Mechanism of Adenovirus E4-ORF3-Mediated SUMO Modifications

Viruses interplay with the host SUMOylation system to manipulate diverse cellular responses. The Ad E4-ORF3 protein forms a dynamic nuclear network to interfere with and exploit different host processes, including the DNA damage and interferon responses. We previously reported that E4-ORF3 is a SUMO E3 ligase. Here, we demonstrate that this activity is a conserved function of evolutionarily diverse human Ad E4-ORF3 proteins and that E4-ORF3 functions directly to promote SUMO conjugations to multiple cellular proteins. Recruitment of cellular substrates into E4-ORF3 nuclear inclusions is required for SUMO conjugation to occur in vivo. We probed the mechanism by which E4-ORF3 functions as a SUMO E3 ligase. Only multimeric, but not dimeric, E4-ORF3 binds to the SUMO E2 conjugation enzyme UBC9 in vitro only in a trimeric complex with SUMO. These results reveal a novel mechanism by which a conserved viral protein usurps the cellular SUMO conjugation machinery.

infection, including the DNA damage response (DDR) and an interferon (IFN) response (2,3). The linear double-stranded Ad genome is sensed in the cytoplasm to stimulate type I IFN production (2). Nuclear Ad DNA is sensed as double-strand breaks to trigger a DDR (3). The DDR and IFN responses significantly inhibit Ad replication if left unabated. Consequently, Ad has evolved multiple mechanisms to interfere with these innate immune pathways.
Many cellular proteins that are relocalized by E4-ORF3 are modified by the small ubiquitin-like modifier (SUMO) (16,17). SUMO modification regulates diverse cellular processes, including transcription, replication, DNA repair, protein localization, and protein stability, mostly by modulating protein-protein interactions (18,19). SUMO conjugation to target lysine residues in substrates follows the paradigm of the ubiquitin conjugation system (18,19). The C-terminal diglycine motif of a processed SUMO is conjugated to a catalytic cysteine residue of the E1-activating enzyme. SUMO is then transferred to the catalytic cysteine of the E2-conjugating enzyme. In humans, there is one heterodimeric SUMO E1-activating enzyme (SAE1/SAE2) and one E2-conjugating enzyme (UBC9). SUMO may be directly transferred from the E2 enzyme to a substrate or the activity of an E3 ligase may promote the process. Only a limited number of SUMO E3 ligases have been identified, and much less is known about them compared with ubiquitin E3 ligases. In humans, three SUMO paralogs, SUMO-1, -2, and -3 are best characterized in this family (18,19). Mature SUMO-2 and SUMO-3 proteins are 97% identical and often referred to as SUMO-2/3. The SUMO-1 protein is ϳ50% identical to SUMO-2/3. SUMO-2/3, but not SUMO-1, can form polymeric SUMO chains via conjugation at SUMO-2/3 lysine residue 11.
We previously demonstrated that the Ad5 E4-ORF3 protein promotes SUMO modifications of multiple cellular proteins and further targets some of them for ubiquitindependent proteasomal degradation (1,12,16,17,20). Using in vitro SUMOylation assays, we showed that the Ad5 E4-ORF3 protein itself functions as a SUMO E3 ligase (20). These findings place the Ad E4-ORF3 protein at the nexus of the cellular SUMOylation system. A growing number of studies have demonstrated that viruses manipulate host SUMOylation pathways (21). Interestingly, it was found that disruption of global SUMO modification in immune cells results in enhanced type I IFN responses (22,23). This result suggests that SUMOylation plays important roles in aspects of antiviral activity as well as virus replication.
In this report, we further characterized E4-ORF3-mediated SUMOylation by analyzing E4-ORF3 proteins from evolutionarily diverse human Ads and if divergent E4-ORF3 target proteins are regulated by distinct mechanisms. Our results demonstrate that in vitro SUMO ligase activity is conserved across human Ad types, and that recruitment of cellular proteins into E4-ORF3 nuclear inclusions is required for their SUMOylation in infected cells. Using in vitro binding assays, we reveal a novel mechanism by which the E4-ORF3 protein interacts with E2 UBC9 and SUMO-3 via formation of a trimeric protein complex. Overall, these results provide mechanistic insights into the activity of a conserved viral protein whose function impacts multiple nuclear signaling pathways by a unique mechanism.
