The adenovirus E4-ORF3 protein functions as a SUMOE3 ligase for TIF-1γ sumoylation and poly-SUMOchain elongationSook-Young Sohna and Patrick Hearinga,1

aDepartment of Molecular Genetics and Microbiology, School of Medicine, Stony Brook University, Stony Brook, NY 11794-5222

Edited by Thomas E. Shenk, Princeton University, Princeton, NJ, and approved May 6, 2016 (received for review March 8, 2016)

The adenovirus (Ad) early region 4 (E4)-ORF3 protein regulatesdiverse cellular processes to optimize the host environment forthe establishment of Ad replication. E4-ORF3 self-assembles intomultimers to form a nuclear scaffold in infected cells and createsdistinct binding interfaces for different cellular target proteins.Previous studies have shown that the Ad5 E4-ORF3 protein inducessumoylation of multiple cellular proteins and subsequent protea-somal degradation of some of them, but the detailed mechanismof E4-ORF3 function remained unknown. Here, we investigate therole of E4-ORF3 in the sumoylation process by using transcriptionintermediary factor (TIF)-1γ as a substrate. Remarkably, we discov-ered that purified E4-ORF3 protein stimulates TIF-1γ sumoylationin vitro, demonstrating that E4-ORF3 acts as a small ubiquitin-likemodifier (SUMO) E3 ligase. Furthermore, E4-ORF3 significantly in-creases poly-SUMO3 chain formation in vitro in the absence ofsubstrate, showing that E4-ORF3 has SUMO E4 elongase activity.An E4-ORF3 mutant, which is defective in protein multimerization,exhibited severely decreased activity, demonstrating that E4-ORF3self-assembly is required for these activities. Using a SUMO3 mu-tant, K11R, we found that E4-ORF3 facilitates the initial acceptorSUMO3 conjugation to TIF-1γ as well as poly-SUMO chain elongation.The E4-ORF3 protein displays no SUMO-targeted ubiquitin ligase ac-tivity in our assay system. These studies reveal the mechanism bywhich E4-ORF3 targets specific cellular proteins for sumoylation andproteasomal degradation and provide significant insight into how asmall viral protein can play a role as a SUMO E3 ligase and E4-likeSUMO elongase to impact a variety of cellular responses.

adenovirus | SUMO | E3 ligase | TIF-1γ | proteasome degradation

Adenoviruses (Ads) are ubiquitous pathogens that infect a widerange of vertebrates. Ad infection is generally associated with

mild disease, but Ads have been increasingly recognized as signif-icant pathogens in infants, the elderly, and immunocompromisedpatients (1). Ads have evolved diverse mechanisms to counteracthost antiviral responses during infection (2). Successful Ad repli-cation relies on functions provided by early region 4 (E4). Thehighly conserved E4-ORF3 protein assembles into a multimericnuclear network, referred to as tracks (3), in infected cells (4, 5).The Ad5 E4-ORF3 protein recruits numerous cellular proteinsinto nuclear tracks including promylocytic leukemia (PML) nuclearbody components (3), the Mre11-Rad50-Nbs1 (MRN) complex (6,7), small ubiquitin-like modifier (SUMO) proteins (8), transcrip-tion intermediary factor (TIF)-1α (9), TIF-1γ (10), and TFII-I (11).This event causes sequestration of the target proteins inhibitingtheir antiviral functions. Relocalization of TIF-1γ and TFII-I byAd5 E4-ORF3 results in their proteasomal degradation (12, 13).We previously showed that Ad5 E4-ORF3 mediates sumoyla-tion of multiple cellular proteins (8, 11, 13) and suggested thatE4-ORF3–induced sumoylation triggers ubiquitination and pro-teasomal degradation of some substrates. Relocalization of targetproteins into E4-ORF3 nuclear tracks is required for substratesumoylation and proteasomal degradation (8, 11, 13).

Protein modification by SUMOs is implicated in diverse cellularprocesses (reviewed in refs. 14 and 15). SUMO proteins are co-valently conjugated to lysine residues in substrate proteins by se-quential enzymatic actions. Three major SUMO proteins (SUMO1–3)have been characterized in vertebrates. SUMO2 shares 97%amino acid homology with SUMO3 (often referred as SUMO2/3)and 50% homology with SUMO1. Only SUMO2/3 can form poly-meric SUMO chains via a consensus SUMO conjugation site atlysine 11. In eukaryotes, one SUMO E1-activating enzyme complex(SAE1/SAE2), one SUMO E2-conjugating enzyme (Ubc9), andonly a small number of SUMO E3 ligases have been identified(14, 15). SUMO E3 ligases promote transfer of SUMO fromthe charged E2 enzyme to substrates. SUMO E1 and E2 en-zymes are sufficient to conjugate SUMO proteins at consensusconjugation sites (Ψ-K-X-E/D; Ψ is a large hydrophobic aminoacid, X is any amino acid), whereas SUMO E3 ligases are thoughtto be important for conjugating SUMO at nonconsensus sites.Some viral proteins function as a SUMO E3 ligase, including theAd5 E1B-55K protein (16, 17) for p53 sumoylation and Kaposi’ssarcoma-associated herpesvirus (KSHV) KbZIP (18) for sumoyla-tion of p53 and the retinoblastoma protein. Recent studies havedemonstrated that poly-SUMO2/3 chains serve as a ubiquitinationsignal for SUMO-targeted ubiquitin ligases (STUbLs) and protea-somal degradation (19). Mammalian STUbLs include RNF4 andRNF111 (Arkadia) and viral STUbLs include varicella zoster virusORF61, herpes simplex virus-1 ICP0, and KSHV K-Rta (19, 20).

