Phosphorylation of the Human Papillomavirus E2 Protein at Tyrosine 138 Regulates Episomal Replication

Papillomavirus (PV) is a double-stranded DNA tumor virus infecting the cutaneous and mucosal epithelium. The PV E2 protein associates with a number of cellular factors to mediate replication of the HPV genome. Fibroblast growth factor receptor 3 (FGFR3) regulates HPV replication through phosphorylation of tyrosine 138 in the HPV E2 protein. Employing a quasivirus infection model and selection for G418 resistant genomes, we demonstrated that Y138 is a critical residue for Brd4 association and that inability to complex with Brd4 does not support episomal replication.

ABSTRACT

The papillomavirus (PV) E2 protein is a critical regulator of viral transcription and genome replication. We previously reported that tyrosine (Y) 138 of HPV-31 E2 is phosphorylated by the fibroblast growth factor receptor 3 (FGFR3) kinase. In this study, we generated quasiviruses containing G418-selectable HPV-31 genomes with phosphodeficient phenylalanine mutant E2 Y138F and phosphomimetic glutamic acid mutant Y138E. We observed significantly fewer early viral transcripts immediately after infection with these Y138 mutant genomes even though E2 occupancy at the viral origin was equivalent to that of wild-type E2. Keratinocytes infected with Y138F quasiviruses formed stable colonies, and the genomes were maintained as episomes, while those infected with Y138E quasiviruses did not. We previously reported that the HPV-31 E2 Y138 mutation to glutamic acid did not bind to the Brd4 C-terminal motif (CTM). Here, we demonstrate that HPV-16 E2 Y138E bound to full-length Brd4 but not to the Brd4 CTM. We conclude that association of E2 with the Brd4 CTM is necessary for viral genome replication and suggest that this interaction can be regulated by phosphorylation of E2 Y138.

IMPORTANCE Papillomavirus (PV) is a double-stranded DNA tumor virus infecting the cutaneous and mucosal epithelium. The PV E2 protein associates with a number of cellular factors to mediate replication of the HPV genome. Fibroblast growth factor receptor 3 (FGFR3) regulates HPV replication through phosphorylation of tyrosine 138 in the HPV E2 protein. Employing a quasivirus infection model and selection for G418 resistant genomes, we demonstrated that Y138 is a critical residue for Brd4 association and that inability to complex with Brd4 does not support episomal replication.

INTRODUCTION

Human papillomaviruses (HPVs) are double-stranded DNA tumor viruses capable of infecting cervical, oral, and anal epithelium. The viral genome replicative program consists of three different stages linked to the differentiation status of the infected epithelium (1, 2). In the initial infection of basal epithelial cells, the viral genome replicates to low copy numbers. In basal and suprabasal cells, PV genomes duplicate and partition to progeny cells in a cell cycle-dependent manner (3, 4). The E2 protein is composed of an approximately 220-amino-acid transactivation domain (TAD), a hinge region, and a DNA binding domain (DBD), which recognizes the inverted palindrome ACCN₆GGT at the origin of replication (ori) (5). The E2 protein is responsible for replication of the virus, activating transcription of the viral early promoter and tethering viral genomes to host mitotic chromosomes in basal cells (6–8). As infected cells migrate into the upper strata in nondividing cells, episomes amplify to hundreds of copies and are packaged into virion particles (9).

The E2 protein undergoes posttranslational modifications (PTMs), including phosphorylation, acetylation, and sumoylation (10–15). Recently, we discovered that the tyrosine residue at position 138 is targeted by the fibroblast growth factor receptor 3 (FGFR3) (16). The mutant with a change of tyrosine to glutamic acid at position 138 (Y138E), which may mimic phosphotyrosine, failed to activate transient origin replication. In contrast, the phenylalanine (F) mutant Y138F induced transient replication at wild-type (WT) levels. Further analysis found that Y138E inhibits association with the C-terminal motif (CTM) of the bromodomain-containing protein 4 Brd4. Brd4 binding to E2 is critical for E1-E2 replication focus formation (17, 18) and E2 stability (19, 20). The function of Brd4 necessary for viral replication is not understood (8).

