|Home | About | Journals | Submit | Contact Us | Français|
The largest isoform of adenovirus early region 1A (E1A) contains a unique region termed conserved region 3 (CR3). This region activates viral gene expression by recruiting cellular transcription machinery to the early viral promoters. Recent studies have suggested that there is an optimal level of E1A-dependent transactivation required by human adenovirus (hAd) during infection and that this may be achieved via functional cross talk between the N termini of E1A and CR3. The N terminus of E1A binds GCN5, a cellular lysine acetyltransferase (KAT). We have identified a second independent interaction of E1A with GCN5 that is mediated by CR3, which requires residues 178 to 188 in hAd5 E1A. GCN5 was recruited to the viral genome during infection in an E1A-dependent manner, and this required both GCN5 interaction sites on E1A. Ectopic expression of GCN5 repressed transactivation by both E1A CR3 and full-length E1A. In contrast, RNA interference (RNAi) depletion of GCN5 or treatment with the KAT inhibitor cyclopentylidene-[4-(4′-chlorophenyl)thiazol-2-yl]hydrazone (CPTH2) resulted in increased E1A CR3 transactivation. Moreover, activation of the adenovirus E4 promoter by E1A was increased during infection of homozygous GCN5 KAT-defective (hat/hat) mouse embryonic fibroblasts (MEFs) compared to wild-type control MEFs. Enhanced histone H3 K9/K14 acetylation at the viral E4 promoter required the newly identified binding site for GCN5 within CR3 and correlated with repression and reduced occupancy by phosphorylated RNA polymerase II. Treatment with CPTH2 during infection also reduced virus yield. These data identify GCN5 as a new negative regulator of transactivation by E1A and suggest that its KAT activity is required for optimal virus replication.
The early region 1A (E1A) gene is the first viral gene expressed in cells upon infection with human adenovirus (hAd). Early in infection the primary E1A transcript is alternatively spliced to yield two predominant isoforms that perform two essential functions required to initiate the viral replication cycle. First, E1A uncouples the cell cycle control program of the host cell, driving it into S phase to provide an optimal cellular environment for viral replication. The smaller major E1A isoform (243 residues in human adenovirus type 5 [hAd5] E1A) is sufficient to override cell cycle progression and drive cells into S phase (16, 34). The other critical function of E1A is to activate transcription of the early viral promoters. Transactivation of the hAd early genes is predominantly mediated by the product of the largest E1A isoform (6, 21, 30, 31). In hAd5 the 13S E1A mRNA codes for a 289-residue protein that differs from the 243R E1A protein by a unique 46-amino-acid (aa) C4 zinc finger domain located within conserved region 3 (CR3), which is essential for viral transactivation (6, 21). In fact, CR3 alone is sufficient to stimulate transcription of an E2 promoter when microinjected into HeLa cells (25). The region of hAd5 E1A spanning residues 139 to 204 (which includes CR3) is critical for the 5- to 20-fold activation of viral transcription necessary for virus growth and is also sufficient for potent activation of a minimal Gal4-responsive promoter as a Gal4-DNA binding domain (DBD) fusion (20, 45).
CR3 can be subdivided into three subdomains: an N-terminal zinc finger region, a promoter-targeting region, and a region known as auxiliary region 1 (AR1) (14, 28, 49). The N-terminal zinc finger subdomain of CR3 activates transcription by interacting with cellular TATA-binding protein (TBP) and MED23 proteins (8, 24, 48). The Mediator component MED23 is absolutely essential to E1A CR3 function, since CR3s from multiple closely related hAds fail to activate transcription in MED23-null MEFs (1, 48). Several additional cellular factors have been implicated in transactivation by E1A CR3. The S8 component of the 19S ATPase Proteins Independent of 20S (APIS), hSUG1, and the p300/CBP and pCAF lysine acetyltransferases (KATs) interact with E1A CR3 and enhance E1A CR3 transactivation (35, 36). Although transactivation by E1A CR3 has been studied predominantly with hAd5, these interactions and their effects on E1A CR3 transactivation are generally conserved across the entire hAd family despite dramatic differences in the magnitude of E1A transactivation exhibited by representative members of each hAd species (1, 35). Moreover, recent competition studies have demonstrated that additional as-yet-unidentified cellular factors not only are involved in E1A transactivation but also are limiting in the cell (1). Recruitment of a functional E1A-containing transcription initiation complex to the early viral promoters requires the C-terminal promoter-targeting subdomain of CR3. This subdomain confers interaction of E1A CR3 with cellular sequence-specific DNA-binding transcription factors (12, 27–29). The precise role of the residues constituting the AR1 subdomain of CR3 in transactivation remains unclear; however, the acidic character of this region is necessary for maximal transactivation by E1A CR3 (49).
Many cellular factors that interact with CR3 also independently bind the N terminus of E1A, which also can function as an activation domain (24, 35–38). This further complicates E1A-mediated activation of transcription. Whether these independent interactions function cooperatively or competitively has not been fully elucidated (36, 55). Of the factors that interact independently with both transactivation regions of E1A, several are KATs (p300/CBP and pCAF) (35, 36). An additional KAT and close relative of pCAF, GCN5, also interacts with the N terminus of E1A, but neither a role for GCN5 in transactivation nor an interaction for GCN5 within CR3 has been demonstrated in mammalian cells (23). In yeasts, which lack pCAF, yeast GCN5 interacts with hAd5 CR3 and is a coactivator of CR3 transactivation (45).
The primary function of GCN5 appears to be as a chromatin-modifying factor (4, 9, 32, 40). GCN5 is a catalytic component of the Spt-Ada-GCN5-acetyltransferase (SAGA) complex in yeast and the Spt-TAFII31-GCN5L (STAGA) complex in mammalian cells. The KAT activity of GCN5 is required to acetylate histone H3 lysine 9 (K9) and K14, and this facilitates transcription elongation by relaxing nucleosomes (4, 9, 32, 40). A definitive role for GCN5 in mammalian transcription has remained elusive, as GCN5-null embryos (GCN5−/−) die very early in development, at 10.5 days postcoitum (d.p.c.) from massive apoptosis. This results from the loss of the deubiquitination activity of GCN5, which leads to genomic instability and telomere crisis (2, 58). In contrast, embryos specifically defective for the GCN5 KAT activity (GCN5 hat/hat) do not exhibit genomic instability or telomere crisis (2); however, they die at 16.5 d.p.c. due to defects in neural tube closure and encephalopathy (11). pCAF-null animals (pCAF−/−) are viable due to compensation by increased levels of GCN5, suggesting some functional redundancy between these closely related KATs (59).
