The nuclear egress complex of Epstein-Barr virus buds membranes through an oligomerization-driven mechanism

During replication, herpesviral capsids are translocated from the nucleus into the cytoplasm by an unusual mechanism, termed nuclear egress, that involves capsid budding at the inner nuclear membrane. This process is mediated by the viral nuclear egress complex (NEC) that deforms the membrane around the capsid. Although the NEC is essential for capsid nuclear egress across all three subfamilies of the Herpesviridae, most studies to date have focused on the NEC homologs from alpha- and beta- but not gammaherpesviruses. Here, we report the crystal structure of the NEC from Epstein-Barr virus (EBV), a prototypical gammaherpesvirus. The structure resembles known structures of NEC homologs yet is conformationally dynamic. We also show that purified, recombinant EBV NEC buds synthetic membranes in vitro and forms membrane-bound coats of unknown geometry. However, unlike other NEC homologs, EBV NEC forms dimers in the crystals instead of hexamers. The dimeric interfaces observed in the EBV NEC crystals are similar to the hexameric interfaces observed in other NEC homologs. Moreover, mutations engineered to disrupt the dimeric interface reduce budding. Putting together these data, we propose that EBV NEC-mediated budding is driven by oligomerization into membrane-bound coats.


Introduction
To achieve successful replication, viruses have evolved strategies to navigate across compartmentalized eukaryotic cells. One of the more unusual mechanisms of traversing intracellular membranes is found in herpesviruses-enveloped DNA viruses that infect multiple animal species, including humans. The family Herpesviridae, which infect mammals, birds, and reptiles are classified into three subfamilies: alphaherpesviruses, betaherpesviruses, and gammaherpesviruses. Among them are nine herpesviruses that infect humans: herpes simplex virus types 1 and 2 (HSV-1 and HSV-2) and varicella-zoster virus (VZV) (alphaherpesviruses); human cytomegalovirus (HCMV) and human herpesvirus types 6A, B and 7 (HHV-6A/B and HHV-7) (betaherpesviruses); and Epstein-Barr virus (EBV) and Kaposi's sarcoma herpes virus (KSHV) (gammaherpesviruses). Together, these viruses infect most of the world's population for life causing a spectrum of diseases ranging from painful sores to blindness to life-threatening conditions in people with weak immune systems [1]. EBV, the focus of this work, causes infectious mononucleosis in adolescents and is associated with several hematopoietic and epithelial cell cancers [2][3][4] and lymphoproliferative disorders in immunocompromised patients, including those with HIV/AIDS or organ transplant recipients [5]. At present, available vaccines only target VZV while therapeutics are suboptimal. Therefore, a better knowledge of the herpesviral biology may pave the way towards effective preventives and therapeutics.
During replication, herpesviral capsids cross several host membrane barriers, traversing three distinct cellular compartments (nucleus, cytoplasm, and TGN/endosomes) while completing their assembly, before exiting the cell as infectious virions [6,7]. This process is termed egress. As the first step in egress, viral capsids, which are assembled and packaged with dsDNA genomes in the nucleus, must get across the double-membraned nuclear envelope, entering the cytoplasm to complete their maturation into infectious virions. Most cellular traffic into and out of the nucleus occurs through the nuclear pores. But the~50-nm opening of the nuclear pore is too small to accommodate the~125-nm capsids of herpesviruses [8]. Instead, capsids utilize a different, more complex escape route, termed nuclear egress, where capsids bud at the inner nuclear membrane (INM) and pinch off into the perinuclear space (primary envelopment). The perinuclear enveloped virions subsequently fuse with the outer nuclear membrane (ONM), releasing the capsids into the cytoplasm (de-envelopment) (reviewed in [9-11]).
Capsid budding at the INM requires the viral nuclear egress complex (NEC), a heterodimer of two conserved viral proteins: UL31, a soluble nuclear phosphoprotein, and UL34, which contains a single C-terminal transmembrane (TM) helix that anchors the NEC in the INM (reviewed in [8]). In many cases, the absence of either UL31 or UL34 causes capsids to accumulate in the nucleus, which reduces the viral titer by several orders of magnitude [12][13][14][15][16][17][18][19][20][21]. Although in rabbit skin cells infected with UL31-null HSV-1, viral titers are only moderately reduced (10-to 50-fold drop) [22], which could be due to the use of a suboptimal alternative egress route such as nuclear envelope breakdown, observed in UL34-null PRV [23].
Regardless, all studies show a reduction in viral titer in the absence of either UL31 or UL34, and thus, both genes are essential for nuclear egress across all three subfamilies of the Herpesviridae.
