|Home | About | Journals | Submit | Contact Us | Français|
The papillomavirus E2 open reading frame encodes the full-length E2 protein as well as an alternatively spliced product called E8^E2C. E8^E2C has been best studied for the high-risk human papillomaviruses, where it has been shown to regulate viral genome levels and, like the full-length E2 protein, to repress transcription from the viral promoter that directs the expression of the viral E6 and E7 oncogenes. The repression function of E8^E2C is dependent on the 12-amino-acid N-terminal sequence from the E8 open reading frame (ORF). In order to understand the mechanism by which E8^E2C mediates transcriptional repression, we performed an unbiased proteomic analysis from which we identified six high-confidence candidate interacting proteins (HCIPs) for E8^E2C; the top two are NCoR1 and TBLR1. We established an interaction of E8^E2C with an NCoR1/HDAC3 complex and demonstrated that this interaction requires the wild-type E8 open reading frame. Small interfering RNA (siRNA) knockdown studies demonstrated the involvement of NCoR1/HDAC3 in the E8^E2C-dependent repression of the viral long control region (LCR) promoter. Additional genetic work confirmed that the papillomavirus E2 and E8^E2C proteins repress transcription through distinct mechanisms.
Papillomaviruses (PVs) infect the squamous epithelia, inducing proliferative lesions. The PV life cycle is tightly linked with the squamous cell differentiation program. In particular, late viral gene synthesis, genome amplification, and virion production occur with the progression of epithelial cell differentiation.
The PVs are small DNA viruses that contain approximately 10 open reading frames (ORFs). Like those of many viruses, PV regulatory proteins are multifunctional. The E2 ORF encodes several viral products (21). The full-length E2 protein is the most studied; it is a prototypic transcription factor, with a C-terminal DNA binding and dimerization domain, an internal hinge region, and an N-terminal “transactivation” domain (34). Initially, E2 was characterized as a transcriptional activator, and subsequently, it was shown to be a potent repressor of the long control region (LCR), which is the viral promoter that controls E6 and E7 expression (21). In addition to its transcriptional activities, E2 functions in concert with the viral E1 helicase to initiate replication of the viral DNA (21). Also, the E2 protein is involved in tethering viral genomes to the host chromatin for maintenance during mitosis (23, 50). Both the C-terminal portion of E2 (E2C), which mediates dimerization and recognition of consensus E2-binding sequences (ACCN6GGT) within the viral LCR, and the N terminus of E2 are required for its various transcription, replication, and genome maintenance activities (9, 13, 21, 33, 35, 38, 45, 64).
In addition to full-length E2, PVs encode an alternatively spliced product of E2 called E8^E2C. This protein contains 12 amino acids from the E8 ORF fused to the C-terminal DNA binding/dimerization domain of E2 (E2C). To date, E8^E2C transcripts have been detected for bovine papillomavirus type 1 (BPV-1) and human papillomavirus types 11, 16, 31, and 33 (HPV-11, -16, -31, and -33); the short segment from E8 exhibits high conservation among the different HPVs for which it has been described and shows partial conservation with the BPV-1 E8 domain (5, 8, 44, 54).
Early studies demonstrated that the N-terminal domain of E2 is necessary for transcriptional repression. Evidence for this came from comparative studies with full-length BPV-1 E2 (E2TA) and a second, truncated BPV-1-specific E2 repressor (E2TR), which is translated from an alternative initiating methionine residue and therefore lacks most of the N-terminal transactivation domain (TAD) (29). Assessment of the transcriptional repression functions of the E2TA and E2TR proteins revealed that those of E2TR were impaired compared to that of full-length E2TA (9). Later studies confirmed these results through the demonstration that amino acid substitution within the N-terminal domain impaired E2-dependent repression activity (9, 13, 38). Interestingly, although they lack the N-terminal domain of full-length E2, which is required for the repression function, the HPV-16 and -31 E8^E2C proteins are able to repress transcription and viral genome replication from the LCR (27, 54-56, 67). Furthermore, the repression activity has been mapped to the E8 ORF, which also has been shown to be a transferable repression domain (55, 56, 67).
HPVs are classified as low or high risk depending on the clinical lesions with which they are associated and the potential for these lesions to progress to cancer. The oncogenic potential of the high-risk HPVs is determined by their E6 and E7 proteins, which inactivate p53 and Rb, respectively (10, 21, 46, 60). Both E2 and E8^E2C are able to induce irreversible senescence in HeLa cells through repression of E6/E7 transcription from the LCR promoter, resulting in the renewal of the cellular p53 and Rb tumor suppression pathways (3, 7, 9, 12, 55, 59). The integration of the HPV genome into the host chromosomes in a manner that disrupts the E2 ORF, and the resulting increase in E6 and E7 expression, is often an important step in cancer progression (4, 43, 47).