In the present study, we report the identification of a second independent conserved GCN5 binding site within CR3, which maps to residues 178 to 188 of hAd5 E1A. GCN5 associated with the viral E4 promoter in an E1A-dependent manner. Moreover, the two interaction sites in E1A cooperate to recruit GCN5 to the transcriptional template. GCN5 functions as a negative regulator of E1A transactivation, as (i) RNA interference (RNAi) depletion of GCN5 increased E1A transcriptional activation, (ii) ectopic expression of GCN5 repressed E1A-dependent transactivation, and (iii) pharmacological inhibition or genetic ablation of the KAT activity of GCN5 relieved the repressive effect on E1A transactivation. Interestingly, KAT inhibition also decreased virus yield. Furthermore, the repressive effect of GCN5 on E1A transactivation correlated not only with a hyperacetylation of cellular histone H3 K9/K14 associated with the viral E4 promoter but also with a reduction in phosphorylation of the RNA polymerase II C-terminal domain (phospho-CTD RNAPII) at the promoter, and this required the KAT activity of GCN5. Collectively, our results indicate that GCN5 provides a new layer of negative regulation to the existing model of E1A transactivation. Perturbation of this repression function reduces virus production, suggesting that a balance between viral gene expression needed for replication and production of toxic gene products is required for optimal virus replication.
Human A549, HeLa, and HT1080 cells, as well as wild-type mouse embryonic fibroblasts (wt MEFs) and homozygous GCN5 KAT-defective (hat/hat) MEFs (11), were maintained at 37°C and 5% CO2 in Dulbecco modified Eagle medium (Wisent) with 10% fetal bovine serum (Gibco) and 100 U/ml of penicillin-streptomycin (Wisent). A549 cells and MEFs were transfected with the FuGENE HD reagent (Roche), according to the manufacturer's directions, in a ratio of 3 μg of total DNA to 9 μl of FuGENE HD per well of a six-well plate. HeLa and HT1080 cells were transfected with the Superfect reagent (Qiagen), according to the manufacturer's directions. Wild-type littermate MEFs and GCN5 hat/hat MEFs were provided by S. Roth-Dent (M. D. Anderson Cancer Center, Smithville, TX) and have been described previously (11).
The Gal4-responsive luciferase reporter vector pGL2-(Gal4)6-Luc and Gal4DBD fusions for each hAd E1A CR3 and the N terminus of E1A (residues 1 to 82) have been described previously (3, 44). The expression vector for enhanced green fluorescent protein (EGFP) fused with hAd5 E1A CR3 Δ178-184 was produced by PCR amplification of the CR3 region using the primers CR3N-F (5′-AGACGAATTCGGTGAGGAGTTTGTGTTA-3′) and CR3C-R (5′-CGCGGATCCATTAGGTAGGTCTTGCAGGCTC-3′), using 13S E1A dl1114 (Δ178-184) as the template with Phusion polymerase (NEB) according to the manufacturer's directions. The PCR product was cloned into pCAN-myc-EGFP with EcoRI and XbaI (20). The expression vector for mGCN5 was generated by PCR using the primers mGCN5-F (5′-TCGGAATTCGCGGAACCTTCCCAGGCCCCAAACC-3′) and mGCN5-R (5′-GACTCTAGACTACTTGTCGATGAGCCCTCC-3′), with Phusion polymerase according to the manufacturer's directions, using a previously described expression vector for mGCN5 as the template (57). The PCR product was digested with EcoRI and XbaI and cloned into the EcoRI and NheI sites of pCMX-FLAG. A correct clone was verified by sequencing. Expression vectors for hemagglutinin (HA)-tagged MED23 (pCS2+human Sur2-HA) (8), HA-tagged human TBP (pcDNA4HA-hTBP) (38), and FLAG-tagged pCAF (pCMX-FLAG-pCAF) (60) have been described previously.
The Gal4 fusion activation assay has been described previously (1). HT1080 cells were chosen for this assay because they provided excellent and consistent transfection efficiency. The full-length E1A activation assay was performed as follows: 24 h prior to transfection, 1.5 × 105 HT1080 cells/well were seeded on six-well plates. Cells were transfected in a 1:3:4 ratio of E4 reporter pGL2-E4v3 to E1A expression vector to GCN5 (either pCMX-FLAG mGCN5 or empty pCMX-FLAG was used as a control) (36). At 6 h posttransfection, cells were washed and fresh medium was added. For activation assays involving the GCN5-specific KAT inhibitor cyclopentylidene-[4-(4′-chlorophenyl)thiazol-2-yl]hydrazone (CPTH2) (Sigma), either the indicated final concentration of CPTH2 or an equal volume of dimethyl sulfoxide (DMSO) (vehicle) was added after washing at 6 h posttransfection. Cells were harvested at 48 h posttransfection and assayed for luciferase activity. Luciferase activity was expressed as the percentage of control ± standard deviation (SD), and the value of wt CR3 or full-length 13S E1A was set to 100%. Mean percent activation of control was compared by Student's t test; significant difference between the two means was set at P values of <0.05.