Our previous study revealed that Ad5 E4-ORF3 functions as a SUMO E3 ligase for TIF-1␥ using an in vitro SUMOylation reaction (20). We also showed that Ad5 E4-ORF3 promotes poly-SUMO-3 chain formation in the absence of substrate. Phylogenetic relationships of the E4-ORF3 proteins of different Ad types are shown in Fig. 1B. The Ad5 E4-ORF3 protein has diverged significantly from those of other Ad species. Therefore, we investigated whether this SUMO E3 ligase activity of E4-ORF3 is conserved among the Ad types. In cultured cells, all the E4-ORF3 proteins tested were able to enhance SUMO-3 conjugation to TIF-1␥ and promote TIF-1␥ degradation (Fig. 1C, pulldown and lysate, respectively). We next analyzed SUMO E3 ligase activities using an in vitro SUMO conjugation assay. Recombinant E4-ORF3 proteins were purified from Escherichia coli and their ligase activities determined in the absence or presence of a substrate, glutathione S-transferase (GST)-tagged TIF-1␥ΔCT (20). As described in our previous report (20), we optimized the reaction conditions such that weak TIF-1␥ SUMOylation was evident with E2 alone. We then incubated reaction mixtures with increasing concentrations of E4-ORF3 proteins. Consistent with the observation in Fig. 1B, the E4-ORF3 proteins of all different Ad types facilitated SUMO-3 conjugation to TIF-1␥ (Fig. 1D), as well as poly-SUMO-3 chain formation (Fig. 1E). The Ad5 E4-ORF3 L103A mutant, which is defective in the formation of polymeric nuclear tracks and subsequent E4-ORF3 functions (6), was used as a negative control. Taken together, our data indicate that SUMO E3 ligase activities of E4-ORF3 for TIF-1␥ SUMOylation and poly-SUMO chain formation are conserved across human Ad species.
TFII-I subnuclear relocalization and SUMOylation are induced by E4-ORF3 from only species C (Ad5) but not by any other Ad types in cultured cells (12). Because E4-ORF3 proteins from five different types showed E3 ligase activities for TIF-1␥ SUMOylation and poly-SUMO-3 chain formation ( Fig. 1D and E), we asked whether the ligase activity of E4-ORF3 for TFII-I SUMOylation is specific for Ad5 in vitro. We performed an in vitro SUMOylation assay of TFII-I using E4-ORF3 proteins from five types and compared SUMOylation levels. All of the E4-ORF3 proteins were able to promote SUMO-3 conjugation to TFII-I in vitro (Fig. 2C), indicating that E4-ORF3 proteins from different Ad types possess E3 ligase activities for TFII-I SUMOylation. These results imply that the defect in induction of TFII-I SUMOylation displayed by E4-ORF3 proteins from Ad types other than species C in vivo is mainly caused by a lack of their ability to relocalize TFII-I into E4-ORF3 nuclear inclusions, not by a lack of E4-ORF3 SUMO E3 ligase activities. Since we found multiple cellular proteins that are targeted for E4-ORF3-mediated SUMOylation, we asked if E4-ORF3 affects SUMOylation of other Ad viral proteins. Ad5 E1B-55K oncoprotein is a well-known SUMO substrate, and SUMO-1 modification at the lysine 104 residue is required for the transforming activity of E1B-55K (24,25). In agreement with previous reports (26,27), E1B-55K was recruited into E4-ORF3 tracks in HeLa cells infected with an E4-ORF6-deficient Ad5 mutant virus (dl355), while it colocalized with viral replication centers in the absence of E4-ORF3 and E4-ORF6 (dl355-inORF3) at 12 hpi (Fig. 3A). Next, to determine SUMOylation levels of E1B-55K, His 6 -tagged SUMO-expressing HeLa cells were infected with dl355 or dl355-inORF3 for 12 h. Since the stable HeLa cell line expressing His 6 -SUMO-1 exhibited lower Ad infectivity than did the stable His 6 -SUMO-3 HeLa cell line, we infected cells with titers indicated in Fig. 3 to obtain similar levels of viral protein expression. Consistent with previous reports, E1B-55K was modified by both SUMO-1 and SUMO-3 in Ad5-infected cells (Fig. 3B). Surprisingly, only SUMO-1, but not SUMO-3, conjugation to E1B-55K was significantly increased by E4-ORF3 (Fig. 3B), indicating that E4-ORF3 enhances SUMO-1 conjugation to E1B-55K in the absence of E4-ORF6.