To study viral genome replication, traditional methods include transfection of HPV genomes into human foreskin keratinocytes (HFKs). Large quantities of DNA are transfected into cells that (i) do not mimic the true nature of HPV infection and (ii) may induce a DNA damage response. Van Doorslaer et al. inserted a neomycin gene cassette into the HPV-18 genome, replacing much of the L1 and L2 capsid genes (21). G418 selection for the cis-linked genome led to increased keratinocyte colonies that maintained replicating episomes. Another advance has been the development of HPV-based pseudoviruses capable of infecting keratinocytes (21–26). HPV genomes can be transfected along with L1 and L2 and packaged into recombinant virus-like particles (VLPs) known as quasiviruses. Quasiviruses only contain a single HPV genome (22) and better reflect the natural process of HPV infection.

The purpose of this study was to determine the requirement for E2 Y138 in genome replication by using quasivirus infection to reproduce the normal pathway by which viral DNA accesses the basal cell nucleus. We demonstrate that this tyrosine residue is critical for association of E2 with the Brd4 C-terminal motif CTM. Our results suggest that tyrosine 138 phosphorylation is a switch to limit long-term episomal replication and that this residue is important for transcription in keratinocytes.

RESULTS

We made E2 Y138 mutations in the HPV-31neo plasmid where a neo cassette under the control of a SV40 promoter was inserted in place of the L1 and L2 gene regions (T. Gilson and E. J. Androphy, submitted for publication). Quasivirus infectivity was confirmed showing equivalent G418-resistant colonies in HEK293 cells. At 2 to 3 days after infection of NIKS (near-duploid immortalized keratinocyte skin) cells, HPV-31 DNA content (Fig. 1A) and E2 binding to the viral origin (Fig. 1B) were not significantly different after infection with Y138F or Y138E quasiviruses compared to WT, demonstrating equal HPV DNA content with each quasivirus infection and similar levels of E2 protein. The genomes with E2 Y138F and Y138E expressed significantly fewer transcripts than cells infected with WT (Fig. 1C). At 2 weeks postinfection, we observed robust colony formation in NIKS cells and HFKs with WT genomes (Fig. 2A). The Y138F infection resulted in fewer colonies than the WT, while the Y138E genomes failed to produce colonies in both cell types. These findings were similar in NIKS cells and two donor-derived HFK lines (D1 and D2). Colonies with Y138F genomes were maintained for long-term survival and analyzed for HPV DNA content, viral transcript levels, and the ability to maintain episomes. The Y138F genome copy number was lower than the WT number in all cell lines tested (Fig. 2B). There were fewer spliced E8̂E2 and E1̂E4 transcripts in the NIKS (Fig. 2C) and HFK D2 (Fig. 2D) Y138F cells than in WT genome lines.

FIG 1.

(A) NIKS cells were infected with HPV-31 WT and mutant quasiviruses for 2 days, and DNA was quantified by real-time PCR with primers to HPV-31 DNA. Values are expressed as means ± standard errors of the means (SEM) (n = 3 per group). (B) NIKS cells were infected with WT, Y138F, or Y138E quasiviruses. Four days postinfection, cells were harvested for ChIP. Endogenous HPV-31 E2 protein was immunoprecipitated with an anti-HPV-31 E2 polyclonal antibody. Relative E2 binding at the HPV-31 ori was measured and normalized to EE IgG. (C) NIKS cells were infected with WT and Y138 mutant quasiviruses and divided into three groups. E1̂E4 and E8̂E2 mRNAs were measured at 3 days postinfection. Values are expressed as means ± SEM. *, P < 0.05 using a 2-way t test (n = 3).

FIG 2.

NIKS cells were infected with HPV-31 WT and mutant quasiviruses and selected with G418 for 2 weeks. (A) Numbers of colonies. (B) Quantification of HPV-31 DNA using real-time PCR in NIKS and HFK D1 and D2 cells infected with HPV-31 WT and Y138F quasiviruses from three sequential passages. (C) E1̂E4, E8̂E2, and E6 transcripts were measured in NIKS cells infected with WT and Y138F genomes from three sequential passages. Values are expressed as means ± SEM. *, P < 0.05, and #, P < 0.1, using a 1-way t test. (D) E8̂E2 and E1̂E4 were measured in NIKS and HFK D2 cells infected with WT and Y138F using endpoint PCR with or without reverse transcriptase (RT). Relative transcripts were normalized to 18S using qPCR, where the value for each WT cell line equals 1.

