal variation in cell lines selected for stable expression of LMP1 and limits artefacts that may arise after long-term expression of the protein.

Rather than using microarrays as a stepping stone towards a detailed, mechanistic examination of just one or a few genes, we have taken advantage of the global nature of array analysis to paint a holistic molecular portrait of cellular pathways and processes that are modulated by LMP1 expression in epithelial cells. Thus we have identified and validated LMP1-induced changes in more than 60 genes that impinge on the processes of cell growth and differentiation, motility, the cell cycle, inflammation and vesicular transport. Many of these gene expression changes are characteristic of inflamed, hyperproliferative epidermis and have been identified as targets of the pro-inflammatory cytokines IL-1α and IL-1β. Our findings indicate a multiplicity of LMP1-induced cellular functions that may play a role in the progression of keratinocytes towards the malignant state through chronic stimulation of NF-κB and AP-1 and the induction of pro-inflammatory cytokines such as IL-1α and IL-1β.

Cell culture and isolation of SCC12F cells containing an inducible LMP1 gene.
The human keratinocyte cell line SCC12F (Rheinwald & Beckett, 1981) was cultured as described previously (Dawson et al., 2000). Stable derivatives (SCC12F-tet-LMP1) that express the LMP1 gene under the control of a tetracycline-inducible promoter were generated as described previously (Dawson et al., 2000; Floettmann et al., 1996). The EBV-positive NPC cell line C666-1 (Cheung et al., 1999) was cultured as described by Stewart et al. (2004). Protein lysates from the LMP1-positive NPC xenograft C15 (Busson et al., 1988) were prepared from snap-frozen material (Young et al., 1988).

Antibodies.
Antibodies used in this study are listed in Supplementary Table S1, available in JGV Online.

Immunofluorescence staining, immunoblotting and flow cytometry.
Indirect immunofluorescence staining, immunoblotting and flow cytometry were performed as described previously (Dawson et al., 1990, 2000, 2003).

Preparation of total RNA for expression array analysis.
Three replicate pairs of uninduced and induced (72 h) SCC12F-tet-LMP1 cultures derived from a single clone were grown on separate occasions. Total RNA was prepared from each culture using Trizol reagent (Invitrogen) followed by a clean-up step using RNeasy columns (Qiagen) according to the manufacturers' instructions. RNA concentration was estimated using a Nanodrop ND-1000 spectrophotometer and RNA quality was assessed by agarose gel electrophoresis. These same three paired replicate RNA preparations were used in all array and quantitative PCR (Q-PCR) experiments. For the amplification studies, RNA was diluted to the required concentration using RNase-free water.

The amplification and labelling protocols are depicted schematically in Supplementary Fig. S1.

One microgram protocol.
Briefly, 1 µg total RNA was used as the template for synthesis of cDNA using a T7 RNA polymerase promoter/(dT)₂₄ oligonucleotide (Proligo) as primer. DNA polymerase I was employed to generate the second strand, yielding double-stranded cDNA containing the T7 promoter at one end. This served as template for in vitro transcription (IVT) using the Affymetrix IVT labelling kit to produce biotinylated antisense RNA (Affymetrix, 2004).

Ten microgram protocol.
This is an earlier Affymetrix labelling protocol. It uses 10 µg total RNA and the Enzo High Yield RNA synthesis labelling kit in place of the Affymetrix IVT kit.

ExpressArt RNA amplification and labelling.
RNA amplification starting from 5 ng total RNA was performed using the ExpressArt kit (AMS Biotechnology) according to the manufacturer's instructions (Supplementary Methods). Note that the final IVT step is the same as that used in the other protocols (Supplementary Fig. S1). In the amplification experiments reported here, as in the 10 µg protocol, the Enzo labelling kit was used.

Hybridization to Affymetrix arrays.
Affymetrix Human Genome Focus arrays were used. Biotinylated RNA was fragmented according to the Affymetrix protocol (Affymetrix, 2004). Hybridization of samples labelled using the Affymetrix IVT kit was performed according to the current protocol, whereas samples labelled with the Enzo kit (10 µg standard and ExpressArt-amplified) were hybridized using an earlier version of the procedure in which DMSO is absent from the hybridization cocktail. Arrays were washed and stained on an Affymetrix FS400 fluidics station and then scanned using an Agilent G2500A GeneArray scanner according to Affymetrix procedures. GCOS software (Affymetrix) was used for instrument control and data acquisition.

