Consecutive Inhibition of ISG15 Expression and ISGylation by Cytomegalovirus Regulators

Interferon-stimulated gene 15 (ISG15) encodes an ubiquitin-like protein that covalently conjugates protein. Protein modification by ISG15 (ISGylation) is known to inhibit the replication of many viruses. However, studies on the viral targets and viral strategies to regulate ISGylation-mediated antiviral responses are limited. In this study, we show that human cytomegalovirus (HCMV) replication is inhibited by ISGylation, but the virus has evolved multiple countermeasures. HCMV-induced ISG15 expression was mitigated by IE1, a viral inhibitor of interferon signaling, however, ISGylation was still strongly upregulated during virus infection. RNA interference of UBE1L (E1), UbcH8 (E2), Herc5 (E3), and UBP43 (ISG15 protease) revealed that ISGylation inhibits HCMV growth by downregulating viral gene expression and virion release in a manner that is more prominent at low multiplicity of infection. A viral regulator pUL26 was found to interact with ISG15, UBE1L, and Herc5, and be ISGylated. ISGylation of pUL26 regulated its stability and inhibited its activities to suppress NF-κB signaling and complement the growth of UL26-null mutant virus. Moreover, pUL26 reciprocally suppressed virus-induced ISGylation independent of its own ISGylation. Consistently, ISGylation was more pronounced in infections with the UL26-deleted mutant virus, whose growth was more sensitive to IFNβ treatment than that of the wild-type virus. Therefore, pUL26 is a viral ISG15 target that also counteracts ISGylation. Our results demonstrate that ISGylation inhibits HCMV growth at multiple steps and that HCMV has evolved countermeasures to suppress ISG15 transcription and protein ISGylation, highlighting the importance of the interplay between virus and ISGylation in productive viral infection.


Introduction
Type I interferons (IFNs) are multifunctional cytokines that represent crucial components of the innate immune response to viral infection. Recognition of viral infection by host cells induces the synthesis of type I IFNs. Secreted IFNs interact with IFN receptors on target cells, triggering a signaling cascade that involves Janus kinase (JAK) and signal transducer and activator of transcription (STAT) families. Activated STAT1 and STAT2 heterodimerize and bind to IFN regulatory factor 9 (IRF9) to form a complex called IFN-stimulated gene factor 3 (ISGF3). This complex translocates into the nucleus and induces ISGs with diverse antiviral activities by binding to IFNstimulated response elements (ISREs) in their promoters (for review [1]).
Although several studies have suggested a general role for ISG15 as an antiviral molecule, proviral effects of ISGylation have also been reported for certain viruses. In Newcastle disease virus (NDV) infection, ISGylation of RIG-I reduced IFN responses as a negative feedback regulation mechanism [45], whereas ISGylation of IRF3 stabilized IRF3 [21]. In addition, enhanced ISGylation by ISG15 overexpression promoted HCV production, while reduced ISGylation by UBE1Lor ISG15-knockdown inhibited HCV production [46,47]. Therefore, the effects of global protein ISGylation appear to vary among the different viruses. In addition, ISG15 is secreted to the extracellular space as a free unconjugated form [48,49]. While secretion of ISG15 in granulocytes is shown to activate T cells and natural killer cells to produce IFNγ in mycobacterial infection [50,51], the role of secreted ISG15 in viral infection is not clear. Furthermore, a role of free ISG15 as a negative regulator that prevents IFNα/β overamplification and auto-inflammation by sustaining UBP43 levels has been suggested in humans but not in mice [52,53].
Human cytomegalovirus (HCMV) is an opportunistic pathogen that causes severe disease complications and pathologies in newborns and immunocompromised individuals [54]. During productive infection, HCMV gene expression occurs sequentially in three phases: immediateearly (IE), early, and late. IE proteins and virion-associated tegument proteins play key roles in initiating viral gene expression and modulating host cell functions. HCMV employs several mechanisms to counteract IFN production and subsequent ISG activation. IE2 and pp65 inhibit IFN production [55][56][57], whereas IE1 suppresses the IFN response by directly binding to STAT2 [58][59][60] and PML [61,62]. HCMV infection results in a decrease in the levels of JAK1 and p48, two components of the type I IFN signaling pathway [63,64]. Modulation of the stability and phosphorylation of STAT proteins during HCMV infection was also reported [65,66].
ISG15 transcription is induced in HCMV infection; however, its regulation during infection, the role of ISGylation in viral growth, and viral targets of ISG15 have not been characterized. In this study, we show that ISG15 expression and ISGylation are initially induced after HCMV infection but later suppressed by viral responses, and that IE1, a viral inhibitor of STAT signaling, plays an important role in reducing ISG15 transcription. By silencing the expression of E1, E2, and E3 ISGylation enzymes and of an ISG15 protease, we also demonstrate that ISGylation inhibits HCMV growth at multiple steps, including viral gene expression and virion release. Furthermore, we show that pUL26, a viral tegument protein, interacts with ISG15, E1, and E3 and is modified by ISG15, which inhibits pUL26 activity to promote viral growth. Moreover, we reveal that expression of pUL26 is able to suppress ISGylation induced by virus infection. Our results indicate that HCMV has evolved countermeasures to suppress ISG15 transcription and protein ISGylation, highlighting the important of the interplay between virus and ISG15 signaling during virus infection.