To test for the presence of episomes, the exonuclease V assay was used and compared to HPV-31 DNA from CIN612-9E cell lines with episomal and integrated genomes (Fig. 3A). In this assay, exonuclease V does not digest double-stranded circular DNA such as mitochondrial DNA or HPV episomes but does digest linear DNA (27, 28). Total DNA was isolated from NIKS and HFK cell lines with WT and Y138F genomes and subjected to exonuclease V digestion followed by quantitative PCR. Actin DNA was completely digested in all samples (Fig. 3B to D). Mitochondrial DNA was resistant to levels previously published (28), confirming enzyme specificity to linear DNA (Fig. 3B to D). Both the WT and Y138F genomes maintained episomes in NIKS (Fig. 3B), HFK D1 (Fig. 3C), and HFK D2 (Fig. 3D) cell lines. Southern blot analysis confirmed the presence of the HPV-31 genome in HFK D2 and NIKS WT and Y138F cell lines (Fig. 3E).

FIG 3.

(A) CIN612 episomal and integrated cell lines served as controls in the exonuclease V digestion assay. (B to D) HPV-31 DNA from NIKS (B), HFK D1 (C), and HFK D2 (D) cells infected with WT and Y138F quasiviruses was quantified using real-time PCR after exonuclease V digestion. (E) Southern blot analysis of keratinocyte cell lines with HPV-31 WT and E2 Y138F genomes in HFK D2 and NIKS cells. The HPV-31neo plasmid (500 pg) was used as a positive control.

HPV-31 genomes with the E2 Y138E phosphomimetic mutation were unable to maintain episomes in keratinocytes. We previously reported that the HPV-31 Y138E E2 protein did not bind to the Brd4 CTM but coimmunoprecipitated with full-length Brd4 (16). The Brd4 CTM has been shown to contact amino acids R37 and I73 within the TAD of E2 (24, 29–32). The HPV-16 E2 R37A/I73A mutant does not bind the Brd4 CTM (29) but does interact with HPV-16 E1 (33, 34) and is also replication defective (17). We compared this double mutant to Y138E for Brd4 binding. The Y138F and Y138E mutations were constructed in HPV-16 E2 for comparison, since E2 proteins from HPV genotypes have various binding affinities to Brd4 (45). HEK293TT cells were transfected with FLAG-HPV-16 E2 WT, Y138F, Y138E, and R37A/I73A mutant constructs along with the glutathione S-transferase (GST)-tagged Brd4 CTM. Y138E and R37A/I73A did not coimmunoprecipitate with the Brd4 CTM (Fig. 4A). To test the ability of the E2 R37A/I73A to bind the full-length Brd4, HEK293TT cells were transfected with FLAG-Brd4 along with HPV-16 E2 FLAG-WT or R37A/I73A constructs. Brd4 coimmunoprecipitated E2 R37A/I73A but to a lesser extent than WT (Fig. 4B). To compare the HPV-16 E2 Y138 mutants, HEK293TT cells were transfected with FLAG-Brd4 along with FLAG-HPV-16 E2 WT, Y138F, Y138E, or R37A/I73A E2 constructs. Brd4 immunoprecipitated all the E2 proteins (Fig. 4C) but less efficiently than R37A/I73A (51% of WT) and Y138E (34% of WT).

FIG 4.

(A) HEK293TT cells were transfected with GST-Brd4 CTM and FLAG-HPV-16 E2 Y138 mutants. HPV-16 E2 was immunoprecipitated with M2 (FLAG) antibodies. Complexes were blotted with GST and M2 (FLAG) antibodies. (B) HEK293TT cells were transfected with full-length FLAG-Brd4 and FLAG-HPV-16 E2 WT or R37A/I73A. Brd4 was immunoprecipitated with Brd4 antibodies and immunoblotted with M2 (FLAG) antibodies. IgG antibodies were used as nonspecific controls. (C) HEK293TT cells were transfected with full-length FLAG-Brd4 and FLAG-HPV-16 E2 Y138 or R37A/I73A mutants. Brd4 was immunoprecipitated with Brd4 antibodies, and bound proteins were immunoblotted with M2 (FLAG) antibodies.

