Cytomegalovirus immediate-early 1 proteins form a structurally distinct protein class with adaptations determining cross-species barriers

Restriction factors are potent antiviral proteins that constitute a first line of intracellular defense by blocking viral replication and spread. During co-evolution, however, viruses have developed antagonistic proteins to modulate or degrade the restriction factors of their host. To ensure the success of lytic replication, the herpesvirus human cytomegalovirus (HCMV) expresses the immediate-early protein IE1, which acts as an antagonist of antiviral, subnuclear structures termed PML nuclear bodies (PML-NBs). IE1 interacts directly with PML, the key protein of PML-NBs, through its core domain and disrupts the dot-like multiprotein complexes thereby abrogating the antiviral effects. Here we present the crystal structures of the human and rat cytomegalovirus core domain (IE1CORE). We found that IE1CORE domains, also including the previously characterized IE1CORE of rhesus CMV, form a distinct class of proteins that are characterized by a highly similar and unique tertiary fold and quaternary assembly. This contrasts to a marked amino acid sequence diversity suggesting that strong positive selection evolved a conserved fold, while immune selection pressure may have fostered sequence divergence of IE1. At the same time, we detected specific differences in the helix arrangements of primate versus rodent IE1CORE structures. Functional characterization revealed a conserved mechanism of PML-NB disruption, however, primate and rodent IE1 proteins were only effective in cells of the natural host species but not during cross-species infection. Remarkably, we observed that expression of HCMV IE1 allows rat cytomegalovirus replication in human cells. We conclude that cytomegaloviruses have evolved a distinct protein tertiary structure of IE1 to effectively bind and inactivate an important cellular restriction factor. Furthermore, our data show that the IE1 fold has been adapted to maximize the efficacy of PML targeting in a species-specific manner and support the concept that the PML-NBs-based intrinsic defense constitutes a barrier to cross-species transmission of HCMV.


Introduction
To combat viral infections, host organisms have developed an intricate defense network comprising the intrinsic, innate, and adaptive immune response. While innate and adaptive defense mechanisms rely on pathogen-induced activation, the intrinsic immune system is conferred by constitutively expressed restriction factors thus mediating a front-line defense against invading pathogens [1]. Since the discovery of the first class of restriction factors targeting retroviral capsids, numerous cellular factors have been identified that restrict diverse steps in the life cycle of viruses [2]. During the evolutionary "arms race", however, viruses have evolved means to evade or directly counteract these antiviral host factors, mainly by expressing antagonistic proteins. The evolutionary pressure that restriction factors and antagonists have exerted on each other resulted in further adaptations at the virus-host interface. Thus, restriction factors are often less effective against viral infections of their natural host but constitute potent barriers to cross-species infections [3].
Specific structures within the cell nucleus termed PML nuclear bodies (PML-NBs) or nuclear domain 10 (ND10) have been shown to play a major role in the intrinsic defense against a variety of viruses, including members of the highly host-adapted herpesvirus family [4]. PML-NBs are dynamic multiprotein complexes that accumulate in distinct foci within the interchromosomal space and have been implicated in cellular key processes such as cell cycle progression, apoptosis, senescence, DNA damage and antiviral responses [5]. PML, the signature protein of PML-NBs, belongs to the immunomodulatory tripartite motif (TRIM) protein family, whose members share an N-terminal domain structure comprising a RING domain, one or two B-Boxes, and a coiled-coil (CC) domain (often subsumed under the term RBCC domain) [6]. Within the N-terminal region, PML additionally harbors target sites for covalent modification with small ubiquitin-like modifier (SUMO) proteins, which enables the interaction with further protein components and, therefore, is essential for PML-NB biogenesis [7,8]. Upon herpesvirus infection, PML-NBs associate with viral genomes as soon as they have entered the nucleus [9,10]. This association blocks viral infection at a very early step, since PML-NB proteins rapidly promote the condensation of herpesviral DNA into transcriptionally inactive heterochromatin [11]. Besides PML, several other PML-NB components including Sp100, hDaxx, ATRX, and MORC3 function as restriction factors and contribute to the repression of viral gene expression in a cooperative manner [12][13][14][15][16].
In order to overcome the PML-NB-based defense, herpesviruses encode antagonistic effector proteins, which employ different strategies to either inactivate single PML-NB components or to disrupt the integrity of the whole structure. The herpes simplex virus type I immediateearly protein ICP0, for instance, disarms PML-NBs in a rapid and efficient way by inducing a widespread proteasomal degradation of SUMO-modified proteins including PML-NB components [17]. In contrast, immediate-early protein IE1 of human cytomegalovirus (HCMV), a ubiquitous β-herpesvirus causing serious disease in immunocompromised individuals, uses a more careful strategy, likely due to the prolonged replication cycle of HCMV. IE1 directly interacts with PML and blocks its SUMOylation in a proteasome-independent manner [18,19]. Since SUMO modification of PML is essential for PML-NB integrity, this results in a dispersal and inactivation of PML-NB foci. Structural characterization of IE1 has shown that it comprises a folded core domain (IE1 CORE ), which mediates the interaction with PML and is flanked by a short disordered region at the N-terminus and a longer disordered region at the C-terminus containing a SUMOylation motif and a STAT interaction site [20][21][22]. Crystallization of the IE1 CORE domain of rhesus cytomegalovirus (RhCMV), as described in a previous publication of our groups, revealed a so-far unobserved femur-like all-α-helical fold with local similarity to the conserved coiled-coil domain of TRIM proteins [22]. Since IE1 CORE efficiently binds to the PML (TRIM19) coiled-coil domain, we proposed that IE1 sequesters PML via structural mimicry using an extended binding surface.
In this study, we present the experimentally determined crystal structures of human and rat cytomegalovirus (RCMV) IE1 CORE . All crystallized IE1 CORE domains share a highly similar, all-α-helical fold. Since we observed that the mechanism of PML-NB disruption is likewise conserved between primate and rodent IE1 proteins, we conclude that cytomegaloviruses have evolved this distinct protein fold to effectively bind and inactivate an important antiviral defense. Closer investigation of the crystal structures revealed slight differences in the helix arrangement of rat compared to primate cytomegalovirus IE1. This correlates with a comparative functional analysis of human and rat cytomegalovirus IE1 showing that neutralization of PML-NBs occurs only in cells of the natural host species but not during cross-species infection. For RCMV, this block of cross-species infection can be alleviated by expression of human IE1 in human host cells. In summary, our data provide evidence that the IE1 fold has been adapted to maximize the efficiency of PML-NB targeting and strengthen the concept that the PML-NBs-based intrinsic defense constitutes a barrier to cross-species transmission of HCMV.