Our current understanding of the NEC function is largely based on the studies of homologs from the alphaherpesviruses HSV-1 and PRV. First, overexpression of PRV NEC in mammalian cells was found to cause formation of capsidless vesicles in the perinuclear space [24]. This demonstrated that UL31 and UL34 were the only viral proteins necessary for nuclear envelope budding. Subsequent in-vitro studies established the intrinsic budding ability of the NEC by showing that purified recombinant NEC from HSV-1 or PRV vesiculated synthetic lipid vesicles in vitro in the absence of any other factors [25,26]. The crystal structures of NEC homologs from HSV-1 [27] and PRV [27,28] provided blueprints for mechanistic studies and, in particular, revealed extensive interactions that stabilize the heterodimer. Finally, cryogenic electron microscopy and tomography (cryo-EM/ET) studies showed that the NEC oligomerizes into hexagonal coats on the inner surface of budded vesicles formed by recombinant HSV-1 NEC in vitro [25,43], in uninfected cells overexpressing PRV NEC [29], and in perinuclear enveloped vesicles purified from HSV-1-infected cells [30]. A high-resolution view of the NEC/ NEC oligomeric interfaces was afforded by the crystal structure of HSV-1 NEC [27], in which the NEC formed a hexagonal lattice with the same geometry and dimensions as the hexagonal membrane-bound coats formed during budding [25,29,30,43]. Follow-up work confirmed that HSV-1 NEC mutations that disrupt oligomeric interfaces reduce budding in vitro [27] and in infected cells [31,32]. Collectively, these findings established the NEC of alphaherpesviruses as a robust membrane-budding machine that forms hexagonal membrane-bound coats (reviewed in [8]).
While less well studied than the alphaherpesvirus homologs, the NECs from the betaherpesviruses HCMV and murine cytomegalovirus (MCMV) are also essential for successful viral replication [33,34]. Insertions or deletions within the conserved regions of MCMV UL53, a UL31 homolog, reduce viral replication [35,36]. Additionally, point mutations in HCMV UL50, a UL34 homolog, designed to interfere with NEC formation, also reduced viral replication, thereby underscoring the importance of complex formation for functionality [37]. The crystal structures of the NEC homolog from the betaherpesvirus HCMV [38,39] revealed structural similarities with the NEC from alphaherpesviruses HSV-1 [27] and PRV [27,28] despite the relatively low sequence identity. Moreover, in one of the structures, the HCMV NEC [39] formed a hexagonal lattice with the same geometry and dimensions as those formed by the HSV-1 [25,27,30,43] and PRV counterparts [29]. Altogether, this implied a common NEC-mediated mechanism of primary capsid envelopment between the alpha-and betaherpesviruses.
In gammaherpesviruses EBV and KSHV, mechanistic studies of the NEC function utilized overexpression in mammalian cells or insect cells infected with recombinant baculovirus, respectively [40,41]. Overexpression of UL34 homologs, EBV BFRF1 [41] or KSHV ORF67 [40], caused formation of proliferations of nuclear and cytoplasmic membranes that resemble stacks or tubules. In contrast, overexpression of both UL34 and UL31 homologs (EBV BFRF1 and BFLF2, KSHV ORF67 and ORF69) produced curved multilayered cisternae at the INM in the case of EBV [41] and perinuclear vesicles in the case of KSHV [40], the latter resembling the perinuclear vesicles observed in mammalian cells overexpressing PRV NEC [24]. These findings implicated a conserved role for the gammaherpesviral NEC in nuclear envelope remodeling. However, neither the intrinsic budding ability nor coat formation have yet been demonstrated for NEC from any gammaherpesvirus. Moreover, the available structural information is limited to EBV BFRF1 bound to a short segment of BFLF2 [42]. To fill in these knowledge gaps, we pursued structural and functional studies of EBV NEC.
Here, we show that the purified, recombinant EBV NEC vesiculates synthetic lipid membranes in vitro and forms membrane-bound coats, which suggests that the intrinsic membrane budding ability is a conserved property of the NECs across Herpesviridae. We also report the most complete crystal structure of EBV NEC to date and show that it resembles known structures of NEC homologs from the other two subfamilies. Importantly, EBV NEC crystals contain five independent, structurally distinct heterodimers in the asymmetric unit. The structural differences among the 5 EBV NEC heterodimers, notably within BFLF2 and at the BFLF2/ BFRF1 interface, for the first time experimentally demonstrate conformational dynamics within the EBV NEC. However, instead of hexamers, the EBV NEC forms dimers in the crystals, and its membrane-bound coats formed in vitro appear different from the hexagonal geometry observed in NEC coats of alphaherpesviruses [25,29,43]. The dimeric interfaces observed in the EBV NEC crystals are similar to the hexameric interfaces observed in other NEC homologs. Moreover, mutations engineered to disrupt dimeric interfaces reduce budding. Therefore, we propose that while the NEC operates as an oligomerization-driven membrane budding machine, the EBV NEC coats have a different, yet unknown geometry, potentially, due to the structural flexibility of the EBV NEC.

