Papillomavirus E2 Protein Induces Expression of the Matrix Metalloproteinase-9 via the Extracellular Signal-Regulated Kinase/Activator Protein-1 Signaling Pathway

Introduction

Papillomaviruses are small dsDNA viruses causing epithelial tumors, including cancers of the female cervix and the head and neck (1–3). In addition, papillomaviruses are believed to be involved in the development of nonmelanoma skin cancer (4–6). The early molecular events of epithelial carcinogenesis induced by papillomaviruses have been extensively studied as far as cell proliferation and immortalization are concerned (7–9). Deregulation of the cell cycle, inhibition of apoptosis, induction of telomerase, and aneuploidy are mostly due to the activities of the viral oncogenes E6 and E7 of high-risk genital papillomaviruses that are necessary risk factors for cervical cancer (10, 11). In contrast, factors contributing to the progression of in situ carcinomas into invasively growing metastasizing cancers are largely unknown.

A suitable animal model to investigate the molecular mechanisms of papillomavirus-induced tumor formation and progression is the New Zealand White rabbit, which develops 4 to 8 weeks after infection with the cottontail rabbit papillomavirus (CRPV) skin tumors that ultimately metastasize within 1 year to the lungs without a need for additional cofactors (12). In papillomas, the viral DNA is maintained as monomeric plasmids in comparable copy numbers as in carcinomas that contain, in addition to monomeric forms, multimeric circles as well as integrated DNA. The presence of extrachromosomal copies in the carcinomas indicates viral DNA replication and the expression of the E1 and E2 proteins (12). Almost all early viral gene products, including the viral transcription and replication factor E2, are required for tumor formation. Noteworthy, we showed previously that the infection of rabbits with CRPV mutant genomes encoding single amino acid exchanges in the transactivation domain of E2 strongly impeded the efficacy of tumor formation (13). Most of the mutant E2 proteins had lost their ability to transactivate artificial promoters with multiple E2-binding sites, but all maintained their DNA-binding and replication capacities (13). On the cellular level, we identified tumor-promoting candidate genes by comparing the gene expression profiles of benign versus malignant rabbit skin tumors, which revealed proteins involved in cell motility, tumor invasion, and metastasis that are up-regulated in the process of CRVP-induced cancer progression (14). The transition from benign to invasive tumors is most likely not only dependent on the neoplastic cells themselves but also strongly affected by the interaction of neoplastic cells with their extracellular matrix (15). Extensive remodeling of the extracellular matrix is mediated through the activity of matrix metalloproteinases (MMP; ref. 16). MMPs are believed today to play a key role in the creation of a tumor-favorable environment due to the activation of growth factors, their antiangiogenic and proangiogenic properties, and the ability to cleave cell-cell and cell-matrix contacts (17, 18). The expression of MMPs is strictly controlled at both transcriptional and post-transcriptional levels and by specific interactions with naturally occurring tissue inhibitors of MMPs (19). Several studies have reported that especially MMP-2 and MMP-9 are up-regulated in cervical neoplasias and suggested a correlation between increased MMP-2/MMP-9 expression and poor prognosis (20, 21). Other studies suggested that MMP-2 and MMP-9 could be involved in the early steps of skin carcinoma development in immunosuppressed patients, which are of high risk to develop skin cancers that are >90% human papillomavirus (HPV) positive (22, 23). The transcription of the MMP-9 and MMP-2 genes is, however, differently controlled. Whereas the promoter of MMP-9 contains cis-elements [activator protein-1 (AP-1) and nuclear factor-κB (NF-κB)] that can be regulated through mitogen-activated protein kinases (MAPK; refs. 24, 25), the promoter of MMP-2 shares no such conserved cis-elements (19). MAPKs are a family of enzymes that transduce signals via several phosphorylation steps into the nucleus (26). The three best characterized members of the MAPK are the stress-activated c-Jun NH₂-terminal kinase, the p38 kinase, and the extracellular signal-regulated kinase (ERK). Growth factors lead to an activation of ERK by upstream regulator proteins, such as Raf, which itself can be activated by Ras, a key mediator of cell proliferation (27). Therefore, activation of ERK and entry into the nucleus can often be found in cancer and has been linked to cervical and other epithelial neoplasia (28, 29).