The domain organization of IE1 is conserved across primates and rodents
The architecture of cytomegalovirus IE1 proteins appears to be evolutionary conserved across species. An in silico disorder prediction of the rodent member rat cytomegalovirus IE1 protein (ratIE1) is in agreement with the presence of a folded core domain that is flanked by a short partially or fully disordered N-terminal segment as well as a disordered extended C-terminal segment as previously observed in the primate IE1 proteins from human (humIE1) and rhesus (rhesIE1) cytomegalovirus ( Fig 1A) [22].
An experimental validation of the in silico prediction via a limited proteolysis digestion of full-length recombinant ratIE1 (residues 1 to 565) yielded a single and stable 45 kDa fragment ( Fig 1B). A mass spectrometry analysis of this fragment revealed that it extends from residues 1 to 392 of the ratIE1 sequence (S1 Fig). It includes the very N-terminal residues of ratIE1, which in case of the homologous rhesIE1 and humIE1 proteins, were prone to digestion in previous experiments and therefore postulated as not being part of the core domains in these proteins (humIE1 CORE , residues 14 to 382; rhesIE1 CORE , residues 36 to 395) [22]. Interestingly, the in silico disorder analysis of ratIE1 anticipated this result since the predicted disorder tendency for the first 16 residues is considerably reduced in ratIE1 versus hum-and rhesIE1 (Fig 1A).
One humIE1 and two ratIE1 variants were produced for further characterization. The two ratIE1 variants, covering residues 1 to 392 and residues 30 to 392, show an almost identical allα-helical secondary structure composition as analyzed by CD spectroscopy (Fig 1C). Moreover, the CD spectra of the two ratIE1 variants are almost indistinguishable from that of humIE1 CORE in agreement with the assumption of a shared core domain in IE1 proteins ( Fig  1C). Of the two ratIE1 variants, only the N-terminally truncated variant yielded protein crystals. This variant, covering residues 30 to 392, is from here-on referred to as the ratIE1 CORE domain. Taken together, ratIE1, humIE1 and rhesIE1 share an approximately 350-residue, allα-helical core domain that is flanked by a short, fully or partially disordered region at the Nterminus and a 110-to 170-residue-long disordered region at the C-terminus.

Crystal structures of ratIE1 CORE and humIE1 CORE
The structure of ratIE1 CORE was solved to 3.4 Å resolution with R work = 21.5% and R free = 24.6% (Table 1). Initial phases were obtained with the MAD technique since any molecular replacement calculations with the previously determined rhesIE1 CORE structure failed [22]. The main chain of ratIE1 CORE could be built from residues 33 to 392, and only three N-terminal residues could not be located in the electron density (Fig 2A). Because of the low resolution of the crystallographic analysis, the correctness of the sequence registration was corroborated by additional experiments. Firstly, we calculated an anomalous difference map with phases derived from the refined structure and amplitudes from the seleno-methionine peak diffraction data set (Table 1). A close inspection of this difference map showed that all eleven peaks with densities above 4.6 σ can be explained by the 12 selenium atoms present in seleno-methionine-substituted ratIE1 CORE (S1 Table). No density peaks above 3.7 σ remain unaccounted for. Secondly, we recorded a long-wavelength 6 keV X-ray diffraction data set from nonsubstituted ratIE1 CORE crystals in order to maximize the anomalous sulfur signal (Table 1). An In silico disorder prediction analysis of human (hum), rhesus (rhes) and rat (rat) cytomegalovirus IE1 sequences using IUPred2A [69]. The disorder score for all three proteins suggest a globular domain with disordered N-and C-termini (scores � 0.5 indicate disorder). (B) Limited proteolysis of recombinant ratIE1. Purified ratIE1 was incubated with subtilisin (1 mU protease per mg ratIE1) for different times, and samples were analyzed by SDS-PAGE and Coomassie blue staining. (C) CD spectroscopy of humIE1 14-382, ratIE1 1-392 and ratIE1 30-392. The spectra were normalized at 207 nm as suggested by Raussens and coworkers [57]. https://doi.org/10.1371/journal.ppat.1009863.g001

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs anomalous difference map calculated with these data showed clear peaks for the sulfur atoms of all 7 cysteine and 10 out of 12 methionine residues present in ratIE1 CORE (S2 Table). Albeit no clear peaks were observed at the sulfur position of methionines 83 and 391 in this map, the positions of the atoms were clearly visible in the analysis of the previous seleno-methionine peak diffraction data set. The crystal structure of humIE1 CORE was solved to 3.2 Å resolution with R work = 22.5%, R free = 26.8% (Table 1). Initial phases were obtained by molecular replacement with program MR-Rosetta and using rhesIE1 CORE as a search model [23]. In the final model, the humIE1 CORE protein chain could be traced from residues 25 to 356 and 363 to 382 ( Fig 2B).

RatIE1, humIE1 and rhesIE1 share a unique fold
RatIE1 CORE and humIE1 CORE share a highly similar overall fold, which bears close resemblance to that of the previously determined rhesIE1 CORE structure (Figs 2 and S2) [22]. All three IE1 CORE proteins display a femur-like structure consisting of α-helices only. These are arranged into two head regions interconnected by a stalk region composed of three to four long α-helices. HumIE1 CORE resembles rhesIE1 CORE more closely than ratIE1 CORE . HumIE1-CORE can be superimposed onto rhesIE1 CORE with an rmsd Cα value of 2.3 Å, while the

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs structures of humIE1 CORE and ratIE1 CORE differ by an rmsd Cα value of as high as 4.6 Å ( Table 2). The structural deviations between these proteins are paralleled by marked differences in sequence identities. While humIE1 CORE and rhesIE1 CORE can be aligned with 24% sequence identity, the sequence identity between humIE1 CORE and ratIE1 CORE amounts to only 22% (Table 2).
It is known from comparative structural biology that the lower the sequence identities between the proteins, the more dissimilar the structures of the respective proteins are and vice versa [24]. However, sequence identities as low as 24 or 22% fall below the cut-off value of 28% that has been derived as a lower limit for safely inferring structural details and overall similarities from sequence identities in proteins of more than 200 residues in length [25,26]. As a consequence of the low sequence identities, humIE1 CORE , rhesIE1 CORE and ratIE1 CORE exhibit marked differences (Figs 2 and S2). Thus, helix H1 is significantly shorter in ratIE1 CORE , and the position of H2 is rotated by approximately 90˚in comparison to humIE1 CORE and rhesIE1-CORE . Furthermore, a kink separates helices H5 and H6 in humIE1 CORE and rhesIE1 CORE , whereas ratIE1 CORE contains one continuous helix termed H5/6. Besides this, the curvature of several helices, namely H3, H6 and H9, also slightly differs between humIE1 CORE and rhesIE1-CORE on one hand and ratIE1 CORE on the other hand. At the same time, ratIE1 CORE has an additional helix H12 at the C-terminus in comparison to humIE1 CORE and rhesIE1 CORE , which consist of eleven helices in total.
A DALI search against the entire protein data bank (PDB, performed in February 2021) unambiguously identifies these three proteins as forming a unique structure family (S3 Table) [27,28]. Additional candidate homologous proteins, as identified by DALI, either display excessively high rmsd Cα values exceeding 8 Å when aligning up to 240 residues or the structural homology is limited to considerably smaller segments of about 100 residues in the compared proteins so that rmsd Cα values of about 3 Å upwards are obtained (S3 Table). This shows that clear structural homology extending over the entire length of the compared protein structures is only detectable within the group of IE1 CORE proteins, but not to any other protein of known structure.