Anti-E1A hybridoma clone M73 was used for immunoprecipitation (IP), Western blotting (WB), and chromatin immunoprecipitation (ChIP) in the form of cell culture supernatant. Anti-pRB hybridoma clone C36 was used to detect pRb by WB. Anti-MED23 was purchased from Novus Biologicals. Anti-myc antibody clone 9E10 hybridoma was a gift from P. White (McMaster University, Hamilton, Ontario, Canada), and supernatant from hybridoma cultures was used for WB and coimmunoprecipitation (Co-IP). Anti-FLAG M2 antibody and anti-FLAG-M2-agarose were purchased from Sigma. Anti-HA clone 3F10 rat monoclonal antibody was purchased from Roche. Rabbit anti-GFP antibody was purchased from Clontech. Rabbit anti-H3 and anti-H3 acetylated K9 and K14 were purchased from Upstate. Rabbit anti-RNA polymerase II CTD repeat YSPTSPS antibody and anti-RNA polymerase II CTD repeat YSPTSPS (phospho-S5) antibody were purchased from Abcam. The antibody against GCN5 was raised against amino acids 1 to 93 of human GCN5, which are not conserved with pCAF. The corresponding portion of the GCN5 cDNA was amplified by PCR and subcloned into pGEXJDK. The plasmid was transformed into BL21 Escherichia coli cells, which were then induced with 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 3 h, and recombinant protein was purified by affinity chromatography using glutathione Sepharose. The purified protein was injected into rabbits, and the antisera were purified by protein A Sepharose chromatography. The purified IgG fraction recognizes GCN5 and does not cross-react with pCAF (Fig. 1B).
Co-IP of GCN5 with full-length E1A from infected HeLa cells was performed as described previously (23). HeLa cells were chosen for this assay to reproduce the interaction of 12S E1A and GCN5, as well as the newly identified CR3 interaction, under the same conditions as reported previously (23). Co-IP of GCN5 with myc-EGFP-fused CR3s was performed as described previously with anti-myc antibody (clone 9E10) (1). Co-IP of myc-EGFP-fused CR3 and CR3 mutants with FLAG-tagged GCN5 was performed as described previously, but using FLAG-M2-agarose (Sigma) (1). Co-IP of HA-tagged MED23, HA-tagged TBP, and FLAG-tagged pCAF with wt and Δ178-184 CR3-myc-EGFP fusions were performed as described previously, but using rabbit anti-GFP antibody. WB analysis was performed as described previously (1).
Validated Silencer select small interfering RNAs (siRNAs) against human GCN5 (5′-GAAGCUGAUUGAGCGCAAtt-3′) and negative-control siRNA2 were purchased from Ambion. siRNA transfections were performed with siLentFect reagent (Bio-Rad), according to the manufacturer's directions, using 20 nM control or GCN5 siRNA. HeLa cells were chosen for this assay because the siRNAs were validated in HeLa cells and potently knocked down GCN5 within a suitable time frame for the assay. Typically, 24 h prior to siRNA transfection 1.5 × 106 HeLa cells were seeded on 10-cm plates. At 24 h post-siRNA transfection, cells were reseeded to six-well plates at 2 × 105 cells per well. At 48 h post-siRNA transfection, cells were transfected again as described previously with the appropriate expression vectors to perform the Gal4 fusion activation assay (1). Luciferase activity is expressed as fold activation above Gal4DBD alone ± SD, where Gal4DBD alone was set to 1. Mean fold activation was compared by Student's t test; a significant difference between two means was set at a P value of <0.05.
ChIP assays were performed in A549 cells to complement the viral growth experiments as described previously (36, 50) using the antibodies described above. PCR for a 300-bp region of the Ad5 E4 promoter was performed as described previously (36). Fold enrichment of the E4 promoter was determined by quantitative PCR (qPCR) of ChIP DNA. The amount of specific promoter DNA immunoprecipitated was expressed as a percentage of input. The percent input was then normalized relative to the negative-control antibody (IgG), which was set to 1 (fold enrichment). Mean fold enrichment results were compared by Student's t test; significant differences were set at P values of <0.05. The ratio of acetylated histone H3 (H3-Ac) relative to total histone H3 (H3) was calculated from the fold enrichment. The ratio of H3-Ac/H3 was analyzed by one-way analysis of variance (ANOVA) using Tukey's posttest. The ratio of phospho-CTD RNAPII to total RNA polymerase II (total RNAPII) was calculated from the fold enrichment. The mean ratios of phospho-CTD RNAPII to total RNAPII were compared by Student's t test; a significant difference between the two means was set at a P value of <0.05. Similarly, the ratio of GCN5 to E1A recruited was calculated using fold enrichment and the mean ratios of GCN5/E1A fold enrichment were compared by Student's t test. (The designation “ns” in the figures indicates a nonsignificant difference between the two means, where P > 0.05.)
Quantitative reverse transcription-PCR (qRT-PCR) assays of hAd E4orf6/7 transcripts were performed as described previously (1). Briefly, human A549 cells or MEFs (wt littermate or hat/hat) were infected with either wt hAd5 or mutant hAd5 virus at a multiplicity of infection (MOI) of 2. For experiments in which CPTH2 was used, after 1 h of adsorption the virus inoculum was removed and cells were replenished with fresh media containing either 50 μM CPTH2 or an equivalent volume of DMSO (vehicle). At 16 h postinfection, total RNA was isolated as described previously and used to generate cDNA (1). Quantitative PCR was performed in triplicate with a 15-μl reaction mixture and 1× iQ-SYBR green SuperMix (Bio-Rad) according to the manufacturer's directions in a MyiQ real-time PCR instrument (Bio-Rad). The primers for hAd5 targets and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were described previously (38). Data were analyzed using IQ5 software (Bio-Rad). E4orf6/7 mRNA levels were normalized to GAPDH levels as an internal control for each sample. E4orf6/7 mRNA levels in cells infected with ΔE1A mutant dl312 were set equal to 1. The mean relative expression levels of E4orf6/7 transcripts were compared by Student's t test; a significant difference was set at a P value of <0.05.
Human A549 cells were infected at an MOI of 5 with either wt (dl309) or ΔE1A (dl312) adenovirus for 1 h. After adsorption, the virus inoculum was removed and cells were replenished with fresh media containing either 50 μM CPTH2 or an equivalent volume of DMSO (vehicle). At 96 h postinfection, cells and media were harvested and subjected to three freeze-thaw cycles, and the yield of virus present was determined by plaque assay on HEK293 cells. The virus yields from duplicate experiments were averaged, and mean virus growth ± SD values were plotted and compared by Student's t test; significant differences were set at P values of <0.05.