To date, little is known about the repression mechanism by which E8^E2C functions. The use of several histone deacetylase complex (HDAC) inhibitors implicated class I HDACs in E8^E2C-mediated transcriptional repression (3). Furthermore, HDAC3 was shown to interact with the minimal repression domain of the wild-type E8 ORF (3). HDAC3 is the core enzymatic component of nuclear receptor corepressor 1 (NCoR1) transcriptional silencing complexes (31, 63). Cellular nuclear and steroid receptors typically interact with the C terminus of NCoR1, while HDAC3 is recruited through an N-terminal NCoR1 SANT domain, otherwise known as the deacetylase activation domain (DAD) (6, 18, 19, 31, 32). NCoR1 repression complexes also contain other proteins, such as TBL1, TBLR1, and GPS2 (14, 20, 63, 65).
In order to elucidate the mechanism of E8^E2C transcriptional repression, we employed a proteomic strategy utilizing the Comparative Proteomic Analysis Software Suite (CompPASS) to identify interacting cellular proteins for repression-competent and repression-impaired E8^E2C proteins. Through this unbiased proteomic strategy, we identified six high-confidence candidate interacting proteins (HCIPs) unique to wild-type (WT) E8^E2C, two of which were NCoR1 and TBLR1. We showed that the NCoR1/HDAC3 complex interacts with WT E8^E2C but not with mutants with defective repression. Through genetic studies with small interfering RNA (siRNA), we confirmed that NCoR1/HDAC3 complexes are utilized by E8^E2C, but not by the full-length E2 protein, for its repression function. Conversely, genetic work demonstrated that several cellular repressors determined in our laboratory to be involved in E2-specific transcriptional repression are not engaged in E8^E2C-mediated repression. In addition to the elucidation of the E8^E2C transcriptional repression mechanism, we provided evidence that distinct viral repression pathways are utilized by the different E2 ORF-encoded gene products. The identification of nonoverlapping repression pathways suggests that viral transcription may be regulated differentially as the abundance of cellular and virally encoded factors involved in these processes change during the squamous epithelial cell differentiation process that regulates the viral life cycle.
Untagged pSG5-based (Stratagene) expression plasmids for HPV-31 WT E8^E2C, KWK-E8^E2C, and d3-12-E8^E2C have been described previously (54, 56). HPV-31 E2 expression plasmids were constructed by PCR amplification of the HPV-31 E2 ORF (nucleotides [nt] 2693 to 3808) from an HPV-31a genomic construct, pLit31, kindly provided by Jason Bodily and Lou Laimins, which was then inserted into the empty pSG5 vector to create pSG5-31E2 (p6108). The pSG5-31E2(HA) (p6109), pSG5-WT-E8^E2C(HA) (p6110), pSG5-KWK-E8^E2C(HA) (p6111), and pSG5-d3-12-E8^E2C(HA) (p6112) constructs were generated by multisite QuikChange mutagenesis (Stratagene) with complementary forward (5′-CCACCACATCGAATTACCCATACGATGTTCCAGATTACGCTTCCAAAACCTGCGCC-3′) and reverse (5′-GGCGCAGGTTTTGGAAGCGTAATCTGGAACATCGTATGGGTAATTCGATGTGGTGG-3′) primers inserted in order to disrupt the EcoRI site found within the hinge domain. The pGL4.20-HPV18LCR-luc construct (referred to below as 18LCR-luc) (p5194) has been described previously (48). The BPV-1 E2TA (p760) and E2TR (p1153) expression plasmids, as well as the parental pCDV1 plasmid (p751), have been described previously (9, 29, 39, 62). pEGFP-N1 (Clontech) was used as a transfection control.
Stable C33A/pSG5(EV), C33A/pSG5-31E2(HA), C33A/pSG5-WT-E8^E2C(HA), C33A/pSG5-KWK-E8^E2C(HA), and C33A/pSG5-d3-12-E8^E2C(HA) pools were generated by cotransfecting NdeI-linearized pSG5 vectors with the empty puromycin-resistant pBABE-puro vector (36) at a 10:1 ratio. Stable C33A/18LCR-luc/16E2-c1 and C33A/18LCR-luc/16E2-c5 cells have been described previously (51). Clonal, stable C33A/18LCR-luc/E8^E2C-c14 and C33A/18LCR-luc/E8^E2C-c8 reporter lines were generated by cotransfecting NotI-linearized puromycin-resistant pGL4.20-18LCR-luc with NdeI-linearized pSG5-WT-E8^E2C at a 1:10 ratio. The HeLa and C33A cell lines were grown in high-glucose Dulbecco's modified Eagle's medium (hgDMEM; Invitrogen) supplemented with 10% standard fetal bovine serum (FBS; HyClone) and 1% penicillin-streptomycin (GIBCO/Invitrogen). All stable C33A cell lines were grown in the presence of 0.75 μg/ml puromycin (Sigma).