Array data analysis.
Data were normalized and processed to give expression values using the RMA protocol (Irizarry et al., 2003). Statistical analysis for significance and fold change was performed by the SAM method (Tusher et al., 2001), the PM-only model of dChip (Li & Wong, 2001) and by rank products (RP) analysis (Breitling et al., 2004). These three methods each use different statistical algorithms for their analysis of significance and consequently yield somewhat different results from the same input data (for discussion see Irizarry et al., 2003; Breitling et al., 2004).

Quantification of LMP1 gene transcription.
LMP1-specific mRNA was measured by Q-PCR (Bell et al., 2006).

Q-PCR measurement of cellular gene expression.
TaqMan gene expression assays (Supplementary Table S2) were selected from the Applied Biosystems website () together with 18S rRNA as a standard control, assembled onto a microfluidics card (Applied Biosystems) and analysed using an ABI Prism 7900HT Sequence Detection System. Individual 50 µl Q-PCRs were performed for S100A2, S100A7, IL1R1, SerpinA3, ATF3 and BIRC3 (see Supplementary Methods).

The relative quantity (RQ) of RNA for each gene in LMP1-positive cells versus LMP1-negative cells was calculated using the method (Livak & Schmittgen, 2001). The mean of the RQs from the three replicate pairs was used in further analysis.

Electrophoretic mobility shift assays (EMSAs).
EMSAs for AP-1- and NF-κB-binding sites were performed as described previously (Eliopoulos et al., 1997; Stewart et al., 2004).

RT-PCR.
Amplification by RT-PCR was performed as described by Eliopoulos et al. (1997). Gene-specific primers are shown in Supplementary Table S3.

Cytokine array and ELISA.
See Supplementary Methods.

Validation of the inducibility of SCC12F-tet-LMP1 cells
SCC12F-tet-LMP1 cells were grown in the presence or absence of doxycycline for 72 h, after which cells were examined for LMP1 protein expression by immunofluorescence (Fig. 1a) and immunoblotting (Fig. 1h). In each case, a robust induction of LMP1 protein expression was observed after removal of doxycycline. In keeping with our previous observations (Dawson et al., 1990), the induction of LMP1 was accompanied by a striking change in the morphology of SCC12F cells (Fig. 1b); cells lost their cuboidal appearance and took on a more fusiform morphology accompanied by an alteration in differentiated characteristics. In the absence of LMP1 induction, SCC12F-tet-LMP1 cells maintained the same morphology as the parental cells (Fig. 1b) and, like the parental cells (Dawson et al., 2000), expressed high levels of involucrin (IVL), a cross-linked envelope protein, and the calcium-dependent cell-adhesion protein E-cadherin (CDH1). However, expression of both of these differentiation-associated antigens was significantly reduced in LMP1-expressing cells (Fig. 1c, d). That these effects were induced as a consequence of LMP1 signalling is consistent with the observation that LMP1 expression was accompanied by increased amounts of nuclear NF-κB and AP-1, as assayed by binding nuclear extracts containing these transcription factors to radiolabelled oligonucleotide probes containing target sequences (Fig. 1e) and, in agreement with previous findings (Dawson et al., 1990, 2000), by the induction of the NF-κB target genes CD40 and ICAM1 (Fig. 1f).

(61K): Fig. 1. Characteristics of SCC12F-tet-LMP1 cells before and after LMP1 induction. SCC12F-tet-LMP1 cells were cultured in the presence or absence of 5 µg doxycycline (Dox) ml–1 for 72 h. Parental SCC12F cells were included as a negative control. (a) Indirect immunofluorescence staining for LMP1. Bars, 20 µm. (b) Phase-contrast micrographs demonstrating the morphological alteration induced by LMP1 in SCC12F cells. Bars, 20 µm. (c) Reduction of E-cadherin expression in LMP1-expressing cells. Bars, 5 µm. (d) Reduction of IVL expression in LMP1-expressing cells. Bars, 20 µm. (e) EMSA demonstrates that LMP1 expression results in increased NF-κB and AP-1 activity in nuclear extracts isolated from induced cells. Uninduced cell lines show constitutively low basal levels of NF-κB and AP-1. (f) FACS analysis demonstrating increased expression of cell surface CD40 (left panel) and ICAM1 (right panel) in uninduced (solid line) and induced (dashed line) SCC12F-tet-LMP1 cells. Numbers shown represent channel numbers, values on an arbitrary scale that are proportional to the total fluorescence of the cell population being measured. The shaded curve shows the isotype control. (g) Levels of LMP1-specific mRNA in SCC12F-tet-LMP1 cells in the presence (light-grey bars) or absence (dark-grey bars) of doxycycline, normalized to that found in the EBV-positive lymphoblastoid cell line X50-7. The level in X50-7 is given a value of 1.00 (dashed line). (h) Immunoblots demonstrating the levels of LMP1 protein in induced and uninduced SCC12F-tet-LMP1 cells compared with the LMP1-negative NPC cell line C666-1, the LMP1-positive NPC xenograft C15 and X50-7. β-Actin was used to confirm equal protein loading.