EBV NEC crystallization and structure determination
For crystallization, we designed a construct EBV NEC195Δ65 composed of residues 1-195 of BFRF1 and residues 66-318 of BFLF2 (Fig 1A), which is similar to those of previously crystallized NEC homologs [27,28,38,39,42]. This construct lacks the membrane-proximal regions (MPRs) in both BFRF1 and BFLF2 as well as the TM anchor of BFRF1 ( Fig 1A). The EBV NEC195Δ65 was expressed in E. coli and purified using affinity and size-exclusion chromatography, similarly to HSV-1 and PRV NEC [27]. Crystals were grown by vapor diffusion in hanging drops, and unusually, they appeared in drops that contained only the protein complex but no reservoir solution and were equilibrated over the reservoir solution (25% PEG 3350, 0.1 M Tris-HCl pH 8.5, 0.2 M Li 2 SO 4 ). The crystals took the P4 3 2 1 2 space group with five NEC heterodimers in the asymmetric unit, termed NEC1-5. The structure was determined by a combination of single anomalous dispersion of the Zn 2+ ion and molecular replacement using the structure of BFRF1 fused to the BFLF2 hook by a flexible linker (rcsb pdb 6t3z [42]) as the search model. The structure was refined to 3.97 Å resolution (R work = 27.0%, R free = 30.8%) ( Table 1). Atomic coordinates and structure factors for the EBV NEC structure have been deposited to the RCSB Protein Data Bank under accession number 7t7i.

Overall structure of the EBV NEC
Like its counterparts in alpha-and betaherpesviruses, the EBV NEC resembles an elongated cylinder that is composed of the globular BFRF1 pedestal topped with the globular C-terminal domain of BFLF2 and the N-terminal two-helix "hook" of BFLF2 that wraps around the BFRF1 pedestal ( Fig 1B). There are five EBV NEC heterodimers in the asymmetric unit, NEC1-5, composed of BFRF1 chains A, C, E, G, and I and BFLF2 chains B, D, F, H, and J ( Fig  1B). Three heterodimers, NEC1 (A/B), NEC2 (C/D), and NEC4 (G/H), as well as BFRF1 of NEC3 (E) are well resolved (89-99% of all residues) whereas NEC5 (I/J) and BFLF2 of NEC3 (F) are less well resolved (75-78% of all residues) (S1 Table). The NEC heterodimers have notable structural differences (Fig 1C-1F), mainly in BFLF2 and the relative orientations of BFRF1 and BFLF2 as described in detail below. They can be superimposed with the root mean square deviations (RMSDs) ranging from 1.16 Å to 2.57 Å (S2 Table).

The BFRF1 structure
The BFRF1 structure is similar to those of the UL34 homologs in alpha-and betaherpesviruses and is composed of two β-sheets arranged into a novel fold of a splayed β-sandwich, previously termed a β-taco [37], decorated with four α-helices (S1A Fig). Helices α1, α2, and α4 are oriented parallel to each other and to the longest axis of the NEC and brace helix α2 of BFLF2 hook. In chains A, C, E, and G, all residues except the last C-terminal residue 195, are resolved. These four chains can be superimposed with RMSDs ranging from 0.72 to 0.79 Å (S2 Table). Several loops in these four chains, notably, the β5/β6 loop, adopt different conformations ( Fig  1D). Additionally, chain E contains the longest α4 helix but lacks the short β6 strand (S2A . BFLF2 is shown in blue while BFRF1 is shown in pink. Grey regions denote residues omitted from constructs including the membrane-proximal regions (MPRs) and the transmembrane (TM) region in darker grey. (B) Crystal structures of the 5 EBV NEC copies in the asymmetric unit with chain IDs. BFLF2 is shown in blue with light blue for α-helices and dark blue for β-strands. BFRF1 is shown in pink with light pink for α-helices and dark pink for β-strands. (C) Ribbon overlay aligned to NEC2 (Chains C+D) for the 5 copies of EBV NEC in the asymmetric unit. NEC1 (teal), NEC2 (gold), NEC3 (light purple), NEC4 (green), and NEC5 (purple). (D) Ribbon overlay aligned to chain A for the 5 copies of BFRF1 in the asymmetric unit. Red oval highlights differences in β5/β6 loop. Chain A (teal), chain C (gold), chain E (light purple), chain G (green), and chain I (purple). (E) Ribbon overlay aligned to the globular domain of B for the 5 copies of BFLF2 in the asymmetric unit. Red oval highlights differences in β2/β3 loop. Chain B (teal), chain D (gold), chain F (light purple), chain H (green), and chain J (purple). (F) Ribbon overlay of the 5 copies of BFLF2 aligned to BFRF1 chain A (hidden) in the asymmetric unit. Same coloring as (E). Images were created in PyMol [44].

PLOS PATHOGENS
Structure and function of EBV NEC Fig). Chain I has several unresolved regions and has the largest RMSDs, 1.01 to 1.14 Å (S2 Table).