Large-scale anti-hemagglutinin (anti-HA) immunoprecipitations (IPs) were performed with stable C33A/pSG5-based expression pools for proteomic identification of HCIPs. Briefly, 4- by 15-cm plates of 85% confluent stable pools were harvested in NP-40 lysis buffer (0.5% NP-40 substitute, 1 mM EDTA, 100 mM NaCl, 50 mM Tris-HCl [pH 8.0]) supplemented with Complete Mini EDTA-free protease inhibitor tablets (Roche). Extracts were sonicated at 35% intensity for 8 s with a Bronson digital sonifier and were clarified for 20 min at 17,000 × g, and the soluble fraction was incubated with anti-HA-agarose resin (Sigma) overnight at 4°C. IPs were washed five times with NP-40 lysis buffer and were then eluted three times, for 30 min each time, at room temperature with 500 μg/ml HA peptide (Sigma). The eluted proteins were concentrated by trichloroacetic acid (Sigma) precipitation, followed by acetone (catalog no. 270725; Sigma) washes as described previously (52). Small-scale anti-HA IPs were performed similarly, except that a single 10-cm-diameter plate of cells was used; the protein content for each IP was normalized; and bound proteins were boiled off in 1× sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer. In order to achieve equivalent input levels of each HA-tagged protein despite differences in E8^E2C expression levels, the KWK- and d3-12-E8^E2C extracts were diluted with a C33A/pSG5(EV) extract prior to the IP.
Equal amounts of protein, as determined with a bicinchoninic acid (BCA) protein assay kit (Pierce/Thermo Scientific) according to the manufacturer's instructions, were separated by SDS-PAGE. Primary antibodies against actin (MAB1501; Millipore), Brd4 (one antibody used by Schweiger et al.  and one from Bethyl Laboratories [A301-985A]), EP400 (A300-541A; Bethyl Laboratories), HA (catalog no. 12013819001; Roche), HDAC1 (ab7028; Abcam), HDAC2 (ab7029; Abcam), HDAC3 (ab7030; Abcam), NCoR1 (A301-145A; Bethyl Laboratories), SETDB1/KMT1E (ab5420; Abcam), SMCX (NB100-55328; Novus Biologicals), and TRIM28/KAP1 (ab10484; Abcam) were used in these experiments. Bound antibodies were detected with horseradish peroxidase-conjugated donkey anti-rabbit secondary antibodies (NA934V; Amersham/GE Healthcare). For actin detection, an Alexa Fluor 680-conjugated goat anti-mouse IgG secondary antibody (A21059; Molecular Probes/Invitrogen) was used prior to visualization using an Odyssey infrared imaging system (LI-COR).
The eluted interacting proteins for IPs using C33A/pSG5(EV), C33A/pSG5-WT-E8^E2C(HA), C33A/pSG5-KWK-E8^E2C(HA), and C33A/pSG5-d3-12-E8^E2C(HA) stable pools were characterized by liquid chromatography-tandem mass spectrometry (LC-MS-MS) in duplicate runs according to previously published protocols (52). Peptides and their corresponding proteins were determined by using Sequest and a target-decoy strategy and were subsequently analyzed using CompPASS in order to identify HCIPs. The total spectral counts (TSC) for each interactor were used to generate raw statistical measures, called Z scores, for each interaction. Normalized D scores (DN scores) are Z scores that have been adjusted to account for the discovery rate, or the percentage of all CompPASS-analyzed immunoprecipitations in which an interactor has been discovered, and the reproducibility of detection within the duplicate injections onto the LC-MS-MS. A DN score of ≥1 and a Z score of ≥3.5 are thresholds above which an interactor is considered a HCIP (52). A final comparative step was performed to identify HCIPs unique to WT E8^E2C (not present in the KWK and d3-12 mutant E8^E2C proteins). Interactome maps of the identified WT E8^E2C-specific HCIPs were generated by cross-referencing reported protein-protein interactions (PPI) found within the MINT, BioGRID, and STRING public databases.
Extracts were harvested using 1× reporter lysis buffer (Promega), subjected to freeze-thawing at −80°C, vortexed, and centrifuged for 5 min at 14,000 × g to recover the soluble material. Extract concentrations were quantified by a BCA protein assay (Pierce/Thermo Scientific) and measured by a SpectraMax 190 microplate reader (Molecular Devices) according to the manufacturer's specifications. Luciferase measurements were obtained by mixing the extract with the luciferase assay reagent (Promega), followed by measurement with an LMax luminometer (Molecular Devices). For each extract, the relative luciferase units (RLUs) were normalized to the extract's protein concentration (in milligrams per milliliter). Fold activation was determined by a final normalization to empty-vector conditions (for DNA transfections) or siRNA control (SiC) conditions (for siRNA knockdowns).