We were interested to identify a molecular basis for the different tumorigenic activity of the E2 transactivation domain mutants and looked for differences between skin tumors induced by CRPV wild-type (WT) genome or by mutant genomes defective in E2. This report provides novel evidence that mutations in the transactivation domain of CRPV-E2 impede not only the formation of new skin tumors in rabbits but also their progression into invasive carcinomas, which we show is linked to the ability of E2 to induce the expression of MMP-9 via the stimulation of the MAPK/ERK pathway resulting in an activation of AP-1.

Materials and Methods

Animal model. New Zealand White rabbits were infected with WT CRPV DNA or mutant E2 genomes with the help of a “helios gene gun” (Bio-Rad, München, Germany) as described previously (13). Tumors were harvested at the indicated time points, snap frozen, and stored at −80°C.

Plasmids. The cloned WT and mutant CRPV genomes and expression vectors for WT and mutant CRPV-E2 and WT HPV31-E2 have been described (13, 30). Mutations in the DNA-binding domain of CRPV-E2 (mutation of amino acid 320/321) were introduced into the respective expression plasmids by exchanging the WT sequence with mutant fragments generated by PCR. The E2-dependent reporter plasmid pC18-SP1-luc has been described (30). The CRPV promoter construct pCRPV-PL,P1,P2,P3-luc was constructed by amplifying CRPV nucleotides 7,328 to 964 by PCR, which added KpnI sites, and then cloning the fragment into KpnI-digested pGL3 basic (Promega, Mannheim, Germany). To generate the MMP-9 luciferase reporter construct (hMMP-9-pGL3), the MMP-9 promoter fragment was excised from plasmid pSV0-MMP-9 (31) with BglII and NcoI and cloned into pGL3 basic (Promega) digested with BglII and NcoI. Introduction of point mutations into the AP-1-binding sites and deletion of the NF-κB-binding site were made with QuickChange Site-Directed Mutagenesis kit (Stratagene, Amsterdam, the Netherlands) using two primers for each mutation, which carried two base changes or lack the complete binding sequence, respectively (AP-1/1: 5′-GAAGCAGGGAGAGGAAGCTTTGTCAAAGAAGGCTGTCAGG and 3′-CCTGACAGCCTTCTTTGACAAAGCTTCCTCTCCCTGCTTC, AP-1/2: 5′-CACACCCTGACCCCTTTGTCAGCACTTGCCTG and 3′-CAGGCAAGTGCTGACAAAGGGGTCAGGGTGTG, NF-κB: 5′-ACAGGGGGTTGCCCCAGTAGCCTTGCCTAGCAG and 3′-CTGCTAGGCAAGGCTACTGGGGCAACCCCCTGT). All mutations were confirmed by DNA sequencing. Reporter plasmids pAP-1-luc, pNF-κB-luc, and pMCS-luc are commercially available (Stratagene).

Tissue cultures and materials. The rabbit epithelial cell line (ATCC CCL68) was maintained as described (13). Normal human foreskin keratinocytes and rabbit keratinocyte cell lines AVS, which contain the CRPV genome, RK1-16E7/ras (30), and CRPV-E6 and CRPV-E7 rabbit keratinocytes, which were obtained after retroviral infection and G418 selection, were maintained in supplemented keratinocyte serum-free medium (Invitrogen, Karlsruhe, Germany). The human squamous cell carcinoma cell line (UM-SCC1) was maintained as described (32).

RNA extraction and quantitative reverse transcription-PCR. Total RNA was extracted from biopsies using TRIzol reagent (Invitrogen) and treated with RQ1 RNase-free DNase (Promega). Quantitative PCR for the detection of MMP-9 transcripts was done with specific primers 5′-CTGGGCAAGGGCGTCGTGGTC-3′ and 5′-CGTGGTGCAGGCGGTGTAGGAG-3′ with SYBR Green in the Sequence Detection System 5700 (Applied Biosystems, Warrington, United Kingdom). To quantify the absolute amount of transcripts, standard curves were generated with dilutions of plasmids containing cloned cDNA fragments. Absolute transcript numbers of MMP-9 were normalized to β-actin transcripts. Following quantitative PCR amplification, melting curves were determined for the amplicons of the plasmid standards and the samples. By comparison of the melting points of the amplicons of the biopsy samples and the plasmid standards, the specificity of the amplicons was verified.

