Revisiting promyelocytic leukemia protein targeting by human cytomegalovirus immediate-early protein 1.

Promyelocytic leukemia (PML) bodies are nuclear organelles implicated in intrinsic and innate antiviral defense. The eponymous PML proteins, central to the self-organization of PML bodies, and other restriction factors found in these organelles are common targets of viral antagonism. The 72-kDa immediate-early protein 1 (IE1) is the principal antagonist of PML bodies encoded by the human cytomegalovirus (hCMV). IE1 is believed to disrupt PML bodies by inhibiting PML SUMOylation, while PML was proposed to act as an E3 ligase for IE1 SUMOylation. PML targeting by IE1 is considered to be crucial for hCMV replication at low multiplicities of infection, in part via counteracting antiviral gene induction linked to the cellular interferon (IFN) response. However, current concepts of IE1-PML interaction are largely derived from mutant IE1 proteins known or predicted to be metabolically unstable and globally misfolded. We performed systematic clustered charge-to-alanine scanning mutagenesis and identified a stable IE1 mutant protein (IE1cc172-176) with wild-type characteristics except for neither interacting with PML proteins nor inhibiting PML SUMOylation. Consequently, IE1cc172-176 does not associate with PML bodies and is selectively impaired for disrupting these organelles. Surprisingly, functional analysis of IE1cc172-176 revealed that the protein is hypermodified by mixed SUMO chains and that IE1 SUMOylation depends on nucleosome rather than PML binding. Furthermore, a mutant hCMV expressing IE1cc172-176 was only slightly attenuated compared to an IE1-null virus even at low multiplicities of infection. Finally, hCMV-induced expression of cytokine and IFN-stimulated genes turned out to be reduced rather than increased in the presence of IE1cc172-176 relative to wild-type IE1. Our findings challenge present views on the relationship of IE1 with PML and the role of PML in hCMV replication. This study also provides initial evidence for the idea that disruption of PML bodies upon viral infection is linked to activation rather than inhibition of innate immunity.


Introduction
Promyelocytic leukemia (PML) bodies, also known as nuclear domain 10, are membrane-less nuclear organelles present in most cells (reviewed in [1,2]). PML bodies are heterogeneous and dynamic, ranging in size between 0.1 and 1.0 μm and typically displaying as five to 30 spherical structures per nucleus interspersed between chromatin. They are composed of the eponymous PML proteins, also known as tripartite motif (TRIM) 19, that self-organize into a shell-like scaffold. The PML scaffold forms, in part by phase separation, around an inner core which sometimes contains nucleic acids [3][4][5][6][7][8][9][10]. Embedded in the scaffold or core are numerous unrelated proteins, most of which associate with PML bodies in a conditional and transient manner [11,12]. Proteins constitutively residing in PML bodies include the six nuclear PML isoforms, the speckled 100 kDa (Sp100) nuclear antigens, the death domain-associated protein (Daxx) and the small ubiquitin-like modifier (SUMO) family members SUMO1, SUMO2, SUMO3 and SUMO5 [13,14]. PML bodies are considered to be SUMOylation hubs, and most proteins associated with these organelles are post-translationally modified by one or more SUMO paralogs [12,15]. While SUMO1 attaches only as a monomer, SUMO2, SUMO3 and SUMO5 can form polymeric chains on target proteins including PML [14,16]. The biogenesis and integrity of functional PML bodies depends on oligomerization and poly-SUMOylation of PML proteins at three major lysine residues. In the absence of SUMOylation, PML proteins may condense into spherical structures unable to recruit other proteins [17,18]. For instance, aggregates referred to as mitotic accumulations of PML protein (MAPPs) are known to form following PML de-SUMOylation and breakdown of PML bodies at the onset of mitosis [19][20][21][22][23].
appears to enhance IE1 SUMOylation as well [57,58]. Finally, C-terminal amino acids 476 to 491 represent the chromatin tethering domain (CTD) [48,59]. This domain contains a nucleosome binding motif (NBM) that targets the acidic patch formed by histones H2A and H2B on the nucleosome surface and thereby mediates association with both interphase and mitotic chromatin [60][61][62]. The CTD peptide adopts an extended, v-shaped conformation with a short α-helix at its C-terminus and has been implicated in regulation of higher-order chromatin structure and viral genome maintenance during hCMV latency [62,63]. In the context of high passage hCMV strains, IE1 has been shown to be required for efficient viral replication in fibroblasts under conditions of low but not high multiplicity of infection (MOI) [64][65][66]. However, low passage hCMV strains deficient in IE1 are substantially attenuated at both low and high MOI in fibroblasts suggesting that the protein serves an even more important role in viral replication than originally thought ( [67] and this work).
Initially characterized as a promiscuous activator of transcription, IE1 has later emerged as an antagonist of intrinsic and innate immune responses to hCMV infection (reviewed in [43,68]). At least two distinct activities contribute to evasion of intrinsic and innate immunity by IE1. First, the STAT binding motif in the viral protein confers complex formation with STAT2 [45,52,53,69]. The IE1-STAT2 interaction results in diminished DNA binding of IFN-stimulated gene factor 3, a trimeric complex of STAT1, STAT2 and IFN regulatory factor 9 required for induction of type I ISGs many of which encode antiviral products. Consequently, IE1 inhibits type I ISG activation conferring relative resistance to the antiviral activities of IFNα and IFNβ on hCMV [45,53,69]. In addition, the STAT binding motif mediates interaction between IE1 and STAT3 resulting in decreased interleukin 6 (IL6)-induced gene expression and increased expression of type II ISGs linked to STAT1 phosphorylation [52, [70][71][72]. Secondly, the IE1 core domain interacts with the coiled-coil domain of PML [49,73]. Complex formation between IE1 and PML results in transient co-localization of IE1 at PML bodies. This interaction is followed by inhibited oligomerization of PML, reduced de novo PML SUMOylation and disruption of PML bodies [56, [74][75][76]. PML targeting by IE1 is thought to promote hCMV replication in at least two ways: by relieving viral transcription from repression mediated by PML proteins or PML bodies (intrinsic immunity) and by inhibiting PMLdependent IFN and ISG expression triggered by viral infection (innate immunity) [34, 74,77,78]. Besides PML isoforms and SUMO paralogs, IE1 has also been shown to target two other resident proteins of PML bodies, Daxx and Sp100A [51, 79,80], most likely to antagonize antiviral restriction via transcriptional repression. While the IE1-Daxx interaction has not been studied in detail, the viral protein was shown to reduce the SUMOylation of Sp100A and to target the cellular protein for proteasomal degradation [51,79].
Numerous studies have investigated how IE1-PML interaction and disruption of PML bodies affect the hCMV productive cycle (reviewed in [42,81]). It has been concluded that PML targeting is a central activity by which IE1 antagonizes intrinsic and innate immunity to facilitate hCMV replication [34,77,78,82]. However, most of these studies were merely of correlative nature and involved IE1 mutants with disrupted core domains resulting in proteins known to be metabolically unstable and predicted to be globally misfolded. For example, many conclusions have relied on IE1 mutant proteins and viruses that replace leucine 174 with proline (L174P) [56,77,83,84]. To our knowledge, a mutation in IE1 that produces a metabolically stable protein and selectively abolishes PML binding without affecting other activities has not been identified. Here we performed systematic clustered charge-to-alanine scanning of IE1 to identify a mutant protein (IE1cc172-176) that fails to interact physically with PML, to inhibit PML SUMOylation and to associate with PML bodies. However, IE1cc172-176 accumulates to wild-type levels in host cell nuclei, and functions unrelated to PML are not affected by the mutation. We utilized IE1cc172-176 to re-evaluate the contribution of PML targeting by IE1 to hCMV replication. Our findings challenge the predominant view that interaction with PML is central to the function of IE1 and provide evidence for antiviral rather than proviral effects of PML body disruption.