Significance

Viruses interplay with the host sumoylation system to manip-ulate diverse cellular responses. The adenovirus (Ad) early re-gion 4 (E4)-ORF3 protein forms a dynamic nuclear network tointerfere with and exploit host processes. Here we provide thefirst direct evidence to our knowledge that the Ad5 E4-ORF3protein is a new type of viral small ubiquitin-like modifier(SUMO) E3 ligase that promotes substrate sumoylation andpoly-SUMO3 chain formation. This explains the interrelation-ship between E4-ORF3–mediated substrate sumoylation andproteasomal degradation. Our study reveals that E4-ORF3functions directly to regulate cellular enzymatic processes.Because E4-ORF3 does not share homology with any knownubiquitin or SUMO E3 ligases, these results confer an impor-tant clue to understand the mechanism by which E3 ligasestransfer SUMOs from charged E2 to selected substrates andmediate chain elongation.

Author contributions: S.-Y.S. and P.H. designed research; S.-Y.S. performed research; S.-Y.S.and P.H. analyzed data; and S.-Y.S. and P.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: patrick.hearing@stonybrook.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1603872113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1603872113 PNAS | June 14, 2016 | vol. 113 | no. 24 | 6725–6730

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0

TIF-1γ (TRIM33) is a ubiquitous nuclear protein that belongsto the tripartite motif (TRIM) family, containing a conservedN-terminal region composed of a RING domain, two B boxes, anda coiled-coil domain. TIF-1γ functions in various cellular processesincluding transcriptional regulation, TGF-β signaling, and DNAdamage repair (reviewed in ref. 21). The RBCC domain of TIF-1γhas ubiquitin ligase activity that targets Smad4 for ubiquitin-dependent proteasomal degradation (22). Functioning as a chromatinreader, the PHD/Bromo domain positions TIF-1γ on chromatin oftarget genes and activates (23) or represses transcription (24). TIF-1γalso acts as an E3 ligase for inhibitory monoubiquitination ofSmad4 (25). TIF-1γ regulates epithelial–mesenchymal transitionby functioning as a SUMO E3 ligase for the transcriptional reg-ulator, SnoN1 (26). Finally, ectopic TIF-1γ expression reduces Adearly and late gene expression, whereas TIF-1γ knockdownaugments viral gene expression (12). These latter results showthat TIF-1γ has antiviral activity and help to explain why TIF-1γ isa target of Ad E4-ORF3 activity.Here we confirm that Ad5 E4-ORF3 is sufficient to mediate

sumoylation and proteasomal degradation of TIF-1γ in vivo.Using an in vitro assay system with purified components, weshow for the first time to our knowledge that the Ad5 E4-ORF3protein is an E3 ligase for TIF-1γ sumoylation and poly-SUMOchain elongation. The ability of the E4-ORF3 protein to formhigher order oligomers is essential for these activities. Theseresults are consistent with the hypothesis that E4-ORF3 proteinfunctions to recruit the SUMO machinery and its substrates inproximity of one another using its polyvalent scaffold to assemblehigher order protein complexes and facilitate SUMO conjugation.

ResultsThe Ad5 E4-ORF3 Protein Is Sufficient to Promote TIF-1γ Sumoylationand Proteasomal Degradation. Our previous study showed thatTIF-1γ sumoylation is induced by infection with wild-type Ad5,but not by an E4-ORF3–deficient virus (11). To further verify therole of E4-ORF3, we infected His6-SUMO3–expressing HeLacells (27) with increasing amounts of Ad5-WT or an Ad-CMV-E4-ORF3 expression vector and examined TIF-1γ sumoylationby capturing His6-SUMO3 with Ni2+-NTA resin (11). SUMO3conjugation to TIF-1γ was increased depending on the E4-ORF3expression level (Fig. 1 A and B) and E4-ORF3 was sufficient toinduce TIF-1γ sumoylation (Fig. 1B). In uninfected cells, basallevels of TIF-1γ and TFII-I sumoylation were detected, whereassumoylation of Nbs1 was undetectable (Fig. 1 A and B). In additionto sumoylation, E4-ORF3 decreased TIF-1γ and TFII-I, but notNbs1, expression levels (Fig. 1B), consistent with previous reports(8, 12). Additionally, we confirmed that E4-ORF3 mediates pro-teasomal degradation of TIF-1γ in HeLa and A549 cells using theproteasome inhibitor MG132 (Fig. 1C). These results demonstratethat E4-ORF3 targets TIF-1γ for sumoylation and proteasomaldegradation and suggest that E4-ORF3 regulates cellular targetproteins through distinct posttranslational modifications.