Transient luciferase reporter assays were performed by seeding 8 × 104 C33A cells in hgDMEM-10% FBS in a 12-well plate format. A 2:1 ratio of Fugene 6 (Roche) was used to cotransfect 20 ng of the indicated pSG5 expression plasmid, 100 ng of the pGL4.20-18LCR-luc reporter, and 880 ng of pSG5(EV) per 12-well plate. The medium was changed 24 h posttransfection, and the transfection efficiency was monitored with the pEGFP-N1 control plasmid. Extracts were harvested 48 h posttransfection for BCA and luciferase assays. For siRNA experiments, a reverse transfection protocol utilizing Dharmafect 2 (Dharmacon/Thermo Scientific) was used to deliver 20 nM each Dharmacon siRNA to 4 × 104 C33A/18LCR-luc/WT-E8^E2C-c14, C33A/18LCR-luc/E8^E2C-c8, C33A/18LCR-luc/16E2-c1, or C33A/18LCR-luc/16E2-c5 cells. In brief, siRNAs were diluted in 1× siRNA buffer (Dharmacon/Thermo Scientific) and were then added to 12-well plates. Dharmafect 2 was preincubated with Opti-MEM (GIBCO/Invitrogen) and added to each well. The siRNAs were complexed with Dharmafect 2/Opti-MEM for 20 min, after which 4 × 104 reporter cells were added to each well. Two HPV-31 E2C-specific siRNAs were generated with the following sense strand sequences: 5′-GGACATGTACAGATGGAAA-3′ (E2C-1) and 5′-CTTAAAAGGTGATGCAAAT-3′ (E2C-3). Dharmacon siGENOME nontargeting siRNA 1 (SiCONTROL [SiC]; catalog no. D-001210-01; Dharmacon/Thermo Scientific) was used as a transfection control. The HPV-16 E2-specific siRNA (16E2-2) has been described previously (51). Dharmacon/Thermo Scientific ON-TARGETplus individual siRNA duplexes for arginase-1 (ARG1) (LU-009922-00), bleomycin hydrolase (BLMH) (LU-005793-00), caspase-14 (CASP14) (LU-004403-00), HDAC1 (LU-003493-00), HDAC2 (LU-003495-00), HDAC3 (LU-003496-00), NCoR1 (LU-003518-00), SETDB1 (LU-020070-00), TBLR1 (LU-012927-00), transglutaminase 3 (TGM3) (LU-010088-01), and TRIM28 (LU-005046-00) were used. Individual Dharmacon/Thermo Scientific siGENOME siRNA duplexes for Brd4 (D-004937-02/05), EP400 (D-021272-02/03), and SMCX/JARID1C (D-010097-02/04) were used. The medium was changed 24 h posttransfection, and transfection efficiency was monitored by siGLO RISC-free control siRNA (D-001600-01; Dharmacon/Thermo Scientific). Extracts were harvested 72 h posttransfection for BCA protein assays, luciferase assays, and immunoblotting. Each experiment was performed in triplicate.
A total of 3.5 × 105 HeLa cells were seeded into 60-mm-diameter plates in hgDMEM-10% FBS. After overnight incubation at 37°C, cells were cotransfected with 2 μg of the indicated pSG5 plasmid along with 1 μg of puromycin-resistant pGL4.20. The following day (24 h posttransfection), cells were split, with 2.5 × 105 cells in a 10-cm-diameter plate. Beginning 48 h posttransfection, the medium was changed to hgDMEM-10% FBS supplemented to contain 0.4 μg/ml puromycin; this medium was replaced every 2 days. On day 11 posttransfection, colonies were fixed for 1 h at room temperature with 4% formalin in Dulbecco's phosphate-buffered saline (DPBS), stained with 0.1% crystal violet for 15 min at room temperature, and then counted manually. The extent of growth suppression for HPV-31 E2 or E8^E2C was determined by normalization to pSG5(EV), while BPV-1 E2TA and E2TR were normalized to their empty parental plasmid, pCDV1.
In order to identify cellular proteins involved in E8^E2C repression, we designed a proteomic strategy using affinity purification of proteins associated with wild-type and mutant forms of E8^E2C proteins that contained a single HA tag imbedded within their hinge regions (Fig. (Fig.1A).1A). The KWK and d3-12 mutant forms of E8^E2C have been described previously; they contain a triple point mutation and a 10-amino-acid deletion of the E8 open reading frame, respectively (Fig. (Fig.1A).1A). Both the KWK and d3-12 mutants have impaired repression activity (55, 56, 67).
Previous studies have established that insertion of tags within the hinge region does not interfere with the E2 or E8^E2C functions (55). Therefore, we engineered expression plasmids with HA tags within the hinge region and verified that the tag did not affect the activities of the individual E2 and E8^E2C constructs by using transient luciferase reporter assays. In these experiments, we examined the abilities of the various E8^E2C proteins to repress a heterologous HPV-18 LCR-luciferase reporter construct (18LCR-luc) in C33A cells, which are HPV-negative cervical cancer cells. These luciferase reporter assays confirmed that WT E8^E2C was repression competent, whereas the KWK and d3-12 forms had impaired repression activity (Fig. (Fig.1B).1B). Furthermore, these experiments demonstrated that the respective repression phenotypes of wild-type and mutant E8^E2C proteins were not affected by the internal HA tag (Fig. (Fig.1B1B).