Human GCN5 (GCN5) was previously shown to interact with hAd5 E1A through residues 26 to 35 in its N terminus (23). We confirmed that 12S (243R) E1A Δ26-35 failed to coimmunoprecipitate GCN5, although this mutant can still coimmunoprecipitate pRb (Fig. 1A, lane 2). Interestingly, E1A Δ26-35 retained the ability to coimmunoprecipitate GCN5 when both the 12S (243R) and 13S (289R) isoforms of E1A were present (Fig. 1A, lane 3). As a control for CR3-specific interactions, we demonstrated that wt 12S E1A and 12S E1A Δ26-35 failed to coimmunoprecipitate the CR3-specific target MED23, whereas the same mutant coimmunoprecipitated MED23 when the 13S (289R) isoform was present (Fig. 1A, lane 3). This suggested that, like MED23, there is an interaction between GCN5 and the CR3 region that is unique to the largest E1A isoform (Fig. 1A). Antibody specificity experiments indicated that the GCN5 antibody does not cross-react with the closely related pCAF protein (Fig. 1B). Using a collection of E1A CR3 mutants, the interaction of GCN5 was mapped to residues 178 to 188, since FLAG-tagged mouse GCN5 (mGCN5) failed to coimmunoprecipitate the Δ178-184 and the Δ180-188 mutants of CR3 (Fig. 1C). CR3 mutants that have been characterized previously to lose interaction with pCAF (Δ139-147), MED23 (H160Y), or TBP (V147L) retained interaction with FLAG-tagged mGCN5 (Fig. 1C) (8, 17, 35, 48). Furthermore, the interaction of mGCN5 with CR3 did not require the zinc finger subdomain, because a point mutant in one of the coordinating cysteine residues (C157S) and a large deletion encompassing the majority of the zinc finger region (Δ140-160) also retained interaction with FLAG-tagged mGCN5 (Fig. 1C). Thus, the interaction of CR3 with GCN5 is independent of pCAF, MED23, and TBP binding as well as the zinc finger subdomain of E1A-CR3 (Fig. 1C). We also profiled the interactions of the Δ178-184 mutant with some of the known cellular binding partners of CR3. Although CR3 Δ178-184 is unable to bind mGCN5, it retains interaction with pCAF. This mutant still binds TBP weakly but does not interact with MED23 (Fig. 1D). Collectively, these experiments indicate that E1A CR3 contains a second independent interaction site for GCN5 that requires at least residues 178 to 188.
CR3 is critical for activation of viral early gene transcription. We hypothesized that the interaction of the largest E1A isoform with GCN5 would be involved in transcription and that GCN5 would be associated with viral promoters during infection. The E4 promoter is potently stimulated by E1A during infection, and we have shown previously that both E1A and p300 can be found to occupy a 320-bp region of this promoter during infection (Fig. 2A) (1, 36). We similarly determined whether GCN5 was recruited to the adenoviral E4 promoter during infection by using ChIP. During wt infection, E1A and GCN5 were found to co-occupy the E4 promoter (Fig. 2B) and GCN5 recruitment required E1A (Fig. 2B). ChIP assays using viruses lacking the N-terminal GCN5 binding site (E1A Δ26-35) or the CR3 binding site (E1A Δ178-184) demonstrated that E1A was still recruited to the promoter, albeit less efficiently. In contrast, GCN5 recruitment was dramatically reduced by the loss of either interaction site in E1A (Fig. 2B). Therefore, both GCN5 binding regions in E1A appear to be necessary for efficient recruitment of GCN5 to the E4 promoter. We quantified the recruitment of GCN5 to the E4 promoter in cells infected with wt, ΔE1A, or E1A Δ178-184 hAd5. Although E1A Δ178-184 is enriched on the E4 promoter greater than 10-fold above background, the recruitment of GCN5 is no different from that of the ΔE1A virus (Fig. 2C). These results confirm that the newly identified interaction site in CR3 is critical for efficient recruitment of GCN5 to the adenoviral E4 promoter.
The ability of the CR3 domains of representative E1As from each hAd species to coimmunoprecipitate GCN5 was tested using cells cotransfected with expression vectors for myc-EGFP-fused E1A CR3 and FLAG-tagged mGCN5. The CR3 regions of E1A proteins representing each hAd species were sufficient for binding to FLAG-tagged mGCN5, although the hAd4 E1A CR3 interaction was weaker (Fig. 3A). To determine the functional role that GCN5 plays in E1A CR3-dependent transactivation, GCN5 was depleted in HeLa cells by specific siRNA. Knockdown was most effective at 20 nM siRNA, based on WB for endogenous GCN5 (Fig. 3C, inset). HeLa cells were transfected with this dose of siRNA and subsequently cotransfected with a Gal4-responsive luciferase reporter and an expression vector for E1A CR3 fused to the Gal4DBD. Surprisingly, the transactivation function of hAd12, hAd5, and hAd40 E1A CR3 increased in cells transfected with GCN5-specific siRNA compared to cells transfected with control siRNA (Fig. 3C). The ability of hAd3 and hAd9 E1A CR3 to activate transcription was not significantly increased upon depletion of GCN5. These were the two weakest activators of transcription in our panel of E1A CR3s (1). Furthermore, these CR3s showed an increase in transcriptional activation in GCN5 siRNA-treated cells compared to control siRNA-treated cells; however, this difference was not significant (P > 0.05). The transactivation function of hAd4 E1A CR3 was unaffected by GCN5 knockdown (Fig. 3C); this may relate to the fact that the interaction between GCN5 and hAd4 E1A CR3 was the weakest of the hAds tested (Fig. 3A). Since depletion of GCN5 resulted in an overall increase in transcriptional activation by a subset of E1A-CR3s, it appears that GCN5 plays a repressive role in E1A CR3-dependent transactivation. The N terminus of E1A also interacts with GCN5 via residues 26 to 35 (23), and it is also capable of activating transcription of a Gal4-responsive promoter when fused to the Gal4DBD (7). However, the transactivation function of the N terminus of E1A was unaffected by GCN5 depletion (Fig. 3C). Thus, although both regions of E1A can bind GCN5, these interactions have different effects on transcriptional activation as Gal4DBD fusions.