CMV IE1 proteins display an identical dimerization mode
All IE1 CORE proteins not only display a similar and unique overall fold but also form highly similar dimeric assemblies. In the ratIE1 CORE and humIE1 CORE crystals, the crystallographic

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs asymmetric units contain a single protein chain. However, in both cases, inspection of the crystal packing interactions reveals the presence of tightly interacting dimers (Fig 3A and 3B). In these dimers, the two protomers are related by crystallographic two-fold symmetry axes ( Fig 3C) and hence, the dimers display C 2 point group symmetry similarly to previously described rhesIE1 CORE [22]. In all IE1 CORE structures, the monomers dimerize via an identical interface, and highly similar crossing-angles are adopted between monomers ( Fig 3C). The cross-species conserved quaternary arrangement is also evident when comparing the superposition of dimers with the superposition of monomers. When superimposing the various dimers, the calculated rmsd Cα values are only marginally higher than the deviations obtained between monomers in support of a conserved quaternary assembly in IE1 CORE proteins ( Table 2). Analysis of all IE1 CORE structures with program EPPIC suggests that the dimeric assembly corresponds to the biologically active unit of IE1 CORE [29]. All remaining protein interfaces observed in the various crystals are classified as mere crystal packing contacts. The sizes of the dimer interfaces are also comparable between IE1 CORE proteins ranging from 2240 to 2430 and 2470 Å 2 in ratIE1 CORE , humIE1 CORE and rhesIE1 CORE (PDB entry 4WIC), respectively. Interestingly, an interface of 2470 Å 2 is only observed in crystals of rhesIE1 CORE before induction of a crystallographic

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs phase transition [30]. A dehydration of rhesIE1 CORE crystals induces a distinct conformational rearrangement in one segment of one protomer of rhesIE1 CORE , and a more extensive dimer interface of about 3070 Å 2 is formed [22,30].
Taken together, the IE1 proteins appear to form a distinct class of proteins characterized by a shared unique tertiary fold and quaternary assembly. At the same time, the sequence identities observed between these proteins map these to the so-called "twilight zone", where inference of structural details from sequence alignments only has to be cautioned [25,26].

The canonical IE1 CORE fold is built from conserved regions of left-and right-handed coiled-coils
The all-α-helical fold of IE1 CORE consists of specific left-and right-handed helix pairings that originate from distinct hydrophobic repeat motifs. The N-terminal head region of ratIE1 CORE , humIE1 CORE and rhesIE1 CORE is formed by helices H3, H7 and H8, and these helices form lefthanded coiled-coils (Fig 4). The sequences of these helices mainly contain heptad repeats. In these 'abcdefg' repeats, hydrophobic residues are displayed at positions a and d and give rise to left-handed helix crossings (

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs whereas the stalk and the adjacent C-terminal head region of ratIE1 CORE are formed by continuous right-handed coiled-coils, this segment is interrupted by a region of left-handed coiled-coils in rhesIE1 CORE and humIE1 CORE (Figs 4 and S3). At the stalk-head transition (H3/ H4), rhesIE1 CORE and humIE1 CORE display an insertion of two hydrophilic residues, which point towards the solvent and locally distort the helix geometry to form a sharp kink. In contrast, H3 and H4 of ratIE1 CORE are separated by a short unstructured linker (Fig 4). The primate CMV proteins further lack one heptad repeat in the middle of H6. Overall, the three IE1 CORE structures show similar helix-pairing arrangements. At the same time, specific differences exist in the hydrophobic repeat patterns between the primate and the rodent IE1 CORE structures.
These repeat patterns appear more conserved than individual amino acids. Program PRO-MALS3D was used to generate a structure-based multiple sequence alignment of the three IE1 CORE proteins (S3 Fig) [32]. Surprisingly, at nine positions only, amino acid types are conserved across all three proteins. Derivation of pairwise sequence identities from the structurebased alignment reveals that in case of rhesIE1 CORE and humIE1 CORE , the observed sequence identity matches that obtained with standard sequence alignment algorithms (22 versus 24%, respectively, Table 2). However, when comparing the structure-derived sequence identity between ratIE1 CORE and either humIE1 CORE or rhesIE1 CORE , sequence identities as low as 9% are obtained for both comparisons. These are considerably lower than the 22% sequence identities obtained with standard sequence alignment algorithms.
Knowledge of the distribution of hydrophobic repeat motifs could help to more reliably model additional IE1-homologous proteins since these distributions are responsible for the topological arrangement of the α-helices in IE1 CORE . To test this, the sequence of the structurally uncharacterized mouse CMV IE1 (murIE1) protein was manually incorporated into the structure-based sequence alignment of rhesIE1, humIE1 and ratIE1 (S3 Fig). The alignment shows that the regions can be readily identified and that these show high similarity to those of the crystallized IE1 proteins. We propose that these conserved repeat patterns can be used to improve the reliability of sequence alignments and the correctness of homology models, in particular in cases, where sequence identities fall within the "twilight zone".

The mechanism of PML-NB disruption is conserved among primate and rodent cytomegaloviruses
Due to the structural conservation of the IE1 core domain, the question arose whether all IE1 homologs use the same molecular mechanism to disrupt the antiviral PML-NBs. To address this issue, several rat PML (ratPML) deletion mutants were generated and analyzed for an interaction with ratIE1 in HEK293T cells ( Fig 5A). Co-immunoprecipitation experiments revealed that ratIE1 binds full-length ratPML and, even more efficiently, the truncated ratPML RBCC protein (Fig 5B, lane 3 and 6). A construct encoding an N-terminally extended ratIE1-CORE protein (ratIE1 1-392) was sufficient for this interaction (Fig 5B, lane 2 and 5), which is in accordance with our previous data on human PML (humPML) and humIE1 [22]. Please note that for all cell-based assays, this N-terminally extended IE1 CORE variant was used since the N-terminus has been proposed to harbor the NLS signal [33]. Deletion of the coiled-coil domain from the ratPML RBCC protein (ratPML RB) abolished the interaction with ratIE1-CORE suggesting that ratPML-NBs are targeted through coiled-coil interactions ( Fig 5C, lane  3). Proper folding of the ratPML RB fragment was confirmed by CD spectroscopy, which also revealed a shared secondary structure composition with the corresponding humPML RB construct (S4 Fig). Moreover, we found that ratPML constructs lacking the RING domain (ratPML BCC and ratPML ΔR) are also not able to bind ratIE1 CORE (Fig 5C, lane 4

