Adenovirus E1A Inhibits SCFFbw7 Ubiquitin Ligase*

The SCFFbw7 ubiquitin ligase complex plays important roles in cell growth, survival, and differentiation via the ubiquitin-proteasome-mediated regulation of protein stability. Fbw7 (also known as Fbxw7, Sel-10, hCdc4, or hAgo), a substrate recognition subunit of SCFFbw7 ubiquitin ligase, facilitates the degradation of several proto-oncogene products by the proteasome. Given that mutations in Fbw7 are found in various types of human cancers, Fbw7 is considered to be a potent tumor suppressor. In the present study, we show that E1A, an oncogene product derived from adenovirus, interferes with the activity of the SCFFbw7 ubiquitin ligase. E1A interacted with SCFFbw7 and attenuated the ubiquitylation of its target proteins in vivo. Furthermore, using in vitro purified SCFFbw7 component proteins, we found that E1A directly bound to Roc1/Rbx1 and CUL1 and that E1A inhibited the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex but not that of another RING-type ubiquitin ligase, Mdm2. Ectopically expressed E1A interacted with cellular endogenous Roc1/Rbx1 and CUL1 and decelerated the degradation of several protooncogene products that were degraded by SCFFbw7 ubiquitin ligase. Moreover, after wild-type adenovirus infection, adenovirus-derived E1A interacted with endogenous Roc1/Rbx1 and decelerated degradation of the endogenous target protein of SCFFbw7. These observations demonstrated that E1A perturbs protein turnover regulated by SCFFbw7 through the inhibition of SCFFbw7 ubiquitin ligase. Our findings may help to explain the mechanism whereby adenovirus infection induces unregulated proliferation.

Ubiquitylation is a post-translational modification that modulates stability, localization, and function of proteins (1). Ubiquitin forms a polyubiquitin chain via an isopeptide linkage to a lysine of the target protein, which then efficiently targets the proteins for degradation via the 26S proteasome. The abundance of several proteins is regulated by the ubiquitin-depend-ent proteolytic pathway (2). Protein ubiquitylation is catalyzed by a cascade reaction involving three enzymes, ubiquitin-activating enzyme (E1), 3 ubiquitin-conjugating enzyme (E2), and ubiquitin ligase (E3). The E3 components are primarily involved in the specific recognition of their target proteins (3). The RING finger is one of the most characterized motifs and is essential for E3 ubiquitin ligase activity (4). The SCF complex is a multisubunit RING finger type ubiquitin ligase that catalyzes the ubiquitylation of various proteins involved in cell cycle control and signal transduction. The SCF complex consists of three invariable components, Roc1/Rbx1 (RING finger protein), Skp1 (adapter protein), and CUL1 (scaffold protein), as well as one variable F-box protein that serves as a receptor for the target proteins (5)(6)(7).
Fbw7 (F-box and WD repeat domain-containing 7, also known as Fbxw7, Sel-10, hCdc4, or hAgo) is an F-box protein that binds to and ubiquitylates some key regulators of cell growth, including cyclin E (8), c-Myc (9, 10), c-Myb (11,12), c-Jun (13), and Notch (14). Given that Fbw7 is responsible for degradation of several protooncogene products, Fbw7 can be considered to be a tumor suppressor. Indeed, mutations in Fbw7 were found in various types of cancers, and Fbw7 ϩ/Ϫ mice have greater susceptibility to radiation-induced tumorigenesis than wild-type mice (15)(16)(17)(18). It has been proposed that F-box proteins often recognize a distinct consensus amino acid (aa) sequence that contains a phosphorylation site. The consensus phospho-binding motif for Fbw7, termed the Cdc4 phosphodegron, is found in its substrates (19). Substrate phosphorylation plays an important role in interaction with an F-box protein for ubiquitylation. It has been reported that neddylation of CUL1 is necessary for the ubiquitin ligase activity of the SCF complex, and CAND1 binds to unneddylated CUL1 to form an inactive complex (20 -23). However, other mechanisms that regulate the activity of the SCF complex are poorly understood.
The proteins encoded by E1A (early region 1A) of human adenovirus type 5 play an important role in viral replication (24). E1A promotes DNA synthesis and immortalization of primary rodent cells, and it cooperates in cell transformation with other oncogene products, including E1B encoded by E1B (early region 1B) of human adenovirus (25). The E1A gene generates two major species of mRNA, 13S and 12S, that encode proteins of 289 and 243 aa, respectively. Although they differ in that the larger protein contains an extra internal 46-aa sequence, both isoforms do possess potent cell transforming activities. E1A protein interacts with various cellular proteins, including the pRb family proteins and p300/CBP, and perturbs cell proliferation and differentiation (26,27). pRb controls G 1 /S progression by interacting with E2F family transcription factors. Interaction of E1A with pRb stimulates E2F-dependent transcription of several growth-promoting genes (28). Therefore, E1A promotes an unregulated S-phase entry and perturbs the cell cycle. p300 and CBP are transcriptional coactivators, which are recruited to various promoters by interacting with diverse transcription factors. E1A modulates p300/CBP-mediated transcriptional activation and chromatin remodeling through direct interaction (29). However, additional uncharacterized cellular polypeptides are found in E1A immunoprecipitates of adenovirus-transformed or adenovirus-infected cells (30). Thus, the mechanism by which E1A exerts its biological activity is not fully understood.