Next, we determined if E4-ORF3 directly functions as a SUMO E3 ligase for Nbs1 and Mre11, which were the first identified SUMO targets affected by E4-ORF3 in cultured cells (17). Nbs1 and Mre11were purified from E. coli (GST-Nbs1 and GST-Mre11) and their SUMOylation levels assessed by incubating with Ad5 E4-ORF3 in in vitro reactions. SUMO-1 and SUMO-3 conjugation to Nbs1 occurred weakly in the absence of E4-ORF3, but the modifications were significantly increased by E4-ORF3 (Fig. 3D, top). In the case of Mre11, no SUMO-modified Mre11 was detected in the absence of E4-ORF3, but weak SUMO-1 modification and pronounced SUMO-3 modification were observed in the presence of E4-ORF3 (Fig. 3D, bottom). These results demonstrate that Ad5 E4-ORF3 indeed functions as a SUMO E3 ligase of Mre11 and Nbs1. Although Ad5 E4-ORF3 still exhibited E3 ligase activity on all of the substrates examined, E4-ORF3-induced SU-MOylation of Mre11, Nbs1, and E1B-55K was much weaker than that of TIF-1␥ and TFII-I. All of these substrates were purified under the same native (nondenaturing) conditions, implying the possible involvement of cellular factors for SUMOylation in vivo.
E4-ORF3 interacts with UBC9 in the presence of SUMO-3. We showed binding of Ad5 E4-ORF3 with UBC9 in cultured cells by coimmunoprecipitation (20). To assess whether this binding is direct or requires other factors, GST pulldown analyses were carried out using recombinant proteins, GST and GST-UBC9, and His 6 -tagged proteins E4-ORF3-WT, -L103A, and -N82A. Point mutations at the highly conserved L103 or N82 residues in E4-ORF3 still allow the formation of an E4-ORF3 dimer but completely block polymer assembly (29,30). Surprisingly, Ad5 E4-ORF3 was not coprecipitated with UBC9 alone (Fig. 5A), indicating that they are not directly associated each other. Strikingly, the addition of SUMO-3 to the binding mixture enabled efficient E4-ORF3-WT pulldown with UBC9 (Fig. 5A). In contrast, the L103A and N82A mutant proteins were completely defective in this binding (Fig. 5A), suggesting that E4-ORF3 polymerization is crucial for this interaction to occur.
Since E4-ORF3 proteins from five different Ad types were found to possess SUMO E3 ligase activities, we next questioned whether the ternary interactions of E4-ORF3, SUMO-3, and UBC9 are conserved among Ad types. We confirmed that E4-ORF3 proteins from Ad 12, 3, 9, and 4 all bound SUMO-3/UBC9 (Fig. 5C), demonstrating that the ability of E4-ORF3 to form ternary interactions with SUMO and UBC9 is conserved across Ad species.
Previous reports have shown that E4-ORF3 recruits SUMO-1 and -2/3 into the nuclear inclusions (7,17). To determine the subcellular localization of UBC9 in the presence of E4-ORF3, E4-ORF3 was expressed in HeLa cells, and endogenous UBC9 localization was analyzed by high-resolution microscopy. UBC9 mostly localized within the nucleus in a diffuse pattern, and E4-ORF3 expression did not make significant changes in this localization. However, colocalization of UBC9 and E4-ORF3 was observed at multiple sites throughout the nucleus in all cells that expressed E4-ORF3 (Fig. 5B).
E4-ORF3 associates with backside SUMO-bound UBC9. We sought to understand the mechanism of how E4-ORF3 associates with UBC9 together with SUMO-3. The C-terminal glycine of SUMO forms a thioester linkage via the catalytic cysteine of UBC9. A second SUMO moiety may bind noncovalently on the backside of UBC9 that is opposite to the active-site cysteine 93 residue (31). For the backside binding interaction, the interfaces are found to be formed by basic amino acids in UBC9 (R13, R17, and H20) and acidic amino acids in SUMO (E67 in SUMO-1, D63 in SUMO-2, and D62 in SUMO-3) (32). Since the GST pulldown assays were conducted under conditions without E1 and ATP, SUMO conjugation to the catalytic cysteine of UBC9 did not occur, suggesting the possibility of a requirement for noncovalent binding between SUMO-3 and UBC9 for interaction with E4-ORF3. To test this possibility, the UBC9 backsidebinding-deficient SUMO-3-D62R mutant was purified from E. coli, and GST pulldown assays were conducted as described above. Remarkably, E4-ORF3 was not able to interact with UBC9 in the presence of SUMO-3-D62R instead of SUMO-3-WT (Fig. 6A), indicating that backside binding by SUMO-3 to UBC9 is essential for the ternary interaction with Ad5 E4-ORF3. To examine the correlation between the E4-ORF3 SUMO E3 ligase activity and ternary interactions, in vitro poly-SUMOylation assays were carried out using the SUMO-3-WT, SUMO-3-D62R mutant, and polymerization-defective SUMO-3-K11R mutant in the presence of Ad5 E4-ORF3 in the absence of substrate. As expected, the SUMO-3-K11R polymerization mutant showed no SUMO conjugation. Poly-SUMOylation of the SUMO-3-D62R mutant occurred poorly in the absence of E4-ORF3, and the addition of increasing amounts of E4-ORF3 exhibited only weak changes in D62R chain formation compared to wild-type SUMO-3 (Fig. 6B). These results support the conclusion that backside binding of SUMO-3 to UBC9 is required for enhanced poly-SUMO chain formation by E4-ORF3.