Next, we performed colony reduction assays to assess the abilities of these tagged HPV-31 E2 and E8^E2C proteins to repress the expression of the HPV-18 oncogenes in HeLa cells. BPV-1 E2TA and E2TR were included in these experiments as controls, because E2TR, a BPV-1-specific truncated form of E2 that lacks most of the N-terminal domain, has an impaired repression function (9). As expected, we observed strong growth suppression with full-length BPV-1 E2 (E2TA) and moderate suppression with full-length HPV-31 E2 (Fig. (Fig.1C).1C). Interestingly, the WT E8^E2C protein produced robust growth suppression to an extent even greater than that of the BPV-1 or HPV-31 full-length E2 protein. In addition, the mutant KWK and d3-12 E8^E2C proteins showed growth suppression activities that were impaired relative to that of WT E8^E2C and were similar to that observed for BPV-1 E2TR. The presence of the HA tag did not affect the growth suppression phenotype for either wild-type or mutant E8^E2C proteins.
Before performing the proteomic analysis to identify interacting partners of the functionally tagged E8^E2C proteins, we assessed the ability of the internal affinity tag to be immunoprecipitated in vivo. To this end, we generated stable C33A cell lines expressing either the empty vector (EV) or the HA-tagged WT, KWK, or d3-12 E8^E2C protein. Stable pools of expressing cells were used for a large-scale, single anti-HA immunoprecipitation (IP). In this procedure, equal amounts of protein extracts were incubated overnight with anti-HA agarose resin. A small amount of preelution resin was analyzed via anti-HA immunoblotting to assess the efficiency of the IP. These immunoblots confirmed that the internal HA tag was recognizable in vivo and that each of the E8^E2C proteins could be immunoprecipitated proportionately to its relative input expression level (Fig. (Fig.1D1D).
Successful immunoprecipitation of the various internally HA tagged E8^E2C proteins confirmed that stable C33A expression cells could be used to identify interacting cellular proteins that may be involved in the E8^E2C repression phenotype. In order to identify those cellular proteins that interacted with WT E8^E2C, but not with the repression-impaired KWK- or d3-12-E8^E2C mutant, we eluted all bound proteins from the large-scale immunoprecipitation (Fig. (Fig.1D)1D) with HA peptide. The eluate for each IP was analyzed using previously described mass spectrometry protocols and the Comparative Proteomic Analysis Software Suite (CompPASS) in order to determine high-confidence candidate interacting proteins (HCIPs) for each bait protein (52). The removal of identified background proteins and interactors found to be associated with repression-impaired mutant E8^E2C proteins (see Table S1 in the supplemental material) produced a list of six HCIPs unique to WT E8^E2C (with normalized D scores [DN scores] given in parentheses): NCoR1 (22.00), TBLR1 (15.14), ARG1 (4.76), BLMH (2.52), TGM3 (1.68), and CASP14 (1.43) (Fig. (Fig.2A).2A). DN- versus Z-score plots were generated for all WT E8^E2C interactors and demonstrated the highest significance for NCoR1 and TBLR1 interactions (Fig. (Fig.2A).2A). In addition to not interacting with KWK- or d3-12-E8^E2C, NCoR1 and TBLR1 were not identified as interactors for any of the other 106 proteins comprising the database against which the E8^E2C IP-MS-MS data were analyzed (see Table S1 in the supplemental material).
We next examined whether any of the six HCIPs specific to WT E8^E2C that we identified had a functional role in repression of the papillomavirus LCR. To address this question, C33A cells were stably transfected with a previously described HPV-18 LCR-controlled luciferase reporter (18LCR-luc) and pSG5-WT-E8^E2C(HA) (48). We chose to use the HPV-18 LCR-controlled reporter in C33A cells because this cell line is known to support both E2- and E8^E2C-dependent repression of the HPV-18 LCR (Fig. (Fig.1B)1B) (40). Single clones were isolated and characterized. Of those with observable WT E8^E2C expression, we selected the C33A/18LCR-luc/E8^E2C-c14 cell line for further genetic studies because the integrated 18LCR-luc reporter was repressed by the E8^E2C protein, as revealed by the increase in luciferase activity following treatment with HPV-31 E2C-specific siRNA 3 (E2C-3) over that with the siGENOME nontargeting siRNA 1 control (SiC) (Fig. (Fig.33).
In addition, the C33A/18LCR-luc/E8^E2C-c14 cells were transfected with four individual siRNA duplexes targeting each of the six HCIPs identified. Cell lysates were harvested 72 h posttransfection to determine the luciferase and protein levels. Each of the four individual siRNA duplexes against NCoR1 significantly increased luciferase activity, although not to the same extent as the positive-control E2C-3 siRNA (Fig. (Fig.3).3). We observed a correlation between the extent of NCoR1 knockdown and the luciferase activity observed in cells treated with three of the four NCoR1-specific siRNAs (Fig. (Fig.3).3). The alleviation of LCR repression with NCoR1 knockdown occurred even in the presence of higher levels of E8^E2C. This indicates that the knockdown of NCoR1 compromises E8^E2C-mediated transcriptional repression.
In contrast to the effects observed with NCoR1-specific siRNAs, introduction of siRNAs targeting the other five HCIPs did not increase luciferase activity over that with SiC (Fig. (Fig.3).3). While many of the HCIPs do not appear to be involved in transcriptional repression of the LCR, this result does not exclude the possibility that the HCIPs may be involved in other E8^E2C functions, such as the regulation of DNA replication.