The BFLF2 structure
BFLF2 structure resembles those of the UL31 homologs in alpha-and betaherpesviruses and is composed of a globular domain and a helical hook. The globular domain of BFLF2 consists of two β-sheets decorated with a helical cap plus several additional helices (S1B Fig). The conserved CCCH-type Zn-binding site, formed by three cysteines and one histidine, is located at the base of the globular domain. Chains B, D, and H are well resolved and lack only the N termini (residues 66-77, 66-76, and 66-77, respectively); chain D also lacks the C terminus, residues 317-318. In contrast, chains F and J have several unresolved regions (S2B Fig). The five BFLF2 copies have several structural differences. First, several loops, such as the β2/β3 loop, adopt different conformations ( Fig 1E). Further, depending on the BFLF2 chain analyzed, a structural element may be missing, or its length may differ. For example, chain F is missing the β3 strand that is found in chains B, D, H, and J whereas chains B, D, F have a longer β1 (S2B Fig). Finally, chain D has a short 3 10 helix preceding helix α1 (S2B Fig).
The most notable difference, however, is the distinct relative orientations of the globular domain and the hook in the five BFLF2 copies (Fig 1E and 1F) that result in somewhat different BFLF2/BFRF1 interfaces across the five EBV heterodimers as described below. Due to these conformational differences, BFLF2 chains superpose onto one another with RMSDs ranging from 1.25 Å to 2.41 Å (S2 Table).

The BFRF1/BFLF2 interactions and conformational dynamics in the EBV NEC
BFLF2 interacts with BFRF1 at two interfaces, the so-called "primary" or hook-in-groove interface that involves the BFLF2 hook, and the "secondary" or globular interface, which involves the globular domain of BFLF2. The hook-in-groove interface was previously visualized in the structure of BFRF1 fused to the BFLF2 hook by a flexible linker with an interface of 1,569 Å 2 and described in detail [42]. Therefore, we will not elaborate on it except to note that the interface varies from 1,403 to 1,727 Å 2 across the 5 heterodimers (S2A and S2B Fig and S5 Table) and contributes roughly 80% of the total BFRF1/BFLF2 interface. The differences in the area buried at the hook-and-groove interface are due mainly to the number of interacting By contrast, the globular domain interface contributes approximately 20% of the total BFRF1/BFLF2 interface, burying 333 to 464 Å 2 across the 5 NEC copies (S5 Table). Across the five EBV NEC heterodimers, the globular domains of BFLF2 adopt distinct orientations relative to BFRF1 (Fig 3A) where they are either more or less tilted towards the BFRF1/hook module, as measured by the angles between D10 BFRF1 /P98 BFRF1 /D287 BFLF2 . The tilt towards the hook increases, with a concomitant decrease in the angle in B<F<H<D (104˚>100˚>94˚> 89˚) (Fig 3A). Chain I could not be analyzed due to D287 BFLF2 being unresolved, but the overall alignment is very similar to chains B and F ( Fig 3A). Major interactions are polar and include a salt bridge D115 BFRF1 -R128 BFLF2 , observed in all NEC copies except in the poorly resolved NEC5, and a hydrogen bond between T158 BFRF1 and E250 BFLF2 in NEC1 and NEC4 ( Fig 3B).
We attribute the differences in tilt across the EBV NEC heterodimers to the differences in the interactions at the globular domain interface, specifically, the presence or the absence of the distal hydrogen bond between T158 BFRF1 and E250 BFLF2 (Fig 3B). We hypothesize that in NEC1, this hydrogen bond keeps BFLF2 anchored to BFRF1 and in an upright position ( Fig  3B). By contrast, in NEC2, the lack of the hydrogen bond ( Fig 3B) allows the globular domain of BFLF2 to tilt away from the BFLF2/BFRF1 interface and towards the BFLF2 hook. This tilted orientation is stabilized by the crystal contacts between chain D and chain H of an NEC symmetry mate.
Analysis of the heterodimeric interfaces revealed no conserved identical residues across EBV NEC and its homologs.

NEC/NEC interactions
Oligomerization of the NEC into the hexagonal "honeycomb" lattice is a major driving force of membrane budding (reviewed in [8]). This lattice has been visualized in the cryo-ET reconstructions of the HSV-1 NEC coats formed in vitro [25,43] and in infected cells [30] as well as in PRV NEC coats formed in cells overexpressing the NEC [29]. Additionally, this lattice was also observed in HSV-1 NEC and HCMV NEC crystals (Fig 4A and 4B). Structure-guided mutagenesis of hexameric and inter-hexameric interfaces in the HSV-1 NEC hexagonal lattice established its significance for budding in vitro [25,27] and nuclear egress in infected cells [31,32].
The EBV NEC did not form hexamers in the crystals, however. Instead, four EBV NEC copies formed two dimers, NEC1/NEC4 and NEC2/NEC3 (Fig 4C and 4D). The dimeric interfaces in EBV NEC crystals are very similar to the hexameric interfaces in the crystals of HSV-1 and HCMV NEC, and when comparing them, we refer to them henceforth as oligomeric. The dimeric interfaces in EBV NEC are larger (759 Å 2 for NEC1/NEC4 and 967 Å 2 for NEC2/ NEC3) than the hexameric interfaces in HSV-1 NEC (613 Å 2 for NEC AB and 572 Å 2 for NEC CD ) and HCMV NEC (760 Å 2 ) (S6 Table). This could be due to a somewhat larger number of residues at the N termini of both BFRF1 and BFLF2 that contribute to the interface ( Fig  5 and S6 Table). Analysis of the oligomeric interfaces revealed only a slight enrichment for the BFLF2 interface of NEC2/NEC3 and NEC1/NEC4 (S7 Table), but no enrichment for BFRF1, possibly, because the interfaces are mainly formed by similar rather than identical residues.