If GCN5 is a repressor of E1A CR3-stimulated transcription and is limiting in the cell, overexpression of GCN5 should decrease E1A CR3-dependent transactivation. To test this, we cotransfected human HT1080 cells with a Gal4-responsive luciferase reporter, an expression vector for hAd5 E1A CR3 fused to the Gal4DBD, and either increasing amounts of an expression vector for FLAG-tagged mGCN5 (pCMX-FLAG mGCN5) or an empty vector (pCMX-FLAG) as a control. Ectopic expression of GCN5 resulted in a dose-dependent repression of E1A CR3 transactivation (Fig. 4A). At the highest dose used (1:1 ratio of pCMX-FLAG mGCN5 to pM-hAd5 CR3), a 50% reduction in E1A CR3 transactivation was observed (Fig. 4A). In a similar manner, overexpression of GCN5 resulted in a 40% decrease in transactivation of the viral E4 reporter by full-length E1A (Fig. 4B). Consistent with the ChIP data, mutations in either the binding site for GCN5 within the N terminus of E1A (E1A Δ26-35) or CR3 (E1A Δ178-184) rendered transactivation of the E4 reporter by full-length E1A significantly less sensitive to overexpression of GCN5 (Fig. 4B). Taken together, these data indicate that there is limited availability of GCN5 in this cell line and that the repressive effect of GCN5 on transactivation of the E4 promoter by full-length E1A requires both binding sites to efficiently interact with and recruit GCN5 to the E4 promoter.
GCN5 has a well-established role as a transcriptional coactivator that acetylates histone H3 at K9 and K14 and facilitates transcriptional elongation in the context of the SAGA complex (4, 9, 32, 40). Although the viral genome is initially devoid of cellular histones, recent reports demonstrate that histones, including H3, bind to viral DNA during the early phases of infection (15, 22, 41). These reports indicated that the H3 associated with the viral genome is acetylated at K9 and K14 during virus infection, likely by GCN5. However, no direct role for E1A or GCN5 in this process was established. We hypothesized that E1A may be retargeting GCN5 to viral templates in order to utilize its KAT activity to regulate viral transactivation. To test this hypothesis, we treated cells with the pharmacological inhibitor cyclopentylidene-[4-(4′-chlorophenyl)thiazol-2-yl]hydrazone (CPTH2) (13). CPTH2 specifically inhibits GCN5-dependent acetylation of histone H3 K14 in vitro and in vivo at micromolar concentrations (13). We first titrated the effect of CPTH2 on hAd5 E1A CR3 transactivation. Human HT1080 cells were cotransfected with a Gal4-responsive luciferase reporter and an expression vector for either Gal4DBD alone or hAd5 E1A CR3 fused to the Gal4DBD. Cells were then treated with either vehicle (DMSO) or a range of concentrations of CPTH2 and assayed for luciferase activity 48 h posttreatment. The fold activation by hAd5 E1A CR3 over Gal4DBD alone treated with DMSO was set to 100%. Treatment of cells with CPTH2 resulted in an increase in E1A CR3-stimulated transcription. This effect peaked at a dose of 50 μM CPTH2 and increased E1A CR3-dependent activation by almost 60% (Fig. 5A). Pharmacological inhibition of GCN5 KAT activity mimicked RNAi depletion of GCN5, indicating that the KAT activity is required to exert the repressive effect on E1A CR3 transactivation. HT1080 cells cotransfected with an E4-responsive luciferase reporter and an expression vector for full-length wt hAd5 E1A (289R) were also treated with CPTH2, and this similarly increased E1A transactivation compared to vehicle control (Fig. 5B). Moreover, mutants lacking either GCN5 binding site on E1A showed no significant change in transactivation in the presence of CPTH2 (P > 0.05), further demonstrating the necessity of both interaction surfaces for proper recruitment and utilization of GCN5 to the E4 promoter by E1A (Fig. 5B).
To further evaluate the repressive role of GCN5's KAT activity in the context of viral infection, we used hat/hat MEFs. In these cells, both copies of GCN5 were replaced with GCN5 harboring a mutation in the catalytic residues of the KAT domain that abolishes KAT activity (11). We measured expression levels of hAd5 E4orf6/7 mRNA during infection by qRT-PCR. The hat/hat MEFs or wt littermate control MEFs were infected at an MOI of 2 with either ΔE1A, wt E1A, E1A Δ26-35, or E1A Δ178-184 hAds. At 16 h postinfection there was a significant increase in E4orf6/7 levels in wt hAd5-infected hat/hat MEFs relative to the wt littermate control MEFs (Fig. 6A). The adenoviruses harboring E1As with mutations in either GCN5 binding site showed no significant difference in E4orf6/7 levels between hat/hat and control MEFs, again indicating that both binding sites for GCN5 on E1A are important for affecting E1A transactivation (Fig. 6A). This result in the context of the viral genome is consistent with results obtained using the E4 luciferase reporter and the small-molecule inhibitor CPTH2 (Fig. 5B). We also tested the effect of CPTH2 on E4orf6/7 expression in the context of viral infection. As seen with the reporter assays, there was a substantial increase in E1A transactivation upon treatment of wt MEFs with CPTH2 (Fig. 6B, black bars). However, CPTH2 treatment did not increase E1A-dependent transactivation of the E4 promoter to the same extent as that observed in the hat/hat MEFs, indicating that CPTH2 is not as effective as the genetic knockout at tolerable doses. Importantly, CPTH2 had no effect on E1A transactivation in the hat/hat MEFs compared to vehicle (Fig. 6B). This strongly suggests that the effects of CPTH2 in this system are specifically mediated by inhibition of GCN5 KAT activity in vivo rather than through effects on other related KATs. Furthermore, these data provide additional in vivo support for a model in which E1A uses the KAT activity of GCN5 to repress transactivation of the viral E4 promoter in the context of infection.