Due to the potential role of NCoR1 in E8^E2C-mediated repression of the LCR, we wanted to confirm the specificity of the NCoR1 interaction with the E8 open reading frame. In addition, we asked whether a specific interaction of HDAC3, the core catalytic component of NCoR1 silencing complexes, could be detected in complex with E8^E2C (31, 63). Because the initial proteomic work only compared interaction profiles among repression-competent and -impaired E8^E2C proteins, we extended this analysis to include the full-length E2 protein. Anti-HA IPs were performed using stable E8^E2C- or E2-expressing C33A cell lines. We confirmed that, in agreement with the mass spectrometry data, NCoR1 interacts with WT E8^E2C, but not with the KWK or d3-12 mutant form of E8^E2C (Fig. (Fig.4).4). Also, HDAC3 was found to bind exclusively to the WT E8^E2C protein; this is consistent with earlier published findings (3). In these experiments, NCoR1 and HDAC3 did not interact with full-length E2 protein. Collectively, these data suggest that NCoR1 complexes are recruited to E8^E2C through interactions that are mediated by the 12 amino acids encoded by the E8 ORF.
Given that both NCoR1 and HDAC3 specifically interacted with WT E8^E2C, we asked whether other cellular proteins recruited to NCoR1 silencing complexes might also play a role in E8^E2C-mediated transcriptional repression. Although HDAC3 is the core histone deacetylase of NCoR1 repression complexes, NCoR1 has been reported to recruit HDAC1/HDAC2 through interactions with Sin3 proteins (2, 16, 17, 28, 37). Previous work has shown that nonspecific class I HDAC inhibitors partially alleviate transcriptional repression by E8^E2C (3). Furthermore, TRIM28 and SETDB1 had been identified as potential interactors of WT E8^E2C in specific glutathione S-transferase (GST) and protein A pulldown experiments (3). Although TRIM28 and SETDB1 have not been implicated in E8^E2C transcriptional repression (3), TRIM28 is known to be recruited to NCoR1/HDAC3 complexes (20, 57). Furthermore, SETDB1 is an H3K9 demethylase recruited by TRIM28 to mediate chromatin condensation (20, 53).
To examine the potential roles of these individual genes in E8^E2C-mediated transcriptional repression of the HPV-18 LCR, C33A/18LCR-luc/E8^E2C-c14 cells were transfected with two different siRNAs against each target gene. Extracts were harvested 72 h posttransfection to determine protein concentrations and luciferase activity. We observed increases in luciferase activity in cells transfected with HPV-31 E2C-specific siRNAs 1 and 3 (E2C-1 and E2C-3) over that with the SiC transfection control (Fig. (Fig.5).5). Similarly, a 2- or 5-fold increase in luciferase activity was observed with knockdown of either HDAC3 or NCoR1, respectively. However, knockdown of HDAC1, HDAC2, SETDB1, or TRIM28 did not alleviate LCR repression in this assay. Furthermore, we repeated this siRNA knockdown experiment in a second reporter cell line, C33A/18LCR-luc/E8^E2C-c8, in order to rule out possible positional effects of LCR integration within any one given clonal cell line used for experimentation. The C33A/18LCR-luc/E8^E2C-c8 cell line confirmed the specific role of HDAC3 and NCoR1 repression of the viral LCR (see Fig. S1 in the supplemental material).
To examine whether NCoR1/HDAC3 repression of the LCR is E8^E2C dependent, or whether NCoR1/HDAC3 might also have a general or full-length E2-specific role in LCR repression, the individual siRNAs targeting NCoR1 and HDAC3 were examined for their abilities to relieve the repression of the 18LCR-luc reporter in a cell line expressing the full-length HPV-16 E2 protein. As expected, the positive-control siRNA targeting HPV-16 E2 and the siRNAs targeting the previously identified E2-dependent cellular repressors Brd4, EP400, and SMCX (51, 61) resulted in 18-, 8-, 9-, and 8-fold activations of luciferase activity, respectively (Fig. (Fig.6).6). However, no increase in luciferase activity was observed in C33A/18LCR-luc/16E2-c1 cells with an NCoR1 or HDAC3 knockdown (Fig. (Fig.6).6). The siRNAs specific for HPV-31 E2C (E2C-1 and E2C-3) also failed to impact HPV-16 E2 protein levels or luciferase activity. This is consistent with our finding that NCoR1 and HDAC3 did not interact with full-length E2 (Fig. (Fig.44).
Again, to ensure that this finding was not unique to this particular HPV-16 E2-expressing cell line, this siRNA experiment was repeated in a second clone, C33A/18LCR-luc/16E2-c5. As shown in Fig. S2 in the supplemental material, NCoR1 and HDAC3 knockdowns had no effect on luciferase activity in this alternative E2 reporter cell line, in contrast to the positive-control 16E2-2 siRNA and the Brd4-05, EP400-03, and SMCX-02 siRNAs. Therefore, NCoR1/HDAC3 complexes do not appear to be involved in repression mediated by the full-length E2 protein. Together, these data indicate that NCoR1 and HDAC3 are involved specifically in E8^E2C-mediated repression of the LCR.