EBV NEC has an intrinsic membrane budding activity
To measure the membrane budding activity of the EBV NEC, we generated and purified a recombinant soluble construct NEC228Δ15-His 8 lacking the first 15 residues of BFLF2 and residues 229-336 of BFRF1, which eliminated the transmembrane domain. In its place, a His 8 - . Polar contact residues for UL31/UL53 (yellow) and UL34/ tag was added to the C terminus of BFRF1 (Fig 6A) to tether the complex to membranes. Histidine tags-when used in conjunction with nickel-chelating lipids-act as membrane anchors [49]. Using the established in-vitro budding assay utilizing giant unilamellar vesicles (GUVs) (Fig 6B and 6C) [25,50,51], we observed that EBV NEC228Δ15-His 8 (Fig 6A) mediated budding at a level comparable to that of HSV-1 NEC220-His 8 (Fig 6D). Therefore, we conclude that the intrinsic membrane budding ability previously reported for alphaherpesviruses HSV-1 and PRV is conserved in EBV, a gammaherpesvirus.
In HSV-1 and PRV, the N-terminal membrane-proximal region (MPR) of UL31 (residues 1-50 in HSV-1), specifically, clusters of basic residues, are required for membrane budding in vitro [50] and nuclear egress in infected cells [52], respectively. To probe the importance of the putative MPR of EBV BFLF2 (residues 1-65) in budding, we generated a series of N-terminal truncations, EBV NEC228Δ24-His 8 , EBV NEC228Δ34-His 8 , and EBV NEC228Δ44-His 8 ( Fig  6A). We found, however, that only the longest truncated construct, EBV NEC228Δ15-His 8 , efficiently budded membranes (15%) whereas the rest, e.g., EBV NEC228Δ24-His 8 , did not ( Fig 6D). Therefore, residues 16-24 of BFLF2 are required for budding. The sole basic cluster in the putative BFLF2 MPR, residues R22/R23, is located within this region. We hypothesize that, by analogy with basic clusters in HSV-1 UL31 MPR [50], this dibasic motif interacts with membranes and is essential for the EBV NEC budding activity.

EBV NEC forms membrane-bound coats inside budded vesicles
The HSV-1 NEC oligomerizes into hexagonal coats during budding events in vitro [25,43] and in infected cells [30]. To investigate if the EBV NEC also forms coats during in-vitro budding, we used EBV NEC215-N31S BFLF2 side by side with HSV-1 NEC220 in membrane-budding experiments and imaged them using cryo-EM. HSV-1 NEC220 is the construct used in prior in-vitro budding experiments and oligomerizes into a hexagonal coat [25,43]. EBV NEC215-N31S BFLF2 is the construct analogous to HSV-1 NEC220 based on secondary structure alignment (S3 Fig) that contains a N31S mutation in BFLF2 introduced to reduce spontaneous proteolytic cleavage of BFLF2 during expression and purification and improve protein yield.
Each protein complex was incubated with large unilamellar vesicles (LUVs) of size and composition previously used for cryo-EM visualization of HSV-1 NEC220 budding [25]. In both EBV and HSV-1 NEC samples, we observed vesicles with thickened membranes containing internal membrane-bound coats (Fig 7A-7D). In contrast, non-budded LUVs had thinner membranes (Fig 7A and 7D). Both the EBV and HSV-1 NEC coats are~11-nm thick in side views (Fig 7B and 7D insets; side view). Our previous cryo-EM experiments with HSV-1 NEC [25,43] established that these coats are composed of a single membrane-bound NEC layer. Therefore, based on our measurements, we hypothesize that the coats formed in the presence of EBV NEC are also composed of a single membrane-bound EBV NEC layer. In both cases, instead of ILVs inside "mother" vesicles, only individual budded vesicles were observed, similarly to the original studies with HSV-1 NEC [25]. We hypothesize that the excess amount of NEC in the cryo-EM experiments resulted in multiple rounds of budding such that the membrane of the "mother" vesicle was depleted, releasing budded vesicles containing internal NEC coats.

PLOS PATHOGENS
Structure and function of EBV NEC . Interface residues were assigned by PDBePISA [46]. Only residues with 30% or more of their surface area buried at the oligomeric interface are shown, in green and mauve in UL31 and UL34 homologs, respectively.  [47] and annotated using Espript [48]. Interface residues were assigned by PDBePISA [46]. Residues with 30% or more of their surface area buried at the oligomeric interface are highlighted in green and mauve in UL31 and UL34 homologs, respectively. Interface residues mutated in this study are marked with a blue star below the sequence. https://doi.org/10.1371/journal.ppat.1010623.g005

PLOS PATHOGENS
Structure and function of EBV NEC hexagonal geometry because their appearance closely resembles the coats formed by the same HSV-1 NEC construct that had hexagonal geometry revealed by 3D averaging [25,43]. In contrast, the top views of the EBV NEC coats (Fig 7B inset, top view) did not resemble the top views of HSV-1 NEC coats (Fig 7D inset, top view). Consequently, the differences in appearance between the HSV-1 and EBV NEC coats leads us to speculate that the EBV NEC oligomerizes into coats of a different geometry, possibly, due to the structural plasticity observed in the EBV NEC crystal structure. Such conformational flexibility may influence the interactions between neighboring NECs enabling the assembly of an alternative NEC lattice. The detailed analysis of the EBV NEC coat geometry is the subject of future work.