Q-PCR was used to determine the levels of LMP1 mRNA in relation to the amount found in cells naturally infected by EBV. In induced cells, the level of LMP1 mRNA was 2.2-, 4.9- and 4.5-fold higher than in the EBV-positive B-lymphoblastoid cell line X50-7 whereas, in uninduced cells, it was 6.7-, 4.2- and 7.1-fold lower (Fig. 1g). LMP1-specific RNA was induced 14-, 20- and 32-fold in the three replicate pairs of RNA used for microarray and Q-PCR analysis (Fig. 1g). To assess levels of LMP1 protein expression in an epithelial cell background, we used the LMP1-negative NPC cell line C666-1 and the LMP1-positive NPC xenograft C15 alongside induced and uninduced SCC12F-tet-LMP1 cells and X50-7. Whilst LMP1 was undetectable in C666-1, a trace amount was observed in uninduced SCC12F-tet-LMP1 cells. Levels in induced cells were comparable to those in C15 and X50-7 (Fig. 1h).

Differentially expressed genes
The three replicate pairs of RNA used in this work were also the subject of a comparative, technical exercise to assess the effect of different workup procedures on microarray results. Affymetrix procedures using 1 µg and 10 µg total RNA and an amplification protocol starting from 5 ng were used (see Methods). Some true (i.e. validated by alternative methods) gene expression changes were identified by each protocol but missed by the other two. Results from the three workup procedures are presented here only in the context of additional, biologically relevant information that was obtained from the combination of methods. For a discussion of the impact of different workup protocols, see Li et al. (2004b). Data from the three protocols were normalized separately with RMA and analysed with SAM using criteria of fold change ≥1.5 and percentage false-positives (pfp) ≤10 %. The 10 µg, 1 µg and 5 ng analyses respectively predicted 100, 77 and 233 differentially expressed genes.

In addition to SAM, we used two other methods of differential gene identification; dChip and RP. dChip generated gene lists that were similar to those obtained using RMA/SAM, whereas the RP analysis was somewhat different. Using RP with pfp ≤10 % and fold change ≥1.5, the 10 µg, 1 µg and 5 ng analyses respectively predicted 140, 109 and 100 differentially expressed genes. The relationship between the differentially expressed gene lists obtained by RP, SAM and dChip on one dataset (10 µg) is shown in Fig. 2. The complete gene list obtained by analysing all three datasets with the three methods is presented in Supplementary Table S4.

(14K): Fig. 2. Relationship between significantly changed genes identified by different procedures, showing differentially expressed genes predicted by SAM, RP and dChip on the 10 µg dataset. Figures in parentheses indicate the total number of differentially expressed genes in a particular analysis.

The differential expression of the majority of genes of interest predicted by microarray analysis was validated by at least one of the following methods: Q-PCR, RT-PCR, immunofluorescence staining, Western blotting, quantitative ELISA or antibody array (Table 1, Figs 3, 4 and 5). When comparing the fold changes obtained by microarray and Q-PCR (Table 1), it was notable that Q-PCR scored a number of genes as having very large fold changes, whereas the values predicted from the array data fell far short of these. Such discrepancies between Q-PCR and microarray measurements are well known (for discussion see Draghici et al., 2006). However, large fold changes at the RNA level did not necessarily lead to similar changes at the protein level. In the three cases (IL8, IL6 and VEGFC) for which we obtained quantitative protein expression data, Q-PCR-determined fold changes of 34.6, 12.0 and 5.5 at the RNA level translated into fold changes of 3.0, 7.1 and 1.4, respectively, at the protein level, as determined by ELISA (Table 1).