Time course of ISG15 expression and protein ISGylation during HCMV infection
The time course of ISG15 expression and protein ISGylation during HCMV infection were investigated with different multiplicity of infections (MOIs). In human fibroblast (HF) cells infected with HCMV (Towne), the levels of ISG15 and protein ISGylation were elevated by 24 h at all MOIs tested (MOIs of 0.2 to 10) (Fig 1A, lanes 2-6). At 48 and 72 h after infection, even greater levels of ISG15 expression and protein ISGylation were observed at relatively low MOIs h. Cells were fixed with methanol and double-label IFA was performed with antibodies for ISG15 and UL112-113. Hoechst stain was used to stain cell nuclei. The images were obtained by confocal microscopy.
(0.2, 0.5, and 1) (Fig 1A, lanes 13-15 and 24-26); however, the levels of free ISG15 and ISG15 conjugates at high MOIs (3 and 10) were much lower than those at low MOIs ( Fig 1A, lanes 16-17 and 27-28). The time course of ISG15 expression and protein ISGylation was also examined in cells infected with UV-inactivated virus (UV-HCMV). In UV-HCMV infection, the levels of ISG15 and ISG15 conjugates were elevated at 24 and 48 h and correlated proportionally with MOI ( Fig 1A, lanes 7-11 and 18 -22). Levels of free ISG15 at 72 h induced by UV-HCMV were lower than those at 48 h, probably due to the termination of signaling ( Fig  1A, compare lanes 29-33 and 18-22). The lack of viral gene expression in UV-HCMV infection was verified by the absence of viral IE protein expression. Collectively, these results comparing HCMV and UV-HCMV infection demonstrate that ISG15 expression and protein ISGylation are initially induced by HCMV infection, but are subsequently suppressed in a manner dependent on viral gene expression.
Notably, we observed a greater induction of ISG15 and protein ISGylation with HCMV than with UV-HCMV at low MOIs (MOIs of 0.2, 0.5, and 1) (Fig 1A, compare lanes 2-4 and 7-9). We hypothesized that this was due to the induction of ISG15 at different level in uninfected cells that surround infected cells at low MOIs. To test this hypothesis and to examine the effect of HCMV infection on ISG15 expression at a single cell level, HF cells were infected with HCMV or UV-HCMV at a low MOI (0.5) and stained for ISG15 and viral UL112-113 proteins. We found that ISG15 expression was reduced in HCMV-infected cells, in which UL112-113 viral replication proteins were expressed at high level; however, it was markedly increased in neighboring uninfected cells compared to mock-infected cells. We also found that the levels of ISG15 in uninfected neighboring cells were higher in HCMV infection than in UV-HCMV infection ( Fig 1B). This result suggests that an indirect effect of virus infection on neighboring uninfected cells is responsible for the greater induction of ISG15 and protein ISGylation by HCMV than by UV-HCMV at low MOIs.

HCMV IE1 suppresses ISG15 transcription
Since HCMV IE1 inhibits the activation of ISRE-containing promoters by sequestering STAT2 [58][59][60] and PML [61], IE1 expression may be responsible for the suppression of free ISG15 and ISG15 conjugate levels during HCMV infection. To test this hypothesis, we first compared the effects of wild-type HCMV, UV-HCMV, and IE1-deleted mutant virus (CR208) infection (MOI of 3) on ISG15 transcription by RT-PCR. The CR208 virus exhibited an MOI-dependent growth pattern with showing a severe growth defect in HF cells at low MOIs but normal growth at high MOIs [67]. All viruses increased ISG15 mRNA levels 12 h after infection; however, ISG15 induction was terminated earlier for HCMV than UV-HCMV, and not appreciably terminated for CR208, which continued to produce high levels of ISG15 transcripts even at a late stage of infection (72 h) (Fig 2A). ISG15 transcription induced by UV-HCMV infection might be gradually decreased due to the termination of the IFN signaling through several negative regulatory mechanisms. However, when cells were infected with CR208 at this high MOI, the replication of virus without IE1 appeared to lead to a robust activation of IFN signaling. This result indicates that IE1 indeed plays an important role in reducing ISG15 transcription during HCMV infection.
We also compared the levels of ISG15 and ISG15 conjugates in HCMV, UV-HCMV, and CR208 infection. When cells were infected with viruses at an MOI of 3, free ISG15 levels were elevated at 12 h by all viruses, whereas the level of ISG15 conjugates markedly increased after 24 h (Fig 2B). The delayed induction of ISGylation is similar to what was observed in IFNβtreated cells [68] and may result from the delayed induction of the ISGylation machinery. Consistent with the results shown in Fig 1A, UV-HCMV induced more ISG15 conjugates than wild-type virus at this MOI ( Fig 2B, compare lanes 3-5 and 8-10). Importantly, CR208 also resulted in greater ISGylation than wild-type virus (Fig 2C. compare lanes 3-5 and 13-15), indicating that IE1 is required for the suppression of ISG15 expression and ISGylation during HCMV infection. Immunoblot analysis demonstrated that IE1 and IE2 failed to be expressed in UV-HCMV infection and that IE1 failed to be expressed and IE2 levels were reduced in CR208 infection under these experimental conditions (Fig 2B and 2C). Although our results show that IE1 is largely responsible for the suppression of ISG15 transcription, it is notable that CR208 infection resulted in slightly lower levels of ISGylation compared to UV-HCMV infection (Fig 2B, compare lanes 10 and 15). Considering that ISG15 transcript levels remained high up to 72 h in CR208-infected cells (Fig 2A), this finding suggests that other viral processes, besides those mediated by IE1, may also be implicated in the downregulation of protein ISGylation. . Cell lysates were also prepared and analyzed by immunoblotting as in Fig 1A (B). (C) Control and IE1-expressing HF cells produced by retroviral vectors were mock-infected or infected with wild-type HCMV, UV-HCMV, or CR208 virus at an MOI of 3, or treated with IFNβ (1,000 U/ml) for 48 h. Immunoblotting was performed with antibodies for ISG15, IE1/IE2, p52 (encoded by UL44) and β-actin. (D) 293T cells were co-transfected with plasmids expressing myc-ISG15 (wild-type, ISG15 GG , or ISG15 AA ), HA-UBE1L (E1), Flag-UbcH8 (E2), or HA-Herc5 (E3) as indicated. At 48 h after transfection, cell lysates were prepared by boiling the cell pellets in sodium dodecyl sulfate (SDS) loading buffer and immunoblotted with anti-myc and anti-β-actin (a loading control) antibodies. (E and F) Co-transfection/ISGylation assays were performed in 293T cells with or without increasing amounts of plasmid expressing IE1 (E) or in control and IE1-expressing HF cells (F). Immunoblotting was performed with anti-myc, anti-IE1, and anti-β-actin antibodies.
The inhibition of ISG15 expression by IE1 was further investigated using IE1-overexpressing HF cells generated using retroviral vectors. Control and IE1-overexpressing HF cells were infected with HCMV, UV-HCMV, or CR208. The results of immunoblot analysis showed that IE1 overexpression suppressed the induction of ISG15 and ISG15 conjugates by both virus infection and IFNβ treatment (as a control) (Fig 2C, compare lanes 2-5 and 7-10), further supporting the critical role of IE1 in reducing ISG15 expression.
Although reduction of protein ISGylation during HCMV infection may be largely attributed to suppression of ISG15 transcription by IE1, it cannot be ruled out that IE1 also affects the ISGylation reaction. Therefore, we studied whether IE1 directly affects ISG15 conjugation reactions using co-transfection/ISGylation assays. To set up co-transfection/ISGylation assays, two different ISG15 forms, an active form with a stop codon immediately after the C-terminal double glycine residues (ISG15 GG ) and an inactive form with the double glycine residues substituted with alanine residues (ISG15 AA ), were employed. In cells transiently co-transfected with ISG15, UBE1L (E1), UbcH8 (E2), and Herc5 (E3), intact ISG15 and ISG15 GG (an active form), but not ISG15 AA (an inactive form), were conjugated to proteins (Fig 2D). When co-transfection/ISGylation assays were performed with or without IE1 overexpression, IE1 expression by co-transfection or retroviral transduction did not inhibit levels of ISG15 conjugates, indicating that IE1 does not inhibit the enzymatic cascade of reactions required for protein ISGylation (Fig 2E and 2F).