DISCUSSION

Our previous studies demonstrated that HPV-31 E2 with tyrosine 138 mutated to glutamate (Y138E) bound E1 and was capable of binding the ori yet was inactive for induction of E1-dependent transient DNA replication. Y138E did not form nuclear replication foci, and it bound to full-length Brd4 but did not associate with its CTM (16). Our transient-transfection DNA replication experiments utilized cytomegalovirus (CMV) promoter-driven E1 and E2 expression vectors with the HPV origin. In the infection model described here using quasiviruses with HPV genomes and a cis-linked selection gene, the native P97 promoter drives endogenous E1 and E2 levels. We hypothesized that the E2 Y138E genome would not be maintained episomally even under G418 selection pressure due to defective Brd4-CTM binding.

Contacts between the Brd4 CTM and E2 are mediated by two highly conserved E2 TAD residues, R37 and I73 (24, 29–32). Recently, the Chiang lab found that both high-risk and low-risk HPV E2 proteins associate differently with phosphorylated Brd4 (7). The Brd4 basic residue-enriched interaction domain (BID) interacts with the DBD of both high-risk and low-risk HPV E2 proteins. The Brd4 phosphorylation-dependent interaction domain (PDID) modulates binding to the DBD of HR E2 proteins, and this association is independent of the BID region (7). Together, the present study and our previous publication demonstrate that the Y138E E2 mutant protein is capable of binding full-length E2 Brd4 but not the CTM (16). In our experiments, the CTM consists of the terminal 138 amino acids of Brd4 (amino acids [aa] 1224 to 1362), which is also the p-TEFb binding domain (aa 1209 to 1362) (36). The transcriptional elongation factor p-TEFb enhances RNA polymerase II processivity (37, 38). The cocrystal structure of HPV-16 E2 with the last 20 amino acids of Brd4 (aa 1343 to 1362) revealed bonds to R37 and I73 (29). This structure would not include other potential contacts between the E2 TAD and Brd4. Our data are consistent with the beta sheet region of E2 including Y138 mediating additional interactions with Brd4. We believe that it is unlikely that Y138 mutant proteins are misfolded, as these maintain the ability to complex with E1, which depends on much of the E2 TAD structure being intact (39, 40).

Binding to the Brd4 CTM increases stability of E2 proteins (20, 41) by preventing E2 nuclear export and cytoplasmic E2 ubiquitylation (19). We observed that as with HPV-16 E2 R37A/I73A (19), protein levels of Y138E were consistently lower than those of the WT. Although these mutants can bind to full-length Brd4 through contacts between the E2 DBD and the Brd4 BID, the inability of the E2 protein to associate with the Brd4 CTM decreases E2 protein stability.

During initial viral replication, Brd4 localizes to E1 and E2 replication foci; however, it is displaced from these foci during genome amplification (18). You’s group also found that Brd4 was recruited to replication foci and that it was essential for genome replication (17). While Brd4 is required for HPV transcription (35) and tethering E2 to chromatin (32, 35, 42), HPV-31 genomes with R37 or I73 mutations were still capable of expressing viral proteins and stably replicated (43). In contrast, the HPV-16 E2 I73A genome was not maintained as an episome in keratinocytes (44), and transient replication with HPV-16 R37A/I73A was impaired in C33A cells (17).