In the present study, we found that E1A interacts with SCF Fbw7 in mammalian cells. Given that E1A and Fbw7 possess opposite functions in cell cycle regulation, we investigated whether E1A and Fbw7 interfere with each other. Although Fbw7 did not affect the stability of the E1A protein, ubiquitylation of the target proteins of SCF Fbw7 was attenuated by coexpression with E1A. Using in vitro purified components, we found that E1A directly targeted Roc1/Rbx1 and CUL1. Indeed, we observed that the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex was disturbed by E1A in vitro. Furthermore, we observed a decelerated turnover of SCF Fbw7 target proteins. These results suggest that E1A can interfere with the turnover of proteins whose stability is regulated by SCF Fbw7 ubiquitin ligase through the inhibition of the SCF Fbw7 complex. Given that loss of function of Fbw7 promotes unregulated cell growth (18,31), our findings have the potential to help explain the mechanism whereby adenovirus infection induces unregulated proliferation.
Cell Culture, Plasmid Transfection, and Adenovirus Infection-All cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics. HEK293 cells were transfected using the calcium phosphate method, whereas HeLa cells and U2OS cells were transfected using FuGENE 6 reagent (Roche Applied Science) or Lipofectamine 2000 reagent (Invitrogen). Adenovirus type 5 was propagated in HEK293 cells. The infectious units for each cell type were determined by immunofluorescence with the anti-E1A antibody. HeLa and U2OS cells were infected at 5 infectious units/cell.
Immunoprecipitation and Immunoblotting-Cells were lysed in immunoprecipitation lysis buffer (0.5% Triton X-100, 300 mM NaCl, 25 mM Tris-HCl (pH 7.5), and protease inhibitor mix). Immunoprecipitation and immunoblotting were done as described previously with some modifications (32,39). For in vivo ubiquitylation assays, cell lysates were denatured by treatment with 2% SDS and 5% 2-mercaptoethanol at 100°C to dissociate non-covalent protein associations prior to immunoprecipitation.
Retroviral Production and Transduction-For retrovirus production, pBabe-puro-based individual expression plasmids were co-transfected with gag, pol, and VSV-G envelope expressing helper plasmids into HEK293T cells using the FuGENE 6 reagent. At 18 h after transfection, the medium was replaced with Dulbecco's modified Eagle's medium supplemented with 2% fetal bovine serum and incubated for an additional 24 h. Viral supernatant was collected and filtered using a 0.45-m syringe filter, supplemented with 10% fetal bovine serum and 10 g/ml Polybrene. Cells were infected with 8 ml of the viral supernatant for 24 h. Infection was repeated once to increase infection efficiency. After infection, cells were selected with 2 g/ml puromycin for 48 h.
Recombinant Proteins and in Vitro Binding Assay-GST and GST-fused Roc1/Rbx1, CUL1, Skp1, and Fbw7 proteins were expressed in Escherichia coli BL21 (DE3) and affinity-purified with glutathione-Sepharose 4B (GE Healthcare). The proteins, with the exception of GST-fused E1A, were eluted with 50 mM glutathione and then dialyzed to remove the glutathione. In the case of GST-fused E1A, GST was removed by cleavage with PreScission Protease (GE Healthcare), and the released E1A was collected. Strep⅐Tag-fused E1A protein was expressed as described above and purified using a Strep⅐Tactin Super Flow column (Novagen). A GST pull-down assay was performed as follows. A total of 10 pmol of Strep⅐Tag-fused E1A was incubated with equal amounts (mol/mol) of GST or GST-fused pro-  . Immunoprecipitation with mouse normal IgG was performed as a control (IP: IgG). Immunoprecipitates were immunoblotted with the indicated antibodies. C, E1A directly interacts with Roc1/Rbx1 and CUL1. Recombinant E1A protein was incubated with the indicated proteins in vitro. GST-fused proteins were pulled down with glutathione-Sepharose beads and immunoblotted with anti-E1A antibody (top) or anti-GST antibody (bottom). Each of the recombinant proteins described here was bacterially expressed and purified independently. D, E1A interacts with cellular endogenous Roc1/Rbx1 and CUL1. HEK293 cells were lysed, and E1A proteins derived from the cell genome were immunoprecipitated with anti-E1A antibody. Coprecipitated endogenous CUL1 or Roc1/Rbx1 was detected by immunoblotting with anti-CUL1 and anti-Roc1/Rbx1 antibodies, respectively. WT, wild type. teins for 1 h at 4°C, and GST-fused proteins were precipitated with glutathione-Sepharose 4B. Precipitates were analyzed by immunoblotting.