The results described above suggested that E1A utilizes GCN5 as a repressor, possibly to fine-tune transactivation of the early viral promoters in order to generate and maintain an optimal environment for virus replication. We determined the consequence of blocking GCN5 KAT activity on virus replication in permissive cells, using CPTH2. Human A549 cells, the diagnostic cell line for hAd isolation and analysis, were infected at an MOI of 5 with wt hAd5 and treated with either vehicle (DMSO) or 50 μM CPTH2. At 96 h postinfection, cells were harvested and freeze-thawed three times, and virus titer was determined on HEK293 cells. Chemical inhibition of the KAT activity of GCN5 resulted in a 1-log-unit reduction in virus growth at 50 μM CPTH2 (Fig. 7). These data indicate that the KAT activity of GCN5 is required for optimal virus growth and suggest that E1A transactivation must be tightly regulated by both coactivators and repressors in order to maximize virus growth.
Recent studies have shown that histone H3 associates with the viral genome during infection and that it is acetylated during the course of infection, presumably by GCN5 (15, 22, 41). Given that the predominant function of GCN5 on cellular chromatin is to acetylate histone H3 on K9 and K14, it is possible that GCN5 is needed to modify cellular histones associated with the hAd genome. Using ChIP analysis, we examined the ratio of K9/K14-acetylated H3 to total H3 on the viral E4 promoter to determine if GCN5 recruitment affected H3 acetylation. We found a significant difference in the ratio of K9/K14-acetylated H3 to total H3, or hyperacetylation of H3 K9/K14, occupying the E4 promoter of cells infected with wt hAd (Fig. 8A). Consistent with our data as a whole, we did not observe hyperacetylated H3 K9/K14 on the E4 promoter in cells infected with ΔE1A or E1A Δ178-184 viruses. There is no significant difference in the ratio of K9/K14-acetylated H3 to total H3 between cells infected with ΔE1A or E1A Δ178-184 viruses (Fig. 8A). Furthermore, infection of wt MEFs with wt hAd results in a similar hyperacetylation of histone H3 K9/K14 associated with the viral E4 promoter that is lost in hat/hat MEFs (Fig. 8B). These results suggests that E1A CR3-mediated recruitment of the KAT activity of GCN5 leads to hyperacetylation of histone H3 on K9 and K14 associated with the viral E4 promoter.
Taken together, our data indicate that E1A recruits GCN5 to the viral E4 promoter through two distinct interactions, resulting in an overall decrease in E1A CR3-dependent transcriptional activation. We also demonstrated that this effect requires the KAT activity of GCN5. On the basis of this information, we hypothesized that the decrease in transcription at the viral E4 promoter would also correlate with decreased occupancy of active RNA polymerase II phosphorylated on its C-terminal domain (phospho-CTD RNAPII) (42). To directly determine if the KAT activity of GCN5 recruited to the E4 promoter by E1A could affect the activation/phosphorylation status of the RNAPII-CTD, we determined the occupancy of phospho-CTD RNAPII and total RNAPII on the viral E4 promoter by ChIP using phospho-CTD RNAPII-specific and pan-RNAPII antibodies, respectively, in wt versus hat/hat MEFs and calculated the relative recruitment of phospho-CTD RNAPII to total RNAPII. There was an almost 2-fold enrichment of phospho-CTD RNAPII on the viral E4 promoter in hat/hat MEFs compared to wt control MEFs infected with wt hAd (Fig. 9A). However, there is no significant difference in the ratio of GCN5 to E1A recruited to the E4 promoter (Fig. 9B). Therefore, the KAT activity of GCN5 recruited to the E4 promoter by E1A increased the local acetylation of histone H3 K9K14 yet ultimately repressed activation by limiting phosphorylation of RNAPII-CTD (Fig. 9C).
The CR3 portion of hAd5 E1A is a potent transcriptional activation module and serves as a paradigm of viral transactivation (5, 16, 34). CR3 is the most highly conserved of the four conserved regions within E1A, yet there are dramatic differences in the potency of the CR3 domains from six representative E1As with respect to transactivation (1). We have previously demonstrated that these differences are independent of the conserved coactivators of E1A CR3: MED23, TBP, SUG1, pCAF, and p300/CBP (1). Moreover, competition experiments suggest that additional cellular factors are involved in E1A CR3 transactivation and may include negative regulators (1). We report here the identification of an interaction surface in E1A CR3 for the cellular KAT GCN5. Furthermore, our experiments indicate that GCN5 is a novel negative regulator of E1A transactivation that may function by altering the acetylation state of the viral genome assembled with cellular histones into chromatin and the phosphorylation status of RNAPII at the viral E4 promoter.
Although the interaction of GCN5 with the N terminus of E1A had been reported previously, these experiments used hAds in a background in which only the 12S (243R) isoform of E1A was synthesized and which thus did not contain CR3 (23). We repeated these interaction experiments using viruses harboring the same mutation, i.e., deletion of residues 26 to 35 of E1A, in a background that synthesizes the larger 13S-encoded (289R) isoform of E1A. In this context, the E1A Δ26-35 mutant retained interaction with GCN5 (Fig. 1A), suggesting that GCN5 also binds to a second site present within CR3 of E1A. We confirmed that CR3 alone is sufficient for interaction with GCN5 (Fig. 3A). The GCN5 interaction site within CR3 was mapped to require at least residues 178 to 188, which excludes the residues required for interaction with SUG1 (Fig. 1C) (38). The promoter-targeting domain of E1A CR3 was originally shown to be residues 183 to 188, which does overlap this newly identified GCN5 binding site (17, 26, 53). However, we have shown using ChIP analysis that E1A Δ178-184 is still recruited to the E4 promoter (Fig. 2B and andC).C). Co-IP of E1A CR3 with GCN5 required residues 178 to 184 of CR3 in reciprocal Co-IP scenarios indicating a robust and specific interaction. Although the Δ178-184 mutant of CR3 appears to be impaired for interaction with MED23 and TBP, we have shown that the GCN5 interaction is independent of interaction with pCAF, MED23, and TBP using key point mutants that disrupt specific interaction with TBP, MED23, and pCAF (V147L, H160Y, and Δ139-147, respectively). Furthermore, the GCN5 interaction is independent of the zinc finger binding subdomain of CR3. Taken together, these data suggest that the Δ178-184 mutant of CR3 may impact the integrity of the zinc finger subdomain, which would explain the defects in MED23 and TBP binding as well as the impaired activation function of this CR3 mutant. Importantly, E1A CR3Δ178-184 retains interaction with the closely related pCAF KAT. Thus, this mutant further differentiates these two closely related KATs, which both interact with E1A yet appear to have opposing roles in E1A transactivation. Perhaps these two KATs are associated with distinct complexes that bind CR3 through distinct interaction surfaces. Despite indications that CR3 is indeed a structured region of E1A, no detailed information on the structure of CR3 is currently available, but the possibility that E1A-CR3 could be recruiting both activating and repressive modules through distinct surfaces is enticing.