ISGylation inhibits HCMV growth
The role of protein ISGylation in HCMV infection was investigated by silencing expression of ISGylation enzymes using RNA interference. HF cells expressing control shRNA or shRNA for UBP43, an ISG15-specific protease, were generated using lentiviral vectors. IFNα treatment of normal and control shRNA (shC)-expressing HF cells induced the expression of UBP43 proteins; however, UBP43-specific shRNAs (shUBP43-1 and shUBP43-2) efficiently suppressed this induction (Fig 3A). UBP43 knockdown enhanced protein ISGylation in cells infected with UV-HCMV that stimulates interferon signaling (Fig 3B), but reduced the production of progeny virions in HCMV-infected cells to nearly 10% of that in control cells (Fig 3C), suggesting that enhanced ISGylation by UBP43 knockdown attenuates HCMV growth. However, UBP43 is also known to negatively regulate IFN signaling by downregulating JAK/STAT signaling [15,69]. Consistently, we found that UBP43 knockdown enhanced STAT2 phosphorylation in HCMV-infected cells ( Fig 3D). Therefore, it is possible that the reduction of HCMV growth in UBP43 knockdown cells is also a result of other aspects of the IFN response.
The effect of ISGylation on HCMV growth was further assessed by depleting Herc5, a major ISG15 E3 ligase in human [6,7]. Herc5 knockdown HF cells were generated using lentiviral vectors expressing shRNAs. qRT-PCR assays confirmed the efficient reduction of Herc5 transcript levels in cells expressing Herc5-specific shRNAs (shHerc5-1 and shHerc5-2) compared to in cells expressing control shRNA (shC) after UV-HCMV infection ( Fig 3E). We found that Herc5 knockdown markedly reduced protein ISGylation as expected (Fig 3F), but increased virus titers by 8-to 11-fold compared to in control cells (Fig 3G), demonstrating that the reduction of ISGylation by Herc5 knockdown facilitates HCMV growth. Similar enhancement of HCMV growth was observed in HF cells depleted of UBE1L (E1) or UbcH8 (E2) (S1 Fig). Collectively, our results with UBP43, Herc5, UBE1L, and UbcH8 knockdown cells demonstrate an inverse relationship between ISGylation and HCMV growth, indicating a general inhibitory role of ISGylation in HCMV infection.
We also investigated whether enhanced ISGylation affects HCMV growth by ectopically expressing ISG15 together with ISGylation enzymes. HF cells co-transfected with plasmids expressing myc-ISG15 GG or myc-ISG15 AA and ISGylation enzymes were infected with HCMV and the production of progeny virions was compared. The results showed that the expression of myc-ISG15 GG and ISGylation enzymes induced higher levels of ISG15 conjugates and reduced progeny virus titers to 50% of control, whereas the expression of ISG15 AA did not significantly affect the levels of ISG15 conjugates or progeny virions (S2 Fig). These results are also consistent with our results using shRNA-expressing cells in which we showed an inverse relationship between levels of ISG15 conjugates and HCMV growth.