The NEC oligomeric interface is important for budding
The dimeric NEC/NEC interfaces observed in EBV crystals closely resemble the hexameric NEC/NEC interfaces observed in crystals of HSV-1 and HCMV NEC. Mutations engineered to disrupt hexameric interfaces in HSV-1 NEC reduce budding in vitro [25,27] and nuclear

PLOS PATHOGENS
Structure and function of EBV NEC egress in infected cells [31,32], confirming the significance of oligomerization for membrane budding.
To probe the importance of the dimeric interface in EBV NEC budding, we mutated interface residues equivalent to those that form hexameric interfaces in HSV-1 NEC. Prior mutational analysis of the hexameric interface in HSV-1 NEC pinpointed the hydrophobic triad (V92 UL34 /V247 UL31 /F252 UL31 ) and a polar residue (T123 UL34 ) near the top margin of the interface along with a polar residue E37 UL34 near the bottom margin of the interface (Fig 8A), which forms a hydrogen bond with T89 UL31 of the neighboring NEC, as important for budding in vitro [25,27] and in infected cells [31,32]. Guided by these mutations, we mutated the equivalent hydrophobic "triad" residues in EBV, L87 BFRF1 , L262 BFLF2 , and F267 BFLF2 , as well as the polar N121 BFRF1 , the equivalent of T123 UL34 (Fig 8B and 8C). Residue E37 UL34 , which has and EBV (B and C) oligomeric interfaces with interacting residues in UL31 and BFLF2 (green) and UL34 and BFRF1 (pink). Middle inset shows a zoom of the total interaction interface. Distance measurements (Å) between EBV NEC S112 BFLF2 and Q43 BFRF1 are shown. Right inset shows a zoom of the hydrophobic triads (V92 UL34 /V247 UL31 /F252 UL31 in HSV-1 and L87 BFRF1 /L262 BFLF2 /F267 BFLF2 in EBV) and polar residue (T123 UL34 in HSV-1 and N121 BFRF1 in EBV) at the top margin of the interface. Crystal structure of the HSV-1 NEC homolog (rcsb pdb 4xzs) was used. All images created in PyMol [44]. (D) In-vitro budding assay. Vesicles contain Nichelating lipids to tether His 8 -tagged NEC to membranes. % budding was determined by counting the number of ILVs after addition of NEC. Background levels of ILVs in the absence of NEC, typically around 10%, were subtracted from all values. All samples contain the N31S BFLF2 mutation. Significance to 215-N31S BFLF2 -His 8 was calculated using an unpaired Student's t-test with Welch's correction (p<0.05 = � , p<0.005 = �� , p<0.0007 = ��� ). In all plots, error bars represent the standard error of the mean (68% confidence interval of the mean) for at least three individual experiments. (E) Table summarizing the levels of in-vitro budding compared to wild-type for EBV mutants and their HSV-1 counterparts. Data for HSV-1 mutants obtained from [27]. https://doi.org/10.1371/journal.ppat.1010623.g008

PLOS PATHOGENS
Structure and function of EBV NEC the strongest non-budding phenotype in HSV-1 NEC, lacks a counterpart in EBV. Instead, we mutated a nearby residue Q43 BFRF1 that is located in an analogous position and forms a hydrogen bond with S112 BFLF2 of the neighboring NEC (Fig 8B and 8C), the equivalent of T89 UL31 (Fig 5A). All mutations were introduced into the EBV NEC215-N31S BFLF2 -His 8 construct, which is identical to the construct used in cryo-EM experiment except for the presence of a Cterminal His 8 -tag in BFRF1.
We found that three out of five EBV mutations, L262W BFLF2 , F267W BFLF2 , and Q43A BFRF1 , reduced budding to <25% of the WT EBV NEC (Fig 8D and 8E). Another mutation, N121Q BFRF1 , reduced budding,~2-fold. Notably, two out of three mutations in the EBV hydrophobic "triad", F267W BFLF2 and L262F BFLF2 , significantly reduced in vitro budding, underscoring the importance of this structural element for budding in both HSV-1 and EBV. Although mutation L87W BFRF1 did not have an apparent effect on budding, we hypothesize that position 87, located in a flexible loop, can accommodate a larger tryptophan side chain.
Therefore, we hypothesize that the dimeric interfaces observed in the EBV NEC crystals are functionally relevant and represent oligomeric interfaces formed during budding. Fig 8E compares the effectiveness of analogous interface mutations in EBV and HSV-1 NEC.