TIF-1γ Sumoylation Is Increased by Ad5 E4-ORF3 in Vitro. Ad5 E4-ORF3 induces sumoylation of multiple cellular proteins (8, 11, 13).It remained unclear whether E4-ORF3 is a SUMO E3 ligase oremploys a cellular E3 ligase. To test this, we prepared recombinantproteins and preformed in vitro sumoylation assays. Purifiedrecombinant GST–TIF-1γ protein was incubated with SUMO E1(SAE1/SAE2), SUMO E2 (Ubc9), and His6-SUMO3 in the pres-ence of ATP and sumoylation levels were determined by Westernblot. TIF-1γ was efficiently conjugated by SUMO3 in vitro in theabsence of a SUMO E3 ligase. We optimized the reaction condi-tions such that weak TIF-1γ sumoylation was evident with SUMOE1 and E2 enzymes alone. We then incubated reaction mixtureswith increasing concentrations of purified Ad5 E4-ORF3 protein.Strikingly, we observed a significantly enhanced ladder of highmolecular weight species of TIF-1γ and SUMO3 in the presence of

wild-type E4-ORF3, whereas only one shifted band was observed inthe control reaction (Fig. 2A), demonstrating that TIF-1γ sumoyla-tion was enhanced by E4-ORF3 in the absence of any cellularSUMO E3 ligase. We also tested the E4-ORF3 L103A mutant,which is defective in formation of nuclear tracks and subsequent E4-ORF3 functions, including the sumoylation of target proteins (7, 8).Interestingly, TIF-1γ sumoylation remained unchanged by additionof E4-ORF3 L103A in the same range of concentrations as wild-type,suggesting that E4-ORF3 self-assembly is required for this activity.Slightly increased higher molecular weight species of SUMO3 wereevident at the highest concentration of L103A (Fig. 2A), showing thatthe mutant protein has residual activity. It is noteworthy that we didnot observe any change in the mobility of the E4-ORF3 protein inthese assays, indicating that E4-ORF3 is not a SUMO substrate,although it contains a consensus sumoylation site at K8. These resultsled us to hypothesize that E4-ORF3 is a viral SUMO E3 ligase.

E4-ORF3 Is a SUMO Ligase for TIF-1γ. A previous study showed thatTIF-1γ has SUMO E3 ligase activity in its C-terminal domains(26). Therefore, we could not exclude the possibility that E4-ORF3stimulates TIF-1γ autosumoylation. To test this, we purified aC-terminal truncated TIF-1γ protein (TIF-1γΔC). TIF-1γΔC con-tains all four known sumoylation sites (K776, 793, 796, and 839) (28)and the RBCC domain, but not the PHD domain and bromodomain

Fig. 1. E4-ORF3 enhances TIF-1γ sumoylation and proteasomal degradation.(A) His6-tagged SUMO3-expressing HeLa cells were infected with increasingamounts of wild-type Ad5 [Ad5-WT, 0, 50, 100, 200, and 400 virus paticlesper cell (P/cell)] and SUMO-conjugated proteins were analyzed at 8 hourspost-infection (hpi) (pull-down). RanGAP1 was used as a control for SUMOcapture. Total cell lysates were analyzed by Western blot (lysate). (B)Recombinant empty Ad-CMV (lanes 1 and 2, 900 P/cell) or Ad-CMV-HA-E4-ORF3 expression vector (lanes 3–5, 100, 300, and 900 P/cell) were used toinfect His6-SUMO3-HeLa cells. At 10 hpi, SUMO conjugates were analyzed asin A. (C) HeLa and A549 cells were infected with empty Ad-CMV (1,000 P/cell)or Ad-CMV-HA-E4-ORF3 (200 and 1,000 P/cell) and treated with 20 μMMG132 at 1 hpi. At 10 hpi (HeLa) or 12 hpi (A549), cells were harvested andprotein levels were determined by Western blot.

6726 | www.pnas.org/cgi/doi/10.1073/pnas.1603872113 Sohn and Hearing

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0

(Fig. 2E). E4-ORF3 has been shown to associate with TIF-1γ via thecoiled-coil domain (10). Similar to the full-length protein, TIF-1γΔCsumoylation was significantly increased by wild-type E4-ORF3, butvery weakly by the L103A mutant (Fig. 2B). The basal level ofsumoylation of TIF-1γ and TIF-1γΔC with E1 and E2 enzymesalone varied slightly between experiments (Fig. 2 A and B, lanes 2),but E4-ORF3 mutant protein L103A did not behave differently withthe two substrates. We examined TIF-1γΔC sumoylation at a higherconcentration of E4-ORF3 L103A protein and observed minor, butdetectable, activity (Fig. S1). This result is consistent with the con-clusion that the L103A mutant protein has residual activity. Thesedata demonstrate that the SUMO E3 ligase activity of TIF-1γ isdispensable for its enhanced sumoylation mediated by the E4-ORF3 protein.

We compared the kinetics of TIF-1γΔC sumoylation in the ab-sence or presence of E4-ORF3 (Fig. 2C). TIF-1γ sumoylation wasobserved as early as 5 min after incubation with E4-ORF3. Ittook 15 min to generate multiply sumoylated TIF-1γΔC in thepresence of wild-type E4-ORF3, whereas 120 min in the absence ofE4-ORF3 and 60 min in the presence of the L103 mutant. After a120-min incubation with E4-ORF3, only multiply sumoylated TIF-1γΔC protein was detected, demonstrating that E4-ORF3 mark-edly accelerates the rate of TIF-1γ sumoylation. We confirmed thatall of the high molecular weight bands were SUMO3 conjugates byincubating the reaction mixture with the SUMO protease, SENP1(Fig. 2D). These results confirm that E4-ORF3 promotes TIF-1γsumoylation along with E1 and E2 enzymes in vitro and suggestthat E4-ORF3 has intrinsic SUMO E3 ligase activity.