Protein analysis. Specimens were crushed in liquid nitrogen and immediately resuspended in buffer containing 1% NP40, 25 mmol/L Tris-HCl (pH 7.5), 25 mmol/L NaCl, 1 mmol/L Na₃VO₄, 10 mmol/L NaF, 10 mmol/L sodium pyrophosphate, 1 mmol/L phenylmethylsulfonyl fluoride, 0.5 mmol/L EGTA, and 10 nmol/L okadaic acid. After melting at 4°C and two centrifugation steps to remove remaining cell debris, total protein amount in supernatants was measured using the Sigma TPRO-562 procedure (Sigma, Taufkirchen, Germany) according to the manufacturer's guidelines. Zymography was done as described previously (24). Quantification of the protein amount was done by densitometric measurement.

Transfected cells were lysed 40 hours after transfection using radioimmunoprecipitation assay (RIPA) buffer containing 50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% NP40, and 0.25% sodium deoxycholate. Tumor or cell lysates were normalized for protein amount and loaded on 10% SDS-PAGE. After electrophoresis, proteins were blotted on polyvinylidene difluoride membranes for 1 hour at 100 V. After blocking in 3% PBS-Tween 20/nonfat dry milk overnight at 4°C, membrane was incubated with antibodies [sc-94-G for ERK1, 1:2,000, Santa Cruz Biotechnology, Santa Cruz, CA, and IM09L for MMP-9, Oncogene Research, Merck Biosciences, Darmstadt, Germany, 1:2,000 for 2 hours (IM09L) overnight at 4°C] at room temperature. Horseradish peroxidase–conjugated antibodies were all purchased from Santa Cruz Biotechnology and used for visualization with Enhanced Chemiluminescence Plus System (Amersham Biosciences UK Ltd., Buckinghamshire, United Kingdom) at 1:2,000 dilutions. Differences in protein expression were measured using densitometric tools. Quantification of the protein amount was done by densitometric measurement.

Samples normalized for 1,000 μg protein were first incubated for 1 hour at 4°C with 20 μL protein A agarose on a rotating device, beads were pelleted by centrifugation, and supernatant was incubated with 4 μg pERK antibody (sc-7383 for pERK, 1:1,000, Santa Cruz Biotechnology) and 20 μL protein A agarose at 4°C overnight. Immunoprecipitates were collected by centrifugation and washed with RIPA buffer. The pellet was then resuspended in 40 μL electrophoresis buffer and boiled for 3 minutes followed by SDS-PAGE with subsequent Western blotting.

In vitro invasion/mobility assay. Cells (n = 150,000) were seeded 16 hours before transfection in six-well plates. Transfection was done using 250 ng of the indicated vectors together with 3 μL LipofectAMINE (Invitrogen). A diluted Matrigel/serum-free medium solution (0.25×, 20 μL) was applied to filters with 8-μm pore size (Becton Dickinson, Bedford, MA) and incubated at 37°C for 45 minutes. Cells were resuspended 24 hours after transfection in Opti-MEM culture medium and 80,000 cells were added to the upper compartment of the invasion chamber containing MMP-9 antibody (IM09L) or control preimmune serum at same concentrations, whereas the lower compartment contained complete CRL growth medium with antibody or preimmune serum at equal concentrations. After incubation for 40 to 50 hours at 37°C, invasion was determined based on the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) activity on the lower side of the filter as a percentage of the total activity in the chamber. For experiments using PD 098059, same concentrations were added in both chamber compartments and volume differences were normalized using DMSO.

Transient reporter assays. CRL cells (∼2.5 × 10⁴) were seeded into the wells of 24-well tissue culture plate in sf1Ep medium supplemented with epidermal growth factor (EGF) and FCS. The next day, cells were cotransfected with 20 ng luciferase reporter plasmid and 2.5 ng pSG5 or the respective E2-pSG5 expression vector DNA in the presence of 1.25 μL LipofectAMINE. Five hours after transfection, the medium was changed to sf1Ep medium without serum and EGF. Luciferase assays were carried out 48 hours after transfection. The cells were washed twice with cold PBS and then lysed by adding 100 μL cold luciferase extraction buffer [0.1 mol/L potassium phosphate (pH 7.8), 1% Triton X-100, 1 mmol/L DTT]. Lysates were cleared by centrifugation (20,000 × g, 5 minutes, 4°C) and 80-μL aliquots of extract were subjected to luminometer analysis as described in the manufacturer's manual. Transient luciferase expression assays were repeated with different plasmid preparations three to five times, each in duplicate measurements.