ISGylation reduces HCMV gene expression and virion release
To investigate the mechanisms by which ISGylation inhibits HCMV growth, we first compared the expression profile of viral proteins in control and Herc5 knockdown cells. When cells were infected with an MOI of 0.2, levels of major IE (IE1 and IE2), early (p52), and late (pp28) viral proteins produced at 24, 48, 72, and 96 h were higher in Herc5 knockdown cells than in control cells ( Fig 4A, left panels). This effect of knocking down Herc5 diminished when cells were infected at higher MOIs. At an MOI of 1, IE1 expression was similar between control and Herc5 knockdown cells, although the levels of IE2, p52, and pp28 were slightly increased in Herc5 knockdown cells (Fig 4A, center panels). At an MOI of 5, levels of viral proteins were comparable between control and Herc5 knockdown cells (Fig 4A, right panels). Progeny virion titers measured in the culture supernatant correlated with the levels of expressed viral proteins ( Fig 4B). These results demonstrate that reduced ISGylation by silencing Herc5 promotes viral gene expression, an effect that is more evident at lower MOIs.
We further investigated whether ISGylation affects the activation of viral promoters using reporter assays (S3 Fig). HF cells co-transfected with ISG15, UBE1L, UbcH8, and Herc5 exhibited substantially increased protein ISGylation, whereas cells co-transfected without Herc5 did not. In HF cells exhibiting enhanced ISGylation by co-transfection, the activity of viral MIE promoter was repressed to 50% of that observed in control cells. Similarly, IE2-mediated activation of viral early [UL112-113 and UL54 (POL)] and late [UL99 (pp28)] promoters was also suppressed in cells showing enhanced ISGylation. In control experiments, ectopic expression of Herc5 alone was not sufficient to increase ISGylation and therefore did not affect viral promoter activation. These results demonstrate that enhanced ISGylation inhibits the activation of viral promoters.
We also assessed whether ISGylation inhibits HCMV virion release as is the case in HIV infection [43]. Control and Herc5 knockdown cells were infected with HCMV at an MOI of 1 or 5 and the levels of cell-associated and extracellular progeny virions produced were compared. The results showed that although more progeny virions were produced in Herc5 knockdown cells than in control cells (Fig 4B), the percentages of cell-associated virions to total virions were lower in Herc5-knockdown cells, 3.5% (MOI 1) and 3.8% (MOI 5), than in control cells, 19.7% (MOI 1) and 17% (MOI 5) (Fig 4C). These results indicate that, as observed in HIV infection, ISGylation inhibits HCMV virion release.

Covalent and non-covalent interaction of pUL26 with ISG15
To identify potential HCMV proteins that interact with the ISG15 pathway, we screened the HCMV ORF library [70] for ISG15-or UBE1L-interacting proteins using yeast two-hybrid assays. Twenty HCMV proteins were identified as potential ISG15-interacting proteins, and five viral proteins (encoded from RL1, UL19, UL21A, UL26, and UL30) were found to interact with both ISG15 and UBE1L. Most of these ISG15 binding and all of UBE1L binding in yeast assays were also detected by co-immunoprecipitation (co-IP) assays (S4 Fig and S2 Table). Among them, the UL26 gene encodes the tegument proteins, p27 and p21, which are produced using two in-frame start codons and are shown to regulate viral gene expression, NF-κB signaling, and virion stability [71][72][73][74]. Since UL26 interacted with both ISG15 and UBE1L and its role in viral growth was relatively well reported compared to others, we further investigated the interaction of pUL26 with the ISG15 pathway. In co-IP assays, UL26-p21 interacted with ISG15 AA , demonstrating that ISG15 can non-covalently interact with UL26 ( Fig 5A). In similar co-IP assays, UL26-p21 also interacted with UBE1L and Herc5, but not UbcH8 (Fig 5B-5D). When co-IP assays were performed using HCMV (Towne)-infected cell lysates, immunoprecipitation with an anti-UL26 antibody co-precipitated unconjugated free ISG15 (Fig 5E), and in a reciprocal experiment, immunoprecipitation with an anti-ISG15 antibody co-precipitated UL26-p21 (Fig 5F), indicating that UL26-p21 interacts with ISG15 during virus infection.
It has been suggested that newly synthesized viral proteins may be broadly modified by ISG15 during virus infection [41]. Therefore, we also investigated whether UL26 is covalently conjugated by ISG15 during HCMV infection. Cell lysates prepared from HCMV (Towne)infected cells were boiled in SDS-containing buffer and then immunoprecipitated with an anti-UL26 antibody. Immunoblotting of the sample with anti-ISG15 antibody revealed a band that is consistent with an ISG15-modified form of UL26-p21 (Fig 5G). To further investigate these non-covalent and covalent interaction of UL26 proteins with ISG15, we generated a recombinant HCMV (Toledo) expressing UL26 proteins tagged at their carboxyl termini with an HA tag (S5 Fig). Compared to the wild-type virus, the recombinant virus produced equivalent amounts of viral major IE (IE1 and IE2) and UL26 proteins, and progeny virions (Fig 5H  and 5I). When co-IP assays were performed with lysates from UL26-HA virus infected cells, immunoprecipitation of UL26 proteins with an anti-HA antibody co-precipitated free ISG15 (Fig 5J). When HA-UL26 virus-infected cells were subjected to co-IP assays to detect ISGylated proteins, bands that are consistent with ISG15-modified forms of UL26 proteins were detected (Fig 5K). Since ISG15 can modify ubiquitin, forming ISG15-ubiquitin mixed chains [75] and a single lysine reside can be poly-ISGylated [20,76], the smear bands of ISGylated UL26 appear to be UL26 proteins that contain ISG15-ubiquitin mixed chains or poly-ISG15 chains. Taken together, our data suggest that the UL26-encoded proteins non-covalently interact with ISG15 and are also covalently modified by ISG15. In a control experiment, we found that some viral proteins, which did not interact with ISG15, UBE1L, and Herc5 in co-IP assays, could be ISGylated in co-transfection/ISGylation assays (S6 Fig), supporting the concept that viral proteins may be broadly ISGylated during infection due to IFN-upregulated expression of Herc5 (E3) in polyribosomes [41].