Clustered charge-to-alanine scanning of IE1 identifies stable core domain mutants
To identify stable mutant proteins that allow for conclusive discrimination between individual activities of IE1, we performed systematic clustered charge-to-alanine mutagenesis across the whole length of the viral protein. Charged residues clustered in the primary sequence are likely to be located on the surface of the folded protein where they may facilitate interactions with other proteins. Conversely, the replacement of clustered charged residues with alanine may disrupt such interactions without significantly affecting the three-dimensional structure and stability of the protein [85][86][87][88][89][90]. A clustered charge is commonly defined as two or more charged residues in a window of five amino acids [87,91]. According to this definition, 24 clustered charges were identified across the 491 amino acids of IE1 (Fig 1). Two to six charged residues within each cluster, selected based on charge density, were changed to alanine by sitedirected mutagenesis (Fig 1B). The resulting 24 clustered charge mutants are referred to as IE1cc6-8, IE1cc21-26, etc with numbers indicating the residues N-and C-terminal of the target site. As expected, most residues targeted by clustered charge mutagenesis are predicted to reside on the surface of the viral protein ( Fig 1A). Following cloning and lentiviral transfer, MRC-5-derived cell lines individually expressing the 24 IE1 mutants in a doxycycline (dox)inducible fashion (TetR-IE1 cells) were generated.
The clustered charge-to-alanine scanning indicates that even small mutations in surface residues of the IE1 core domain tend to interfere with protein accumulation to normal levels. Yet, this analysis identifies mutations both in-and outside the core domain that result in stable proteins.
A subset of IE1 core domain mutants are defective for PML-related activities during interphase and mitosis Next, we conducted immunofluorescence microscopy to compare the intracellular localization of wild-type IE1 and the 24 clustered charge mutants. All mutants except for IE1cc21-26 resembled the wild-type protein in localizing predominantly, if not exclusively, to the nucleus (Fig 3A and 3B
These experiments identify residues in the IE1 core domain required for compromising the integrity or preventing the formation of PML complexes during interphase and mitosis by adversely affecting IE1 levels, PML interaction or both.