DISCUSSION
SUMO E3 ligases increase the rate of SUMO transfer from the E2-conjugating enzyme UBC9 to substrates by recruiting SUMO-charged UBC9 and substrates into a complex and stabilizing SUMO-charged UBC9 in an activated closed conformation (18,19). The activities of known cellular SUMO E3 ligases fall into two categories. Members of the Siz/PIAS family of SUMO E3 ligases contain a RING domain which interacts with SUMO-charged UBC9. Siz/PIAS proteins are thought to function in an analogous manner to RING-type ubiquitin E3 ligases and function as adaptors to promote SUMO-UBC9 and substrate interaction to facilitate conjugation. RANBP2 and ZNF451 are SUMO E3 ligases that lack a RING domain. These proteins bind SUMO and SUMOcharged UBC9 and promote SUMO conjugation by serving as adaptor proteins between substrates and SUMOylation components. In this study, we investigated how the Ad E4-ORF3 protein functions as a SUMO E3 ligase and define parameters associated with substrate specificity.
The Ad5 E4-ORF3 protein has been shown to relocalize SUMO and SUMOylated substrates into nuclear inclusions to facilitate SUMO conjugation in infected cells (17). The structure of the E4-ORF3 protein has been solved (30), and E4-ORF3 does not have a RING domain. E4-ORF3 residues 103 to 107 are a predicted SIM, but these sequences are not critical for E4-ORF3 SUMO ligase activity (20). Here, we showed that the E4-ORF3 protein does not directly bind either SUMO or UBC9 alone. Instead, E4-ORF3 binds UBC9 only in a trimeric complex with backside-bound SUMO (Fig. 5 and 6). Charging of UBC9 via thioester linkage of SUMO to the catalytic cysteine residue is not required for these interactions to occur. Furthermore, our results using E4-ORF3 N82A and L103A mutants revealed that E4-ORF3 polymerization is crucial for this interaction. Based on these results, we propose a model for a trimeric interaction of E4-ORF3 with UBC9 and SUMO (Fig. 6D). This trimeric interaction could be due to a conformational change of UBC9 by SUMO backside binding. However, this possibility seems unlikely since backside binding of SUMO to UBC9 does not promote any significant change in UBC9 structure (33,34). Rather, we think it is more likely that E4-ORF3 binds to a newly Sohn and Hearing ® emerged interface created by UBC9 and backside-bound SUMO. We hypothesize that E4-ORF3 functions to recruit the SUMO machinery, including UBC9 (Fig. 5B), and its substrates in proximity of one another using its polyvalent scaffold (29,30) to assemble higher-order protein complexes.
To elucidate the characteristics of E4-ORF3-mediated SUMO conjugation and conservation of function, we analyzed E4-ORF3 proteins from different Ad species and divergent substrate proteins. The in vitro E3 ligase activity of Ad5 E4-ORF3 was conserved across the Ad types tested (species A to E) (Fig. 1), even though Ad5 E4-ORF3 shares only 50% or less amino acid sequence identity with E4-ORF3 of other Ad types. In infected cells, SUMO conjugation to proteins occurred only when substrates were recruited into E4-ORF3 nuclear inclusions generated by each Ad type (12,16,17,20). Why the E4-ORF3 proteins of different Ad species differ in their abilities to relocalize certain cellular proteins such as Nbs1, Mre11, and TFII-I (7, 12) is not understood structurally or functionally. We previously suggested that relocalization of cellular proteins by E4-ORF3 is a prerequisite for their SUMOylation (17). Since the SUMOylation-deficient TIF-1␥ protein was efficiently recruited into E4-ORF3 inclusions (Fig. 4), SUMOylation of target proteins is not a requirement for E4-ORF3mediated relocalization, confirming that relocalization is followed by SUMO conjugation. Collectively, our data demonstrate that both E4-ORF3 polymerization, which is essential for its in vitro SUMO E3 ligase activity, and substrate recruitment into the nuclear scaffold are required for E4-ORF3-mediated SUMOylation in infected cells.