Results

High expression of MMP-9 in cottontail rabbit papillomavirus–induced skin lesions depends on wild-type E2 transactivator protein sequence. We have shown previously that mutations in the transactivation domain of the CRPV-E2 transcription/replication protein strongly reduces but not completely abolishes the ability of CRPV to induce tumors in New Zealand White rabbits (13). Long-term follow-up experiments (14-19 months) with rabbits carrying mutant CRPV-E2–induced papillomas showed in the meantime that mutant E2-induced papillomas do not progress into invasive squamous cell carcinomas in contrast to WT CRPV-induced lesions (Table 1; P < 0.001 with power > 0.90). To address the underlying mechanism, we first analyzed the influence of two representative mutations in the E2 NH₂-terminal domain (I73A and R37K; Fig. 1) on the activity of the early viral promoters but found no significant difference to WT E2 with respect to their transactivation capacity (pCRPV-PL,P1,P2,P3-luc; Fig. 1). This excludes that the loss in tumor induction efficiency we observed earlier (13) is primarily due to altered regulation of the viral promoters by mutant E2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Activation of a reporter construct containing all four promoters of CRPV (pCRPV-PL,P1,P2,P3-luc; gray columns) or a reporter construct carrying four E2-binding sites in front of a minimal promoter (pC18-SP1-luc; black columns) by CRPV-E2 and two mutant E2 constructs (E2/I73A and E2/R37K). Rabbit keratinocytes (CRL6502) were cotransfected with expression vectors for WT and mutant E2 and the respective luciferase reporter construct. Fold activation of the reporter constructs pCRPV-PL,P1,P2,P3-luc and pC18-SP1-luc in comparison with the cotransfection with the empty expression vector pSG5. Columns, mean of three independent experiments; bars, SD.

However, we found that the ability to stimulate artificial promoters with four E2-binding sites (pC18-SP1-luc; ref. 30) differed at least 3-fold (Fig. 1) between WT E2, mutant R37K, and mutant I73A, suggesting that changes in the cellular gene expression that maybe caused by WT E2 and the mutant R37K but not by the I73A mutant could play a role in the tumorigenesis caused by CRPV. Indeed, in earlier gene expression profiling studies with CRPV-infected New Zealand White rabbits, we showed that during tumor progression several cellular genes involved in tumor invasion and cell motility become up-regulated (14). To identify the molecular basis for the different tumorigenic activity of the E2 transactivation domain mutants, we looked for differences between skin tumors induced by CRPV WT genome or by mutant genomes. Because transgenic mouse models of papillomavirus-associated carcinogenesis have already shown an important role for MMPs, especially MMP-9, in tumor establishment and growth (15), we first investigated the role of gelatinases as a subgroup of MMPs in CRPV-dependent tumorigenesis. For this, we employed zymographical techniques that combine SDS-PAGE with the detection of the enzymatic activity of gelatinases. This revealed that indeed three papillomas and four carcinomas (Fig. 2A) and a pulmonary metastasis caused by WT CRPV contained a 92-kDa gelatinase, which is absent in normal uninfected skin. In addition, a 72-kDa gelatinase was found at equal levels in normal skin as well as in CRPV-induced tumors. The identical molecular weight of both 92- and 72-kDa proteins with the known human MMP-9 and MMP-2, respectively, in combination with the gelatinase activity suggests that the enzymes detected represent their rabbit homologues. Interestingly, papillomas induced by a CRPV genome encoding either I73A or I73L mutant E2 protein expressed by far less MMP-9 in contrast to WT E2 or the mutant R37K-induced papillomas (Fig. 2A). To analyze if the observed increase in MMP-9 protein level is mediated by transcriptional induction, we next determined the level of MMP-9 transcripts by quantitative real-time PCR with RNA extracted from WT CRPV-E2 and mutant (I73A and R37K) CRPV-E2-induced tumors. The absence of MMP-9 transcripts in I73A mutant E2 tumors in comparison with WT E2- and R37K-induced lesions (Fig. 2B) showed a correlation between increased transcript levels of the MMP-9 gene and the presence of a WT CRPV-E2 amino acid sequence at position 73.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Zymographic detection of gelatinase activity in CRPV-induced skin lesions and in normal uninfected skin with SDS-PAGE containing 0.1% (w/v) gelatin. A, white bands indicate gelatinolytic activity from different protein extracts of WT CRPV-induced skin lesions [papilloma (Pap), carcinoma (Ca), or metastasis (Met)] or papillomas induced by mutant (mt) CRPV-E2 genomes (I73L, I73A, and R37K) that were normalized for protein amount. As a control, serum-free tissue culture supernatant from the human squamous cell carcinoma cell line UM-SCC1 (Control) was used. B, measurement of MMP-9 transcripts by quantitative real-time PCR in total RNA extracted from rabbit skin tumors induced by WT CRPV genome (WT CRPV-E2) or by two CRPV genomes with mutations in the 5′ end of the E2 gene (CRPV-E2/R37K and CRPV-E2/I73A). The amount of MMP-9-specific transcripts in 50 ng total RNA is given in relation to the amount of β-actin transcripts.