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs (B) Efficient interaction of ratIE1 CORE with ratPML in co-immunoprecipitation analysis. HEK293T cells were co-transfected with expression plasmids encoding FLAGtagged ratIE1 or ratIE1core (residues 1-392) and Myc-tagged ratPML variants. After cell lysis, immunoprecipitation was performed with an anti-FLAG antibody. Coprecipitated ratPML proteins (IP), precipitated ratIE1 proteins, and proteins within the cell lysate (input) were analyzed by Western blotting as indicated. (C) Binding of ratIE1 CORE to ratPML requires both the coiled-coil and the RING domain. HEK293T cells were co-transfected with expression plasmids encoding FLAG-tagged ratPML variants and Myc-tagged ratIE1 CORE (residues 1-382) as indicated. Upper two panels: Western blot detection of ratIE1 and ratPML after immunoprecipitation using an anti-FLAG antibody. Lower two panels: detection of ratIE1 and ratPML in cell lysates before precipitation (input). (D) Inhibition of ratPML SUMOylation by ratIE1 expression. HEK293T cells were transfected with expression plasmids encoding Myc-ratPML, HA-SUMO2 and FLAG-ratIE1 as indicated. After cell harvest, ratPML and SUMOylated ratPML were visualized by Western blotting using anti-Myc and anti-HA antibodies, respectively. Expression of IE1 was analyzed with an anti-FLAG antibody and β-actin was included as internal control. (E) Impact of RCMV infection on ratPML SUMOylation. Rat embryonic fibroblast (REF) cells were infected with PLOS PATHOGENS CMV co-speciation evolved a distinct IE1 structure to target PML-NBs However, comparatively low expression levels of such constructs in lysate and precipitation samples hint to a possible requirement of the RING domain for proper folding and solubility of ratPML.
Next, we examined whether ratIE1 induces a loss of ratPML SUMOylation and disruption of ratPML foci. Transfection experiments using HEK293T cells showed that ratIE1 is sufficient to induce a loss of SUMOylated ratPML (Fig 5D). To verify this result in the context of infection, rat embryonic fibroblast (REF) cells were either not infected (mock) or infected with RCMV and were analyzed for the SUMOylation state and localization of ratPML at immediate-early times. While SUMOylated forms of ratPML were still detectable at 4 hours postinfection (hpi) and showed comparable levels as in non-infected cells, we observed a clear loss of ratPML SUMOylation beginning at 8 hpi ( Fig 5E). In parallel with the depletion of SUMOylated ratPML, the intracellular localization of ratPML and ratIE1 changed from a dot-like to a nuclear diffuse staining pattern (Fig 5F). Since these data match previous findings on rhesIE1 and murIE1, which also abrogate PML SUMOylation and induce a dispersion of PML, we conclude that the molecular mechanism underlying PML-NB disruption is conserved across cytomegalovirus species and relies on the unusual fold of the IE1 core domain [22,34].

PML-NBs are not disrupted during cross-species infection
Due to the structural similarity of primate and rat CMV IE1, we next investigated whether IE1 proteins can counteract the PML-based defense during cross-species infection. As shown in Fig 6A, we found that HCMV is capable of entering REF cells and initiate humIE1 expression. However, humIE1 did not localize to nuclear foci, but was distributed throughout the nucleus and did not affect the integrity of ratPML-NBs. In line with this observation, no interaction of humIE1 CORE with ratPML RBCC was detected in co-immunoprecipitation analysis (Fig 6B,  lane 3), suggesting that humIE1 is neither able to bind nor to disrupt PML-NBs in rat cells. In a vice versa experiment, we infected primary human fibroblast (HFF) cells with RCMV. We observed no colocalization of ratIE1 with humPML at 4 h after RCMV infection, suggesting that it does not target PML-NBs in human cells (Fig 6C). At later stages, however, ratIE1 was recruited to large, nuclear domains resembling viral pre-replication compartments. Since PML-NBs were found adjacent to but not colocalizing with these structures (Fig 6C, panel 4) and since no interaction of ratIE1 CORE with humPML could be detected (Fig 6D, lane 2), it can be assumed that not humPML but another cellular or viral protein is responsible for recruiting ratIE1 into nuclear domains. Taken together, these data suggest that PML-NBs are not disrupted by IE1 upon cross species infection and point to a contribution of the PML-based intrinsic defense to the species barrier.

IE1 induces PML-NB dispersal in a species-specific manner
In order to analyze the cross-species activity of IE1 homologs in absence of other viral proteins, we performed a set of experiments using transduced fibroblasts. Lentiviral vectors were utilized to establish human fibroblast (HFF) and rat fibroblast (REF) cells with doxycyclineinducible expression of FLAG-tagged humIE1 or FLAG-tagged ratIE1 as well as control cells. Subsequent immunofluorescence analysis of HFF cell populations in absence or presence of doxycycline revealed a clear dispersal of PML foci upon humIE1 expression (Fig 7A, panel 4), RCMV at an MOI of 1.5 or mock infected, and were harvested at indicated times for Western Blot analysis of ratPML (upper panel), ratIE1 (middle panel), and β-actin (lower panel) as loading control. (F) Impact of RCMV infection on ratPML-NB integrity. REF cells were infected with RCMV at an MOI of 0.7 or mock infected, and were harvested at indicated times for immunofluorescence analysis of ratIE1 (left panel) or ratPML (right panel

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs whereas ratIE1 did neither colocalize with nor disrupt PML-NBs (Fig 7A, panel 6). Quantification of PML foci per cell nucleus corroborated this finding by showing a sharp decline of PML foci in doxycycline-treated HFF/humIE1, while induction of ratIE1 expression did not alter the number of PML-NBs (Fig 7B). In accordance, we observed that humIE1, but not ratIE1, is able to inhibit the SUMOylation of PML in HFF cells (Fig 7C). Equivalent results were obtained in REF cells since only expression of ratIE1 and not humIE1 resulted in dispersal of PML foci (Fig 7D and 7E) and loss of PML SUMOylation (Fig 7F). Overall, our data suggest that the slight structural differences observed in the core domain of primate and rodent IE1 Species-specific binding of IE1 proteins to ratPML in coimmunoprecipitation analysis. HEK293T cells were co-transfected with expression plasmids coding for the TRIM motif of ratPML fused to a myc-tag (ratPML RBCC) and either FLAG-ratIE1 CORE (residues 1-392), FLAG-humIE1 CORE (residues 1-382) or an empty plasmid (pcDNA3). Afterwards, immunoprecipitation was performed with an anti-FLAG antibody. Left panels: Western blot detection of precipitated IE1 proteins and co-precipitated ratPML RBCC (IP). Right panels: detection of IE1 proteins and ratPML RBCC in cell lysates before precipitation (input). (C) Analysis of PML-NB integrity in human fibroblasts after RCMV infection. HFF cells were infected with RCMV-E (MOI = 0.5) or mock infected. Cells were fixed at indicated times for immunofluorescence analysis of humPML and ratIE1. Cell nuclei were visualized by DAPI staining. (D) Species-specific binding of IE1 proteins to humPML in co-immunoprecipitation analysis. HEK293T cells were co-transfected with expression plasmids encoding myc-tagged humPML and either FLAG-ratIE1 CORE (residues 1-392), FLAG-humIE1 CORE (residues 1-382) or an empty plasmid (pcDNA3). After immunoprecipitation of IE1 with an anti-FLAG antibody, coprecipitated humPML (left panels) as well as proteins in the lysate before precipitation (right panels) were detected by Western blotting. https://doi.org/10.1371/journal.ppat.1009863.g006

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs proteins represent evolutionary adaptations to the respective host and result in species-specific targeting of PML-NBs.