We demonstrate that HPV-31 E2 Y138F binds to the Brd4 CTM, whereas Y138E cannot. Y138F + E1 stimulated transient DNA replication and supported stable episomal maintenance, while Y138E was defective for both activities. These data suggest that the Brd4 CTM is important for mediating E2 replication. Transcript levels with both E2 Y138F and Y138E genomes were reduced; however, the relationship to Brd4 is not as clear. The HPV-31 E2 R37A/I73A and the HPV-16 R37A and I73A mutants were reported to be inactive in transient-transcription reporter assays (34, 45). Instead of JQ1 and similar Brd4 inhibitors, we and others have used p-TEFb inhibitory compounds and found that transcription was impaired but not abrogated, implying that the E2 transcriptional activity involves factors in addition to the Brd4 CTM and subsequent release of p-TEFb (16, 46). It is also possible that the interaction between Brd4 and Y138F may inhibit Brd4-mediated transcription, perhaps by failing to release from this complex. More studies are needed to discern the specific regulatory role of the Brd4 CTM and pTEF-b in viral transcription and replication.

Our data suggest that the beta sheet region of the TAD of E2 that includes tyrosine 138 is essential for binding to the Brd4 CTM and is critical for viral episomal replication. We predict that phosphorylation of tyrosine 138 by the FGFR3 kinase inhibits E2 binding to the CTM and thereby regulates PV replication after initial infection. The activity served by E2 binding to CTM in viral replication remains to be determined.

MATERIALS AND METHODS

Plasmids and antibodies.

The following plasmids were used: codon-optimized HPV-16 E2 WT and R37A/I73A (from A. McBride) were cloned into the HindIII and BamHI sites of pCDNA3. The following antibodies were used: mouse anti-FLAG M2, rabbit anti-Brd4 (Cell Signaling), and rabbit anti-GST. Y138 mutations were made in pCDNA3-HPV-16 E2 using the Q5 site-directed mutagenesis kit from New England BioLabs and confirmed by sequencing.

Cell lines.

All cell lines were cultured at 37°C and at 5% CO₂. HaCaT (from N. Fusenig), J2-3T3 fibroblast feeders (from Howard Green), and HEK293TT (from J. Schiller and C. Buck) were cultured in Dulbecco’s modified Eagle medium with 10% fetal bovine serum (Peak Serum) and penicillin-streptomycin (100 U/ml; Life Technologies). J2-3T3 fibroblast feeders were transfected with pCDNA3 using Lipofectamine 2000 (Invitrogen) and selected with 500 μg/ml G418 to generate G418-resistant feeders. NIKS cells were grown in F medium on J2-3T3 G418-resistant fibroblast feeders. HFKs (donors D1 and D2) were maintained in keratinocyte serum-free medium (KSFM) containing human recombinant epidermal growth factor (EGF), bovine pituitary extract (BPE), and penicillin-streptomycin (100 U/ml; Life Technologies). Following infection, cells were cultured in F medium with J2-3T3 G418-resistant cells. Episomal and integrated CIN612 cell lines (from L. Laimins) were grown in E medium with J2-3T3 feeders.

Production of quasiviruses and infection of keratinocytes.

The QuikChange mutagenesis kit (Agilent) was used to create the Y138F and Y138E mutations of E2 in the pBR322HPV31neoH plasmid, and these were confirmed by sequencing. The E2 mutations were subcloned into the pBR322HPV31neoHC plasmid using EcoRI and PvuI. The plasmid was named HPV-31neo (Gilson and Androphy, submitted). Wild-type and mutant genomes were digested overnight with HindIII, and the 8-kb HPV-31 genome was extracted from the backbone using the PureLink Quick Gel extraction kit (Invitrogen). Genomes were ligated overnight with T4 DNA ligase (New England Biolabs).

Quasiviruses were produced as previously described (26) with the following modifications. HEK293TTF cells (from R. Roden) were transfected with 12 μg of ligated genomes and 12 μg of HPV-16 codon-modified L1 and L2 genes (pShell L1L2; from J. Schiller) using 2 μg/ml polyethylenimine (PEI). Cells were harvested 48 h later for quasivirus isolation by washing in phosphate-buffered saline (PBS) with 10 mM MgCl₂ and lysing in PBS containing 10 mM MgCl₂, 0.5% Triton X-100, 25 mM ammonium sulfate (pH 9), and 5 mM CaCl₂. Lysates were incubated at 37°C for 48 h for maturation (47) and spun at 7,000 rpm for 10 min.