IE1cc172-176 encodes a stable protein selectively defective for physical and functional interaction with PML
The immunofluorescence experiments described above identified eight IE1 mutant proteins (IE1cc112-114, cc134-138, cc161-165, cc172-176, cc258-263, cc286-292, cc307-310, cc318-319) that fail to co-localize with PML and are severely defective for PML body disruption as well as inhibition of MAPP formation. However, only IE1cc172-176 produced normal protein levels (Fig 2A) suggesting that the other seven mutations may impose general rather than PML-specific functional defects on IE1. Accordingly, only IE1cc172-176 but neither of the other mutants with PML-related phenotypes were comparable to the wild-type protein in regulating STAT signaling. This conclusion was drawn based on the induction of STAT1-and inhibition of STAT2-or STAT3-stimulated host genes (S2 Fig). To further characterize the IE1cc172-176 mutant, we compared this protein to previously published IE1 mutants including IE1dl410-420, L174P, L130G/I132G/L133G (also known as YL3) and Y315G/V316G/L317G (also known as YL4). IE1dl410-420 lacks 11 amino acids from the 'acidic domain' that encompass the STAT binding motif, and is selectively inactive for regulating IFN-and IL6-type signaling [52]. This deletion was also combined in a double mutant with IE1cc172-176 (IE1cc172-176/dl410-420). The IE1 L174P core domain mutant is broadly defective for all tested functions including PML binding and co-localization, inhibition of PML SUMOylation, PML body disruption and inhibition of PML-mediated transcriptional repression [34,56,77,83]. The core domain mutants IE1 L130G/I132G/L133G and Y315G/V316G/L317G have both been shown to be defective in PML co-localization, PML body disruption, inhibition of PML SUMOylation and functional complementation of infected cell protein 0 (ICP0), a herpes simplex virus 1 protein that shares functional similarities with IE1 [98,99]. In addition, the core domain mutant IE1dl291-320 was included in the some of TetOderived plasmids expressing only the HA tag (w/o) or HA-tagged forms of the indicated wild-type (wt) and clustered charge mutant IE1 proteins. Indirect immunofluorescence staining was performed using mouse anti-HA and rabbit anti-PML combined with goat anti-mouse Alexa Fluor 488 and goat antirabbit Alexa Fluor 594 antibodies. Images from interphase cells showing the typical localization of IE1 and PML are presented along with merge images (Leica DMRX microscope, 63× objective). (C) TetR (w/o) and TetR-IE1 cells expressing the indicated wild-type (wt) or clustered charge mutant IE1 proteins were treated with dox for 24 h. Indirect immunofluorescence staining was performed using mouse anti-HA (IE1) and rabbit anti-PML combined with goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 antibodies. 4 0 ,6-diamidino-2-phenylindole (DAPI) was used to stain DNA. Merge images were taken using a Keyence BZ-9000 microscope (40× objective). Representative images are shown in S1 Fig. The percentage of cells exhibiting predominantly disrupted or intact PML bodies was determined from at least two fields of view (>100 cells) based on manual inspection aided by ImageJ software (National Institutes of Health) [134]. the assays. IE1dl291-320 is generally defective in both PML-and STAT-related functions (S3 Fig) [34 , 73].
Immunofluorescence microscopy confirmed that all tested IE1 mutants localized to the nucleus but only IE1dl410-420 was able to disrupt PML bodies like the wild-type protein ( Fig  5A and S3C Fig). We further observed a less efficient reduction in high molecular weight PML species compared to wild-type IE1 in all mutants other than IE1dl410-420 ( Fig 5B). Finally, binding to PML above background was exclusively observed for the wild-type and IE1dl410-420 proteins, but for none of the other mutants ( Fig 5C). Out of all mutants defective in these assays, only IE1cc172-176 and cc172-176/dl410-420 were present at levels comparable to the wild-type protein, while the levels of IE1 L174P, L130G/I132G/L133G and Y315G/V316G/L317G were substantially lower ( Fig 5C). Relative comparison of metabolic stability by temporally controlled IE1 expression further demonstrated that all tested mutants except for IE1cc172-176 exhibit markedly accelerated turnover compared to the wild-type protein (S4 Fig).
To our knowledge, IE1cc172-176 is the first mutant suited to provide specific information about the contribution of PML-related activities to IE1 function and hCMV replication, as it fails to interact with PML but produces a stable protein fully active for other functions. We therefore focused our further analyses on the phenotype of IE1cc172-176.

IE1 can undergo mixed SUMO chain formation and is SUMOylated at nucleosomes independent of PML binding
It has been established that IE1 can undergo post-translational modification by covalent attachment of a single SUMO1 or SUMO3 moiety to K450 [53-58, 83, 100, 101]. Furthermore, a recent report concluded that PML acts as an E3 ligase for IE1 SUMOylation [58]. Thus, we expected that IE1cc172-176 would exhibit reduced SUMOylation relative to the wild-type protein. However, increased levels of high molecular weight forms most likely corresponding to mono-SUMOylated proteins were detected for IE1cc172-176 and cc172-176/dl410-420 compared to wild-type IE1 ( Fig 5B). Further analysis revealed that IE1cc172-176 and cc172-176/ dl410-420 are not only hyper-modified by SUMO monomers, but also form polymeric chains that include SUMO1 and SUMO2 (Fig 5C). Hyper-SUMOylation appeared to result specifically from lack of PML binding, since it was not observed in IE1 proteins other than those carrying the cc172-176 mutation (Figs 5B, 5C and 6A). As expected, any detectable SUMO conjugation to IE1 was abolished when K450 was replaced by arginine in either a single mutant (IE1 K450R) or a double mutant (IE1cc172-176/K450R) ( Fig 6A). However, unexpectedly, proteins lacking the CTD (IE1dl476-491, cc172-176/dl476-491) required for nucleosome binding turned out to be SUMOylation-deficient as well. Consistent with a link between chromatin targeting and SUMO modification, IE1dl476-491 exhibited reduced localization to chromatin, lack of PML binding (IE1cc172-176) resulted in enhanced chromatin association, and an intermediate phenotype was observed for a double mutant (IE1cc172-176/dl476-491). SUMOylated forms of IE1 were detected across all nuclear compartments including nucleoplasm, chromatin and nuclear matrix (Fig 6B).
These results reveal that SUMOylation of IE1 is nucleosome-based, occurs independent of PML binding and can involve mixed polymeric chains. and TetR-IE1 cells expressing the indicated wild-type (wt) or clustered charge mutant IE1 proteins were treated with dox for 24 h. Indirect immunofluorescence staining was performed using mouse anti-HA and rabbit anti-PML combined with goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 antibodies. DAPI was used to stain DNA. Images from mitotic cells showing the typical localization of IE1, IE2 and PML relative to DNA are presented along with merge images (Keyence BZ-9000 microscope, 100× objective). https://doi.org/10.1371/journal.ppat.1008537.g004

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited

IE1 targets PML and Sp100 via distinct activities
Besides PML and SUMO1-3, Sp100A is another constituent of PML bodies targeted by IE1. It has been proposed that IE1 binds to Sp100A, interferes with Sp100A SUMOylation and targets the protein for proteasomal degradation [51, 79,102]. It is unclear whether the effects IE1 exerts on PML and Sp100A are linked or not. We performed immunofluorescence analyses to monitor how wild-type and mutant IE1 proteins affect the localization and accumulation of Sp100. Sp100 was predominantly identified as nuclear dots both in the absence of IE1 as well as in the presence of mutants that accumulate to low levels and are thought to be broadly defective (IE1 L174P, L130G/I132G/L133G or Y315G/V316G/L317G). In contrast, little Sp100 was detected when wild-type IE1, IE1cc172-176, dl410-420 or cc172-176/dl410-420 were expressed ( Fig 7A). Results consistent with these observations were obtained from immunoblotting. Reduced levels of Sp100 were detected in cells expressing wild-type IE1, IE1cc172-176, dl410-420 or cc172-176/dl410-420, but not in those expressing IE1 L174P, L130G/I132G/ L133G or Y315G/V316G/L317G. The reduction in Sp100 levels appeared to affect preferentially higher molecular weight (SUMOylated) forms of Sp100, most likely both Sp100A and Sp100B ( Fig 7B).
These results indicate that Sp100 targeting by IE1 is a distinct activity unrelated to PML or STAT binding. Thus, IE1 targets the two major structural constituents of PML bodies via physically separable and functionally distinct mechanisms.