ISGylation of UL26 regulates protein stability and inhibits UL26 activities to suppress NF-κB signaling and promote viral growth
The UL26 ORF from the Towne strain contains three lysine residues (K54, K136, and K169) and K54 and K169 are conserved in human CMVs (Fig 6A). To determine the ISGylation sites of UL26 proteins, we performed co-transfection/ISGylation assays using wild-type UL26-p21 or its mutants in which lysine residues were replaced with arginines. The results showed that the K54R, K136R, and K169R mutants were still ISGylated; however, the K136/169R double mutant was not ISGylated, indicating that K136 and K169 are the major ISGylation sites (Fig 6B). When HF cells expressing wild-type UL26-p21 or K136/169R mutant were generated by retroviral vectors, expression level of the K136/169R mutant protein was higher than that of the wild-type protein (Fig 6C), suggesting that ISGylation of UL26 proteins may regulate protein stability.
The SUMO fusion proteins were often used to study the function of SUMO modification of proteins [77][78][79][80][81]. To investigate the effect of UL26 ISGylation on its activity, we used the UL26-p21(K136/169R)-ISG15 AA fusion protein as a surrogate for the ISG15-modified form of UL26-p21. We produced the ISG15 fusion to the lysine mutant form of UL26 to compare the activities of ISGylation-defective UL26 and its ISG15 fusion form. HF cells expressing the K136/169R mutant or the UL26-p21(K136/169R)-ISG15 AA fusion protein were generated by retroviral vectors (Fig 6D). UL26 proteins have been shown to inhibit TNFα-induced NF-κB activation [74]. UL26-p21(K136/169R) moderately inhibited TNFα-induced NF-κB activation but UL26-p21(K136/169R)-ISG15 AA did not, suggesting that ISGylation of UL26 inhibits its activity to downregulate NF-κB signaling (Fig 6E). We further investigated the effect of UL26 ISGylation on its role in viral growth. When control and UL26-expressing cells were infected with the UL26-deleted mutant HCMV (AD169) [72] and the levels of progeny virus titers were compared, pre-expression of the K136/169R mutant significantly increased the growth of mutant virus but its ISG15 AA fusion protein did not (Fig 6F). Together, these results indicate that ISGylation of UL26 inhibits its activities to suppress NF-κB signaling and promote the growth of UL26-deleted mutant virus.

UL26 inhibits protein ISGylation
Since influenza virus NS1B antagonizes ISGylation in human cells via direct binding to ISG15 [3,82], we tested whether UL26 can regulate protein ISGylation. We found that expression of UL26-p21 reduced the levels of ISG15 conjugates in cells co-transfected with UBE1L (E1), UbcH8 (E2), Herc5 (E3), and ISG15 GG , an active form of ISG15 (Fig 7A). In a similar assay, ISGylation of charged multivesicular body protein (CHMP) 5, a component of the endosomal sorting complex required for transport (ESCRT) machinery, was inhibited by UL26-p21 expression (Fig 7B). When control and UL26-p21-expressing HF cells were infected with UV-HCMV, the levels of ISG15 conjugates were significantly reduced in UL26-p21-expressing cells compared to control cells (Fig 7C). This effect of UL26 was also found with the K136/ 169R mutant protein (Fig 7D), which still interacted with ISG15 AA in co-IP assays (Fig 7E), suggesting that the inhibitory effect of UL26 on ISGylation is not dependent on its own ISGylation. In control experiments, viral proteins such as pUL85, pUL71, and IE2 did not inhibit ISGylation of cellular proteins in co-transfection/ISGylation assays (S6 and S7 Figs).
To further investigate the inhibitory effect of UL26 on ISGylation during HCMV infection, we compared ISGylation levels between cells infected with wild-type and UL26-deleted mutant viruses. With an MOI of 0.2, we found that the ISGylation levels were initially increased until 48 h but gradually decreased at 72 h and 96 h, and that the UL26-deleted virus infection  Expression levels of myc-ISG15, SRT-CHMP5, and β-actin in whole cell lysates were also determined by immunoblotting. (C) Control and UL26-p21-expressing HF cells produced by retroviral vectors were infected with UV-HCMV at an MOI of 1. The samples were prepared at the indicated time points and the levels of ISG15, UL26 and β-actin were determined by immunoblotting. The relative levels of ISGylated UL26 protein over unmodified protein (normalized with β-actin) are also shown as graphs. (D) Control or UL26 (wild-type or mutant)-expressing HF cells were infected with UV-HCMV and immunoblotting was performed as described in (C). (E) 293T cells were co-transfected with plasmids expressing SRT-UL26-p21 (wild-type or K136/ 169R) or myc-ISG15 AA as indicated. At 48 h after transfection, cell lysates were immunoprecipitated with anti-SRT antibody. Immunoprecipitated samples and whole cell lysates were detected by immunoblotting. (F) HF cells expressing control shRNA (shC) or shRNA specific for IE1 (shIE1) produced by retroviral transduction were infected with wild-type or ΔUL26 mutant virus at an MOI of 5 for the indicated time points. Cell lysates were immunoblotted with antibodies for ISG15, viral proteins (IE1, IE2, and UL26), and β-actin. (G) HF cells were pre-treated or not with IFNβ (100 U/ml) for 24 h and then showed only minimally increased levels of ISGylation at late times of infection compared to wild-type virus infection (S8 Fig, compare lanes 5-7 and 11-13). We reasoned this minimal effect of UL26 to the suppression of ISG15 transcription by IE1. Therefore, to minimize the effect of IE1 expression we performed a similar experiment in cells expressing shRNA for IE1 (shIE1). Control shRNA (shC) and shIE1-expressing HF cells, which were produced by retroviral vectors, were infected with wild-type and UL26-deleted mutant viruses at an MOI of 5. The results of immunoblotting showed that the expression of shIE1 substantially reduced the IE1 protein accumulation compared to control cells (Fig 7F, compare lanes 2-5 and 7-10), and that in shIE1-expressing cells the ISGylation levels were substantially increased at 24 h after infection and gradually decreased at 48 h and 72 h under these experimental conditions (Fig 7F, lanes  8-10). Notably, we found that the UL26-deleted virus less effectively reduced the ISGylation levels than wild-type virus at the late phase of infection (Fig 7F, compare lanes 8-10 and 13-15). These results demonstrate an inhibitory effect of UL26 on ISGylation during virus infection.
Since HCMV growth was suppressed by IFN-induced ISGylation and UL26 inhibited ISGylation, we tested whether UL26-deleted mutant virus is more susceptible to type I IFN treatment than wild-type virus. We found that while the titers of wild-type virus were reduced by 12-fold by IFNβ pre-treatment, those of UL26-deleted virus were reduced by 40-fold, indicating that UL26-deleted virus less effectively overcomes the IFNβ-mediated anti-viral responses than wild-type virus (Fig 7G). Overall, our results demonstrate that, in addition to IE1 that suppresses ISG15 transcription, UL26 plays an important role in evading the ISG15-associated antiviral responses by inhibiting protein ISGylation (Fig 7H).