In Vitro Ubiquitylation Assay-GST-HA-tagged Roc1/Rbx1 and the His 6 -FLAG-tagged CUL1 C terminus region (corresponding to aa 324 -776) were simultaneously expressed in E. coli BL21 (DE3) and co-purified with glutathione-Sepharose 4B, as described in a previous report (38). The resultant Roc1/Rbx1-CUL1 complex was incubated with 50 ng of E1 (Boston Biochem), 300 ng of UbcH5c (Boston Biochem), and 1 g of ubiquitin (Sigma) in a ubiquitylation buffer (50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 2 mM NaF, 2 mM ATP, and 0.6 mM dithiothreitol) together with or without recombinant E1A for 60 min at 37°C. The reactions were terminated and denatured in Laemmli's sample buffer at 100°C and analyzed by immunoblotting.

RESULTS
E1A Interacts with Fbw7 in Vivo-We previously reported that adenovirus E1A gene products were polyubiquitylated and degraded in a proteasome-dependent manner in mammalian cells (32). However, the ubiquitin ligase that targets E1A for ubiquitin-dependent degradation has not been identified. We noticed that there was a possible Cdc4 phosphodegron sequence (11,19) within the 13S E1A gene product but not within the 12S product (aa residues 183-191 of 13S E1A). Fbw7/Cdc4 recognizes the Cdc4 phosphodegron and binds to the substrates via a WD40 repeat domain. Thus, we first hypothesized that Fbw7 mediates the ubiquitin-dependent degradation of 13S E1A protein as a ubiquitin ligase. To address this hypothesis, we investigated whether Fbw7 interacts with E1A in mammalian cells. The expression plasmid for HA-Fbw7 was cotransfected into HEK293 cells together with the plasmid for 4ϫFLAG-13S or -12S E1A. When FLAG-tagged 13S E1A was immunoprecipitated with anti-FLAG antibody, Fbw7 was coprecipitated. However, unexpectedly, Fbw7 was also coprecipitated with 12S E1A (Fig. 1).
Next, to analyze the effect of Fbw7 on 13S E1A protein stability, we chased protein levels of E1A by immunoblotting after inhibiting de novo protein synthesis in the cells with cycloheximide (CHX chase assay). The result indicated that coexpression of Fbw7 has no apparent effect on the degradation of 13S E1A (supplemental Fig. S1). Taken together, Fbw7 physically interacts with E1A without affecting the turnover of E1A.
E1A Attenuates Ubiquitylation of SCF Fbw7 Target Proteins in Mammalian Cells-Although E1A usually affects the level of protein expression by its action at the transcriptional level through the interaction with pRb and p300/CBP (21,40,41), it has been shown that E1A alters certain protein expression lev-els by affecting regulation mechanisms of the protein stability (42,43). Thus, we supposed that the ubiquitylation activity of the SCF Fbw7 complex is affected by E1A. To test this hypothesis, we assessed the effect of E1A on c-Myb ubiquitylation, which is mediated by the SCF Fbw7 complex and then followed by proteasomal degradation (11,12) in mammalian cells (in vivo ubiquitylation assay). Myc epitope-tagged c-Myb (c-Myb-Myc), FLAG-ubiquitin, and HA-Fbw7 expression plasmids were cotransfected into HEK293 cells with or without a His 6 -Xpresstagged E1A (hereafter referred to as Xpress-E1A) expression plasmid. To dissociate proteins noncovalently associated with c-Myb, cell lysates were incubated under denaturing conditions, and subsequently c-Myb-Myc was immunoprecipitated with anti-Myc antibody. The immunoprecipitates were subjected to immunoblotting using an anti-FLAG antibody to evaluate the ubiquitylation levels of c-Myb ( Fig. 2A, Denature, IP: ␣Myc). Consistent with previous studies, coexpression of Fbw7 enhanced c-Myb ubiquitylation (lanes 3 and 4). Importantly, when E1A was coexpressed, c-Myb ubiquitylation was markedly diminished (lane 5).
We also confirmed that the ubiquitylation level of c-Myb protein was attenuated by coexpression of E1A in U2OS cells, which do not express any viral oncoproteins (supplemental Fig. S2).
Substrate specificity of SCF ubiquitin ligase is altered by the substrate recognition subunit F-box protein. To verify whether E1A specifically affects SCF Fbw7 , we assessed the effect of E1A on IB␣ ubiquitylation, which is mediated by the SCF Fbw1/␤TrCP complex (34,44). Although the ubiquitylation of IB␣ is facilitated by Fbw1 (Fig. 2D, lanes 3 and 4), E1A did not affect it (Fig. 2D, lane 5). Taken together, these results suggested that E1A specifically inhibited the E3 ubiquitin ligase activity of SCF Fbw7 .