Fig. 2. In vitro sumoylation of TIF-1γ with E4-ORF3.(A) GST-tagged recombinant TIF-1γ protein (100 nM)was incubated with 50 nM E1, 250 nM E2, 50 μMHis6-SUMO3, and the indicated concentrations ofHis6-E4-ORF3-WT or His6-E4-ORF3-L103A proteins at37 °C for 60 min. Reaction mixtures were analyzedby Western blot with anti–TIF-1γ, anti-SUMO2/3, andanti–E4-ORF3 antibodies. (B) GST–TIF-1γΔC (100 nM)was used as a substrate and sumoylation was analyzedas described in A. (C) GST–TIF-1γΔC was incubatedwith (+WT and +L103A) or without (−) His6–E4-ORF3for the indicated time periods. (D) GST–TIF-1γΔC wassumoylated at 37 °C for 60 min and incubated with10 nM SENP1 catalytic domain at 20 °C for 10 min. Atotal of 3 μM His6–E4-ORF3 was used in C and D.(E) Schematic representation of TIF-1γ and TIF-1γΔC.RBCC contains the RING finger, B box, and coiled-coildomain; PB contains the plant homeodomain (PHD)and bromodomain.

Fig. 3. Poly-SUMO3 chain assembly with E4-ORF3.(A) The indicated concentrations of His6–E4-ORF3-WT or His6–E4-ORF3-L103A proteins were added toreaction mixtures described in Fig. 2A and incubatedat 37 °C for 60 min. Products were analyzed byWestern blot with anti-SUMO2/3 and anti–E4-ORF3antibodies. The two images in the panel separated bya white space were obtained from the same sourcefile. (B) The E1/E2/His6-SUMO3 mixture was incubatedwith (+E4-ORF3) or without (−) His6–E4-ORF3 for theindicated time periods and products were analyzed byWestern blot with anti-SUMO2/3 antibody. (C) E1/E2/His6-SUMO3 and His6–E4-ORF3 were incubated at37 °C for 60 min and further incubated with 10 nMSENP1 catalytic domain at 20 °C for 10 min. (D) E1 orE2 enzyme was left out of the reaction mixtures de-scribed in Fig. 2A, incubated at 37 °C for 60 min. Atotal of 1.5 μM His6–E4-ORF3 was used in B–D.

Sohn and Hearing PNAS | June 14, 2016 | vol. 113 | no. 24 | 6727

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0

E4-ORF3 Promotes Poly-SUMO3 Chain Formation. In anti-SUMO2/3blots of in vitro sumoylation products, we observed SUMOconjugates of lower molecular weight than unmodified substrateproteins in the presence of E4-ORF3 (Fig. 2 A and B). SUMOE3 ligases are generally thought to induce poly-SUMO chainformation (often referred as SUMO E4 elongase activity), butthe detailed mechanisms are not well understood. Recent studiesidentified E4 elongase activities of plant SUMO E3 ligasesPIAL1 and 2 (29) and human SUMO E3 ligase ZNF451 (30). Todetermine if E4-ORF3 has SUMO E4 elongase activity, weperformed an in vitro sumoylation assay without substrate protein.Remarkably, E4-ORF3 dramatically promoted poly-SUMO3 chainformation depending on the E4-ORF3 concentration (Fig. 3A) andincubation time (Fig. 3B). In contrast, the L103A mutant showedno effect in this assay (Fig. 3A). SUMO E1 and E2 enzymes aloneinduced the formation of polymeric SUMO3 chains consistent witha previous report (31). The addition of SENP1 resulted in break-down of E4-ORF3–induced SUMO3 chains into free SUMO3(Fig. 3C). We removed E1 or E2 enzyme from the reaction anddetected only free SUMO3 protein (Fig. 3D). These data dem-onstrate that E4-ORF3 functions as a SUMO E4 elongase.

E4-ORF3 Induces the Initial Acceptor SUMO3 Conjugation to TIF-1γ asWell as SUMO3 Chain Elongation. We next sought to determine ifE4-ORF3 promotes the initial SUMO3 conjugation to lysineresidues in the substrate protein, chain elongation of SUMO3conjugates already linked to substrate, or both. We generated amutant SUMO3, K11R, which is unable to form polymeric chains(31) and tested it for in vitro sumoylation with E4-ORF3. TIF-

1γΔC was incubated with SUMO3 wild-type or K11R for 60 minor 180 min. In the absence of E4-ORF3, minimal conjugation ofwild-type SUMO3 was observed at 60 min and several highermolecular weight bands were evident at 180 min. TIF-1γΔC wasconjugated to SUMO3 K11R at 180 min in the absence of E4-ORF3, indicating that TIF-1γΔC is modified at multiple lysineresidues by only SUMO E1 and E2 enzymes. E4-ORF3 increasedTIF-1γΔCmodification by the SUMO3 K11Rmutant as well as bywild-type SUMO3 (Fig. 4A), but the high molecular weight ladderobserved with wild-type SUMO3 was not evident with K11R. Wealso examined SUMO3 chain formation using the K11R mutant inthe absence of substrate. The SUMO3 K11R mutant was unableto form a poly-SUMO chain in the presence or absence of E4-ORF3 (Fig. 4B). These results demonstrate that E4-ORF3 me-diates polymeric SUMO3 chain assembly via the consensus K11conjugation site but not at nonconsensus sites in SUMO3. Thisresult confirms that all of the SUMO conjugates from the reactionwith K11R in Fig. 4A correspond to mono-SUMO conjugation ofTIF-1γΔC. These results demonstrate that E4-ORF3 promotesinitial SUMO3 conjugation to TIF-1γ as well as SUMO3 chainextension on multiply monosumoylated TIF-1γ protein.