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited

Fig 7. Distinction between Sp100 degradation and PML interaction by IE1. (A) TetR (w/o) and
TetR-IE1 cells expressing the indicated HA-tagged wild-type (wt) or mutant IE1 proteins were treated with dox for 24 h. Indirect immunofluorescence staining was performed using mouse anti-HA (IE1) and rabbit anti-Sp100 combined with goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 antibodies. DAPI was used to stain DNA. Images from interphase cells showing the typical staining pattern of IE1 and Sp100 relative to DNA are presented along with merge images (Keyence BZ-9000 microscope, 100× objective). (B) TetR cells expressing firefly luciferase and TetR-IE1 cells expressing the indicated HA-tagged wild-type (wt) or mutant IE1 proteins were treated with dox for 48 h. Whole cell extracts prepared in buffer (pH 8.0) with NEM were subjected to immunoblotting for Sp100, IE1 (mouse anti-HA) and GAPDH. https://doi.org/10.1371/journal.ppat.1008537.g007

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited

PML targeting by IE1 is dispensable for hCMV replication
The ability to counteract PML has been considered to be crucial for IE1 function and hCMV replication in human fibroblasts, especially at low MOIs. As pointed out above, this perception is largely based on the analysis of metabolically unstable and presumably misfolded mutant proteins including IE1 L174P. To revisit the impact of PML targeting on hCMV replication, TetR-IE1 cell lines expressing wild-type or mutant IE1 and IE1-negative control cells were infected at two different MOIs (0.001 or 0.5 PFU/cell) with IE1-deficient hCMV (gTBdlIE1) expressing the enhanced green fluorescent protein (EGFP). Infection was monitored by quantifying extracellular viral DNA ( Fig 8A) and imaging fluorescence emission from EGFP ( Fig  8B). As expected, little (MOI = 0.5) or no (MOI = 0.001) replication by gTBdlIE1 was observed in the absence of IE1, but replication was boosted by up to five logs in the presence of wildtype IE1. IE1 L174P, L130G/I132G/L133G or Y315G/V316G/L317G failed to demonstrate any compensatory effect on gTBdlIE1 replication at either MOI consistent with the low protein

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited levels and broad functional defects observed with these mutants. In contrast, attenuated mutant virus replication was rescued to varying degrees by IE1cc172-176, dl410-420 or cc172-176/dl410-420. In fact, gTBdlIE1 replication only modestly differed between cells expressing wild-type IE1 or IE1cc172-176 with roughly one log difference at the lower and almost no difference at the higher MOI. The replication phenotypes linked to lack of PML binding (IE1cc172-176) or STAT binding (IE1dl410-420) appeared to be additive in the double mutant (IE1cc172-176/dl410-420). In contrast to the IE1-PML interaction, the IE1-STAT interaction contributed more significantly to viral replication at the higher rather than the lower MOI. The MOI-dependent phenotype of the IE1dl410-420 mutant may result from increased IFN secretion when a larger number of cells are infected.
These findings indicate that the IE1-PML interaction aids hCMV replication to some extent when cells are infected by single virus particles. However, PML targeting is neither crucial for IE1 function nor for hCMV replication in human fibroblasts, not even at low MOIs.

PML targeting by IE1 is linked to activation rather than inhibition of antiviral gene expression
Previous work established that IE1 inhibits type I IFN signaling and subsequent induction of ISGs by targeting STAT2 [45, 53, 69]. More recent reports suggested that PML interaction contributes to inhibition of type I IFN signaling and ISG induction by IE1 as well [34,77]. However, the latter studies relied on IE1 L174P and mutants with large deletions (IE1 1-382, IE1dl291-320) known or expected to produce proteins that are metabolically unstable, globally misfolded, broadly inactive or all of the above (S4 Fig) [ Our initial analysis of IE1cc172-176 in TetR-IE1 cells revealed that this mutant was at least equally efficient as the wild-type protein in inhibiting the induction of OAS1, a prototypical ISG, triggered by IFNα (S1 Fig). To expand on this observation, fibroblasts were induced to express wild-type IE1, IE1cc172-176, dl410-420, cc172-176/dl410-420 or luciferase (control) and infected with gTBdlIE1 for six hours (Fig 9). IE1cc172-176 expression resulted in significantly reduced induction of transcripts from the IFNB1, IFNL1, CCL5 and TNF genes relative to the wild-type protein following infection. Conversely, expression of IE1dl410-420 led to a marked increase in infection-related induction of the IFNB1, IFNL1, CCL5, TNF and PML genes. This increase was greatly diminished when PML and STAT binding was jointly abolished (IE1cc172-176/dl410-420).
To investigate these findings further, mutant hCMV strains expressing IE1cc172-176 (TBIE1cc172-176 and gTBIE1cc172-176) were generated and compared with corresponding wild-type and revertant viruses. As expected, IE1cc172-176 expressed from the mutant virus did not co-localize with PML, neither at PML bodies nor at mitotic chromatin (Fig 10A), and both TBIE1cc172-176 and gTBIE1cc172-176 were severely defective for PML body disruption (S5 Fig). Consistent with the results from our trans-complementation assays, the replication kinetics of gTBIE1cc172-176 were slightly delayed at low MOI (0.005 or 0.5 PFU/cell) compared to the wild-type and revertant viruses, but only by a factor of less than three (Fig 10B  and S6 Fig). At a higher MOI (1 PFU/cell), the mutation had no significant effect on viral replication ( Fig 10B). We subsequently compared the kinetics of transcript accumulation related to IFNB1, the major type I IFN expressed in fibroblasts, and OAS1 in cells infected with gTBIE1cc172-176 or a revertant virus for six to 24 hours. Again, IFNB1 and OAS1 mRNA levels were lower in the mutant compared to the revertant virus infection at several time points (Fig 10C).
These results confirm that the interaction between IE1 and PML is dispensable for hCMV replication and demonstrate that efficient inhibition of ISG expression by IE1 does not depend PLOS PATHOGENS PML targeting by hCMV IE1 revisited on this interaction. Instead, the findings suggest that the IE1-PML interaction is linked to enhanced rather than reduced expression of cytokines and ISGs. These observations are consistent with the idea that the disruption of PML bodies by IE1 contributes to the antiviral and proinflammatory response during hCMV infection.