CRPV-E2 and HPV31-E2 induce expression and secretion of MMP-9 in vitro. To further investigate if the E2 protein is involved in the regulation of MMP-9 expression, we studied in tissue culture experiments rabbit keratinocytes that were immortalized by the full-length genome of CRPV (AVS), by the single action of CRPV E7, or both CRPV E6 and E7 for the presence of MMP-9 by zymography and Western blot techniques. This analysis revealed low levels of MMP-9 in keratinocytes immortalized by E6/E7 as already observed earlier by others (33), which were significantly higher in cells containing the full-length genome that additionally encodes the E2 gene (Fig. 3A). To explore if the induction of MMP-9 is a common property among the papillomavirus E2 proteins, we expressed high-risk HPV31-E2 protein in normal human keratinocytes and observed by zymography analysis a de novo expression of MMP-9 (Fig. 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Permanent rabbit keratinocyte cell lines containing expression vectors for CRPV E7, CRPVE6/E7, or the whole genome of CRPV (AVS) show different expression levels of MMP-9 protein. A, concentrated serum-free tissue culture supernatants of the indicated cell lines were used for SDS-PAGE combined with Western blot analysis using MMP-9 antibody (top) or for SDS-PAGE with gelatin zymography (bottom). Columns, relative amount of MMP-9 protein in comparison with each other. Concentrated tissue culture supernatant from UM-SCC1 cells served as control for the zymography. B, HPV31-E2 induces expression of MMP-9 in human primary keratinocytes. Cells were transfected with 300 ng of a HPV31-E2 expression vector (lane 2). Tissue culture supernatant was collected 50 hours after transfection. Samples normalized for protein amount were analyzed by zymography. Concentrated tissue culture supernatant from UM-SCC1 cells (lane 1) or from nontransfected normal human keratinocytes (lane 3) served as control. C, Matrigel-based invasion/mobility assay after transient transfection of rabbit keratinocytes (CRL6502) with different E2 constructs. Rabbit keratinocytes were transfected with 250 ng of the indicated constructs. Twenty-four hours after transfection, 80,000 cells were seeded in Matrigel-covered Boyden chambers containing membranes with 8-μm pores. Culture medium was supplemented with MMP-9 antibody (1.5 μg/mL) or preimmune serum, and invasion was determined 40 to 50 hours later densitometrically after MTT vital staining. Invasion is expressed as percentage of cells on the lower side of the membrane versus total cell number. Assays were repeated in triplicate. *, P ≤ 0.05, Student's t test.

To investigate if E2-transfected cells show invasive properties that can be suppressed using a MMP-9 antibody, we did Matrigel-based in vitro invasion/mobility assays with rabbit keratinocytes transiently transfected with expression vectors encoding the WT CRPV-E2 and the I73A as well as the mutant R37K E2. This showed a 2-fold increase of invasiveness for cells expressing the WT E2 protein in comparison with cells transfected with the empty expression vector, whereas cells transfected with the I73A mutant E2- or WT E2-transfected cells and treated with a MMP-9 antibody behaved not differently from cells transfected with the empty vector (Fig. 3C). In contrast, cells transfected with the mutant R37K E2 expression vector showed no significant difference in their invasive potential in comparison with WT E2-transfected cells. These experiments indicate thus far that E2 on its own is able to induce MMP-9 in CRPV-induced rabbit tumors and in cell lines and that MMP-9 is secreted from the transfected cells and is active with respect to the degradation of extracellular matrix.