The interaction of GCN5 with CR3 was conserved across all six hAd species, suggesting that the functional role of GCN5 is also conserved (Fig. 3A). There is a dichotomy of conservation at the primary amino acid level among hAds in the region required for GCN5 interaction (Fig. 3B). Residues 178 to 181 of E1A are poorly conserved among the six representative hAds in our panel. However, residues 182 to 188 share a much greater degree of conservation. Residue 182 in species A through D is phenylalanine; however, in species E and F it is cysteine, which may contribute to the poor interaction of GCN5 with hAd4 CR3. Residues 183 through 188 are very highly conserved among E1As. S185 and S188 are phosphorylated in vivo and are critical for E1A-dependent activation of transcription for the E4 promoter (54).
The primary function of the largest E1A proteins is to activate early viral gene expression by recruiting the cellular transcription machinery to the early viral promoters (5, 16, 34). Thus, GCN5 would be expected to be present at the viral promoter with E1A, and indeed GCN5 is associated with the viral E4 promoter in an E1A-dependent manner (Fig. 2B and andC).C). Interestingly, the recruitment of GCN5 to the E4 promoter requires both GCN5 binding sites within E1A, demonstrating for the first time a physical cooperation between the N terminus of E1A and CR3 to recruit a cellular protein to the viral E4 promoter (Fig. 2B). It should be noted that although an E1A mutant lacking residues 26 to 35 (the N-terminal GCN5 binding site) can still coimmunoprecipitate GCN5 (Fig. 1A), this N-terminal binding site is also required to efficiently recruit GCN5 to the viral E4 promoter (Fig. 2B and andC).C). Furthermore, we demonstrated that both binding sites are required for GCN5 to exert its repressive effect on E1A transactivation of the E4 promoter from both plasmid and viral genomic templates (Fig. 4B, ,5B,5B, and and6A).6A). Despite the large body of evidence that both the N terminus of E1A and CR3 are transactivation domains, there is very little direct evidence demonstrating that the N terminus of E1A and CR3 can cooperate to interact with and recruit cellular proteins to a viral promoter (55). Several cellular proteins that bind to CR3 also interact with the N-terminal CR1 domain of E1A, including TBP, pCAF, SUG1, and p300/CBP (24, 35, 36, 38). There is also evidence that the N terminus of E1A and CR3 can cooperate functionally. For example, both regions can activate transcription when fused to a heterologous DBD (5, 16, 34) and CR3 can synergize with the N terminus to activate the hAd E2 promoter, presumably via recruitment of TBP and sequestration of pRb (55). In essence, the interaction of GCN5 with two distinct regions of E1A suggests a mechanism of cooperative recruitment to the viral promoters that may also apply to other proteins that interact with multiple regions of E1A. Similarly, mechanisms of cooperative recruitment of factors to the viral promoters may explain why both these regions of E1A are required for efficient activation and regulation of early viral gene expression (47, 55).
RNAi depletion of GCN5 resulted in an increase in E1A CR3 transactivation and provided the first indication that GCN5 is a true negative regulator of E1A transactivation (Fig. 3C). This role of GCN5 as a negative regulator was conserved among a subset of our panel of representative CR3s of each hAd species, with the exception of hAd4 E1A CR3 (species D), which also had the weakest physical interaction with GCN5 (Fig. 3A). Perhaps hAd4 E1A CR3 does not require the same level of negative regulation that the other CR3s display as a consequence of its weaker intrinsic activation function. Depletion of GCN5 resulted in an increase in transcriptional activation (Fig. 3C), whereas depletion of the other cellular factors required by CR3 (MED23, p300 SUG1, or pCAF) results in a decrease in transactivation (1). This gain of function phenotype suggests that GCN5 is part of another layer of transcriptional control that is recruited to promoters by E1A to optimize early gene expression. This is also a rare example where removing a cellular binding partner of CR3 results in enhanced E1A transactivation. The same phenotype has been previously demonstrated by depleting CtBP with siRNA or sequestering CtBP with the C terminus of E1A (10). In the context of infection, such a paradigm may be required for regulation of the hAd E4 transcription unit, as many of the gene products of the E4 region are toxic and stimulate cellular defenses and/or antagonize survival pathways (46, 54). Therefore, it appears that E1A utilizes GCN5 to help fine-tune transactivation of the E4 promoter. This pathway could similarly regulate expression of other viral early genes as well.
If GCN5 negatively regulates E1A CR3-dependent transactivation and is available in limited quantities, then ectopic expression of GCN5 should exacerbate the negative effect on CR3 transactivation. This was indeed the case, since overexpression of GCN5 resulted in a decrease in transactivation by CR3 fused to a heterologous DBD, and also by full-length E1A acting on an E4-responsive reporter (Fig. 4). It is somewhat counterintuitive that CR3 alone is still affected by GCN5, given the requirement for both sites to recruit GCN5 to the viral E4 promoter. However, it is possible that the interaction with the CR3 binding region is stronger than the N-terminal interaction (Fig. 1A), and that may be why GCN5 affects CR3 transactivation as a Gal4DBD fusion. We cannot exclude that artificial targeting of CR3 to a promoter through the Gal4DBD renders the GCN5 binding site more accessible or indirectly enhances GCN5 recruitment by direct targeting of CR3 to the promoter. However, the effect of GCN5 overexpression again highlights the cooperative nature of the GCN5 interactions with full-length E1A, as mutation of either binding site rendered full-length E1A nonresponsive to the repressive effects of GCN5 on E4 promoter activation (Fig. 4B).