Table 1. Summary of microarray and validation data Genes are organized by ontology. FC, Fold change; IF, immunofluorescence analysis; WB, Western blotting; Ab, antibody; U, upregulated; D, downregulated; ND, not detected; NS, not significant in this analysis. *, Determined using antibody MRP8/14, which is specific for the S100A8/9 complex. ELISA fold changes are means of triplicate determinations. –, Assay not performed.

(82K): Fig. 3. Validation of differentially regulated genes by RT-PCR, Western blotting and immunofluorescence staining. Differentially regulated genes were detected by RT-PCR followed by agarose gel electrophoresis (a) or by immunoblotting (b) and alterations in cytokine and chemokine expression were detected by cytokine antibody array (c). GAPDH, Glyceraldehyde-3-phosphate dehydrogenase. In (c), positive (1A, 1B, 2A, 2B, 7L and 8L) and negative (1C, 1D, 2C, 2D, 7K and 8K) controls are included. Cytokines (IL6, IL8, CCL2 and CCL5) and chemokines (CXCL1) induced in response to LMP1 expression are indicated (boxed wells). The format of the antibodies on the cytokine membrane is indicated below. (d) Indirect immunofluorescence analysis of the expression of CCL20, RAB27A, INHBA and PLAU in induced and uninduced SCC12F-tet-LMP1. Bars, 20 µm.

(81K): Fig. 4. LMP1 affects the expression of genes associated with cell-cycle control and proliferation, demonstrated by indirect immunofluorescence analysis of the expression of TP53, CCNE2, MCM5 and PTTG1 in induced and uninduced SCC12F-tet-LMP1 cells. Bars, 10 µm (MCM5) or 20 µm (TP53, CCNE2 and PTTG1).

(92K): Fig. 5. LMP1 decreases expression of genes associated with the terminal differentiation process, demonstrated by indirect immunofluorescence analysis of the expression of EVPL, SNAI2, KLK6 and S100P in induced and uninduced SCC12F-tet-LMP1 cells. Bars, 10 µm.

LMP1 induces changes in keratinocyte gene expression that are characteristic of inflamed, hyperproliferative epidermal disorders
Examination of the lists of differentially expressed genes revealed alterations in genes involved in diverse cellular processes such as cell growth and differentiation, cell motility, the cell cycle and inflammation. Of particular interest was that many fell into the categories of inflammatory and wound-response genes (Table 1, Figs 3 and 6). Increased expression of many of these classes has been reported in hyperproliferative and inflamed epidermis and is consistent with previous observations demonstrating that LMP1 induces hyperproliferative lesions in transgenic murine epidermis (Stevenson et al., 2005; Wilson et al., 1990). The induction of several of these genes (IL6, IL8, VEGF, CXCL1, 2, 3, CSF2, CCL5, CCL20, BIRC3, TNFAIP3, NFKBIA and IER3) (Table 1, Fig. 3) by LMP1 is most likely attributable to activation of NF-κB. Indeed, several of the differentially expressed genes have been identified previously as putative NF-κB targets (Table 1).

(94K): Fig. 6. LMP1 alters the subcellular distribution of IL-1α (IL1A), the soluble decoy receptor IL1R2 and the IL-1α/β targets S100A8/9 and PLAUR, demonstrated by indirect immunofluorescence analysis in induced and uninduced SCC12F-tet-LMP1 cells. Bars, 20 µm. Insets show higher magnification of the same fields.

In addition to the broad range of cytokines, chemokines and anti-apoptotic proteins, LMP1 expression was also associated with increased expression of a number of genes known to be induced in response to keratinocyte wounding (Table 1). We detected alterations in the profiles of IL-1 and TGF-β family members (Figs 3 and 6), with increased expression of IL-1β and activin A (inhibin-βA homodimers), biological response modifiers that function to promote keratinocyte proliferation and wound healing (Bamberger et al., 2005; Zhang et al., 2005).