E2 regulates the MMP-9 promoter in trans. As the induction of the MMP-9 protein in CRPV-induced tumors occurs at the level of transcription (Fig. 2B), we next did transient reporter assays with a luciferase construct driven by the well-characterized −670-bp upstream enhancer of the human MMP-9 promoter (31, 34), which contains no E2-binding sites (ACCN6GGT). Rabbit keratinocytes transfected with WT CRPV-E2, four CRPV mutant E2 expression constructs, HPV31-E2, and a construct encoding the DNA-binding–deficient E2 protein of CRPV together with the human MMP-9 promoter luciferase construct showed a 4-fold stimulation of the promoter by the WT E2 proteins, which was lower with the I73A, I73L, and R37A mutant but not significantly reduced with the mutant R37K (Fig. 4A). Interestingly, the DNA-binding–deficient mutant CRPV-E2 320/321 showed only reduced activation of the MMP-9 promoter, indicating that E2 does not need to bind to the promoter region of MMP-9 for its effect. A similar level of induction as was observed with the expression vector for HPV31 (Fig. 4A) could be found for HPV8-E2 and HPV6-E2 (data not shown). To analyze if the endogenous rabbit MMP-9 promoter is responsive to E2-mediated transactivation, we established stable rabbit keratinocyte lines that express either the WT CRPV-E2 protein or the I73A mutant and determined the level of endogenous rabbit MMP-9 transcripts in these cells employing quantitative real-time reverse transcription-PCR. This analysis clearly showed a 10-fold increase of MMP-9 transcripts in stable WT E2-containing cells compared with cells harboring the I73A mutant E2 or the empty expression vector (Fig. 4B). This suggests that the rabbit MMP-9 promoter within its genomic context is activated like the human MMP-9 promoter-reporter construct by the E2 protein.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Activation of the human MMP-9 promoter by WT and mutant CRPV-E2. A, CRL6502 cells were cotransfected with a luciferase reporter construct carrying the −670-bp MMP-9 promoter fragment (31) and expression vectors for WT CRPV-E2, the mutant constructs CRPV-E2/R37K, CRPV-E2/R37A, CRPV-E2/I73L, and CRPV-E2/I73A, a DNA-binding–deficient mutant E2 for CRPV (CRPV-E2-320/321), and WT HPV31-E2 (HPV31-E2). Fold activation of the MMP-9 promoter reporter construct (black columns) by these constructs in comparison with the cotransfection with the empty expression vector pSG5. Columns, mean of three independent experiments; bars, SD. *, P ≤ 0.05, compared with CRPV-E2 (Student's t test). B, measurement of MMP-9 transcripts with quantitative real-time PCR in total RNA extracted from permanent rabbit keratinocyte lines containing expression vectors for WT E2, the mutant E2 I73A, or just the empty expression vector pSG5. The amount of MMP-9-specific transcripts in 50 ng total RNA is given in relation to the amount of β-actin transcripts measured.

E2 activates the matrix metalloproteinase-9 promoter via activator protein-1 sites and induces cell invasiveness through activation of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase 1. It has been well described that the transcriptional activation of the MMP-9 promoter is highly dependent on AP-1 and NF-κB signaling pathways (31, 34). We therefore generated MMP-9 reporter constructs mutated in AP-1- and NF-κB-binding sites alone or in combination and used those together with a CRPV-E2 expression vector for transient reporter assays (Fig. 5). These experiments revealed a reduction of MMP-9 promoter activation by E2 when either the proximal or both the distal and the proximal AP-1 sites have been deleted (Fig. 5A). A deletion of the distal AP-1-binding site by itself had no effect. This suggests a requirement for an intact promoter-proximal AP-1-binding site for E2-mediated activation. A mutation of the NF-κB site by itself showed no loss of E2-mediated activation. To investigate if AP-1-binding sites are not only required but also sufficient to mediate E2 activation, we used reporter constructs with multimerized AP-1 or as a control multimerized NF-κB-binding sites for WT CRPV-E2 and the I73A mutant and tested those together with the WT E2 and the four mutant E2 constructs using transient reporter assays (Fig. 5B). WT E2 was able to stimulate reporter gene expression 8-fold when using multimerized AP-1-binding sites, whereas multimerized NF-κB sites were only marginally activated by WT CRPV-E2 and the mutant I73A. Whereas the mutant R37K transactivated multimerized AP-1 sites as good as WT E2, all other mutant E2 constructs revealed a strongly reduced ability. These data clearly show that WT E2 is able to stimulate promoters solely through the presence of multiple AP-1 sequence motifs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