Discussion
PML bodies have fascinated cell biologists and molecular virologists due to their beauty and apparent involvement in many cellular processes including viral infection. Major evidence for a role of PML bodies and their proteins in the cellular defense against viruses has come from the identification and study of viral antagonists. The hCMV IE1 protein is one of the most prominent antagonists of PML bodies and interacts with several constituents of these organelles including isoforms of SUMO, Sp100 and PML. Binding to PML and disruption of PML bodies have been considered pivotal to IE1 function and hCMV replication, at least upon low MOI infection. However, this perception has relied largely on the analysis of mutant proteins and viruses affecting the IE1 core domain. Given the disruptive and destabilizing nature of many published IE1 mutants and the lack of mutants specifically targeting PML interaction, we conducted systematic clustered charge-to-alanine scanning along the viral protein. To our knowledge, this is a novel approach that has not been pursued before to explore IE1 function. Phenotypic screening of the mutant library resulted in the identification of IE1cc172-176, a stable protein selectively inactive for physical and functional interaction with PML. Further investigation of IE1cc172-176 led to several unexpected new findings relevant to PML function, IE1 activity and hCMV replication with general implications for mechanisms of antiviral defense including innate immune activation.
Deletions in the putatively disordered N-terminal domain (amino acids 1-24), encoded by exon 2 of the major immediate gene, did not adversely affect IE1 protein accumulation in a   previous study [46]. Likewise, the clustered charge mutants IE1cc6-8 and IE1cc21-26 produced slightly higher rather than lower protein levels (Fig 2A). Although the N-terminal domain contains a nuclear localization signal (NLS) [46,48], all IE1 mutants resembled the wild-type protein in localizing to the nucleus (Fig 3A and 3B)