Discussion
The NEC is essential for capsid nuclear egress across the three subfamilies of the Herpesviridae. But to date, most structural and mechanistic studies have focused on the NEC homologs from alpha-and betaherpesviruses leaving the NECs from gammaherpesviruses less well characterized. Here, we undertook structural and functional studies of NEC from EBV, a gammaherpesvirus. First, we demonstrated that the purified, recombinant EBV NEC vesiculates synthetic membranes in vitro and forms membrane-bound coats, which suggests that the intrinsic membrane budding ability is a conserved property of the NECs across Herpesviridae. Second, we presented the most complete crystal structure of the EBV NEC to date, which allowed us to catalog the conserved and subfamily-specific features across NEC homologs from all three subfamilies for the first time. Moreover, we presented evidence that the EBV NEC structure is intrinsically dynamic and hypothesize that such structural plasticity may be important during lattice formation. However, instead of hexamers, the EBV NEC forms dimers in the crystals, and its membrane-bound coats formed in vitro do not resemble the hexagonal coats formed by NEC from alphaherpesviruses [25,43]. The dimeric interfaces observed in the EBV NEC crystals are similar the hexameric interfaces observed in other NEC homologs, and mutations engineered to disrupt dimeric interfaces reduce budding. We hypothesize that membrane budding by the EBV NEC is driven by its oligomerization into membrane-bound coats but that its structural flexibility may enable formation of coats of a different geometry.

The intrinsic membrane budding ability is a conserved NEC property
Previous studies showed that NECs from alphaherpesviruses PRV and HSV-1 budded synthetic membranes in vitro [25,26,50,51] in the absence of any additional factors. Although the NEC homologs from other subfamilies are expected to vesiculate the nuclear envelope around the capsid during nuclear egress, the intrinsic membrane budding ability had not been formally demonstrated until the present study. By demonstrating that the purified, recombinant EBV NEC vesiculates synthetic membranes in vitro, we established the intrinsic membrane budding ability as a conserved property of NECs across the Herpesviridae subfamilies.

PLOS PATHOGENS
Structure and function of EBV NEC least, in-vitro, in EBV, only residues 1-15 can be removed without disrupting budding. We hypothesize that the differences in the required MPR length is due to the number and the location of dibasic motifs, which have been shown to be important for HSV-1 budding [50]. HSV-1 has five distinct basic clusters in the UL31 MPR whereas the EBV BFLF2 MPR only has one cluster (residues R22/R23) (S3B Fig). These findings suggest that the importance of the BFLF2 MPR and, possibly, basic clusters for budding is a conserved property of alpha-and gammaherpesvirus subfamilies.

Conserved and subfamily-specific structural features of NEC homologs
The crystal structures of the NEC homologs from the three divergent Herpesviridae subfamilies reveal remarkable structural similarities despite the relatively low sequence identity. The observed differences are limited to the length of the secondary structure elements, the length and the conformation of several loops, and the relative orientations of UL31 and UL34 homologs within the complex.
Importantly, the five EBV NEC heterodimers have notable structural differences, mainly, within BFLF2 and at the BFLF2/BFRF1 interface. The conformational differences among the five EBV NEC heterodimers (Fig 1E) are comparable to the differences among the NEC homologs (Fig 2E), demonstrating the remarkable structural plasticity of the EBV NEC.

Conformational dynamics across the NEC homologs
Although the structural plasticity observed among the crystal structures of the EBV NEC is unusual among the NEC homologs, conformational dynamics may be shared by some NEC homologs to some extent. A recent molecular dynamics (MD) simulation study of the NEC crystal structures from HSV-1, PRV, and HCMV revealed differences in the intrinsic dynamics across the homologs, with HCMV NEC being more dynamic than HSV-1 or PRV NEC [53]. Whereas the "primary" hook-in-groove interface was very rigid in all three NECs homologs throughout all simulations, the globular domain of HCMV UL50 (UL31 homolog) twisted around the vertical axis to a much greater extent. These differences were attributed to the "secondary" UL31/UL34 interface.
In HSV-1 NEC, the secondary interface contains two salt bridges, R167 UL34 -D104 UL31 (termed "proximal" due to its proximity to the UL31 hook) and R158 UL34 -D232 UL31 (termed "distal" relative to the UL31 hook) (Fig 3C). Throughout MD simulations, both salt bridges in HSV-1 were highly stable resulting in a more rigid secondary interface and domain orientation. The distal salt bridge appears to be important for function because the double mutation R158A UL34 /R161A UL34 reduced viral titers, suggesting that these charged residues are involved in NEC function [54]. The mutation does not interfere with NEC localization [54], nor does it result in a dominant-negative phenotype [31], implying that it does not block NEC/NEC interactions. By contrast, in HCMV, there is a single, proximal salt bridge E147 UL50 -R102 UL53 and a distal hydrogen bond between N139 UL50 and Q103 UL53 instead of a distal salt bridge (Fig 3C). Both the salt bridge and the hydrogen bond in HCMV were less stable than the two salt bridges in HSV-1 NEC, suggesting that the more flexible HCMV domain orientation is due to a less stable secondary interface. Importantly, when a second, distal salt bridge was introduced computationally, the HCMV NEC structure became much less dynamic in the MD simulations, which suggests that two salt bridges increase the conformational stability. While the HSV-1 and HCMV NEC heterodimers differed in domain orientation flexibility on their own, in the context of a hexagonal arrangement the domain twisting was drastically reduced [53]. The authors speculated that the increased HCMV flexibility may influence events prior to nuclear egress and hexamer formation such as interactions with NEC-associated proteins.