E1A Directly Targets Roc1 and CUL1-Fbw7 recognizes phosphorylated substrates via a WD40 repeat domain and binds to Skp1 through the F-box to recruit the remainder of an SCF ubiquitin ligase complex (45). To determine whether these domains are required for the interaction between Fbw7 and E1A, expression plasmids for the WD40 repeat deletion (⌬WD) or the F-box deletion (⌬F) mutant Fbw7 (Fig. 3A) were transfected into HEK293 cells, and a coimmunoprecipitation assay FIGURE 4. Effects of E1A on the SCF Fbw7 complex. A, E1A does not affect the interaction between CUL1 and Fbw7. U2OS cells were cotransfected with the indicated plasmids. FLAG-CUL1 was immunoprecipitated with anti-FLAG antibody. Coprecipitation of HA-Fbw7 was assessed by immunoblotting with anti-HA antibody. B, E1A does not affect the interaction between CUL1 and Roc1/Rbx1. Interaction between CUL1 and Roc1/Rbx1 in U2OS cells was examined by a coimmunoprecipitation assay, as described in A. C, E1A forms a ternary complex with CUL1 and Fbw7. HEK293 cells were cotransfected with indicated plasmids. Cells were lysed, and HA-Fbw7 was immunoprecipitated with anti-HA antibody. Immunoprecipitated HA-Fbw7 was eluted by competition with 0.15 mg/ml HA epitope peptide for 30 min twice so as to not disrupt protein-protein interactions (1st IP: ␣HA, Non-denaturing elution). The eluents were subjected to immunoprecipitation with anti-FLAG antibody (2nd IP: ␣FLAG). The immunoprecipitates were immunoblotted with the indicated antibodies. D, E1A forms a ternary complex with CUL1 and Roc1/Rbx1. Ternary complex formation of FLAG-CUL1, HA-Roc1/Rbx1, and Xpress-12S E1A was assessed by a coimmunoprecipitation assay as described in C.
Because Fbw7 interacts with E1A in an F-box-dependent manner, it is likely that E1A directly targets the F-box and competes with Skp1 for binding to Fbw7. However, interaction between Fbw7 and CUL1 were still observed when E1A was overexpressed in HEK293 cells (supplemental Fig. S3), suggesting that E1A does not disrupt the association of Fbw7 with Skp1. To determine the direct target of E1A, Strep⅐Tag-fused E1A and GST-fused Roc1/Rbx1, Skp1, CUL1, and Fbw7 were expressed in E. coli and purified independently. Binding of E1A to each GST-fused protein was examined using the GST pulldown assay. As shown in Fig. 3C, Roc1/Rbx1 and CUL1 still coprecipitated with E1A in a purified cell-free system, whereas Skp1 and Fbw7 as well as the control GST did not. In addition, Roc1/Rbx1 and CUL1 exogenously expressed in HEK293 cells were coprecipitated with E1A (data not shown). Thus, we concluded that Roc1/Rbx1 and CUL1 are direct targets of E1A and that E1A indirectly interacts with Fbw7 via the Roc1/Rbx1-CUL1 complex.
Interaction of E1A with Endogenous Roc1 and CUL1-To confirm the interaction of E1A with endogenous Roc1/Rbx1 and CUL1 in mammalian cells, cell lysate prepared from HEK293 cells was subjected to immunoprecipitation using an antibody against E1A or control immunoglobulin, followed by immunoblotting with antibodies against CUL1 and Roc1/Rbx1. Both CUL1 and Roc1/Rbx1 were coprecipitated with E1A in the case of precipitation with anti-E1A but not with the control (Fig. 3D), indicating that 293 genome-derived E1A interacts with endogenous CUL1 and Roc1/Rbx1. It was indicated that CUL1 was detected as double bands by immunoblotting and that the slower migrated band corresponded to the neddylated form of CUL1 (47)(48)(49). To examine whether E1A preferentially binds to the neddylated form of CUL1, HEK293 cells were transfected with control or Xpress-E1A expression plasmid. Cell lysates were subjected to immunoprecipitation using the Omni probe, which reacts with the His 6 -Xpress tag, followed by immunoblotting with an anti-CUL1 antibody. E1A preferentially coprecipitated the upper band only once among the three replicate experiments (supplemental Fig. S4A) and did not the other two times (supplemental Figs. S4, B and C). Furthermore, 293 genome-derived E1A coprecipitated both of them (Fig. 3D). Thus, we concluded that E1A interacted with CUL1 without remarkable selectivity between the neddylated and unneddylated forms.
Effect of E1A on SCF Complex Formation-To explain the mechanism by which E1A negatively regulates the SCF Fbw7 complex or how the specificity was determined, it is important to determine whether E1A interferes with the formation of the SCF complex. As mentioned above, coexpression of E1A did not interfere with the interaction of CUL1 with Roc1/Rbx1, Fbw7, and Skp1 in HEK293 cells (supplemental Fig. S3). Using U2OS cells that do not naturally express E1A, we confirmed this result more robustly. As shown in Fig. 4A, immunoprecipi-tation with anti-FLAG antibody coprecipitated HA-Fbw7 in the presence of FLAG-CUL1 (lanes 1 and 2). Coprecipitation of HA-Fbw7 was still observed under conditions when E1A was coexpressed (lane 3). Thus, it was believed that E1A did not A, schematic diagram of E1A constructs. CR1 and CR2, conserved regions in adenovirus serotypes. A summary of Roc1/Rbx1 binding is shown on the right. B and C, interactions of E1A deletion mutants with Roc1/Rbx1 (B) or with CUL1 (C) were assessed by coimmunoprecipitation assays. HEK293 cells were cotransfected with indicated plasmids. Then His 6 -Xpress-tagged E1A constructs (B) and HA-tagged CUL1 (C) were immunoprecipitated (IP) with the Omni probe and anti-HA antibody, respectively. Immunoprecipitates were analyzed by immunoblotting (IB) with the indicated antibodies.
interfere with the interaction between CUL1 and Fbw7 in U2OS cells. Based on the results from a similar coimmunoprecipitation experiment, it was also believed that E1A did not interfere with the interaction between CUL1 and Roc1/Rbx1 (Fig. 4B).