A, activation of the WT and mutant MMP-9 promoter by CRPV-E2. Rabbit keratinocytes (CRL6502) were cotransfected with an expression vector for CRPV-E2 and luciferase reporter constructs carrying the WT (MMP-9) or mutant MMP-9 promoter constructs. The positions of binding sites for transcription factors NF-κB and AP-1 are indicated relative to the transcription start site. Fold activation of the different MMP-9 promoter reporter constructs (black columns) by CRPV-E2 in comparison with the cotransfection with the empty expression vector pSG5. Columns, mean of three independent experiments; bars, SD. *, P ≤ 0.05, compared with the WT promoter construct, MMP-9 (Student's t test). B, activation of two luciferase reporter constructs carrying multimerized AP-1- or NF-κB-binding sites, respectively. Rabbit keratinocytes (CRL6502) were cotransfected with expression vectors for WT CRPV-E2, CRPV-E2/I73L, CRPV-E2/I73A, CRPV-E2/R37K, and CRPV-E2/R37A and the AP-1 reporter construct (black columns) or expression vectors for WT CRPV-E2 and CRPV-E3/I73A and the NF-κB reporter construct (gray columns). Fold activation of the different reporter constructs by CRPV-E2 expression vectors in comparison with the cotransfection with the empty expression vector pSG5. Columns, mean of three independent experiments; bars, SD. C, a chemical inhibitor of MEK1 is able to block E2-induced Matrigel invasion of rabbit keratinocytes (CRL6502). The invasion assay was done as described in Fig. 3C. MEK1 inhibitor PD 098059 was added to medium at the indicated concentrations ranging from 0 to 50 μmol/L. *, P ≤ 0.05, Student's t test. ▪, AP-1; , NFkB.

E2 uses the ERK pathway for MMP-9 promoter activation. Finally, we addressed the question if the MAPK/ERK pathway as a known inducer of AP-1 is required for E2-dependent induction of MMP-9 expression and in vitro invasion. Thus, we did a Matrigel-based invasion/mobility assay, where we transiently transfected rabbit keratinocytes with CRPV-E2 and treated them afterward with increasing concentrations of PD 098059, which inhibits the phosphorylation of MAPK/ERK kinase 1 (MEK1), the immediate upstream activator of ERK1 (35). We observed a dose-dependent reduction of the E2-dependent in vitro invasion brought about by the selective inhibition of the ERK pathway (Fig. 5C). These data show that E2 probably causes activation of MEK/ERK signaling leading through stimulation of AP-1-binding sites to the induction of MMP-9 expression.

CRPV-E2 induces ERK1 expression. Considering that AP-1 is regulated through numerous signal transduction pathways, including the ERK cascade, we went back to study WT E2 and the very limited material available from mutant E2-induced rabbit tumor biopsies for the expression level and activation status of the ERK1/2 kinases known to be involved in the activation of AP-1. Interestingly, we found significant differences in both expression and phosphorylation status of ERK1 (Fig. 6). Activated forms of ERK2 were not detectable possibly due to the low expression level. Normal skin and papillomas induced by the I73A mutant E2 CRPV genome, shown before to contain low to undetectable amounts of MMP-9 transcripts and protein, contained less of the phosphorylated form of pERK1 in comparison with a WT E2-induced papilloma, a R37K CRPV genome-induced papilloma and a WT E2-induced carcinoma that were shown before to contain abundant levels of MMP-9. Because the basic expression of ERK1 was found to be not significantly different between WT E2-induced papillomas and normal skin, we did transient transfections with a WT E2 expression vector in rabbit keratinocytes and observed an induction of ERK1 expression in comparison with cells transfected with the empty expression vector. This finding was consistent with all previous results obtained and suggests that E2 induces MMP-9 expression possibly through altering the expression and activity of ERK kinase isoforms.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

ERK expression and ERK1 activity in different CRPV-induced skin lesions and transiently transfected rabbit keratinocytes. Western blot analysis with protein extracts derived from a papilloma and a carcinoma induced by WT CRPV genomes or from papillomas induced by CRPV genomes mutated within the E2 gene (R37K and I73A) was done using an antibody recognizing ERK1 and ERK2 isoform. M, a molecular weight marker. Immunoprecipitation/Western blot analysis of extracts shown in (A). Extracts were immunoprecipitated with a pERK1 antibody. Immunoprecipitates were separated by SDS-PAGE, blotted onto a membrane, and probed with the ERK1/2 antibody used in (A). CRL6502 cells were transfected using 250 ng CRPV-E2 expression vector (E2) or the empty vector (pSG5), Cell extracts were prepared, normalized for protein amount, and then used for Western blotting analysis with an antibody specific for ERK1. All experiments were done twice with identical results.