Interaction Between E4-ORF3 and the SUMO Machinery Proteins. Inagreement with a recent report (32), we did not see relocalizationof the SUMO E2 enzyme Ubc9 into E4-ORF3 nuclear tracks inAd5-infected HeLa cells. However, E4-ORF3 would only need torecruit a subpopulation of Ubc9 to use its catalytic activity. Weexamined Ubc9–E4-ORF3 interaction by coimmunoprecipitation.Lysates containing Myc-tagged Ubc9 and HA-tagged E4-ORF3were prepared by transient transfection and viral vector infectionof HeLa cells, respectively. Interestingly, we could pull down Myc-Ubc9 with wild-type E4-ORF3, but not with the L103A mutant(Fig. 5A), demonstrating that E4-ORF3 binds exogenously expressedUbc9 in cells.E4-ORF3 has a consensus sumoylation site at position K8 and

is predicted to have a SUMO-interaction motif (SIM) (33) at

Fig. 4. In vitro sumoylation assays with SUMO3 (K11R). (A) GST–TIF-1γΔCwas incubated in reaction mixtures described in Fig. 2A with 10 μM of His6-SUMO3 or His6-SUMO3 (K11R) at 37 °C for 60 min or 180 min in the presenceor absence of 3 μM His6–E4-ORF3. (B) Poly-SUMO chain formation reactionswere performed as described in Fig. 2A using 50 μM SUMO proteins and1.5 μM His6–E4-ORF3 at 37 °C for 60 min or 180 min.

Fig. 5. Interaction between E4-ORF3 and SUMOmachinery proteins. (A) HeLacells were transfected with a Myc-Ubc9 expression vector and infected with arecombinant empty Ad-CMV, Ad-CMV-HA-E4-ORF3-WT, or Ad-CMV-HA-E4-ORF3-L103A expression vector. At 24 hpi, cells were lysed and Myc-Ubc9 wasimmunoprecipitated using anti-Myc antibody or control IgG. Protein levelswere analyzed by Western blot. (B) His6–SUMO3-expressing HeLa cells wereleft uninfected or infected with dl355 (ΔE4-ORF6), dl355/inORF3 (ΔE4-ORF6/ΔE4-ORF3), dl355/L103A (ΔE4-ORF6/E4-ORF3 point mutant L103A), or dl355/DL(ΔE4-ORF6/ E4-ORF3 point mutant D105A/L106A) for 8 h. SUMO conjugateswere analyzed as described in Fig. 1A. (C) HeLa cells were infected with theindicated viruses for 8 h. E4-ORF3 and TIF-1γ were immunostained with anti–E4-ORF3 and anti–TIF-1γ antibodies and visualized by fluorescence microscopy.

6728 | www.pnas.org/cgi/doi/10.1073/pnas.1603872113 Sohn and Hearing

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0

residues 103–107 (LIDLH). As demonstrated earlier, we did notobserve any SUMO3 conjugation to E4-ORF3 in vitro, indicatingthat E4-ORF3 itself is not a SUMO substrate. The SUMO E4elongase ZNF451 uses tandem SIMs for activity (30). The predictedE4-ORF3 SIM contains the L103 residue, which is crucial forE4-ORF3 polymer assembly (7), and residues D105/L106, which arenot required for E4-ORF3 self-assembly and sequestration of PMLand SUMOs, but are absolutely required for relocalization of theMRN complex and TFII-I (7, 8, 13). We examined whether thepredicted SIM in E4-ORF3 is critical for sumoylation of TIF-1γusing the in vivo sumoylation assay. As expected, the L103Amutant was unable to enhance TIF-1γ sumoylation. Interestingly,the D105A/L106A mutant could relocalize TIF-1γ into nucleartracks and induce sumoylation (Fig. 5 B and C), indicating that thepredicted SIM is not required for E4-ORF3 SUMO ligase activityor recruitment of TIF-1γ into its network structure.

E4-ORF3 Does Not Have STUbL Activity in Vitro. TIF-1γ sumoylationand proteasomal degradation is induced by E4-ORF3 (Fig. 1)(12). We recently showed that E4-ORF3 targets TFII-I forsumoylation and proteasomal degradation (13). However, itremained unclear whether E4-ORF3 itself is a STUbL or uses acellular STUbL for degradation of sumoylated proteins. Weexamined ubiquitination of poly-SUMO3 chains in vitro. Becausethe cellular STUbL RNF4 (34) and a viral STUbL, K-Rta (20),prefer UbcH5a as a ubiquitin E2 enzyme for ubiquitination ofpoly-SUMO2, we adopted the same reaction conditions to assayE4-ORF3 activity. Heterologous recombinant poly-SUMO3 chains(di- to octa-SUMO3 molecules) were incubated with ubiquitin,ubiquitin E1 (UBE1), and ubiquitin E2 (UbcH5a) in the absence

or presence of E4-ORF3 (Fig. 6A). RNF4 was used as a positivecontrol of STUbL activity. RNF4 strongly increased ubiquitinationof poly-SUMO3 chains, whereas E4-ORF3 had no effect, even atthe five times higher protein concentration than RNF4, indicatingthat E4-ORF3 does not exhibit STUbL activity in these reactionconditions. We examined interaction between E4-ORF3 and poly-SUMO3 chains, a property of STUbLs. Recombinant poly-SUMO3chains were incubated with His6–E4-ORF3, His6-RNF4, or controlBSA and His6-tagged proteins were captured using Ni2+-NTA resin(Fig. 6B). Poly-SUMO3 chains were efficiently pulled down byRNF4, but not by E4-ORF3, indicating that E4-ORF3 is unableto bind poly-SUMO3 chains. Overall, these data suggest thatE4-ORF3 is unlikely to have STUbL activity and may recruit acellular STUbL to selectively target substrate proteins for ubiq-uitination and subsequent proteasomal degradation.