To probe more deeply into this matter, we examined whether CUL1, Fbw7, and E1A were contained in an identical complex. Myc-CUL1, HA-Fbw7, and 4ϫFLAG-E1A were coexpressed in HEK293 cells (Fig. 4C, bottom). HA-Fbw7 was immunoprecipitated with anti-HA antibody and was eluted under native conditions so as not to disrupt protein-protein interactions. Coprecipitations of Myc-CUL1 and 4ϫFLAG-E1A were confirmed as shown in Fig. 4C, middle. Then the "4ϫFLAG-E1A interacting with HA-Fbw7" was immunoprecipitated with anti-FLAG antibody. As shown in Fig. 4C, top, Myc-CUL1 was coprecipitated, indicating that CUL1, Fbw7, and E1A were contained in an identical complex. A similar coimmunoprecipitation experiment also indicated that CUL1, Roc1/Rbx1, and E1A were contained in an identical complex (Fig. 4D). Thus, both CUL1-Fbw7-E1A and CUL1-Roc1/Rbx1-E1A, respectively, were detected as a ternary complex. Taken together, these results suggested that E1A binds to an intact SCF Fbw7 complex.
Analysis of the Region within E1A Required for Interaction with Roc1 and CUL1-We next determined the region within E1A that was necessary for interaction with Roc1/Rbx1 and CUL1. For this purpose, we constructed expression plasmids for deletion mutants of E1A (Fig. 5A). Individual E1A mutants or wild-type E1A were coexpressed with HA-Roc1/Rbx1 in HEK293 cells to examine the interaction with Roc1/Rbx1 using the coimmunoprecipitation assay. As shown in Fig. 5B, E1A ⌬CR1, ⌬80 -119, ⌬CR2, and ⌬C104 mutants coprecipitated with HA-Roc1/Rbx1, whereas the ⌬N39 mutant did not. This indicated that the essential region of E1A for the interaction with Roc1/Rbx1 was located within the 39 aa of the N terminus. To examine whether the N-terminal portion of E1A is also necessary for interaction with CUL1, the wild-type or ⌬N39 mutant of E1A was coexpressed with HA-CUL1 in HEK293 cells. When HA-CUL1 was immunoprecipitated with the anti-HA antibody, wild-type E1A but not ⌬N39 was coprecipitated (Fig. 5C). This result indicated that the N-terminal por- tion of E1A was required for the association with the SCF complex. Indeed, the E1A ⌬N39 mutant failed to attenuate the ubiquitylation of c-Myb induced by Fbw7 in vivo, implying that E1A attenuated the ubiquitylation through direct interaction with the SCF Fbw7 complex (Fig. 2A, lane 6).
E1A Interferes with Roc1-CUL1 Complex-mediated Ubiquitin Polymerization in Vitro-The observation that overexpressed E1A attenuates ubiquitylation of SCF Fbw7 target proteins in vivo (Fig. 2) prompted us to investigate whether E1A directly inhibits the ubiquitin ligase activity of the SCF Fbw7 complex. It has been shown that Roc1/Rbx1 specifically binds to the C-terminal region of CUL1 (spanning aa 324 -776). Bacterially expressed Roc1/Rbx1 and CUL1-(324 -776) assemble into the complex and exert ubiquitin polymerization activity (38). Using this system with minor modifications (see "Experimental Procedures" and supplemental Fig. S5), we examined the effect of E1A on the polyubiquitin chain elongation activity of the Roc1/Rbx1-CUL1 complex in vitro. As shown in Fig. 6A, incubation of the purified Roc1/Rbx1-CUL1-(324 -776) complex with E1, E2, ubiquitin, and ATP resulted in the formation of a polyubiquitin chain (compare lane 3 with lane 4). This polyubiquitin chain formation was inhibited by recombinant E1A in a dose-dependent manner (lanes 5-7; the molar ratios of E1A to Roc1/Rbx1-CUL1 were 0.2, 2, and 20, respectively). Consistent with the binding activity (Fig. 5B), the addition of E1A ⌬N39 resulted in a remarkably reduced effect on Roc1/ Rbx1-CUL1-mediated polyubiquitin chain elongation (lanes 8 -10). In addition, we found that activation of E2 by E1 (50) was not affected by E1A (supplemental Fig. S6). Taken together, these results suggested that E1A directly inhibited the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex. Next, to examine whether the inhibition of ubiquitin ligase activity is specific for the Roc1/Rbx1-CUL1 complex, we performed an in vitro ubiquitylation assay mediated by Mdm2, a well characterized RING finger type ubiquitin ligase (51). Similar to the Roc1/ Rbx1-CUL1 complex, when purified GST-Mdm2 (39) was incubated with E1, E2, ubiquitin, and ATP, formation of a polyubiquitin chain was observed (Fig. 6B, lanes 1-3). The presence of E1A did not impact the polyubiquitin chain formation (Fig. 6B, lanes 4 -6, molar ratios of E1A to Mdm2 were 0.2, 2, and 20, respectively). In addition, ubiquitylation of p53 by Mdm2 was also not affected by E1A (supplemental Fig. S7). Thus, these results suggest that E1A selectively interferes with the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex.