We previously reported that E4-ORF3 promotes SUMO-2 or SUMO-3 conjugation to substrates with a higher efficiency than SUMO-1 in infected cells with Nbs1, Mre11, and, to a lesser extent, TFII-I (12,17). Here, we found that E4-ORF3 preferentially formed a ternary interaction with SUMO-3/UBC9 compared to SUMO-1/UBC9 (Fig. 6), consistent with these results. A greater extent of SUMO-3, compared to SUMO-1, conjugation to these substrates was recapitulated in vitro ( Fig. 2 and 3), indicating that preferential binding of E4-ORF3 to SUMO-3/UBC9 may represent the basis for these results. We also observed E4-ORF3-mediated SUMO-1 conjugation to the Ad5 E1B-55K protein in vitro and enhancement of E1B-55K SUMO-1, but not SUMO-3, conjugation by E4-ORF3 in infected cells (Fig. 3). This is the first demonstration of the involvement of the E4-ORF3 protein in the regulation of E1B-55K SUMOylation, a process that is critical for certain activities of this protein (24,25). Interestingly, E4-ORF3 cooperates with the Ad E1A and E1B proteins to transform rodent cells, and the E4-ORF3 protein was found to interact with E1B-55K (35). Since SUMO-1 modification of E1B-55K is required for its transformation properties (24), it seems likely that the role of E4-ORF3 in transformation may be to function as a SUMO E3 ligase and promote E1B-55K SUMOylation.
As we and others previously reported (12,15,20), E4-ORF3-mediated SUMOylation results in proteasomal degradation of TFII-I and TIF-1␥ but not Mre11 and Nbs1. The basis for this discrimination in the outcome of E4-ORF3-induced SUMOylation of cellular substrates in not clear at this time. In this study, we confirmed that E4-ORF3 functions as a SUMO E3 ligase of Mre11 and Nbs1, although in vitro SUMOylation levels of Mre11 and Nbs1 were much lower than those of TIF-1␥ and TFII-I, implying a possibility for the involvement of cellular factors. The types of SUMO modification that occur with different substrates targeted by E4-ORF3, e.g., mono-versus poly-SUMOylation, may explain the different outcomes of the process. A wide range of different DNA and RNA viruses impact host protein SUMOylation (36). However, only a few viral proteins have been described as SUMO E3 ligases. Kaposi's sarcoma-associated herpesvirus (KSHV) encodes a SUMO E3 ligase, K-bZIP, that is SIM dependent and specific for SUMO-2/3 modification (37). The Ad E1B-55K protein is a SUMO E3 ligase that is SUMO-1 specific (38). E1B-55K regulates SUMOylation by an unknown mechanism. The Ad E4-ORF3 protein functions as a SUMO E3 ligase by an entirely novel mechanism. It binds a backside-bound SUMO-UBC9 complex and tethers this complex with a number of different cellular substrates through higher-order protein-protein interactions. E4-ORF3 directs both SUMO-1 and SUMO-2/3 conjugation and regulates cellular proteins involved in a range of functions, including IFN signaling, the DDR, and transcriptional regulation. monoclonal antibody, Arnold Levine (Princeton University) for the anti-DBP mouse monoclonal antibodies, and Peter van der Vleit (University Utrecht) for the anti-DBP rabbit antibody. We thank Ruth Rimokh (University Lyon) for the SUMOylation-deficient TIF-1␥ mutant construct. We are very grateful to David Ornelles (Wake Forest University) for advice regarding the E1B protein expression vector in E. coli and Justin Gardin (Stony Brook University) for designing the reconfigured E1B vector. We thank Guo-Wei Tien (Stony Brook University) for help with the high-resolution microscopy and members of our laboratory for informed discussions. We thank Aniska Chikhalya for critical review of the manuscript. This work was supported by NIH grant CA122677 to P.H. The funding agency had no role in the design of the study, data collection or interpretation, or the decision to submit the work for publication. The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. S