DiscussionWe investigated the role of the Ad5 E4-ORF3 protein in theregulation of protein sumoylation. Using in vitro reactions withpurified proteins, we found that E4-ORF3 dramatically stimu-lates SUMO conjugation of TIF-1γ as well as poly-SUMO chainformation (Figs. 2–4). These results demonstrate that Ad5 E4-ORF3 is a viral SUMO E3 ligase and E4 elongase. This is thefirst demonstration to our knowledge of a viral SUMO E4elongase. Only a limited number of SUMO E3 ligases have beenidentified and much less is known about them compared withubiquitin E3 ligases. The ubiquitin E3 ligases are divided intotwo major classes: the RING and HECT families (34). TheRING family enzymes transfer ubiquitin directly from the E2enzyme to the substrate, whereas the HECT family proteins forman intermediate thioester bond between a cysteine residue andubiquitin, and then transfer ubiquitin to the substrate. One of thebest-characterized SUMO E3 ligase families, the Siz/PIAS pro-teins, transfer SUMO proteins through a similar mechanism tothat used by ubiquitin RING-type E3 ligases. Another SUMO E3ligase, RanBP2, differs from RING-type and HECT-type ubiq-uitin ligases and its E3 domain is proposed to enhance sumoy-lation by tethering the substrate in an optimal alignment with theE2–SUMO complex to promote catalysis (35). Most SUMO E3ligases are known to provide interfaces between substrates andthe SUMO machinery through their RING domains, SIMs, orboth. The structure of the E4-ORF3 protein has been solved (4).Based on this structure, E4-ORF3 does not have features asso-ciated with SUMO or ubiquitin ligases. E4-ORF3 residues 103–107 are a predicted SIM, but these sequences are not critical forE4-ORF3 SUMO ligase activity (Fig. 5).E4-ORF3 regulates posttranslational modification of target

proteins by distinct mechanisms during Ad infection. Mre11 andNbs1 sumoylation is triggered by E4-ORF3 (8) and these pro-teins are then targeted for proteasomal degradation by otherviral proteins, E1B-55K and E4-ORF6, in association with acellular ubiquitin E3 ligase complex (6). On the other hand, E4-ORF3 is sufficient to mediate sumoylation and proteasomaldegradation of TFII-I and TIF-1γ in infected cells (11–13).Mre11 and Nbs1 sumoylation is transient (8), whereas TFII-I(13) and TIF-1γ (Fig. 1) remain sumoylated until their proteinlevels are undetectable. A separation of function mutant, E3-ORF3 D105A/L106A, relocalizes and stimulates sumoylation ofTIF-1γ (Fig. 5) but not Mre11, Nbs1, and TFII-I (7, 8, 13). Oneof the shared characteristics among the E4-ORF3 substrates isthat relocalization into nuclear tracks is a prerequisite forsumoylation. We suggest one of the important roles of E4-ORF3–mediated sumoylation is selectively targeting substrate proteins,such as TFII-I and TIF-1γ, for proteasomal degradation by taggingthem with poly-SUMO chains. A previous study showed thatprotein levels of Nbs1 and Mre11 were decreased in cells in-fected with an E4-ORF3–deficient adenovirus (6), indicating that

Fig. 6. In vitro STUbL assay. (A) Mixtures of di- to octa-SUMO3 chains (50 ng)were incubated with 5 μM HA-ubiquitin, 100 nM His6-UBE1, and 200 nMUbcH5a in the presence of 0.5 or 2.5 μM His6–E4-ORF3 or 0.5 μM His6-RNF4 at37 °C for 90min. Ubiquitination of SUMO chains was analyzed byWestern blotusing anti-SUMO2/3 and anti-ubiquitin antibodies. Input His6-tagged proteinswere visualized using anti-His antibody. (B) His6–E4-ORF3, His6-RNF4, or BSAwas incubated with recombinant poly-SUMO chains and pulled down usingNi2+-NTA beads. Bound proteins were analyzed by Western blot with anti-bodies against SUMO2/3, RNF4, and E4-ORF3. A total of 5% of the input re-action is shown in lane 1.