LMP1 also significantly induced members of the S100 protein family, calcium-regulated proteins that control fundamental biological processes such as cell-cycle progression, cell motility and differentiation. S100A7 is overexpressed in psoriasis and, like S100A8 and S100A9, plays a key role in epidermal wound responses (Eckert et al., 2004). Four S100 family members (S100A2, S100A7, S100A8, S100A9) were upregulated, whereas expression of S100P was significantly reduced (Table 1, Fig. 5). Indirect immunofluorescence staining, using an antibody specific for the S100A8/9 complex (MRP8/14), confirmed not only increased levels in LMP1-expressing cells, but also an altered intracellular distribution (Fig. 6).

In addition to the aforementioned growth factors and cytokines, LMP1 influenced expression of other soluble factors that participate in wound healing, cell motility and cell-matrix interactions. LMP1 induced expression of several proteases and their specific inhibitors which modulate the interdependent processes of cell migration and matrix proteolysis as part of a global programme of wound healing: TIMP1, a metalloproteinase inhibitor associated with tumour growth and metastasis in head and neck squamous carcinoma (Ruokolainen et al., 2005), the serine protease urokinase plasminogen activator (PLAU), a metalloprotease, ADAMTS1, and a number of serine proteinase inhibitors (serpins A3 and B7) (Table 1, Fig. 3d). In contrast, LMP1 expression was associated with decreased expression of ECM1 (Table 1), a secreted glycoprotein that binds to heparan sulphate proteoglycan, and transgelin, an actin cross-linking protein involved in cell shape change whose expression is downregulated during cellular transformation (Shields et al., 2002). Another gene implicated in cell invasion, and whose expression was also increased by LMP1, was RAB27A (Fig. 3d), a member of the Ras GTPase family. RAB27A plays a critical role in regulating membrane traffic and its upregulation by LMP1 may participate in increased secretion of cytokines and growth factors (Johnson et al., 2005).

LMP1 induces the expression of genes involved in cell proliferation and cellular stress responses
The induction of LMP1 was accompanied by increased expression of genes involved in cell-cycle regulation, suggesting that the cellular phenotype induced by LMP1 is attributable in part to increased cell proliferation. We found increased expression of MCM5 and MCM6, DNA licensing factors that regulate DNA replication during S phase, CCNE2 (cyclin E2), a cell-cycle regulatory protein that functions as the rate-limiting step in late G1–S phase transition, and geminin (GMNN), an inhibitor of DNA replication initiation that prevents endo-reduplication (Montanari et al., 2005) (Table 1). In addition, LMP1 induced expression of pituitary tumour-transforming gene (PTTG1), a putative oncogene that plays a key role in mitotic regulation and whose deregulated expression is frequently observed in ovarian, prostate and thyroid tumours (Kim et al., 2006; Tfelt-Hansen et al., 2006) (Table 1, Fig. 4). Expression of high mobility group (HMG) box proteins including HMGA2, HMGB1 and HMGB3 was also induced in response to LMP1 expression (Table 1). HMG proteins are DNA-binding proteins with pleiotropic functions whose deregulated expression has been shown to be oncogenic in certain cell systems (Ulloa & Messmer, 2006; Yamada & Maruyama, 2007).

In contrast, we found that LMP1 decreased expression of ID1 (Fig. 3b), a dominant-negative helix–loop–helix protein, and BTG1 (Table 1), an antiproliferative gene family member that, like ID1, is induced in response to TGF-β (Iwai et al., 2004). LMP1 expression was also associated with inhibition of TRIB3, ATF3, ATF4, ASNS and ASS (Table 1), factors that play a key role in the cellular stress response (Jousse et al., 2007).

Interestingly, a significant reduction in expression of the tumour-suppressor protein TP53 was observed in response to LMP1 expression (Figs 3b and 4). This repression was correlated with increased expression of CKS2, a protein that constitutes an essential component of the cyclin/cyclin-dependent kinase complexes that contribute to cell-cycle control (Rother et al., 2007), and a reduction in expression of DDB2, a gene involved in global genomic repair, whose is expression is induced by TP53 (Prost et al., 2007) (Table 1).