Given the specific utilization of NCoR1/HDAC3 in E8^E2C-mediated repression of the LCR, we examined next whether cellular proteins that are implicated in LCR repression by full-length E2 might also be engaged by E8^E2C. For this experiment, we examined Brd4, EP400, and SMCX, which we have determined to be involved in E2-dependent transcriptional repression of the LCR (51). siRNA knockdowns against these cellular targets were performed in the C33A/18LCR-luc/E8^E2C-c14 reporter cell line. While we observed an increase in luciferase activity with the positive-control HPV-31 E2C-specific siRNAs over that with SiC transfection, we did not observe activation of luciferase activity with either Brd4 or SMCX knockdown (Fig. (Fig.7).7). These data indicate that Brd4 and SMCX are engaged exclusively by full-length E2 to mediate repression of the LCR.
In contrast to the findings for these two target genes, we did observe a modest increase in luciferase activity with one of the siRNAs targeting EP400 (Fig. (Fig.7).7). However, when we repeated this siRNA knockdown experiment with a second cell line, C33A/18LCR-luc/E8^E2C-c8, we failed to detect an effect of the EP400-03 siRNA in alleviating E8^E2C repression of the LCR (see Fig. S3 in the supplemental material). We detected a slight, reproducible effect of the Brd4-05 siRNA in the C33A/18LCR-luc/E8^E2C-c8 cell line (see Fig. S3 in the supplemental material). In neither E8^E2C reporter cell line did we observe a role for SMCX in E8^E2C-mediated repression (Fig. (Fig.7;7; see also Fig. S3 in the supplemental material). Collectively, these data suggest that E8^E2C does not utilize the previously identified E2-dependent cellular repression factors.
The papillomavirus E2 ORF is a critical regulatory gene that is involved in the viral processes of DNA replication, genome maintenance, and transcription (21). In addition to the full-length protein encoded by the E2 ORF, the alternative splice variant E8^E2C also mediates transcriptional repression of the viral promoter (55, 56). Previous studies have suggested that E2 and E8^E2C might repress by engaging different cellular repressor complexes, because their activities have been mapped to their disparate N-terminal domains (9, 13, 55, 56, 67). The proteomic studies presented in this article were designed to identify the cellular factors that mediate E8^E2C repression and to test whether E2 and E8^E2C function through the engagement of different cellular factors.
Six HCIPs for HPV-31 WT E8^E2C were identified in the proteomic analysis. The top two HCIPs, NCoR1 and TBLR1, form a well-characterized transcriptional corepression complex whose catalytic activity is mediated through NCoR1 recruitment and activation of HDAC3 (6, 20, 31, 63). The identification of a WT E8^E2C interaction with NCoR1 complexes through this unbiased proteomic approach is consistent with previously published work suggesting HDAC3 involvement in E8^E2C repression (3).
An interactome map of the identified WT E8^E2C-specific HCIPs was generated. The interactome indicated networking among four of the six HCIPs (Fig. (Fig.2B,2B, yellow squares). Level 1 interactors, proteins that associate with at least two of the HCIPs, are included in the interactome map (Fig. (Fig.2B).2B). Although caspase-14 (CASP14) and transglutaminase 3 (TGM3) were not found in this interactome analysis, a recent study reported that these proteins function sequentially in a pathway with bleomycin hydrolase (BLMH) to mediate proteolytic processing events in terminally differentiating keratinocytes (25). Thus, the clustering of these other HCIPs in a pathway important for keratinocyte differentiation, and hence the viral life cycle, suggests that these proteins may be important for other E8^E2C functions, such as DNA replication control.
Alternatively, the identified HCIPs may modulate E8^E2C stability. Although siRNA knockdown of BLMH, CASP14, and TGM3 did not alleviate repression of the LCR, it did result in a slight increase in steady-state E8^E2C levels (Fig. (Fig.3).3). Specifically, we observed an increase in steady-state E8^E2C protein levels with two of the four BLMH siRNAs, two of four CASP14 siRNAs, one of the four arginase-1 (ARG1) siRNAs, and two of four TGM3 siRNAs (Fig. (Fig.3).3). This may suggest that E8^E2C stability is partially dependent on association with the identified high-confidence candidate interaction proteins. Therefore, these HCIPs may modulate E8^E2C stability in vivo. Interestingly, CASP14 has been reported to be lost with the progression of cervical cancer and therefore is a potentially interesting gene for future investigation of HPVs and cancer (26). Along these lines, three of the four siRNA duplexes targeting NCoR1 caused increases in steady-state E8^E2C levels (Fig. (Fig.3).3). This may suggest that E8^E2C stability is partially dependent on association with NCoR1. This would be consistent with the noticeably higher steady-state levels of the KWK- and d3-12-E8^E2C proteins, which did not interact with NCoR1 (Fig. (Fig.1D1D).