E1A Alters the Protein Stabilities Regulated by SCF Fbw7 in Mammalian Cells-The above results prompted us to investigate whether E1A alters the protein stabilities that are regulated by SCF Fbw7 . We first examined the degradation rate of c-Myb exogenously expressed in HeLa cells by the CHX chase assay (Fig. 8A). The degradation rate of c-Myb was accelerated by coexpression of Fbw7 (Fig. 8A, top and second from the top panels; half-lives under these conditions were Ͼ240 min and 9.3 min, respectively). Coexpression of E1A with Fbw7 attenuated the acceleration of c-Myb degradation (Fig. 8A, third from the top panel; half-life under this condition was 28 min), suggesting that E1A inhibits the facilitation of the protein degradation by Fbw7. Similar results were observed in the degradation of cyclin E (Fig. 8B; half-lives when cotransfected with control, Fbw7, and both Fbw7 and E1A plasmids were Ͼ60, 36, and 59 min, respectively) and c-Myc (Fig. 8C; half-lives when cotransfected with control, Fbw7, and both of Fbw7 and E1A plasmids were 26, 13, and 25 min, respectively).
To clarify the role of E1A in a more physiological setting, E1A was ectopically expressed with a retrovirus vector in HeLa cells, and the effect on the degradation rate of endogenous c-Myc was assessed by the CHX chase assay. We found that the degradation of c-Myc protein was prominently retarded by the infection with the E1A expression retrovirus compared with the infection with a control virus. In contrast, degradation of cyclin D1 and p27, targets of SCF Fbx4/␣B-crystallin and SCF Skp2 (53)(54)(55), respectively, were not retarded by E1A (Fig. 9A, top). A similar result was observed in U2OS cells (Fig. 9A, bottom). We also confirmed that infection of U2OS with retrovirus expressing E1A resulted in accumulation of endogenous c-Jun and cyclin E, other targets of SCF Fbw7 . This effect seemed to be, at least partially, independent of the pRb-E2F pathway, because a similar effect was observed in pRb-inactive Saos-2 cells (Fig. 9B). In addition, SCF Fbw1/␤TrCP -mediated degradation of IB␣ in response to TNF␣ occurred normally in the presence of E1A (Fig. 9C). These observations suggest that E1A specifically alters the turnover rates of the target proteins of SCF Fbw7 in mammalian cells.
Finally, we tried to determine whether infection of wild-type adenovirus 5 affects the stability of c-Myc. As shown in Fig. 9D, the expression of E1A protein was detected in HeLa cells infected with human adenovirus 5. We confirmed that E1A associated with endogenous Roc1/Rbx1. Infection with adenovirus 5 decelerated the degradation rate of endogenous c-Myc protein in both HeLa cells and U2OS cells (Fig. 9E). Taken together, these observations suggest that adenovirus perturbed the regulation of several oncogenic protein stabilities through the inhibition of the SCF Fbw7 ubiquitin ligase complex by E1A.

DISCUSSION
In studies of protein degradation systems mediated by SCF ubiquitin ligase, a large number of analyses have been made on the modifications within the substrate in response to degradation signals and their recognition by F-box proteins. However, regulation of the activity of SCF ubiquitin ligase itself is poorly understood.
This is the first study to report that E1A inhibits SCF ubiquitin ligase. However, two major issues, the mechanism by which E1A inhibits the activity of ubiquitin ligase and the determinant of the specificity of E1A, remain unclear.
Mechanism by which E1A Inhibits the Activity of SCF Fbw7 -When addressing the mechanism by which E1A inhibits the activity of Roc1/Rbx1-CUL1 complex, at first, we neglected the possibility that E1A inhibited the E1-mediated activation of E2 in an E3-independent manner. As shown in supplemental Fig.  S6, ubiquitin was transferred to E2 in an E1-dependent manner

E1A Inhibits SCF Fbw7
either in the presence or absence of E1A. Therefore, E1A did not affect the E1-mediated activation of E2 in vitro. This observation is in accordance with the fact that E1A had no effect on the ubiquitin ligase activity of Mdm2. Thus, it seemed that E1A affected the process posterior to E2 activation.