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited accumulation except for IE1cc340-342, suggesting that mutations in predicted Helix11 but not Helix10 may be tolerated (Figs 1 and 2A). The low protein levels observed for many of the IE1 core domain mutants may partly relate to defects in promoter autoregulation or mRNA stability (Fig 2B). Some of the mutations, especially those in Helix8 and Helix9, might interfere with homo-dimerization potentially resulting in shortened protein half-lives in case IE1 dimers are more stable than monomers [50]. However, many mutations likely cause disruption to the overall fold as illustrated by the three-dimensional structure of the hCMV IE1 core modelled on the orthologous domain of rhCMV IE1 [49, 106]. The 11 α-helices and other elements of the IE1 core domain are arranged in coiled-coil bundles forming one single structural unit. Consequently, many mutations targeting the core domain will generate misalignment of secondary structure elements disrupting the fold. For example, L174 is located in the hydrophobic core within predicted Helix5. This residue stabilizes the fold between Helix5, Helix9 and Helix10. Mutation to proline is predicted to disrupt the stabilizing interactions and to induce helix redirection explaining the broad functional defects linked to IE1 L174P. A similar situation applies to IE1 L130G/I132G/L133G and Y315G/V316G/L317G. In contrast, the charged residues (amino acids 172-173 and 175-176) replaced in IE1cc172-176 are expected to be on the protein surface and unlikely to participate in interactions between helices (Fig 11)  IE1cc172-176 produces a metabolically stable and broadly active protein (Figs 2-10, S1 Fig  and S4 Fig).
PML-related activities are generally preserved in most C-terminal and some N-terminal IE1 mutants, but not in mutants involving the core domain [48,73,77,103]. Although the IE1-PML interaction has been proposed to involve an extended interface across the core domain [49, 106], our study identifies predicted Helix5 comprising 18 residues (amino acids 164-181, REMWMACIKELHDVSKGA) to be a critical structural element in the IE1-PML interaction. To our knowledge, IE1cc172-176 is the only existing mutant that specifically probes the interaction with PML. This mutant accumulates to normal steady-state levels (Fig  2A), localizes to the nucleus (Figs 3A, 3B, 5, 6B, 7A and 10A and S2 Fig), associates with chromatin (Figs 4, 6B and 10A), undergoes SUMOylation (Figs 5B, 5C, 6 and 7B) and retains the ability to regulate STAT1-, STAT2-and STAT3-dependent gene expression differentially (S1 Fig). However, this mutant does not co-localize with PML, neither at PML bodies during interphase nor at MAPPs and chromatin during mitosis (Figs 3A, 4 and 10A). Moreover, IE1cc172-176 neither specifically binds to PML nor significantly inhibits PML SUMOylation (Fig 5B  and 5C). These observations strongly suggest that IE1cc172-176 lost any affinity for PML and consequently lacks all PML-related functions. This conclusion seems to clash with the fact that IE1cc172-176 is largely but not completely inactive for PML body disruption (Fig 3 and S2  Fig). Rather than implying residual PML binding by IE1cc172-176 below the detection limit of our assays, we propose that IE1 employs at least two mechanisms for disrupting PML bodies. The major mechanism depends on PML binding while some PML body disruption by IE1 does not seem to require this interaction. Whether degradation of Sp100 or more indirect effects linked to IE1 expression result in the disruption of PML bodies in a subset of cells remains to be determined.
Given the key role ascribed to PML in antiviral intrinsic immunity, it comes as a surprise that replication of an IE1cc172-176 mutant is only slightly attenuated relative to the wild-type virus (Figs 8 and 10B and S6 Fig). While the lack of mutant virus attenuation at MOIs �1 may be due to saturation of the PML-based intrinsic repression system [9,107], the minor replication defects at MOIs �0.5 are less readily explained. We propose that this finding reflects redundant mechanisms employed by hCMV in dealing with PML bodies and their restriction factors. In fact, a number of hCMV proteins other than IE1 including IE2, LUNA, pUL3, pUL35, pUL80a, pUL82 (pp71), pUL83 (pp65), pUL97 and pUS32 have been shown to target PML bodies [75,[108][109][110][111][112][113][114][115][116][117]. Having said that, our findings suggest that interactions with STAT family members via the SBM contribute more significantly to IE1 function during productive hCMV infection than PML interaction, especially at higher MOIs. This conclusion aligns with the idea that the IE1-STAT2 interaction promotes hCMV replication by conferring resistance to IFNα and IFNβ [45, 53, 69].
Besides mediating intrinsic immunity, PML has been implicated in activating innate immunity (reviewed in [29, 30]). Certain nuclear PML isoforms are positive regulators of IFN synthesis, and PML may directly promote induction of some ISGs [29, 31-34]. Moreover, it has been proposed that IE1 inhibits ISG activation in part by interacting with PML [34, 77]. We therefore expected that IE1cc172-176 would be deficient in inhibiting ISG expression compared to wild-type IE1. However, we observed lower instead of higher levels of IFN, TNF and ISG expression with the mutant relative to the wild-type protein upon hCMV infection (Figs 9  and 10). This finding is consistent with the idea that disruption of PML bodies is linked to activation rather than inhibition of antiviral gene expression. Activation of antiviral genes following disruption of PML bodies might be mediated by PML isoforms redistributed from the organelles into the nucleoplasm. Alternatively, viral genomes released from repression imposed by PML bodies and physically liberated from these organelles might trigger the activation of nuclear DNA sensors such as IFN-inducible protein 16 (IFI16) or cyclic GMP-AMP synthase (cGAS) to induce an antiviral response. However, these are merely speculations at this point that need to be experimentally tested. Nonetheless, we propose that disruption of PML bodies may link relief from intrinsic immunity to induction of innate immunity.
IE1 has been identified in physical complexes with not only PML [73] but with at least four other constituent proteins of PML bodies including Daxx [80], Sp100A [79], SUMO1 [56] and SUMO3 [58]. Our results demonstrate that IE1 targets Sp100 irrespective of PML binding (Fig  7). Thus, proteasomal degradation of Sp100 is unlikely a consequence of PML body disruption but rather presents a distinct IE1 activity that deserves further attention. Several groups have reported that a proportion of IE1 is modified at residue K450 by single moieties of SUMO1 or SUMO3, two of five known SUMO paralogs [53-58, 83, 94, 100]. We added SUMO2 to this list (Figs 5C and 6A). SUMO2 and SUMO3 share around 95% sequence identity and are often referred to jointly as SUMO2/3. SUMO2/3 forms poly-SUMO chains since residue K11 confers conjugation to itself or other SUMO paralogs. SUMO1, which exhibits only around 50% sequence identity with SUMO2/3, lacks K11 and is therefore conjugated only as a monomer or as a terminator at the end of a poly-SUMO chain [16,118]. Based on our immunoblotting results (Figs 5C and 6A), we propose that mixed polymeric chains of SUMO2/3 terminated by SUMO1 form at residue K450 of IE1cc172-176. Previous work has shown that SUMOylation of IE1 interferes with binding to STAT2 [53]. It is tempting to speculate about differential effects mono-and poly-SUMOylation may have on IE1 protein interactions beyond STAT2. Although we cannot rule out that poly-SUMOylation is a feature specific to IE1cc172-176, the chains likely exist on the wild-type protein as well but exceed the limit of detection only in the hyper-SUMOylated mutant. Hyper-SUMOylation may result from a difference in subnuclear localization between IE1cc172-176 and the wild-type protein. IE1 usually localizes across several nuclear compartments including nucleoplasm, matrix and chromatin (Fig 6B). A dynamic equilibrium likely exists between the nuclear locations of IE1, which may in part be determined by IE1-PML complex formation in the nucleoplasm and nucleosome binding at chromatin. In this scenario, chromatin association by IE1 would increase on disruption of PML binding. This prediction is consistent with the results from our subnuclear fractionation analysis demonstrating decreased nucleoplasmic and increased chromatin-associated localization of IE1cc172-176 compared to wild-type IE1 (Fig 6B). In turn, increased chromatin association correlates with enhanced SUMOylation and detection of SUMO chains on IE1cc172-176. Conversely, IE1 exhibiting lack of nucleosome binding (IE1dl476-491) and decreased chromatin association is not detectably SUMO-modified (Figs 5C and 6 [119]. SUMOylation by PIAS1 reduces the binding affinity of HMGN2 to nucleosome core particles [119]. Given that both HMGN2 and IE1 target histones H2A-H2B on the nucleosome surface, we are tempted to speculate that PIAS1-mediated SUMOylation might interfere with chromatin association by the viral protein. In fact, our observation that SUMOylated IE1 forms at nucleosomes but localizes throughout the nucleus (Fig 6B) implies a mechanism releasing the viral protein from chromatin.
The findings presented in this study challenge predominant views about IE1, a pivotal viral protein with a crucial function in hCMV replication and pathogenesis. Several of our conclusions should also be more broadly relevant to cellular events beyond hCMV infection given the versatile roles of PML and PML bodies in health and disease (reviewed in [120,121]).