LMP1 represses the expression of genes involved in keratinocyte terminal differentiation
The increased rate of keratinocyte proliferation observed in inflammatory and hyperplastic skin diseases results in abnormal terminal differentiation and impaired epidermal barrier function. We identified a number of genes whose expression is known to be increased in differentiating cells but which were significantly repressed by LMP1. Notably, members of the kallikrein family (KLK5, 6, 7, 8 and 10) were the most significantly downregulated by LMP1 (Table 1, Fig. 5). A significant downregulation in S100P levels was also observed (Fig. 5), which contrasted with other S100 protein family members (S100A2/7/8/9), whose expression was markedly induced by LMP1. Intriguingly, another viral oncoprotein, human papillomavirus 16 E7, also downregulates expression of S100P in epithelial cells (Hellung Schønning et al., 2000).

We also observed significant downregulation of DSG3 (Fig. 3b), a member of a family of desmosomal cell-adhesion molecules that maintain tissue integrity and have tumour-suppressor function. Deregulated expression or the loss of individual desomosome components is frequently observed in oral squamous cell carcinomas and associated with increased metastatic potential (Imai et al., 1995).

LMP1 repressed the transcription factors KLF4, SNAI2 and CEBPG (Table 1, Figs 3a and 5) that regulate keratinocyte barrier function and integrin and cell-adhesion molecule expression. In keeping with a reduction in the differentiation capacity of LMP1-expressing SCC12F cells, we found a significant reduction in expression of SCEL and PPL (a member of the envoplakin family of cross-linked envelope proteins), which serve as substrates for transglutaminase and are incorporated into the cornified envelope of terminally differentiated keratinocytes (Table 1, Fig. 3a).

LMP1 alters the intracellular distribution of proteins in the absence of changes in transcriptional expression
The identification of functional families of genes whose expression was influenced by LMP1 prompted us to investigate related genes that the microarray analysis had not identified as being differentially regulated at the RNA level (Table 2). Thus, FACS or cytokine array analysis revealed upregulation at the protein level of CD40, ICAM1 (Fig. 1f) and the inflammatory cytokine CCL2 (Fig. 3c). Immunofluorescence staining showed that the differentiation-associated cross-linked envelope proteins IVL and EVPL and the cell-adhesion protein CDH1 were downregulated (Figs 1 and 5). Although expression of mRNA for IL-1α remained unchanged, further investigation by immunofluorescence staining revealed a marked difference in the subcellular localization of IL-1α in cells expressing LMP1 (Fig. 6). Here, strong nuclear staining of IL-1α was observed in uninduced cells, whereas diffuse cytoplasmic staining was observed in cells induced to express LMP1, findings which suggest altered processing and nuclear targeting of pre- and pre-pro IL-1α in LMP1-expressing cells (reviewed by Apte & Voronov, 2008). Similarly, the transcriptionally unchanged PLAU receptor (PLAUR) also exhibited intercellular redistribution in LMP1-expressing SCC12F cells (Fig. 6). Compared with uninduced cells, where pronounced cytosolic vesicular staining was observed, strong membrane-associated staining was observed in LMP1-expressing cells. Virtually identical patterns were observed for S100A8/9, where the diffuse cytosolic staining observed in uninduced cells was redistributed to the membrane in response to LMP1 expression. More subtle changes were observed for IL1R2, where the diffuse cytosolic staining observed in uninduced cells became intense in perinuclear regions of LMP1-expressing cells (Fig. 6).

Table 2. Genes that appeared unchanged at the RNA level but which showed changes at the protein level Genes are organized by ontology. RD, Cellular redistribution; see Table 1 for other abbreviations.

The generation of an inducible expression system in the immortalized, non-tumorigenic keratinocyte cell line SCC12F has allowed us to analyse changes in cellular gene expression that occur in response to transient LMP1 expression. The SCC12F keratinocyte cell line was chosen because expression of LMP1 in these cells induces striking effects on cell morphology and differentiation (Dawson et al., 1990, 2000). The ability of SCC12F cells to respond to growth and differentiation stimuli has allowed us to identify the impact of LMP1 expression in the context of a keratinocyte cell line that more closely approximates normal epithelial cells cultured in vitro.

We provide evidence that the ability of LMP1 to induce profound phenotypic changes in SCC12F cells is accompanied by an equally profound global change in host-cell gene transcription programmes and demonstrate that LMP1 activates genetic programmes that are involved in inflammation and wound healing, an effect that is achieved in part through the induction of NF-κB- and AP-1-regulated cytokines, chemokines and growth factors. The fact that many of the genes identified in response to LMP1 expression are also upregulated in inflamed, hyperproliferative epidermal disorders suggests that LMP1 enforces a chronic inflammatory wound-healing response, inducing keratinocytes to proliferate, become more motile and lose certain differentiation-associated characteristics. More importantly, our results provide a molecular view that links inflammatory and wound-healing pathways to the oncogenic behaviour of LMP1 and provide a molecular basis to link chronic inflammation and wound healing with tumour development (Dvorak, 1986).