Our data indicate that specifically HDAC3, and not HDAC1 or HDAC2, is involved in the repression of the LCR. Interestingly, NCoR1 siRNAs induced a greater activation of luciferase activity in the C33A/18LCR-luc/E8^E2C-c14 reporter line than did knockdown of the core catalytic component HDAC3 (Fig. (Fig.5).5). This result is consistent with the partial relief of the E8^E2C repression function by chemical HDAC inhibitors (3) and the dual repression function of NCoR1. In vitro experiments show that, in addition to the recruitment and activation of HDAC3, NCoR1 can bind directly to hypoacetylated histones and is thus postulated to spread along the chromatin to reinforce the localized silencing in a feed-forward mechanism of repression (15, 58, 63).
Despite the fact that TBLR1 is a known component of NCoR1 repression complexes, its knockdown in the secondary functional assays did not relieve E8^E2C-mediated repression in the C33A/18LCR-luc/E8^E2C-c14 cell line (Fig. (Fig.3).3). Such a result might be explained by the presence of a redundant homologue, called TBL1, or the fact that both of these adaptor proteins mediate the turnover of the NCoR1 repression complexes in response to activated kinase pathways (22, 24, 41, 42).
Confirmatory anti-HA IPs and Western blotting showed that NCoR1/HDAC3 complexes specifically interact with WT E8^E2C, not with mutant forms of E8^E2C whose repression functions are deficient (Fig. (Fig.4).4). In addition, NCoR1/HDAC3 failed to interact with full-length E2, which is consistent with the specificity for E8 (Fig. (Fig.4).4). This specificity was corroborated by subsequent genetic work in which siRNA knockdown of NCoR1/HDAC3 alleviated repression of the LCR only in the presence of E8^E2C, not in the presence of full-length E2 (Fig. (Fig.55 and and6;6; see also Fig. S1 and S2 in the supplemental material). siRNA knockdown experiments targeting NCoR1, HDAC3, and cellular factors identified for the repression function of the full-length E2 protein confirmed the largely nonoverlapping pathways of LCR repression by E8^E2C and E2.
The modest effect seen with the EP400-03 siRNA in the C33A/18CLR-luc/E8^E2C-c14 reporter line was not observed in the C33A/18LCR-luc/E8^E2C-c8 cell line, and conversely, the Brd4-05 siRNA effect seen in the C33A/18LCR-luc/E8^E2C-c8 cell line was not detected in the C33A/18LCR-luc/E8^E2C-c14 reporter line (Fig. (Fig.7;7; see also Fig. S3 in the supplemental material). The nonreproducible, subtle effect of Brd4-05 or EP400-03 siRNA knockdown in one of the two E8^E2C cell lines suggests that these two cellular factors are not involved in E8^E2C-mediated repression. Furthermore, siRNA knockdown of SMCX showed no alleviation of E8^E2C-mediated repression in either reporter line (Fig. (Fig.7;7; see also Fig. S3 in the supplemental material). Therefore, these data support a model in which these two viral proteins possess distinct transcriptional repression activities encoded within their unique N-terminal repression domains that mediate their interactions with specific cellular complexes.
Brd4 has been implicated in the regulation of full-length E2 protein stability (11, 30, 66). Also, it was reported previously that Brd4 knockdown could increase steady-state E2 levels (51). Here we showed that Brd4 knockdown not only increased full-length E2 levels (Fig. (Fig.6;6; see also Fig. S2 in the supplemental material), but also increased steady-state E8^E2C levels (Fig. (Fig.7;7; see also Fig. S3 in the supplemental material). Therefore, Brd4 may be able to regulate the stability of these two E2 ORF-encoded viral factors independently of a direct interaction requiring amino acids located within the E2 N-terminal domain, which is absent in E8^E2C (1, 49, 64).
We have demonstrated that the interaction of E8^E2C with NCoR1 influences the transcription of the viral promoter. Under physiological conditions, NCoR1 is recruited through direct interaction with the cellular nuclear and steroid receptor superfamily of transcription factors in order to mediate the transcriptional regulation of various cellular promoters (6, 18, 19, 31, 32, 41, 42). Currently, we are exploring the possibility that E8^E2C expression may play a role in virus-host cell interactions through the perturbation of normal NCoR1 modulation of steroid and nuclear receptor-regulated promoters.
This study provides insight into the mode of E8^E2C-mediated transcriptional repression while highlighting the nonoverlapping mechanisms of E2 and E8^E2C repression. Such a finding suggests that repression of the viral LCR may be regulated differently through the viral life cycle depending on the abundance of the various E2 ORF-encoded proteins or their uniquely engaged cellular repression factors.
We thank members of our laboratories who contributed to discussions pertaining to this research.
This work was supported by National Institutes of Health grants T32CA009361 (to M.L.C.P. and J.A.S.), GM054137 and AG011085 (to J.W.H.), DFG-SFB773 (to F.S.), and P01CA050661 (to P.M.H.).
Published ahead of print on 24 February 2010.
†Supplemental material for this article may be found at http://jvi.asm.org/.