PLOS PATHOGENS
PML targeting by hCMV IE1 revisited Where indicated, cells were treated with 1,000 U/ml recombinant human IFNα A/D (R&D Systems, 11200) for 24 h.

hCMV mutagenesis and infection
Wild-type virus of the low passage hCMV strain TB40/E (TBwt) was derived from the bacterial artificial chromosome (BAC) TB40-BAC4 [128]. A modified version of this BAC with an SV40-EGFP-BGH PolyA cassette inserted between the US34 and TRS1 genes [129] was used to generate the EGFP expressing TB40/E wild-type virus gTBwt. Mutant BACs encoding no IE1 (gTBdlIE1) or IE1 with amino acids 172, 173, 175 and 176 changed to alanine (pTBIE1cc172-176 and pgTBIE1cc172-176) and the revertant BAC pgTBrvIE1cc172-176 were generated by markerless en passant mutagenesis as described [130] using plasmid pLAY2 and oligonucleotide primers listed in Table 1. The identity and integrity of each BAC were verified by Sanger sequencing of the modified region and restriction fragment length analysis following digestion with EcoRI. Viruses were reconstituted and virus stocks produced upon electroporation of MRC-5 cells with BAC DNA following standard protocols. Viruses (g)TBwt, (g) TBIE1cc172-176 and gTBrvIE1cc172-176 were grown on normal MRC-5 fibroblasts while gTBdlIE1 was produced on MRC-5 cells with constitutive expression of wild-type IE1 following transduction with pLKO.DCMV.TetO.cIE1-derived lentivirus [72].
Titers of wild-type and revertant virus preparations were determined by plaque assay on MRC-5 cells in 6-well plates. For plaque assays, infections were performed in triplicates for 16 h with 800 μl inoculum before cells were overlayed with 4 ml Dulbecco's Modified Eagle's Medium supplemented with 3.7 g/l sodium hydrogen carbonate, 1% [w/v] methyl cellulose (Sigma-Aldrich, M0262), 2% [v/v] fetal calf serum, 100 units/ml penicillin and 100 μg/ml streptomycin. Plaques of EGFP-positive viruses were counted at day 9 post infection using an EVOS FL Cell Imaging System (Thermo Fisher) with Plan Achromat 4× Objective and EVOS Light Cube for GFP. Foci of three or more fluorescent cells were counted as plaques. Plaques formed by EGFP-negative viruses were counted at day 12 post infection in phase contrast mode after staining of cell monolayers with 0.5% [w/v] methylene blue in 70% [v/v] methanol. For quantification of infectious viral genome equivalents, 5 × 10 5 MRC-5 cells were seeded onto 6-well dishes and infected 40 h later in triplicates with 800 μl 1:10-diluted virus stock in growth medium supplemented with 50 U/ml benzonase nuclease (Sigma-Aldrich, E1014) to remove free viral DNA. Plates were incubated for 1 h at 37˚C with occasional rocking before the virus inoculum was removed and cells were washed in 5 ml phosphate-buffered saline (PBS) with 10 mM ethylenediaminetetraacetic acid (EDTA). Cells were dislodged and extracellular virus was removed by treatment with 500 μl trypsin (0.5 g/l) and EDTA (0.5 mM) solution for 5 min at 37˚C. Cells were transferred to 2-ml tubes using 1.3 ml PBS/EDTA and collected by centrifugation for 5 min at 800 g. Cell pellets were subjected to an additional washing step with 1.8 ml PBS/EDTA before the supernatant was removed and cells were resuspended in 200 μl PBS/EDTA and transferred to 1.5-ml tubes for DNA isolation and qPCR analysis.
To monitor replication of gTBdlIE1 on trans-complementing cell lines, confluent MRC-5 cultures on 12-well dishes were infected at the indicated MOI by applying 300 μl virus dilution supplemented with dox. At 16 h (day 1) after infection, the inoculum was removed, cells were washed twice with 2 ml growth medium and further incubated in 1 ml growth medium with dox. On day 2 post infection and every other day thereafter, half of the growth medium was replaced and viral replication was assessed by qPCR-based relative quantification of hCMV DNA from culture supernatants. To compare replication of wild-type, mutant and revertant gTB viruses, 2.5 × 10 4 MRC-5 cells were seeded in a volume of 100 μl onto 96-well plates and infected 24 h later at the indicated MOI by adding 100 μl virus dilution. EGFP fluorescence was measured in a Tecan Infinite M200 PRO microplate reader immediately after virus addition and at day 4, 8, 12 and 16 post infection. The following instrument settings were used: measurement mode, fluorescence intensity bottom; excitation wavelength, 483 nm; excitation bandwidth, 9 nm; emission wavelength, 535 nm; emission bandwidth, 20 nm; gain, 80; integration time, 20 μs; flashes, 4 × 7. The signal at day 0 was considered background and was subtracted from all other measurements. For RNA analyses, 2.5 × 10 5 MRC-5 cells were seeded onto 12-well dishes and infected 72 h later at the indicated MOI by applying 300 μl virus dilution. After 2 h, 700 μl pre-warmed growth medium was added and the cells were harvested at the indicated times post infection. For immunocytochemistry, 5 × 10 5 MRC-5 cells were seeded in 6-well dishes and infected 16 h later at the indicated MOI by applying 800 μl virus dilution. After 2 h, 2 ml 37˚C growth medium was added and cells were fixed at the indicated times post infection.

DNA and RNA analyses
To determine steady-state mRNA levels by RT-qPCR, total RNA was isolated from fibroblasts cultured in 12-well dishes using the RNeasy Mini Plus Kit according to Qiagen's instructions. First-strand cDNA was synthesized at 50˚C using the AffinityScript Multiple Temperature cDNA Synthesis Kit and oligo(dT) primers (Agilent Technologies). First-strand cDNA was diluted 10-fold in sterile ultrapure water, and 5 μl was used for real-time PCR as described in detail in previous publications [52,124]. The sequences of oligonucleotide primers used for RT-qPCR are shown in Table 1.
For viral genome quantification, total DNA was extracted from cells grown in 6-well dishes or from 200 μl pre-spun (4000 g, 10 min) culture supernatant using a DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer's spin-column protocol for animal blood and cells. DNA was released from the column in two consecutive steps with 100 μl (supernatant samples) or 200 μl (cell samples) elution buffer. DNA from infected cells was diluted 10-fold, and 5-μl samples were subjected to real-time PCR with primers shown in Table 1.