Sohn and Hearing PNAS | June 14, 2016 | vol. 113 | no. 24 | 6729

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0

E4-ORF3–mediated sumoylation of these substrates does not affectproteasomal degradation of these proteins.Recent studies have demonstrated that SUMO chains act as a

degradation tag upon various cellular stress responses and thatSTUbLs mediate ubiquitination of proteins conjugated withpoly-SUMO chains. In our in vitro ubiquitination assay (Fig. 6),we could not detect any STUbL-like activity of E4-ORF3. Thisresult suggests several possibilities. First, E4-ORF3 may be aSTUbL that has different preference for the ubiquitin E2 en-zyme from RNF4. There are ∼40 E2 enzymes in the humangenome (34). Second, E4-ORF3 may recruit a cellular STUbL topolysumoylated substrates such as TIF-1γ and TFII-I. Third, an-other Ad protein may function as a STUbL. E4-ORF3 is sufficientto induce degradation of target proteins, so this possibility isunlikely. Because E4-ORF3–mediated sumoylation does notalways induce proteasomal degradation, we speculate that acellular STUbL may be involved to target selective proteins fordegradation.Previously, we proposed that relocalization of target proteins

into E4-ORF3 nuclear tracks is required for sumoylation andsubsequent degradation in cells (8, 11, 13). Here we show thatE4-ORF3 self-assembly is critical for SUMO E3 ligase activityin vitro, further substantiating that E4-ORF3 function requireshigher-order multimerization, which creates dynamic interfaceswith ubiquitin-like proteins, substrates, and enzymes (4, 5). Thisstudy provides the first direct evidence to our knowledge that E4-ORF3 plays a role in cellular enzymatic processes. We hypothe-size that E4-ORF3 functions in an analogous manner to RanBP2to recruit the SUMO machinery and its substrates in proximity of

one another, using its polyvalent scaffold to assemble higher orderprotein complexes. Finally, we still do not exclude the possibilitythat a cellular SUMO E3 ligase may be involved in E4-ORF3–mediated sumoylation for some substrates. In a recent report,exogenously expressed PIAS3, a known cellular SUMO E3 ligase,was shown to colocalize with E4-ORF3 (32). Further studies arerequired to completely understand the molecular mechanism(s) ofE4-ORF3 activity.

Materials and MethodsViruses. The following viruses were used in these studies: wild-type Ad5, dl355(ΔE4-ORF6), dl355/inORF3 (ΔE4-ORF6/ΔE4-ORF3), and dl355/pmL103A anddl355/pmD105A/L106A (ΔE4-ORF6/point mutations in E4-ORF3) (7). E4-ORF6plays no role in Ad5-induced protein sumoylation (8). E1-replacement virusesthat express HA-tagged E4-ORF3 of Ad5 were described previously (7).

In Vitro Sumoylation/Desumoylation Assay. SUMO3 conjugation reactionswere performed using 100 nM substrate protein, 50 μM SUMO3, 20–50 nME1, 150–250 nM E2, and 0.1–10 μM E4-ORF3 in a sumoylation reaction buffer(20 mM Hepes pH 7.5, 0.5 mM EGTA, 1 mM DTT, 0.5% Tween-20, 2 mM ATP,and 5 mM MgCl2) at 37 °C for the indicated time periods. For the desumoy-lation assay, 10 nM SENP1 catalytic domain (amino acids 419–644) was addedto the sumoylation reaction mixture and incubated at 20 °C for 10 min. Poly-SUMO3 chain assembly/disassembly reactions were performed as describedabove minus substrate protein.

ACKNOWLEDGMENTS. We thank Dr. Ronald Hay (University of Dundee) forproviding His6-SUMO cell lines, Dr. Thomas Dobner (Heinrich-Pette Institute)for the anti–E4-ORF3 rat monoclonal antibody, and members of our laboratoryfor informed discussions. This work was supported by NIH Grant CA122677(to P.H.).

1. Lion T (2014) Adenovirus infections in immunocompetent and immunocompromisedpatients. Clin Microbiol Rev 27(3):441–462.

2. Berk A (2013) Adenoviridae. Fields Virology (Lippincott Williams & Wilkins, Philadelphia,PA), pp 1704–1731.

3. Doucas V, et al. (1996) Adenovirus replication is coupled with the dynamic propertiesof the PML nuclear structure. Genes Dev 10(2):196–207.

4. Ou HD, et al. (2012) A structural basis for the assembly and functions of a viralpolymer that inactivates multiple tumor suppressors. Cell 151(2):304–319.

5. Patsalo V, et al. (2012) Biophysical and functional analyses suggest that adenovirus E4-ORF3 protein requires higher-order multimerization to function against promyelo-cytic leukemia protein nuclear bodies. J Biol Chem 287(27):22573–22583.

6. Stracker TH, Carson CT, Weitzman MD (2002) Adenovirus oncoproteins inactivate theMre11-Rad50-NBS1 DNA repair complex. Nature 418(6895):348–352.

7. Evans JD, Hearing P (2005) Relocalization of the Mre11-Rad50-Nbs1 complex by theadenovirus E4 ORF3 protein is required for viral replication. J Virol 79(10):6207–6215.

8. Sohn S-Y, Hearing P (2012) Adenovirus regulates sumoylation of Mre11-Rad50-Nbs1components through a paralog-specific mechanism. J Virol 86(18):9656–9665.

9. Yondola MA, Hearing P (2007) The adenovirus E4 ORF3 protein binds and reorganizesthe TRIM family member transcriptional intermediary factor 1 alpha. J Virol 81(8):4264–4271.

10. Vink EI, Yondola MA, Wu K, Hearing P (2012) Adenovirus E4-ORF3-dependent re-localization of TIF1α and TIF1γ relies on access to the Coiled-Coil motif. Virology422(2):317–325.

11. Sohn S-Y, Bridges RG, Hearing P (2015) Proteomic analysis of ubiquitin-like post-translational modifications induced by the adenovirus E4-ORF3 protein. J Virol 89(3):1744–1755.