Expression of humIE1 allows RCMV replication in human cells
Having shown that ratIE1 is not able to affect PML-NBs in human cells, we asked whether inactivation of their antiviral activity by providing humIE1 in trans results in RCMV particle production. Since PML-NBs are known to block the initiation of lytic replication, we first investigated the effect of humIE1 on RCMV immediate-early gene expression. As shown in Fig  8A (lane 4), low multiplicity infection of human fibroblasts with RCMV yielded detectable levels of ratIE1 only when humIE1 expression was induced by doxycycline treatment. This result was confirmed by immunofluorescence analysis, which showed considerably more ratIE1-positive cells after infection of doxycycline-induced HFF/humIE1 compared to control HFF, thus suggesting a contribution of PML-NBs to the block of RCMV infection in human cells (Figs 8B and S5). Since previous studies reported that the main block of cytomegalovirus replication in cross-species infection occurs after IE gene expression at the stage of viral DNA replication, we investigated the effect of humIE1 expression on RCMV particle production [35]. For this, supernatant from RCMV-infected HFF/control and HFF/humIE1 was harvested and titrated on REF cells. We found that humIE1 expression indeed stimulates the release of infectious RCMV virions from human fibroblasts suggesting that RCMV can cross the species barrier with help of the HCMV IE1 protein (Fig 8C). In accordance, multistep growth curve analysis showed that RCMV replication occurs in humIE1-expressing HFF, but not in control HFF (Fig 8D). Expression of humIE1 CORE , which contains the PML binding region but lacks STAT and histone binding sites, also promoted RCMV replication in HFF (HFF/humIE1 CORE ), albeit to lower maximum titers (Fig 8D). Analogous experiments using human fibroblasts depleted for PML (HFF/shPML) as well as control cells (HFF/shControl) likewise revealed an enhanced initiation of RCMV gene expression (Figs 8E and S5) as well as RCMV replication (Fig 8F) in absence of PML, thus further substantiating the role of PML-NBs as a barrier for cross-species infection. Since click-labeling of incoming RCMV genomes in HFF cells revealed a clear colocalization of viral genomes with PML-NBs, but no expression of ratIE1, we hypothesize that PML-NBs target RCMV genomes to induce an efficient transcriptional repression (Fig 8G). Finally, we investigated whether HCMV can replicate in rat fibroblasts that overexpress ratIE1. We observed a significantly increased initiation of HCMV gene expression in REF/ ratIE1 compared to control cells (Figs 8H and S5). Titration of the supernatants on fresh HFF cells revealed that only few infectious HCMV particles were released from REF/ratIE1 cells, irrespective of whether HCMV laboratory strain AD169 (Fig 8I) or HCMV strain TB40/E ( Fig  8J) were utilized. However, no infectious HCMV particles could be detected after infection of REF/control cells (Fig 8I and 8J). In summary, these experiments indicate that disruption of

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs PML-NBs by ratIE1 increases HCMV IE gene expression in rat fibroblasts but, in contrast to the significant release of RCMV from human fibroblasts expressing humIE1, results only in low-level HCMV replication.

Discussion
CMVs are known for their distinct species-specificity resulting from long-term co-evolution with their mammalian hosts [36][37][38]. Restriction factors can act as important drivers of viral adaptation since they serve as barriers for cross-species transmission necessitating the rapid evolution of viral evasive mechanisms during co-speciation [39,40]. While signatures of positive selection are mainly detected utilizing bioinformatics tools for multiple sequence alignments, protein 3D structures have only recently been integrated into the analysis of adaptive evolution [41]. In this study, we report and compare the structures of human, rhesus and rat cytomegalovirus immediate-early 1 (IE1) proteins, which function as antagonists of the intrinsic cellular defense conferred by PML nuclear bodies (PML-NBs). Our data demonstrate that the IE1 CORE domains of primate and rodent cytomegaloviruses form a unique protein class and display high structural similarity, strongly suggesting that this distinct fold represents an evolutionary adaptation to efficiently bind and neutralize PML-NBs.
Surprisingly, the high structural similarity of IE1 CORE domains is paralleled by an unexpectedly high dissimilarity at the sequence level. Of the three available IE1 CORE structures, humIE1 CORE and rhesIE1 CORE exhibit the closest structural similarity and share 22% sequence identity as deduced from a structure-based alignment ( Table 2). In contrast, ratIE1 CORE displays only 9% of structurally identical residues in comparison to the primate proteins. This value differs significantly from the 22% obtained from pure sequence-based alignments ( Table 2). Moreover, the 9% approach the value of 6.5% that is obtained when aligning sequences generated by randomly scrambling the amino acid sequence of ratIE1. The discrepancy between structure and sequence-derived identities raises the question of whether the sequence has been registered correctly in the experimentally derived structure since sequence registration errors cannot per se be ruled out at resolutions lower than 3 Å [42]. In case of ratIE1 CORE , the correct sequence registration has been corroborated by two additional anomalous scattering experiments, thereby rendering the possibility of registration errors highly unlikely. The low structure-derived sequence identity in combination with the pronounced