Protein analyses
For indirect immunofluorescence microscopy, cells were grown on precision cover glasses (Marienfeld Superior, No. 1.5H), fixed with pre-chilled methanol for 20 min at -20˚C and processed as described [52,72].
Whole cell lysates were prepared in 10 mM PIPES-NaOH, pH 7. Insoluble material was removed by centrifugation at 20,000 × g, 4˚C for 10 min, and equal sample volumes were used for SDS polyacrylamide gel electrophoresis and immunoblotting as described [45,52,55,124].
For co-immunoprecipitation experiments, 5 x 10 6 293T cells were seeded onto 10-cm dishes and transfected 24 h later with 10 μg pCMV-PML and 10 μg pLKO.DCMV.TetO IE1 expression plasmid or empty vector using the calcium-phosphate co-precipitation technique [131]. At 48 h post transfection, cells were cross-linked by treatment with 1% [v/v] formaldehyde for 10 min at room temperature. To stop cross-linking, glycine was added to a final concentration of 125 mM, and samples were incubated for another 5 min at room temperature. Cells were washed and harvested with 10 ml ice-cold PBS and lysed in 500 μl 10 mM PIPES-NaOH, pH 7. For subnuclear fractionation, approximately 5×10 6 MRC-5 cells were gently resuspended in 200 μl freshly prepared CSK buffer (10 mM PIPES-NaOH, pH 6.8, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 1% [v/v] protease inhibitor cocktail, 20 mM IAA, 20 mM NEM) with 0.1% [v/v] Igepal CA-630. Nuclei were spun down at 1,300 × g, 4˚C for 1 min, and the supernatant was collected as cytoplasm fraction. The pellet was washed with 300 μl detergent-free CSK buffer before the nucleoplasm was isolated with 200 μl CSK buffer containing 0.5% [v/v] Triton X-100. The nuclear pellet remaining after another 1-min spinning step at 1,300 × g, 4˚C was washed with 300 μl 50 mM Tris-HCl, pH 7.5, 5 mM CaCl 2 and resuspended in 200 μl of the same buffer containing 100 units micrococcal nuclease (Thermo Fisher, 88216) to solubilize chromatin. Samples were sonicated for 10 sec in a Bioruptor Pico to improve the accessibility of chromatin for the enzyme before tubes were incubated for 10 min at 30˚C with shaking. After that, samples were supplemented with 0.5% [v/v] Triton X-100 and sonicated again for 3 min (30 sec ON/30 sec OFF). The soluble chromatin fraction was removed after spinning at 10,000 × g, 4˚C for 5 min, and the insoluble pellet material was washed once with 200 μl CSK buffer containing 0.5% [v/v] Triton X-100 before it was resuspended as matrix fraction in 200 μl 2× protein sample buffer.

Protein modeling and structure visualization
The Normal Modelling Mode provided with the Protein Homology/analogY Recognition Engine 2.0 (Phyre2) [132] was used to model the three-dimensional structure of the hCMV PLOS PATHOGENS PML targeting by hCMV IE1 revisited (Towne) IE1 protein (GenBank AAR31448). 70% of the sequence, corresponding to the IE1 core domain (amino acids 25 to 378), were modelled with 100% confidence by RhCMV IE1/ rhUL123 (PDB 4WID) [49], the single highest scoring template. Protein structures were rendered and annotated using UCSF Chimera 1.13.1 (Resource for Biocomputing, Visualization, and Informatics) [133]. Relative mRNA levels were determined by RT-qPCR for IE1 and typical STAT1-and STAT3-responsive genes (CXCL10 and SOCS3, respectively) and normalized to TUBB. Data presented are means and standard deviations of two biological and two technical replicates. (B) Cells were treated with IFNα for 24 h. Relative mRNA levels were determined by RT-qPCR for IE1 and a typical STA-T2-responsive gene (OAS1) and normalized to TUBB. Data presented are means and standard deviations of two biological and two technical replicates. (C) Indirect immunofluorescence staining was performed using mouse anti-IE1 and rabbit anti-PML combined with goat antimouse Alexa Fluor 594 and goat anti-rabbit Alexa Fluor 488 antibodies. DAPI was used to stain DNA. Individual and merge images were taken using a Keyence BZ-9000 microscope (40× objective). (EPS) S4 Fig. Metabolic stability of wild-type and mutant IE1 proteins. TetOne-IE1 cells with tightly controlled inducible expression of the indicated HA-tagged wild-type (wt) or mutant IE1 proteins were treated with dox (1 μg/ml) for 12 h. Cells were then either collected (0 h) or washed three times with prewarmed growth medium and incubated for another 12,24,36,48 or 60 h in the absence of dox. Whole cell protein extracts were prepared and analyzed by quantitative immunoblotting using rat anti-HA and mouse anti-TUBA combined with goat anti-rat IRDye 800CW (green) and goat anti-mouse IRDye 680RD (red) antibodies. (EPS)

S5 Fig. Disruption of PML bodies in cells infected with IE1cc172-176 mutant hCMV. (A)
MRC-5 cells were infected with gTBdlIE1, gTBwt, gTBIE1cc172-176 or gTBIE1rv172-176 at an MOI of 2 PFUs/cell for 16 h. Indirect immunofluorescence staining was performed using mouse anti-IE1 and rabbit anti-PML combined with goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 antibodies. DAPI was used to stain DNA. Individual and merge images were taken using a Keyence BZ-9000 microscope (40× objective). (B) MRC-5 cells were infected with TBdlIE1, TBwt or TBIE1cc172-176 at an MOI of 2 PFUs/cell for 16 h. Indirect immunofluorescence staining was performed using mouse anti-IE1 and rabbit anti-PML combined with goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 antibodies. DAPI was used to stain DNA. Images were taken using a Keyence BZ-9000 microscope (40× objective). (C) The percentage of nuclei exhibiting predominantly disrupted or undisrupted PML bodies was determined for at least 100 cells using images acquired as described in (