GCN5 has both a KAT activity and a deubiquitinating activity in the SAGA complexes of both yeast and mammalian cells (4, 9, 32, 40). Importantly, we have demonstrated the functional effect of GCN5 on E1A transactivation using multiple experimental systems of increasing sensitivity. Depleting GCN5 by siRNA indicated that GCN5 was a negative regulator of E1A transactivation but could not distinguish which function of GCN5 was responsible, since all of its functions, including its KAT and deubiquitinating activity, were depleted along with GCN5. However, both pharmacological inhibition and targeted genetic ablation of KAT function clearly demonstrated that the repressive effect on E1A transactivation requires the KAT activity of GCN5. Although pharmacological inhibition was not as effective as genetic ablation, we observed that CPTH2 is exquisitely specific to GCN5, as this inhibitor had no effect on E1A transactivation in hat/hat MEFs (Fig. 6B), which retain expression of the closely related KAT pCAF. Both the Gal4DBD-responsive and E4 reporters, as well as direct measurement of the transcriptional output of the E4 promoter via E4orf6/7 mRNA transcript levels during infection, demonstrated that loss of GCN5 or GCN5 KAT activity resulted in an increase in E1A transactivation and that both GCN5 interaction sites in E1A were required to affect transactivation of the E4 promoter (Fig. 4B, ,5B,5B, and and6A).6A). Our results indicate that the KAT activity of GCN5 is required to negatively affect E1A transactivation on both plasmid and viral genomic templates (Fig. 5 and and66).
The hAd genome is not packaged into the virion in complex with cellular histones but has been shown to associate with cellular histones following infection. Indeed, genome-wide analysis has revealed acetylated histones across the hAd genome (15). These include H3, H3.1, and H3.3, which are acetylated at K9 and K14, presumably by GCN5, and also at K18, which is presumably acetylated by p300/CBP (15, 22, 41), yet no direct role for E1A or GCN5 in this process was demonstrated in either report. Consistent with previous reports, we show here that E1A is not required for H3 deposition and K9/K14 acetylation, as cells infected with hAd with deletion of E1A (ΔE1A) still acquire both H3 and K9/K14-acetylated H3 (H3-Ac) on the viral E4 promoter (Fig. 8) (22, 41). In the presence of E1A (cells infected with wt hAd), there is a dramatic increase in the ratio of H3-Ac to total H3, or hyperacetylation of H3 on K9 and K14 at the E4 promoter compared to cells infected with E1A deletion virus (ΔE1A) or cells infected with a virus that cannot recruit GCN5 to the E4 promoter (E1A Δ178-184) (Fig. 8). Furthermore, we demonstrated that GCN5 is recruited to the viral E4 promoter in an E1A-dependent manner (Fig. 2). The consequence of E1A-dependent recruitment of GCN5 is an increased acetylation of histone H3 at the viral E4 promoter. Therefore, our data demonstrate for the first time recruitment by E1A of a chromatin-modifying enzyme that acts on cellular histones bound to the viral genome to influence viral transcription.
Although acetylation of histones is generally associated with increases in gene expression (18, 52, 56), there are examples where increased protein acetylation is associated with transcriptional repression, including the mouse mammary tumor virus promoter (19, 43, 51). It is also well established that histone deacetylases can function as coactivators at some genes (33, 61). Furthermore, genome-wide analysis of gene expression in yeast demonstrated that approximately 80 genes (5.2% of genes analyzed) showed increased expression as a result of GCN5 deletion (19). One of these genes elevated in the gcn5 deletion background is ARG1. The KAT activity of Gcn5p is specifically required for repression of ARG1, and ChIP analysis demonstrated that decreased histone H3 acetylation at the ARG1 promoter is correlated with increased expression (39). This is the same phenomenon that we observe with the viral E4 promoter during infection. Thus, it appears that hyperacetylation of histone H3 and potentially other proteins by GCN5 can be used as a mechanism for fine-tuning transcription. Of particular note, GCN5 can acetylate lysine 48 in the catalytic core of the CDK9/pTEFb complex. This inhibits its ability to phosphorylate RNAPII CTD and reduces transcriptional elongation (42). Mechanistically, the recruitment of GCN5 to the viral promoter by E1A could acetylate factors such as CDK9 and reduce phosphorylation of RNAPII CTD, as depicted in Fig. 9C. This model is entirely consistent with our experiments using hat/hat MEFs, in which we conclusively demonstrated that the GCN5 KAT activity was essential for limiting the phosphorylation status of RNAPII and reducing activation by E1A at the E4 transcription unit.
The net effect of eliminating the KAT function of GCN5 on the hAd replication cycle was reduced virus yield (Fig. 7). The deletion or inhibition of GCN5 KAT activity coincides with an increase in E4 promoter activity, suggesting that a balanced level of E1A transactivation is required for optimal virus growth (Fig. 6). This is not unexpected, as increased transcription of the E4 promoter is known to enhance premature killing of infected cells (54). Thus, the adenoviral E4 promoter may have multiple levels of regulation. Not only does GCN5 appear to limit E1A stimulation of E4 promoter activity, but the E4 transcription unit also encodes the E4orf4 protein, which interacts with cellular PP2A to dephosphorylate S188 of E1A, completing a negative-feedback loop for E1A transactivation (54). Therefore, the negative effect of GCN5 on E1A-dependent transactivation may help provide the optimal intracellular milieu for virus replication by fine-tuning expression of the viral transcription program needed for the viral replication cycle.
We thank Sharon Roth-Dent for providing hat/hat and wt littermate MEFs and comments on the manuscript. We extend our utmost gratitude to P. White for the anti-myc clone 9E10 hybridoma and to P. Hearing for the GCN5 IP protocol. We also thank A. Ablack and B. Goulet for insightful discussions and critical review of the manuscript.
J.N.G.A. was supported by an OGSST award, and G.J.F. was supported by an OGS award from the Ontario Ministry of Training. G.T. was supported by a CGS award from the Canadian Institutes of Health Research. M.C. was supported by an award from the Strategic Training Program in Cancer Research and Technology Transfer. This work was supported by a grant from the Canadian Institutes of Health Research awarded to J.S.M.
Published ahead of print 23 May 2012