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs structural differences explain why it was not possible to solve the structure of ratIE1 via molecular replacement and underline the importance of the experimental structure determination for obtaining an accurate ratIE1 CORE structure with all amino acids reliably allocated and surface patches correctly assembled.
The experimental structure of ratIE1 CORE shows that despite rmsd Cα deviations of up to 4.4 Å, the fold of ratIE1 CORE still remains similar to the primate variants regarding size and overall architecture of the proteins. Thus, we assume that strong positive selection was responsible for the evolution of a distinct IE1 fold in cytomegaloviruses that inactivates the intrinsic cellular restriction instituted by PML-NBs. Furthermore, one may speculate that the unexpected dissimilarity at the sequence level could be due to genetic divergence driven by strong immune selection pressure since IE1 is known to serve as a major target of the cytotoxic T-cell (CTL) response [43]. Although antigenic adaptation to antibody responses is a well-known process, selection pressure imposed by CTL immunity has only recently been recognized as an additional important player shaping long-term viral evolution [44].
Our results show that primate and rodent CMV IE1 proteins not only share a distinct fold but they also use a conserved mechanism to inactivate PML-NBs. In previous studies, we demonstrated that humIE1 directly interacts with humPML via its core domain, which abrogates the de novo SUMOylation of PML followed by the disassembly of PML-NBs [19]. Similarly, the core domain of ratIE1 binds to ratPML and also induces loss of SUMOylation as well as dispersal of PML-NBs in rat cells (Fig 5). Importantly, binding of both humIE1 and ratIE1 to the respective PML proteins depends on coiled-coil interactions, and this correlates with a conserved composition of left-and right-handed coiled-coil motifs that defines the helix pairing and hence the topology of the canonical IE1 CORE fold (Fig 4). Thus, although the sequence identities among the investigated IE1 proteins from primate and rodent CMV IE1 CORE are low, a high degree of similarity is observed within the coiled-coil repeat motifs across host species orders. We therefore propose that this knowledge will help identifying and validating further members of the IE1 CORE family, which might share even lower sequence identities.
Despite an overall similar tertiary fold and quaternary assembly, distinct differences in the helix arrangements and protein surface properties can be observed between primate and rodent IE1 CORE structures (Fig 2). While humIE1 and rhesIE1 exhibit the highest similarity corresponding to functional exchangeability of the respective proteins in context of a CMV infection [22], ratIE1 fails to disrupt PML-NBs in human cells. This correlates with a lack of interaction between ratIE1 CORE and humPML. Likewise, humIE1 does not bind ratPML nor disrupt PML-NBs in rat cells. This is in accordance with previous studies demonstrating that PML-NBs in mouse cells are not disrupted by HCMV and MCMV cannot redistribute PML-NB components in human cells [45]. These data show that rodent and human IE1 proteins function in a species-specific manner strongly suggesting that the IE1 proteins of cytomegaloviruses must have co-evolved with their respective binding partner PML from the host organism.
Interestingly, the sequence dissimilarities observed in the IE1 proteins are not matched by those observed in PML proteins since the latter show a very high degree of sequence conservation. While the sequence identity between the coiled-coil domains of humPML and rhesPML is 95%, the sequence identity between these primate domains with the equivalent ratPML domain is still as high as approx. 70%. The sequence dissimilarities of the IE1 proteins and the sequence similarities of the PML proteins suggest that the IE1 proteins are likely to contribute most to the IE1-PML interaction specificity profile and thereby control the species specificity of CMVs in antagonizing PML-NB-mediated intrinsic immune defense. A recent study has reported that mutation of four adjacent surface residues within helix 5 of humIE1 is able to abrogate its interaction with humPML [46]. While this segment displays a lysine residue in humIE1 (Lys172) and rhesIE1 (Lys187), a glycine residue is displayed at this position in ratIE1 (Gly187). Clearly, this substitution could contribute to the observed interaction profile. However, a comprehensive mapping of the IE1-PML binding interface has not been achieved yet and will require further structural and functional investigations.
Cellular restriction factors and their antagonization by viral effector proteins can act as decisive factors for cross-species transmission of viruses. This has extensively been studied for antiretroviral restriction factors like APOBEC3 deaminases which need to be degraded by adapted lentiviral vif proteins in order to allow cross-species infection [47]. In contrast, the mechanisms limiting cross-species infection of cytomegaloviruses are less well understood. For instance, a very recent study reported that the replication of rhesus CMV in human cells is considerably enhanced upon duplication of a terminal genomic region which enhances expression of the protein kinase R (PKR) antagonist rTRS1 encoded by rhesus CMV [48]. This indicates that translational inhibition instituted by human PKR serves as a barrier against cross-species infection with rhesus CMV, which can be overcome by adaptive gene amplification [49]. Furthermore, Jurak and Brune published that infection of human cells with MCMV triggers the intrinsic apoptosis pathway thus abrogating productive viral infection [50]. Expression of the HCMV encoded bcl-2 homolog UL37x/vMIA, however, alleviates this block indicating that induction of apoptosis may contribute to the inhibition of cross-species infections of rodent CMVs. Here, we report that PML-NBs also serve as a distinct barrier against cross-species infections of rodent CMVs. While ratIE1 was not able to affect PML-NBs in human cells, we found that inactivation of their antiviral activity by providing humIE1 in trans not only resulted in RCMV immediate early gene expression but also in the release of significant amounts of viral particles indicating unrestricted productive infection of human cells by RCMV (Fig 8). Since both the expression of humIE1 CORE and the shRNA mediated depletion of PML in human fibroblasts were sufficient to elicit productive RCMV replication, we conclude that the PML-NB disrupting activity of IE1 plays a major role to allow for permissive infection (Fig 8D-8F). Of note, a perfect colocalization of viral DNA with PML-NBs after infection of human fibroblasts with RCMV was observed suggesting efficient silencing of RCMV gene expression in the absence of PML-NB antagonization (Fig 8G). HCMV infection of rat fibroblasts expressing ratIE1 also resulted in increased initiation of viral gene expression, however, only low release of viral particles was observed indicating an additional barrier for productive HCMV infection in rodent cells (Fig 8H-8J). This is in accordance with previously published results on MCMV demonstrating that knocking-down of PML-NB components significantly increases viral protein production in cross-species infection experiments, however, does not result in productive infection [34,45]. Interestingly, while PML-NB disruption was sufficient for productive RCMV infection of human fibroblasts, MCMV was shown to require additional HCMV gene functions to cross the human species barrier [35]. In summary, while emerging evidence supports the view that cytomegaloviruses have to counteract multiple hurdles to infect the cells of other species, our study provides strong evidence that PML-NBsbased defense contributes as an important barrier against cross-species infections. Furthermore, our data support the concept that long-term co-speciation of cytomegaloviruses has evolved a distinct IE1 fold that has been adapted to maximize the efficiency of PML-NB targeting.

Oligonucleotides and expression plasmids
All oligonucleotide primers used in this study were purchased from Biomers GmbH, Eurofins Genomics GmbH or Metabion GmbH and are listed in S4 Table. Expression plasmids

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs encoding HCMV IE1 (humIE1) 14-382 for prokaryotic expression or humIE1 1-382 for eukaryotic expression were generated as described previously [22]. The codon-optimized RCMV IE1 (ratIE1) template cDNA (strain RCMV-E, sequence based on Uniprot K7XWE8) was obtained from Biocat GmbH gene synthesis service. Prokaryotic expression plasmids coding for full length ratIE1 and the variants 1-392 and 30-392 were generated by PCR amplification of codon-optimized sequences and insertion into pGEX-6P-1 (GE Healthcare), resulting in GST-tagged fusion proteins. Eukaryotic expression plasmids for co-immunoprecipitations were generated via PCR amplification of respective fragments from plasmids containing the ratIE1 cDNA (a kind gift of Sebastian Voigt, Berlin, Germany) or ratPML cDNA (sequence based on Uniprot F1M589), which was synthesized by Biocat GmbH. The PCR products were inserted into pHM971 (pcDNA3-FLAG) or pHM1580 (pcDNA3-Myc) [51]. The plasmid encoding Myc-tagged human PML, isoform VI, was described previously [52]. Prokaryotic expression plasmids coding for ratPML 1-207 and humPML 20-255 in pGEX-6P-1 were generated as above. For humPML 20-234, a stop codon was inserted into the humPML 20-255 construct by means of site-directed mutagenesis. For transduction experiments, the lentiviral vector pInducer20 (a gift from Stephen Elledge; Addgene plasmid # 44012; http://n2t.net/ addgene:44012; RRID:Addgene_44012) was modified by site-directed mutagenesis of the cisrepression sequence (CRS) within its promoter region as the CRS leads to transcriptional repression during HCMV infection [53,54]. Mutagenesis was performed with primers c-CRSmut and nc-CRS-mut and resulted in plasmid pInducer20-CRSmut. FLAG-tagged HCMV IE1 and RCMV IE1 sequences were amplified by PCR with primers listed in S4 Table and were inserted into pInducer20-CRSmut by a combined BP/LR Gateway recombination reaction using pDONR221 (Invitrogen) as entry vector. Lentiviral pLVX-shRNA1-based vectors containing a control shRNA or a shRNA directed against PML were generated as described previously (see S4 Table for target sequences) [55].