Discussion
Our analysis with UV-HCMV and IE1-deleted mutant virus demonstrates that ISGylation is induced by HCMV infection and that IE1 plays a central role in downregulating ISGylation by reducing ISG15 transcription. The latter is consistent with previous findings that IE1 represses transcription of ISGs by sequestrating STAT2 [58][59][60] and PML [61,62]. Notably, although IE1 effectively reduced ISG expression, the level of ISGylation during HCMV infection largely depended on the MOI.
Our data provide evidence for the antiviral roles of ISGylation during HCMV infection. To discern the effects of free ISG15 expression and protein ISGylation on virus replication, we used HF cells in which a specific ISGylation enzyme was depleted by shRNA. The data consistently showed an inverse relationship between the level of ISG15 conjugates and HCMV growth; i.e., enhanced ISGylation by UBP43 knockdown decreased viral growth, while reduced ISGylation by depletion of UBE1L, UbcH8, or Herc5 increased viral growth. Therefore, we conclude that protein ISGylation in general inhibits HCMV growth. Free ISG15 has also been shown to inhibit the replication of certain viruses [34,35,44]. In our analysis, however, overexpression of ISG15 GG with UBE1L, UbcH8, and Herc5 led to a mild reduction of HCMV growth, whereas ISG15 AA , an inactive form, did not significantly affect viral growth, suggesting that expression of free ISG15 prior to HCMV infection may minimally affect viral growth in cultured cells. It should be noted that given the role of free ISG15 in stabilizing UBP43 (or USP18), a negative regulator of IFN signaling [53], overexpression or depletion of ISG15 might affect both positive and negative activities of ISG15 to viral growth. infected with wild-type or ΔUL26 mutant virus at an MOI of 1. The production of progeny virions in the culture supernatant at 6 days after infection was measured by infectious center assays. The results shown are the mean values and standard errors of four independent experiments. (H) Summary for the HCMV strategy that consecutively inhibits ISG15 transcription and protein ISGylation. IE1 inhibits ISG transcription, while UL26 inhibits protein ISGylation.
doi:10.1371/journal.ppat.1005850.g007 HCMV replication was inhibited by ISGylation at multiple steps. First, the expression of viral genes was inhibited under conditions where the level of ISG15 conjugates was increased. Viral regulators that promote viral gene expression may be a direct target of ISGylation. However, we could not observe ISGylation of IE1, which is responsible for the activation of MIE promoters, or IE2, a strong transactivator of viral early and late genes. ISGylation of cellular proteins that in particular play a role in innate immune responses may affect viral gene expression. Notably, ISGylation of IRF3 increases its stability and enhances IRF3-mediated transcriptional activation during Sendai virus infection [16,21] and ISGylation of 4EHP, an mRNA 5' cap structure-binding translation suppressor, plays a role in IFN-induced innate immune response [9]. Second, HCMV virion release was inhibited by ISG15 conjugation. The inhibitory role of ISG15 expression in the budding process of enveloped viruses has been demonstrated in retrovirus infection. ISG15 expression inhibits ubiquitination of HIV Gag and tumor susceptibility gene-101 (Tsg101) proteins, leading to disruption of their interaction [26]. ISG15 conjugation of CHMP5, 2A, and 6 in the ESCRT machinery causes release of vacuolar proteinsorting 4 (VPS4) from the membrane, leading to inhibition of virion release [43]. The ESCRT machinery seems to be involved in the process of HSV-1 and HCMV maturation [83][84][85][86]. Therefore, ISGylation may affect the HCMV maturation process by targeting components involved in the ESCRT machinery.
Although the antiviral role of ISG15 expression and ISGylation has been demonstrated in several viruses, studies on viral targets for ISGylation and viral strategies that interfere with the ISG15-mediated antiviral functions are limited to a few examples. ISGylation of NS1A of influenza A virus inhibits virus replication by interfering with NS1A nuclear import [42]. KSHV vIRF1 is ISGylated but it role on vIRF1 function is not known [39]. In the present study, we demonstrated that HCMV pUL26 is a target for ISGylation. UL26 ISGylation appears to regulate protein stability by competing with ubiquitination. Our analysis using the UL26-p21-ISG15 fusion protein demonstrated that ISGylation inhibits the activities of UL26-p21 to downregulate the TNFα-mediated NF-κB activation and to complement the growth defect of UL26-deleted virus. Therefore, ISGylation of pUL26 is thought to inactivate its function, suppressing HCMV growth. Given the notion that newly synthesized proteins are targeted extremely broadly, perhaps stochastically, by Herc5 in polyribosomes [41], several other HCMV proteins may be ISGylated during infection. We indeed observed that some viral proteins, which did not interact with ISG15, UBE1L, and Herc5 in co-IP assays, were ISGylated in our co-transfection/ISGylation assays. However, we think that ISGylation occurs in a manner dependent on the context of each protein. Furthermore, since several HCMV proteins were bound to ISG15 and/or UBE1L in yeast and co-IP assays, it is also likely that other viral proteins besides pUL26 may affect ISGylation. Identification of more HCMV proteins that interact with the ISGylation system and studies on their functional relevance are warranted.
A few examples for viral regulation of ISGylation have been described. The NS1 protein of influenza B virus non-covalently binds to ISG15 and inhibits ISGylation [3]. Similarly, the vaccinia virus E3 protein interacts with ISG15 and disrupts its antiviral activity [36]. Nairoviruses and arteriviruses have shown to encode ovarian tumor domain (OTU)-containing proteases that hydrolyze ISG15 from target proteins [87] and severe acute respiratory syndrome (SARS) coronavirus encoded the papain-like protease that cleaves both ubiquitin and ISG15 [88]. Recently, KSHV-encoded vIRF1 was shown to interact with Herc5 [39]. In this study, we demonstrated that HCMV UL26-p21 is able to non-covalently bind to ISG15, UBE1L, and Herc5, and inhibit ISGylation. Whether binding all of ISG15, UBE1L, and Herc5 is critical for pUL26 to inhibit ISGylation is not clear and needs to be further investigated. Comparative analysis of wild-type and UL26-deleted mutant viruses demonstrate that de novo expression of UL26 in virus-infected cells correlated with reduced accumulation of ISG15 conjugates. In addition to IE1, the presence of additional viral functions that downregulate protein ISGylation was prompted by our observation that UV-HCMV infection resulted in a higher level of ISG15 conjugates than IE1-deleted mutant virus at late times (Fig 2B). UL26 deletion mutant virus shows a moderate growth defect at low MOIs [72,89]. Notably, the UL26 virus (in AD169 strain) showed a reduced plaque formation at low MOIs and this defect could be rescued by IE1 overexpression [72]. More importantly, like IE1-deleted virus [58][59][60], the growth of UL26-deleted virus was more sensitive to pre-treatment of type I IFNs [74] (and this study), suggesting a role of UL26 in antagonizing type I IFN response. It is likely that a common ISG15-targeting mechanism is shared by influenza B virus, vaccinia virus, and HCMV. HCMV also encodes a tegument protein pUL48 that contains a deubiquitinating protease domain; however, its activity was specific for ubiquitin and did not cleave ISG15 [90].
In this study, we demonstrate that the cellular ISGylation system is a critical part of cellular innate immune response against HCMV infection. We also provide evidence that HCMV has developed several strategies to disarm the ISGylation-mediated antiviral activity; IE1 reduces ISG15 transcription and UL26 inhibits protein ISGylation. This interplay of HCMV with the cellular ISGylation system may be critical for the virus to successfully establish a persistent infection.