LMP1 expression was also accompanied by increased expression of genes that participate in cell growth and whose activity regulates cell proliferation, findings which suggest that the cellular phenotype induced by LMP1 is due in part to effects on cell proliferation. Whether these effects are direct or mediated through the action of cytokines or growth factors requires further study. LMP1 expression was associated with increased expression of cyclin E2 and MCM5 and 6, which regulate G1 to S phase transition and participate in DNA replication during S phase (Maiorano et al., 2006; Payton & Coats, 2002). In addition, LMP1 increased expression of PTTG1, a putative oncogene that regulates the cell cycle and whose transcriptional activity is regulated by ERK-MAPK (Pei, 2000, 2001). A role for PTTG1 in regulating facets of the angiogenic response has been established through its regulation of the expression and activity of VEGFC (Kim et al., 2006), a potent angiogenic factor that was also induced by LMP1 (Table 1, Fig. 3b). An intriguing, yet consistent observation was our finding that LMP1 downregulated expression of the tumour-suppressor gene TP53. Although the relevance of LMP1-mediated repression of TP53 is unknown, recent studies have demonstrated reduced TP53 expression in psoriatic epidermis (Johansen et al., 2005) and keratinocytes wounded in vitro (Vollmar et al., 2002), suggesting that TP53 may be downregulated during wound-healing responses as part of a strategy to inhibit cell loss through apoptosis. Of particular interest was the striking reduction in ID1 expression in cells induced to express LMP1. Although recent studies have described the induction of ID1, 2 and 3 by LMP1 in the C33A epithelial cell line (Everly et al., 2004; Li et al., 2004a), our findings can be explained on the basis that ID1 is a transcriptional target of activin A which, in this context, is associated with transcriptional repression (Rotzer et al., 2006).

A hallmark of hyperproliferative epidermis is aberrant terminal differentiation induced as a consequence of increased cell proliferation (Nickoloff, 1999). We observed significant changes in families of genes whose expression is induced as a consequence of cellular differentiation but which were significantly repressed by LMP1. These included the tissue kallikreins (KLK5, 6, 7, 8 and 10), a subgroup of serine proteases that regulate the final stages of terminal differentiation (Borgono et al., 2007), DSG3, a component of desmosomes that functions to maintain tissue integrity (Chidgey, 2002), and KLF4, a transcription factor that orchestrates epidermal barrier function. In addition, we observed decreased expression of EVPL and PPL, SCEL and IVL, proteins that are deposited beneath the plasma membrane and are cross-linked by calcium-activated transglutaminases to form the cornified envelope. The ability of LMP1 to repress the expression of a number of differentiation-specific genes supports our previous observations, showing that LMP1 modulates keratinocyte differentiation (Dawson et al., 1990, 2000).

The phenotypic consequences of LMP1 expression in SCC12F cells are superficially similar to those of cytokine-stimulated or wounded keratinocytes (Singer & Clark, 1999). Although the mechanism(s) by which LMP1 induces these effects is not completely defined, it is probably linked to chronic or sustained activation of the NF-κB and MAPK signalling pathways. Studies from transgenic mice have shown that constitutive activation of the NF-κB and ERK-MAPK pathways is sufficient to induce epidermal hyperproliferation (Chen et al., 2000; Gutschalk et al., 2006; Hobbs et al., 2004; Kim et al., 2006; Klement et al., 1996). As pro-inflammatory cytokines play a key role in the pathophysiology of psoriasis and other hyperproliferative lesions (Nickoloff et al., 2006), it is possible that the induction of IL-1α/β, IL-6, IL-8, the CXCL chemokines (CXCL1, 2, 3) and GM-CSF contributes to LMP1-induced effects, as these factors not only serve to recruit leukocytes to epithelial sites but also influence keratinocyte migration and proliferation (Groves et al., 1995; Haase et al., 2001; Reiland et al., 1999; Rheinwald & Beckett, 1981). The role of IL-1α/β in these effects is particularly intriguing given that these cytokines are overexpressed in psoriatic skin (Mee et al., 2006) and are capable of inducing hyperproliferative epidermal lesions in transgenic mice (Groves et al., 1995). Microarray analysis comparing transcription profiles derived from psoriatic skin and IL-1α-stimulated keratinocytes has revealed a striking level of overlap (Mee et al., 2007), suggesting that expression of IL-1α/β is causally linked to these disorders.