Recombinant protein production and purification
All variants of IE1 and PML were recombinantly produced in E. coli BL21(DE3) cells (Novagen) as GST-tagged fusion proteins. LB or TB media (Carl Roth) were inoculated with transformed E. coli cells and shaken at 37˚C. Media were supplemented with 100 μg/mL of ampicillin as well as 50 μM ZnCl 2 for the expression of humPML. Overexpression was induced by adding 0.1 mM IPTG and shaking at 20˚C over night. Seleno-methionine labeled ratIE1 30-392 was produced in auto-inducing PASM-5052 medium as described in the literature [56].
All purification steps were performed at 4-8˚C. All chromatography buffers contained 5 mM DTT and either 1 mM EDTA (IE1 variants) or 25 μM ZnCl 2 (PML variants). Cell pellets were resuspended in PBS buffer and lysed by sonication. Fusion proteins were captured using the aforementioned affinity media. GST-fusion proteins were cleaved with a GST-tagged human rhinovirus 3C protease and purified using a second affinity chromatography step. Proteins were then concentrated using Vivaspin 20 centrifugal concentrators (5 kDa molecular weight cutoff, Sartorius Stedim) and purified using a 26/600 Superdex 200 prep grade column (GE Healthcare) pre-equilibrated in 25 mM TRIS/HCl, 150 mM NaCl, 5 mM DTT, pH 7.4. The samples were eluted with an isocratic gradient of 1.2 column volumes of the same buffer at a flow rate of 34 cm/h (3 mL/min).
For crystallization, surface lysine residues of ratIE1 30-392 were chemically methylated. This was performed after the second affinity chromatography step. The buffer was exchanged to 25 mM Na-HEPES, 150 mM NaCl, 5 mM DTT, 1 mM EDTA, pH 7.5 and 20 μL of a 1 M borane dimethylamine complex and 40 μL 1 M formaldehyde were added per mL of protein

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs solution. After two hours of incubation on ice, the addition of borane dimethylamine complex and formaldehyde was repeated. After additional two hours, another 10 μL of dimethylamine borane complex per mL of solution were added and the reaction was incubated over night. The reaction was quenched by adding 125 μL 1 M TRIS/HCl pH 7.5 per mL solution. The methylated protein samples were purified by size exclusion chromatography as described above.

Limited proteolysis
Full-length ratIE1 was incubated at 22˚C with 1 mU subtilisin (Sigma-Aldrich) per mg ratIE1 in the presence of CaCl 2 . 15 μL samples were taken at timepoints between one and 128 min and immediately mixed with 5 μL 4x SDS PAGE loading buffer and boiled at 95˚C for 5 min. A gel band was excised and analyzed by mass spectrometry.

Circular dichroism spectroscopy
Circular dichroism spectra were recorded between 185 and 260 nm using a Jasco J-815 spectropolarimeter (Jasco, Tokyo, Japan). Protein samples were dialyzed twice against 10 mM KH 2 PO 4 pH 7.5 and set to a concentration of 5 μM. Measurements were conducted at 20˚C using a quartz cuvette with a path length of 0.1 cm, scan speed of 20 nm/min, band width 1 nm, data integration time 1 s and data pitch 0.1 nm. All measurements were accumulated eight times and corrected for the sample buffer. The spectra were normalized at 207 nm as suggested by Raussens and coworkers [57].

Protein crystallization
Prior to crystallization, all proteins were dialyzed against an at least 500-fold volume of 25 mM TRIS/HCl, 10 mM DTT, 1 mM EDTA, pH 7.4 (set at 20˚C) and concentrated to 20 mg/mL. Initial crystallization screening was performed using commercial screens and the sitting drop vapor diffusion method. Diffraction quality crystals were obtained in the hanging drop setup.

Crystallographic data collection
Data of native and seleno-methionine derivatized ratIE1 30-392 crystals were collected at 100 K at beamline 14.2 at the BESSY II synchrotron ( [58]; Helmholtz Zentrum Berlin). The protein crystallized in the space group P6 5 22 and the crystal diffracted to 3.4 Å. In addition to a native dataset, two-wavelength MAD data were collected from seleno-methionine derivatized crystals (peak, inflection point). For sequence validation, 6 keV anomalous sulfur data were collected from native crystals at the P13 beamline at DESY ( [59]; Deutsches Elektronen-Synchrotron, Hamburg). Native data of humIE1 14-382 crystals were collected at the P13 beamline at DESY ( [59]; Deutsches Elektronen-Synchrotron, Hamburg). The protein crystallized in the space group C222 1 and the crystal diffracted to 3.2 Å. All diffraction data were processed using XDS and scaled and merged with XSCALE and XDSCONV [60].

Structure determination and refinement
The structure of ratIE1 30-392 was solved using two-wavelength MAD data from five merged seleno-methionine peak and one inflection point dataset. Anomalous data to 4 Å were used for data preparation (SHELXC), substructure search (SHELXD) and initial chain tracing (SHELXE) [61] as implemented in the HKL2MAP package [62]. The initial phases allowed for the calculation of an anomalous map from the peak data to 5 Å using FFT [63] from the CCP4 suite [64]. This map was used to place the selenium atoms of 11 out of 12 seleno-methionine residues. Based on the methionine positions, the model was built by hand and refined against the native 3.4 Å data using multiple iterations of phenix.refine [65], phenix.rosetta_refine [66] and manual fitting in COOT [67]. Atomic displacement parameters were refined using two Bfactors per amino acid residue and three TLS groups per molecule. For further validation of the built sequence, anomalous maps were calculated using phases from the refined model and merged 6 keV X-ray data from seven crystals to a resolution of 5 Å, as well as the aforementioned seleno-methionine peak data.
The dataset from humIE1 was found to be moderately anisotropic. For ellipsoidal truncation, unmerged data were processed using the STARANISO server [68]. The structure of humIE1 14-382 was solved with MR-Rosetta using rhesIE1 CORE (PDB: 4WID; [22]) as a search model [23]. A poly-alanine model was placed in the obtained electron density. Sequence and structure alignments with the published rhesus CMV IE1 structure (PDB: 4WID) were used to place the sequence. Refinement was conducted as described above. Atomic displacement parameters were refined using two B-factors per amino acid residue and eight TLS groups per molecule. Data collection and refinement statistics are summarized in Table 1.

Bioinformatic analyses
Sequence based disorder prediction was performed using IUPred2A using the default setting "IUPred2 long disorder" [69]. The dimer interface areas were calculated using the EPPIC server [29]. Pairwise structure-based sequence identities and RMS deviations were calculated using DALI [70]. For comparison of the dimers, coordinates files for humIE1 and ratIE1 were generated from the molecule in the asymmetric unit and a symmetry related molecule using COOT [67]. The multiple structure-based sequence alignment was generated with PRO-MALS3D [32] using the abovementioned structures. Structure-based sequence identities were determined from the structurally equivalent areas of the alignment. Sequence alignments without structural information were obtained using Clustal Omega [71]. Crystal structure illustrations were generated with PyMol and UCSF Chimera [72,73]. The angles between IE1 monomers were determined with the "Axes/Planes/Centroids" tool implemented in Chimera. Axes were placed through all atoms in each monomer.

Cells and virus infections
HEK293T cells were cultivated in Dulbecco's minimal essential medium (DMEM) containing glutamine and supplemented with 10 % fetal calf serum and penicillin-streptomycin (Sigma). Primary human foreskin fibroblasts (HFF), which were prepared from human foreskin tissue, and rat embryonic fibroblasts (REF), which were obtained from Sebastian Voigt (Berlin, Germany) were maintained in Eagle's minimal essential medium (MEM) supplemented with 7 % fetal calf serum (Sigma), glutamax (Gibco), and penicillin-streptomycin (Sigma). HFF and REF cells were infected with either the HCMV strains AD169 and TB40/E or the RCMV English isolate (RCMV-E, kindly provided by Sebastian Voigt) at specified multiplicities of infection (MOI). Viral titers were determined by IE1 fluorescence. To this end, HFF or REF cells were infected with various dilutions of virus stocks. Cells were incubated for 24 h for titration of AD169 or TB40/E and 8 h for titration of RCMV-E, and were subsequently fixed and stained with a monoclonal antibody against humIE1 or ratIE1. The number of IE1-positive cells was determined and used to calculate viral titers.