Cell culture, transfection, and virus
Human foreskin fibroblast (HF) (ATCC) and human embryonic kidney (HEK) 293T cells (ATCC) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin (100 U/ml), and streptomycin (100 μg/ml). DNA transfection of 293T cells was performed using the N,N-bis-(2-hydroxyethyl)-2-aminoethanesulfonic acid-buffered saline (BBS) version of the calcium phosphate procedure. Electroporation of HF cells was conducted using a Microporator MP-100 (Digital Bio), as described previously [59]. Stocks for the parent Towne virus and the CR208 mutant virus with IE1-deleted were prepared in ihfie1.3 cells as described previously [59]. The HCMV (Towne strain)-GFP virus was grown in HF cells after electroporation with the bacmid DNAs. Wild-type and UL26-deleted HCMVs (AD169 strain) were previously described [72] and grown in UL26-expressing HF cells. Wild-type and UL26-HA-expressing HCMVs (Toledo strain) generated in this study were grown in HF cells. To produce UV-inactivated HCMV (UV-HCMV), the virus stock was irradiated with UV light three times at 0.72 J/cm 2 using a CL-1000 Crosslinker (UVP).

Yeast two-hybrid assays
Yeast AH109 (MATa) cells were transformed with plasmid expressing the GAL4-DNA-binding (DB)-ISG15 (TRP + ) or GAL4-DB-UBE1L (TRP + ) fusion protein. Y187 (MATα) cells were transformed with plasmid expressing the GAL4-activation domain (A)-HCMV ORF (LEU2 + ) fusion proteins. Each transformant was selected on plates lacking tryptophan (SC-Trp) or leucine (SC-Leu). Trp + and Leu + transformants were mated with each other on YPD plates. Diploid cells (a/α) were selected on plates lacking both tryptophan and leucine (SC-TrpLeu). Trp + Leu + colonies were tested for their growth on plates that lack tryptophan, leucine and histidine (SC-TrpLeuHis). Cells expressing bait and prey that interact with each other grow on SC-TrpLeuHis. Cells expressing both GAL4-DB-ISG15 and GAL4-A-UBE1L were used as a positive control, whereas cells expressing GAL4-DB-ISG15 and GAL4-A only were used as a negative control.

Production of the recombinant virus expressing UL26-HA proteins
The Toledo-BAC clone encoding UL26 proteins with a C-terminal HA tag was produced by using a counter-selection BAC modification kit (Gene Bridges). The scheme for bacmid mutagenesis is described in S5 Fig and the LMV primers used for mutagenesis are listed in S1 Table.

Antibodies
Mouse monoclonal antibody (MAb) 810R, which detects epitopes present in both IE1 and IE2, was purchased from Chemicon. Mouse MAbs against UL44 (p52) and UL99 (pp28) were obtained from Virusys. Anti-β-actin and anti-α-tubulin mouse MAbs were purchased from Sigma. Anti-HA rat MAb 3F10 and anti-myc mouse MAb 9E10, conjugated with peroxidase or labeled with fluorescein isothiocyanate (FITC), were purchased from Roche. Anti-ISG15 (F-9) and anti-STAT2 mouse MAbs were obtained from Santa Cruz. Mouse MAb against SRT epitope was previously described [91]. UBP43 antibody was previously described [92]. Rabbit polyclonal Ab (PAb) for STAT2 (C-20) and STAT2 phosphorylated at Tyr689 were purchased from Santa Cruz and Upstate, respectively. Rabbit PAb for ISG15 was kindly provided by Chin Ha Chung (Seoul National University, Seoul, Republic of Korea).