To prepare extracellular matrix (ECM) for infections, HaCaT cells were seeded and grown to confluence. Cells were washed in PBS and lysed in ECM buffer containing 0.5% Triton X-100 and 20 mM ammonium hydroxide in PBS. The ECM was washed 3 times with PBS; F medium and virus stock were added and incubated at 37°C. Four hours later, medium was removed and replaced with keratinocytes in F medium. Forty-eight hours later, 250 μg/ml of G418 was added. Keratinocyte infections with WT and mutant genomes were completed at least five independent times.

HPV-31 DNA quantification after viral infection.

NIKS cells were infected with quasivirus stocks for 2 days and trypsinized, and DNA was isolated using the DNeasy blood and tissue kit (Qiagen) per the manufacturers’ instructions. Fifty nanograms of the total DNA was quantified by real-time PCR using CFX96 Bio-Rad C3000 thermocycler with SsoFast EvaGreen Supermix. HPV-31 long control region (LCR) and actin primers were previously described (48, 49).

For exonuclease V digestion, cell pellets were resuspended in PBS and DNA was isolated using the DNeasy blood and tissue kit from Qiagen. Five hundred nanograms of DNA was treated with exonuclease V (New England Biolabs) or left untreated for 4 h at 37°C. The enzyme was inactivated at 70°C for 20 min. Fifty nanograms of the digested or the undigested DNA was quantified by real-time PCR. The human mitochondrial DNA primers and the exonuclease V protocol were previously described (27). Percent exonuclease V-resistant DNA was calculated as 2^{−(digested CT – undigested CT)} × 100.

Quantification of viral transcripts.

RNA was isolated using TRIzol (Invitrogen) and was converted to cDNA using SuperScript III reverse transcriptase (Life Technologies). HPV-31 E1̂E4 transcripts were measured by real-time PCR (50). HPV-31 E8̂E2 transcripts were amplified using the primers 5′-GTGGAAACGCAGCAGATGGTA-3′ and 5′-TTCGATGTGGTGGTGTTGTTG-3′. HPV-31 E6 transcripts were amplified using the primers 5′-AAGACCGTTGTGTCCAGAAGA-3′ and 5′-CTGTCCACCTTCCTCCTATGTT-3′. Gene expression was normalized to actin or 18S cDNA (27).

Chromatin immunoprecipitation.

Four days postinfection, NIKS cells were treated with 1% formaldehyde, and chromatin immunoprecipitation (ChIP) assays were performed using a ChIP-IT kit (Active Motif). Rabbit HPV-31 E2 (48) and EE epitope antibodies were used. Real-time PCR with primers to the HPV-31 LCR4 was performed using Sso Fast Evagreen Mastermix (Bio-Rad).

Southern blotting.

Southern blotting was performed as described previously (11). Total DNA was isolated from NIKS (WT and Y138F) and HFKD2 (WT and Y138F) cell lines using phenol-chloroform-isoamyl alcohol extraction. At least 15 μg of DNA was digested with HindIII overnight and run on a 1% agarose gel. The gel was treated with 0.1 N HCl and washed in water, and genomic DNA was transferred overnight with alkaline transfer buffer (0.4 N NaOH and 1 M NaCl) and UV cross-linked onto a Hybond nylon membrane. The ³²P dCTP HPV-31 DNA hybridization probe was generated from the HPV-31neo plasmid using the random primer DNA labeling kit (TaKaRa). The hybridization was carried out in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) with 1% sodium dodecyl sulfate (SDS), 1× Denhardt’s solution containing 10 mg Ficoll, 10 mg polyvinylpyrrolidone, and 10 mg bovine serum albumin with salmon sperm. Blots were washed with decreasing concentrations of buffer ranging from 2× SSC–0.1% SDS to 0.5× SSC–0.1% SDS and developed on a PhosphorImager screen. Images were captured using a Typhoon 9410 scanner (GE Healthcare).

Coimmunoprecipitations and immunoblotting.

Cells were transfected with PEI (2 mg/ml), and coimmunoprecipitations were performed as described previously (16). Proteins were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). Gels were transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore), blocked in 5% Tris-buffered saline (TBS)–Tween (0.1%)–milk, and probed with antibodies. Chemiluminescence substrates (Thermo Scientific) were used to detect the antibody signal.