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs

Lentiviral transduction and selection of stably transduced cells
For the generation of HFF and REF cells with doxycycline-inducible expression of humIE1 or ratIE1, replication-deficient lentiviruses were generated using pInducer20-based expression constructs. For this purpose, HEK293T cells seeded in 10 cm dishes (5 x 10 6 cells/dish) were transfected with a pInducer20-based vector together with packaging plasmids pLP1, pLP2, and pLP/VSV-G using the Lipofectamine 2000 reagent (Invitrogen). Viral supernatants were harvested 48 h after transfection, clarified through a 0.45-μm filter and stored at -80˚C. HFF or REF cells were incubated for 24 h with lentivirus supernatants in the presence of 7.5 μg/mL polybrene (Sigma-Aldrich). Stably transduced, IE1-expressing cell populations were selected by adding 500 μg/mL geneticin (Invivogen) to the cell culture medium containing 7 to 10% tetracycline-free fetal bovine serum (Clontech). IE1 expression was induced by addition of 500 ng/mL doxycycline (Sigma-Aldrich). To generate PML-knockdown and control cells, HFF were transduced with lentiviral supernatants produced from pLVX-shRNA-based expression constructs, followed by selection with 5 μg/mL puromycin (Invivogen).

Co-immunoprecipitation
HEK293T cells were seeded in 6-well plates (7 x 10 5 cells/well, 2 wells per sample) and, one day later, were transfected with 2 to 4 μg of plasmid DNA per well using the TurboFect transfection reagent (Thermo Fisher Scientific). 48 h after transfection, the cells were lysed for 25 min at 4˚C in 800 μL of CoIP buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 5 mM EDTA, 0.5% NP-40, 1 mM PMSF, 2 μg/mL of aprotinin, 2 μg/mL of leupeptin, and 2 μg/mL of pepstatin). After centrifugation, aliquots of each sample were taken as input controls and the remaining supernatant was incubated with anti-FLAG antibody M2 (Sigma-Aldrich) coupled to protein-A-sepharose beads for 2 h at 4˚C. The sepharose beads were collected by centrifugation and washed four times with 1 mL CoIP buffer. Finally, immunoprecipitated proteins were recovered by boiling in 4x SDS sample buffer and protein complexes were analyzed by SDS-PAGE and Western blotting.

Indirect immunofluorescence
HFF or REF cells grown on coverslips in six-well plates (3 x 10 5 cells/well) or 12-well plates (1.2 x 10 5 cells/well) were washed twice with PBS at specified times after virus infection or after doxycycline treatment. Cells were fixed with a 4% paraformaldehyde solution for 10 min at room temperature (RT) and washed twice, before permeabilization was achieved by incubation with 0.1% Triton X-100 in PBS for 10 min on ice. Cells were washed again with PBS over a time period of 5 min and incubated with the appropriate primary antibody diluted in 1% FCS in PBS for 30 min at 37˚C. Excessive antibodies were removed by washing three times with PBS, followed by incubation with the corresponding fluorescence-coupled secondary antibody diluted in 1% FCS in PBS for 30 min at 37˚C. The cells were mounted with DAPIcontaining Vectashield mounting medium (Vector Laboratories) and analyzed using a Zeiss Axio Observer Z1 with an Apotome.2. The images were processed and exported with the ZEN 2 software and assembled using CorelDraw 2018. For quantification of PML foci, Z-series images of 50 cell nuclei per sample were taken and the number of PML dots was assessed in maximum intensity projection images (0.3 μm distance). Following antibodies were used for immunofluorescence detection: mAb FLAG M2 (Sigma-Aldrich), mAb HCMV IE1 p63-27 [74], mAb RCMV IE1 (kindly provided by Sebastian Voigt), mAb PML 5E10 (kindly provided by Roel van Driel) to detect rat PML, and pAb PML A301-167A (Bethyl Laboratories) to detect human PML.

RCMV DNA labeling with ethynyl-modified nucleosides and detection by click chemistry
In order to produce labeled RCMV stocks, RCMV-E was grown in REF cells in the presence of 5μM EdC. EdC-containing medium was replaced every 24 h until a strong cytopathic effect was observed. Supernatants from infected cells were clarified by centrifugation at 2000 rpm for 15 min and then pelleted by ultra-centrifugation at 17000 rpm for 3 h at 4˚C. Pellets were rinsed with medium, before they were resuspended and passed through a 20 gauge syringe needle for several times. In order to visualize viral DNA in combination with antibody staining, HFF infected with EdC -labeled RCMV were fixed with 4% PFA for 10 min and quenched with 50 mM ammonium chloride and 50 mM glycine in PBS for 5 min at RT. Cells were washed twice with PBS, permeabilized with 0.1% TritonX100 in PBS for 15 min at 4˚C, and stained with antibodies as described above. After washing cells twice with PBS, copper-catalyzed click reaction was performed by incubating the cells for 90 min at RT with freshly prepared labeling solution containing 10 μM Alexa Fluor 488 Azide (Invitrogen), 10 mM sodium ascorbate, 1 mM copper (II) sulfate, 10 mM aminoguanidine, and 1 mM THPTA in PBS. Cells were washed twice with PBS for 5 min, before coverslips were mounted on microscope slides using Vectashield Antifade Mounting Medium with DAPI (Vector laboratories, Maravai Life-Sciences, San Diego, CA, USA) and sealed with nail polish.

Multistep growth curve analysis
HFF cells were seeded in triplicates or quadruplicates into 12-well dishes at a density of 1.2 × 10 5 cells/well and were treated with doxycycline for one day, before they were infected with RCMV at an MOI of 0.01. Supernatants from infected cells were harvested at indicated days after inoculation and cells were provided with fresh medium containing doxycycline. After proteinase K treatment, all samples were analyzed for the amount of genome copy numbers by quantitative real-time PCR using an Agilent AriaMx Real-time PCR System together with the corresponding software Agilent Aria 1.5 (Agilent Technologies, Inc, Santa Clara, CA, USA). For quantification of the RCMV DNA load, a sequence region within the gB gene locus was amplified using primers 5'RCMV-and 3'RCMV-gB along with the hydrolysis probe RCMV-gB FAM/TAMRA (S4 Table)

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs consisted of an initial step of 3 min at 95˚C followed by 40 amplification cycles (10 s at 95˚C, 30 s 60˚C). Viral genome copy numbers were subsequently calculated using the sample-specific C T value set into relation to the standard serial dilutions. The spectra were normalized at 207 nm as suggested by Raussens and coworkers [57]. The spectra suggest that the humPML and ratPML RB segments share a highly similar secondary structure composition and that both protein variants are properly folded.

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs were subjected to immunofluorescence staining of humIE1 in order to determine the initiation of lytic gene expression. All values are derived from triplicate samples and represent mean values ± SD. P-values were calculated using two-tailed Student's t-test. �� , p � 0.01; ��� , p � 0.001. (TIF) S1 Table. Peak assignment in an anomalous difference map calculated with the selenomethionine peak data. (DOCX) S2 Table. Peak assignment in an anomalous difference map calculated with a long-wavelength data set (2.0 Å, 6 keV).

PLOS PATHOGENS
CMV co-speciation evolved a distinct IE1 structure to target PML-NBs