Neddylation of CUL1 is a known modification that is required for activation of the SCF ubiquitin ligase complex (20). It was of interest to determine whether E1A inhibited the neddylation of CUL1. As shown in supplemental Fig. S8, coexpression of E1A did not alter the neddylation level of CUL1 in HEK293 cells. This result is consistent with the finding that E1A inhibited the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex in an in vitro reconstitution assay that was absent from the neddylation system. Thus, we believe that E1A represses the ubiquitin ligase activity without affecting the neddylation of CUL1. In addition, E1A interacts with CUL1 without remarkable selectivity between the neddylated and unneddylated forms of CUL1.
Furthermore, we examined the effects of E1A on the formation of the SCF complex. Evidence from the coimmunoprecipitation assay indicated that E1A did not disrupt the SCF complex. E1A was contained in an identical complex between CUL1 and Roc1/Rbx1 as well as between CUL1 and Fbw7. Therefore, E1A binds to the intact SCF Fbw7 complex and inhibits ubiquitylation of the substrate, whereas E1A does not interfere with the formation of the SCF complex.
Next, we assumed that E1A would repress the recruitment of E2 to the Roc1/Rbx1-CUL1 complex and examined it using the coimmunoprecipitation assay. Probably because the interaction between E2 and Roc1/Rbx1-CUL1 was transient and unstable (47,56), we failed to detect its presence (data not shown). Recently, Saha and Deshaies (56) displayed this transient interaction by taking advantage of a developed assay based on fluorescent resonance energy transfer. In the future, application of this system to our study may provide more concrete answers.
It has been reported that SCF dimerization facilitates ubiquitin ligase activity. The prevailing view is that the dimerization is mediated by F-box proteins, including Fbw7 and Fbw1/ ␤TrCP, and provides a preferable conformation for attachment of ubiquitin to the substrates or facilitates interaction with the substrates (57,58). Recently, Chew et al. (59) reported that CUL1 dimerized in a substrate recognition subunit-independent fashion. E1A may affect the CUL1 dimerization process. Further studies are necessary to elucidate the mechanism by which E1A represses the ubiquitylation activity of SCF ubiq-uitin ligase. Taking into consideration the mechanism by which the specificity was determined, we further discuss this problem below.
Determinant of the Specificity of E1A-Using an in vitro purified system, we showed that the ubiquitin ligase activity of Mdm2, which contains the RING finger motif, was not affected by E1A. Thus, E1A possesses target specificity among the RING-type ubiquitin ligases. Indeed, we found that the RING finger motif within Roc1/Rbx1 was not required for the interaction with E1A (supplemental Fig. S9, A and B). We next examined whether E1A targeted other cullin-based ligases. All cullins examined in this study interacted with Roc1/Rbx1 in mammalian cells, suggesting that Roc1/Rbx1 forms a Roc1/ Rbx1-cullin ubiquitin ligase complex with each class of cullin. On the other hand, E1A selectively interacted with CUL1 in the cells. Therefore, it was likely that E1A specifically interfered with the ubiquitin ligase activity of the Roc1/Rbx1-CUL1 complex but not with that of the other cullin-Roc1/Rbx1 complexes. Although purified recombinant Roc1/Rbx1 interacted with E1A in vitro, Roc1/Rbx1 coprecipitated with CUL2, CUL3, CUL4A, and CUL5 did not interact with E1A in the cells. We speculate that except for CUL1, the E1A binding sequence within Roc1/Rbx1 might be masked by other cullins that associate with Roc1/Rbx1. Alternatively, as shown in supplemental Fig. S9, C and D, the C-terminal 131 aa residues of CUL1 were required for interaction with E1A. This observation is consistent with the fact that aa 650 -776 of the E1A-binding domain in CUL1 is only 40% conserved in other cullins, such as CUL2 and CUL3. Therefore, E1A selectively binds to CUL1 and inhibits SCF-type ubiquitin ligase.
Most importantly, E1A specifically inhibits the activity of SCF Fbw7 but not the other SCF ubiquitin ligases, including SCF Fbw1/␤TrCP . SCF ubiquitin ligase contains the growth-promoting subtype SCF Skp2 (6). Adenovirus may acquire the more preferable form of the E1A gene to replicate the viral genome during the process of natural selection.