Immunoblot analysis and indirect immunofluorescence assay (IFA)
For immunoblot analysis, cells were washed with phosphate-buffered saline (PBS) and total cell lysates were prepared by boiling the cell pellets in sodium dodecyl sulfate (SDS) loading buffer. Equal amounts of the clarified cell extracts were separated on a SDS-polyacrylamide gel or Gradi-Gel II (Elpis biotech, Republic of Korea) and electroblotted onto nitrocellulose membranes. The blots were blocked by incubation for 30 min at room temperature with PBS plus 0.1% Tween 20 (PBST) containing 5% nonfat dry milk. After being washed with PBST three times, the blots were incubated with the appropriate antibodies in PBST for 1 h at room temperature. After three 5 min washes with PBST, the blots were incubated with horseradish peroxidase-conjugated goat anti-mouse IgG or anti-rabbit IgG (Amersham) for 1 h at room temperature. The blots were then washed three times with PBST, and the protein bands were visualized with enhanced chemiluminescence system (Amersham).
For IFA, cells were fixed in ice-cold methanol for 5 min and rehydrated in cold PBS. Then, the cells were incubated with appropriate primary antibodies in PBS at 37°C for 1 h, followed by incubation with appropriate secondary antibodies at 37°C for 1 h. The mounting solution containing Hoechst and anti-fade reagent (Molecular Probes) was used. For double-labeling, two different antibodies were incubated together. Slides were examined and with a Carl Zeiss LSM710Meta confocal microscope system.

Infectious center assay
The diluted samples were used to inoculate a monolayer of 4 × 10 4 HF cells in a 24-well plate. At 24 h post infection, cells were fixed with 500 μl of cold methanol for 10 min. The cells were then washed three times in phosphate-buffered saline (PBS) and incubated with anti-IE1 rabbit polyclonal antibody in PBS at 37°C for 1 h, followed by incubation with phosphatase-conjugated anti-rabbit immunoglobulin G (IgG) antibody in phosphate-buffered saline (PBS) at 37°C for 1 h. Finally, the cells were gently washed in PBS and treated with 200 μl of developing solution (nitroblue tetrazolium/5-bromo-4-chloro-3-indolylphosphate) at room temperature for 1 h. The positively stained cells were counted for at least three to five separate fields per well under a light microscope (× 200 magnification).

Co-immunoprecipitation (co-IP) assays
Co-transfected 293T cells (8 × 10 5 ) or virus-infected HF cells were harvested and sonicated in 0.7 ml co-IP buffer [50 mM Tris-Cl (pH 7.4), 50 mM NaF, 5 mM sodium phosphate, 0.1% Triton X-100, containing protease inhibitors (Sigma)] by a microtip probe (Vibra-Cell; Sonics and Materials, Inc., USA) for 10 s (pulse on: l s, pulse off: 3 s). Cell lysates were incubated with appropriate antibodies. After incubation for 16 h at 4°C, 30 μl of a 50% slurry of protein A-and G-Sepharose (Amersham) were added and then the mixture was incubated for 2 h at 4°C to allow adsorption. The mixture was then pelleted and washed 7 times with co-IP buffer. The beads were resuspended and boiled for 5 min in loading buffer. Each sample was analyzed by SDS-PAGE and immunoblotting with appropriate antibodies.

Statistical analysis
Statistical significances were determined using the Student's t-test and are indicated by Ã P<0.05, ÃÃ P<0.01, or ÃÃÃ P<0.001.  [93] was a gift from Hua Zhu (UMDNJ-New Jersey Medical School, Newark, NJ, USA). The rpsL-neo cassettes were PCR-amplified using LMV1766/ 1767 primers containing homology arms consisting of 50 nucleotides upstream and downstream of the target region plus 24 nucleotides homologous to the rpsL-neo cassette. The amplified rpsLneo fragments with homology arms were purified and introduced into E. coli GS243 containing wild-type Toledo-BAC for recombination by electroporation using a Gene Pulser II (Bio-Rad). The intermediate Toledo-BAC constructs containing the rpsL-neo cassette were selected on Luria Broth (LB) plates containing kanamycin. Next, the rpsL-neo cassette was replaced by annealed oligo DNAs (LMV1768/1769) consisting of only homology arms (50 nucleotides upstream and downstream of the target region). The ΔUL26 Toledo-BAC was selected on LB plates containing streptomycin. The mutated regions were amplified by PCR and sequenced to verify the desired mutation. The Toledo-BAC encoding UL26-HA was generated from the mutant Toledo-BAC. First, the rps-neo cassettes flanked by homology arms were inserted again into the mutant Toledo-BAC. Next, DNA fragments containing the wild-type UL26 gene with a HA tag at its Cterminus were PCR amplified by 2-steps using LMV1805/1812 and LMV1805/1772. The amplified UL26-HA gene was then inserted into the Toledo-BAC containing the rps-neo cassette by homologous recombination. The LMV primers used for mutagenesis are listed in S1 Table. (B) The regions containing the UL26 ORF from Wt, ΔUL26, and UL26-HA bacmid DNAs were PCR amplified with LMV1764/1765. (C) Wt, ΔUL26, and UL26-HA bacmid DNAs were digested with BglII and the digestion patterns were compared via agarose gel electrophoresis. The bands corresponding to 5,226 and 5,253 bp from wild-type and UL26-HA bacmids, respectively, and a band of 4,660 bp from ΔUL26 bacmid were indicated as arrowheads. Comparative co-transfection/ISGylation assays for UL26 and IE2 were performed in 293T cells with or without increasing amounts of plasmids expressing SRT-UL26-p21 or SRT-IE2 IE1 as in Fig 2D. Cell lysates were prepared and immunoprecipitated with anti-SRT antibody, followed by immunoblotting with anti-myc antibody. Whole cell lysates were immunoblotted with anti-SRT antibody to determine the expression levels of UL26-p21 and IE2, or with anti-myc antibody to determine the effect of UL26-p21 or IE2 expression on ISGylation. (TIF) S8 Fig. Comparison of ISGylation between wild-type and ΔUL26 virus infected cells. HF cells were mock-infected or infected with wild-type or ΔUL26 mutant virus (Ad169) at an MOI of 0.2. Cell lysates were immunoblotted at the indicated time points with antibodies for ISG15, viral proteins (IE1, IE2, and UL26), and β-actin. (TIF) S1 Table. PCR primers used for bacmid mutagenesis. (TIF) S2 Table. Summary of the HCMV proteins that interacted with ISG15 and UBE1L in yeast two-hybrid assays and co-IP assays.