Discussion

In this study, we provide evidence for a novel function of the papillomavirus E2 protein. Our data show that E2 induces the expression and secretion of the MMP-9. Regulation of a cellular gene by E2 has been shown previously in HPV8-E2 that suppresses β₄ integrin expression in human keratinocytes through specific binding in the promoter region and displacement of cellular factors (36). In MMP-9, however, binding of E2 to the promoter region is not required for induction, which is mediated predominantly through the presence of an AP-1 site in close proximity to the transcription start site (31, 34). In contrast, expression of the MMP-2 gene, which lacks AP-1-binding sites in the promoter region, remains unaltered in rabbit epithelium by the infection with CRPV. Furthermore, a multimerized AP-1 site by itself was shown to be sufficient to allow E2-mediated activation of a minimal promoter. The stimulation of AP-1 activity by E2 seems to function through the MAPK/ERK pathway, as a chemical inhibitor of MEK1 (PD 098059) prevents E2-mediated infiltration of E2-transfected cells. This is further supported by the increased level of ERK1 expression in CRPV-E2-transfected rabbit keratinocytes.

All four mutant CRPV-E2 investigated show a dramatic loss of tumor induction efficiency in the rabbit, and the few papillomas that developed also revealed a deficiency in tumor progression. However, only three of the mutant CRPV-E2 constructs (I73A, I73L, and R37A) are no longer competent to induce MMP-9 expression, whereas the mutant R37K behaves like WT CRPV-E2 with regard to stimulation of artificial promoters, MMP-9 activation, induction of cell invasiveness using Matrigel-based assays, and stimulation of ERK activity in tumor cells. This indicates first that induction of MMP-9, which was shown by others to be involved in not only late-stage carcinogenesis but also in angiogenesis and tumor cell proliferation (15), is required but not sufficient for CRPV-induced tumorigenesis. Indeed, previous experiments with mice transgenic for HPV16 and additionally devoid of MMP-9 showed only a reduction of 50% in squamous cell carcinoma incidence in comparison with HPV16/MMP-9-proficient mice (15), which indicates that MMP-9 only partially contributes to epithelial carcinogenesis. It seems that functions of E2 that are lost by the mutant R37K rather do not play a role in tumor progression linked to MMP-9 but during other important events of tumor formation.

Activation of cellular transcription factors like AP-1 and NF-κB by viral proteins has also been described in the latent membrane protein-1 (LMP1) protein of the EBV, conferring to transfected C33A keratinocytes an invasive phenotype by increased MMP-9 expression, which is mediated mainly through the NF-κB and to a lesser extent through SP1 and AP-1 sites within the MMP-9 promoter (37). The transcription factor AP-1 plays a crucial role in various cellular processes, such as proliferation, wound healing, differentiation, and neoplastic transformation, as well as in the activation of papillomavirus genes through highly conserved AP-1 sites in the enhancer controlling early gene expression of papillomaviruses (38, 39).

Therefore, activation of the ERK/AP-1 signal pathway by E2 proteins of papillomaviruses or by LMP1 of EBV may not only contribute to the tumorigenic potential of these viruses but also could constitute in papillomaviruses an autocrine stimulation mechanism important for the very early steps following successful infection of the target cell. In CRPV, we showed before that epithelial stem cells in the hair follicle represent the primary target cells, which express immediately after infection high levels of viral transcripts encoding E6, E7, and E2 (40). In addition, wound healing is necessary for a successful infection with CRPV, which is known to cause a temporary up-regulation of AP-1 (c-fos) in the keratinocytes surrounding the wound (41). This could provide a favorable environment for the induction of the viral enhancer that depends strongly on AP-1 transcription factors. Whereas AP-1 expression usually ceases within 72 hours after the wounding event in uninfected keratinocytes (42), the combined expression of E6, E7, and especially E2, shown to be expressed at the earliest detectable stage of infection (40), could enforce sustained AP-1 activity (43), which enables MMP expression, thereby allowing the infected cells to migrate out of the hair follicle, and further provides a cellular milieu supporting continuous viral gene expression and epithelial tumor development.

These findings open the possibility that E2 is involved in modulating not only viral but also cellular gene expression. During the early stages of a papillomavirus infection, expression of E2 would create a tumor growth-favorable and in part autocrine-stimulating situation. In persistent infections, however, the permanent activation of AP-1 activates cellular proteins, like MMPs and calpain-2 (44), found previously by us to be up-regulated during tumor progression (14), which could contribute to increased cell motility and finally tumor invasion (44).