The induction of the TGF-β-family member activin A by LMP1 is also noteworthy, as this cytokine is overexpressed in hyperproliferative epidermal disorders, is induced in response to keratinocyte wound healing in vitro (Bamberger et al., 2005; Werner & Alzheimer, 2006; Zhang et al., 2005) and is a putative target of IL-1α (Keelan et al., 1998). The induction of activin A may be central to the ability of LMP1 to induce a hyperproliferative wound response, as transgenic mouse studies show that epidermal targeting of activin A induces such lesions (Munz et al., 1999). That LMP1 induced expression of multiple S100 family proteins (S100A2, A7, A8 and A9) (Table 1), all of which are established targets of IL-1α/β and activin A action, further strengthens a role for these cytokines in LMP1-mediated effects (Eckert et al., 2004; Haase et al., 2001; Thorey et al., 2001).

As the transcriptional changes induced by LMP1 in SCC12F cells bear a striking similarity to those of IL-1α/β-stimulated normal keratinocytes, it is possible that the induction of these cytokines may be central to the effects of LMP1 on keratinocyte behaviour. Evidence in support of this comes from our observations showing that the expression and subcellular distributions of various components of the IL-1α/β signalling pathway are altered in LMP1-expressing cells. Thus, in addition to increased expression or processing of IL-1α/β, the expression and/or subcellular localization of the IL-1 decoy receptor IL1R2 and the putative IL-1α/β targets activin A, PLAU, PLAUR, S100A8/9 and GM-CSF were all modified in response to LMP1 (Figs 3d and 6). At this stage, it is not clear whether LMP1 usurps the IL-1 signalling pathway to stimulate changes in host-gene transcription or whether these changes occur as a result of autocrine IL-1α/β stimulation.

By usurping the CD40/TNF signalling pathways, LMP1 activates the NF-κB and MAPK pathways in a constitutive manner (Eliopoulos & Young, 2001). Although chronic stimulation of these pathways is essential for LMP1 to transform B cells, this activation manifests itself as an uncontrolled inflammatory wound response when expressed inappropriately in keratinocytes. Both wound repair and carcinoma formation are characterized not only by increased keratinocyte proliferation and aberrant differentiation but also by extracellular matrix remodelling, increased cell motility and vascularization (Hanahan & Weinberg, 2000). It is noteworthy that LMP1 induced expression of several proteases (PLAU, ADAMTS1) and VEGF, all of which participate in tissue remodelling and whose inappropriate expression may facilitate cell invasion. A crucial difference between transient and chronic activation of these pathways is that chronic or inappropriate activation is potentially oncogenic. The ability of LMP1 chronically to stimulate all of these processes may allow for deregulated cell growth, tumour cell invasion, angiogenesis and, ultimately, metastasis.

While the relationship between inflammation, wound healing and cancer is widely accepted, the molecular mechanisms that link inflammation and cancer remain unclear (Coussens & Werb, 2002). Here, we have identified key changes in gene expression programmes that are induced by LMP1 in the non-tumorigenic keratinocyte cell line SCC12F. Our results provide a molecular view that links inflammatory and wound-healing pathways to the oncogenic behaviour of LMP1. LMP1 coordinates pathways common to wound healing and inflammation but, through chronic stimulation of these pathways, converts them into a potentially oncogenic signal.

We thank Dr K. Toellner for advice and assistance with Q-PCR assays and Drs G. Krupp and P. Scheinert for help with the amplification protocol. This work was supported by Cancer Research UK.

Footnotes

^†, Christopher W. Dawson ^†, Wenbin Wei, John D. O'Neil, Suzanne E. Stewart †These authors contributed equally to this work. ‡Present address: University of Edinburgh Medical School, Division of Pathway Medicine, Chancellors Building, 49 Little France Crescent, Edinburgh EH16 4SB, UK.

A supplementary figure, supplementary methods and four supplementary tables are available with the online version of this paper.