Tworkowski et al. (60) reported that E1A stabilized c-Myc protein via interaction with p400. Since the N-terminal region of E1A is required for interaction with p400 (61), it is not easy to experimentally distinguish between the effect of E1A on SCF Fbw7 and on p400. Because several substrates for SCF Fbw7 , such as c-Myc, c-Myb, c-Jun, and cyclin E, were stabilized by E1A, we believe that E1A has inhibitory activity against SCF Fbw7 . However, we cannot deny the involvement of p400 on the stabilization of c-Myc through their binding with each other. Based on the observation that E1A forms a ternary complex with CUL1 and Fbw7, it was believed that specific competition against F-box protein could not explain the specificity of the negative effect of E1A. As shown in supplemental Fig. S9, we found that E1A bound to the N-terminal portion (aa 23-42) of Roc1/Rbx1 and the C-terminal portion (aa 650 -776; from helix 28 to the C-terminal end) of CUL1 to form the ternary complex. CUL1 binds to the S1 (␤-strand 1; aa 1-23) region of Roc1/Rbx1 (62,63). E1A bound to loop 1 (aa 23-42) behind the S1 region of Roc1/Rbx1. Because E1A bound to the C-terminal portion (aa 650 -776; from helix 28 to the C-terminal end) of CUL1, the E1A binding region of CUL1 is believed to be helices 29 -31 of CUL1. When E1A binds to SCF Fbw7 , E1A could effectively inhibit the ubiquitin ligase activity of Roc1/Rbx1. In contrast, we speculate, because of the conformation of the complex, E1A may not possess effective interactions with SCF ligases with other F-box proteins, such as Fbw1, to inhibit their activity. Namely, in the case of attachment of ubiquitin to the substrates via SCF Fbw7 , it seems that the C-terminal portion of E1A inserts in the space between E2 and the substrate to interfere with the ubiquitin transfer from E2 to the substrate. On the contrary, in the case of SCF Fbw1 , E1A could not inhibit ubiquitin transfer from E2 to its specific substrates, whereas aa 1-40 of E1A also binds to Roc1/Rbx1 and CUL1. Because there may be some differences in the three-dimensional space containing the F-box protein, substrates, and E2, E1A may not able to locate in the region between E2 and the substrates. E1A inhibited the substrate-independent polyubiquitin chain elongation activity of the CUL1-Roc1/Rbx1 complex without Skp1 and the F-box protein. We speculate the reason for this is that there are no three-dimensional spatial obstacles. Analysis of the high order structure of the E1A-SCF Fbw7 complex may provide an answer as to why the specificity is yielded.
Although we investigated effects of E1A on some kinds of SCF complexes, further studies are required in order to better understand the specificity of the inhibitory activity of E1A. However, we conclusively demonstrated that E1A has selective inhibitory activity against SCF Fbw7 .
Biological Significance of Inhibition of SCF Fbw7 -Small DNA tumor viruses, such as human adenoviruses, papillomaviruses, and SV40, rely on the host cell DNA replication machinery for replication of the viral genomes. Because the host DNA replication machinery becomes available during S phase, productive infection depends upon the ability of the virus to induce S phase entry of the host cell from a quiescent state. The E1A gene is essential for the replication of human adenovirus. E1A promotes unregulated DNA synthesis and possesses potent cell transforming activity, so repression of Fbw7, which down-regulates several oncogenic proteins and is considered to be a potent tumor suppressor, is consistent with the function of the E1A gene. Indeed, conditional inactivation of Fbw7 in the T cell lineage of mice caused thymic hyperplasia as a result of c-Myc accumulation, and the mice eventually developed thymic lymphoma (31). In the present study, we proved that, after adenovirus infection, adenovirus-derived E1A bound to endogenous Roc1/Rbx1 and attenuated degradation of the endogenous target protein of SCF Fbw7 . This is direct evidence that SCF Fbw7 is a bona fide target of adenovirus in the host cell.
E1A induces the accumulation of p53 protein and apoptosis through the activation of the ARF-Mdm2-p53 pathway in the host cell (64 -66). Thus, destabilization of p53 is also an important process to utilize the host cell replication machinery. Querido et al. (67) reported that adenovirus E1B 55K and E4orf6 interacted with the Roc1/Rbx1-CUL5 ubiquitin ligase complex in a coordinated fashion, behaving as a substrate recognition subunit so that polyubiquitylation and degradation of p53 was facilitated. Since E1A did not interact with the Roc1/ Rbx1-CUL5 complex, we consider that inhibition of SCF Fbw7 is concomitant with the utilization of the Roc1/Rbx1-CUL5 complex to degrade the growth-suppressive protein p53.
The SCF complexes, which are mainly active from late G 1 to G 2 phase, are involved in the proteolysis of the core components of the cell cycle machinery. The anaphase-promoting complex/cyclosome (APC/C), which is active from the M to G 1 phase, is another macromolecular ubiquitin ligase complex that targets various proteins important for the cell division process. Turnell et al. (43) showed that APC/C interacted with a histone acetyltransferase p300/CBP and stimulated its activities. They proposed that E1A interfered with the activity of APC/C through the interaction with p300/CBP. Interestingly, E1A that lacked the N-terminal region failed to interact with p300/CBP (68). These observations suggest that the N-terminal region is essential for interference with both the SCF complex and APC/C.
It has been shown that SV40 large T antigen, an oncoprotein that is produced by SV40 DNA tumor virus, interacts with Fbw7 and interferes with Fbw7-driven substrate turnover by competition with the substrate for Fbw7 binding (69). Since direct binding of E1A to Fbw7 was not observed, it is unlikely that E1A competes with the substrates for binding to Fbw7. However, repression of Fbw7 may be a common event to stimulate the DNA replication machinery of host cells by such DNA tumor viruses. Thus, it implies that regulation of several protooncogene products by SCF Fbw7 at the post-translational level is essential for the maintenance of normal cell cycle regulation.