12. Forrester NA, et al. (2012) Adenovirus E4orf3 targets transcriptional intermediaryfactor 1γ for proteasome-dependent degradation during infection. J Virol 86(6):3167–3179.

13. Bridges RG, Sohn S-Y, Wright J, Leppard KN, Hearing P (2016) The adenovirus E4-ORF3protein stimulates SUMOylation of general transcription factor TFII-I to direct pro-teasomal degradation. MBio 7(1):e02184–e15.

14. Everett RD, Boutell C, Hale BG (2013) Interplay between viruses and host sumoylationpathways. Nat Rev Microbiol 11(6):400–411.

15. Gareau JR, Lima CD (2010) The SUMO pathway: Emerging mechanisms that shapespecificity, conjugation and recognition. Nat Rev Mol Cell Biol 11(12):861–871.

16. Muller S, Dobner T (2008) The adenovirus E1B-55K oncoprotein induces SUMOmodification of p53. Cell Cycle 7(6):754–758.

17. Pennella MA, Liu Y, Woo JL, Kim CA, Berk AJ (2010) Adenovirus E1B 55-kilodaltonprotein is a p53-SUMO1 E3 ligase that represses p53 and stimulates its nuclear exportthrough interactions with promyelocytic leukemia nuclear bodies. J Virol 84(23):12210–12225.

18. Chang P-C, et al. (2010) Kaposi’s sarcoma-associated herpesvirus (KSHV) encodes aSUMO E3 ligase that is SIM-dependent and SUMO-2/3-specific. J Biol Chem 285(8):5266–5273.

19. Sriramachandran AM, Dohmen RJ (2014) SUMO-targeted ubiquitin ligases. Biochim

Biophys Acta 1843(1):75–85.20. Izumiya Y, et al. (2013) Kaposi’s sarcoma-associated herpesvirus K-Rta exhibits SUMO-

targeting ubiquitin ligase (STUbL) like activity and is essential for viral reactivation.

PLoS Pathog 9(8):e1003506.21. Massagué J, Xi Q (2012) TGF-β control of stem cell differentiation genes. FEBS Lett

586(14):1953–1958.22. Dupont S, et al. (2005) Germ-layer specification and control of cell growth by Ecto-

dermin, a Smad4 ubiquitin ligase. Cell 121(1):87–99.23. Xi Q, et al. (2011) A poised chromatin platform for TGF-β access to master regulators.

Cell 147(7):1511–1524.24. Agricola E, Randall RA, Gaarenstroom T, Dupont S, Hill CS (2011) Recruitment of TIF1γ

to chromatin via its PHD finger-bromodomain activates its ubiquitin ligase and

transcriptional repressor activities. Mol Cell 43(1):85–96.25. Dupont S, et al. (2009) FAM/USP9x, a deubiquitinating enzyme essential for TGFbeta

signaling, controls Smad4 monoubiquitination. Cell 136(1):123–135.26. Ikeuchi Y, et al. (2014) TIF1γ protein regulates epithelial-mesenchymal transition by

operating as a small ubiquitin-like modifier (SUMO) E3 ligase for the transcriptional

regulator SnoN1. J Biol Chem 289(36):25067–25078.27. TathamMH, Rodriguez MS, Xirodimas DP, Hay RT (2009) Detection of protein SUMOylation

in vivo. Nat Protoc 4(9):1363–1371.28. Fattet L, et al. (2013) TIF1γ requires sumoylation to exert its repressive activity on TGFβ

signaling. J Cell Sci 126(Pt 16):3713–3723.29. Tomanov K, et al. (2014) Arabidopsis PIAL1 and 2 promote SUMO chain formation as

E4-type SUMO ligases and are involved in stress responses and sulfur metabolism.

Plant Cell 26(11):4547–4560.30. Eisenhardt N, et al. (2015) A new vertebrate SUMO enzyme family reveals insights

into SUMO-chain assembly. Nat Struct Mol Biol 22(12):959–967.31. Tatham MH, et al. (2001) Polymeric chains of SUMO-2 and SUMO-3 are conjugated to

protein substrates by SAE1/SAE2 and Ubc9. J Biol Chem 276(38):35368–35374.32. Higginbotham JM, O’Shea CC (2015) Adenovirus E4-ORF3 targets PIAS3 and together

with E1B-55K remodels SUMO interactions in the nucleus and at virus genome rep-

lication domains. J Virol 89(20):10260–10272.33. Zhao Q, et al. (2014) GPS-SUMO: A tool for the prediction of sumoylation sites and

SUMO-interaction motifs. Nucleic Acids Res 42(Web Server issue):W325–W330.34. Metzger MB, Hristova VA, Weissman AM (2012) HECT and RING finger families of E3

ubiquitin ligases at a glance. J Cell Sci 125(Pt 3):531–537.35. Tatham MH, et al. (2005) Unique binding interactions among Ubc9, SUMO and

RanBP2 reveal a mechanism for SUMO paralog selection. Nat Struct Mol Biol 12(1):

67–74.36. TathamMH, et al. (2008) RNF4 is a poly-SUMO-specific E3 ubiquitin ligase required for

arsenic-induced PML degradation. Nat Cell Biol 10(5):538–546.

6730 | www.pnas.org/cgi/doi/10.1073/pnas.1603872113 Sohn and Hearing

Dow

nloa

ded

by g

uest

on

Dec

embe

r 27

, 202

0