Structure and function of EBV NEC
The EBV NEC structure is more similar to that of HCMV because it has a proximal salt bridge, D115 BFRF1 -R128 BFLF2 , in four heterodimers (except the incomplete NEC5) but no distal salt bridges (Fig 3B). NEC1 and NEC4 also have a distal hydrogen bond between T158 BFRF1 and E250 BFLF2 . The structural dynamics of the EBV NEC, observed experimentally, correlates with the structural dynamics of the HCMV NEC heterodimer revealed in the MD simulations. However, because these structural dynamics are greatly reduced once HCMV NEC is arranged into a hexagon, we hypothesize that the conformational plasticity of EBV NEC may influence its oligomerization into a coat. Further studies are required to understand the influence of EBV NEC structural flexibility on coat formation.

The ability of the NEC to oligomerize is conserved
The ability of the EBV NEC to oligomerize is supported by our cryo-EM images, the EBV NEC dimers observed in the crystals, and by the similarities between the dimeric interfaces observed in the EBV NEC crystals and the hexameric interfaces observed in HSV-1 NEC [27] and, to a lesser extent, HCMV NEC [39]. Moreover, mutations engineered to disrupt the dimeric interface reduce budding to the extent similar to interface mutations in HSV-1 NEC [27]. Comparisons revealed conserved structural elements across these oligomeric interfaces, notably, a hydrophobic triad at the membrane-distal margin of the interface. Mutations designed to increase the size of the hydrophobic side chain in the triad so as to disrupt oligomerization reduced HSV-1 budding in vitro (F252Y UL31 , V247F UL31 , and V92F UL34 ) [27], and in infected cells (V247F UL31 ) [32]. We found that two out of three corresponding mutations in EBV NEC, F267W BFLF2 and L262F BFLF2 , significantly reduced in vitro budding, stressing the importance of the hydrophobic triad at the oligomeric interface for budding in both HSV-1 and EBV. We propose that the dimeric interfaces observed in the EBV NEC crystals are functionally relevant and represent interfaces formed during budding.
Previous work established that the NECs from alphaherpesviruses PRV and HSV-1 oligomerize into hexagonal coats on the inner surface of budded vesicles formed in vitro (HSV-1) [25,43], in uninfected cells overexpressing the NEC (PRV) [29], and in perinuclear enveloped vesicles purified from infected cells (HSV-1) [30]. Despite the predominance of hexagonal coats, both NEC pentamers [43] and heptamers [30] have also been observed for HSV-1 under certain conditions suggesting HSV-1 NEC can arrange into alternative oligomeric assemblies. Given that a purely hexagonal arrangement cannot yield a closed sphere, it was proposed that incorporation of pentamers or hexamers into a hexameric NEC coat could generate curvature. A similar strategy is utilized by the poxvirus protein, D13L, which forms a curved lattice composed of mainly hexamers, incorporated with pentamers and heptamers [55]. It remains unknown exactly how HSV-1 NEC curvature is achieved.
Given structural similarities between the NEC homologs, we anticipated that EBV NEC would form hexamers in the crystals and hexagonal membrane-bound coats during budding in-vitro. Instead, we found that EBV NEC formed dimers in the crystals and that its membrane-bound coats appeared different than the HSV-1 NEC coats. We hypothesize that the high degree of structural plasticity observed among the five EBV NEC heterodimers in the crystals could potentially influence NEC oligomerization and coat geometry. One possibility is that the EBV NEC plasticity results in the formation a pseudo-hexagonal lattice containing multiple irregular defects, reminiscent of the immature hexameric Gag lattice of HIV capsids [56,57]. As hypothesized for HSV-1 NEC, such irregular defects within the EBV NEC coat could serve to accommodate curvature around the capsid. An alternative possibility is that the EBV NEC coats have an alternative, previously unobserved NEC arrangement, e.g., stacks of NEC dimers.

Structure and function of EBV NEC
We propose that membrane budding by the EBV NEC is driven by its oligomerization into membrane-bound coats but that its structural flexibility enables formation of coats of a different geometry. Future high-resolution cryo-ET reconstructions will uncover the precise structural arrangement of EBV NEC within the membrane-bound coats.
Fractions containing EBV NEC were concentrated to~5 mg/mL using Amicon Ultra-4 50 kDa cutoff centrifugal filter units (EMD Millipore) and stored at -80˚C to avoid aggregation and degradation observed during storage at 4˚C. Protein concentration was determined by absorbance measurements at 280 nm using theoretical extinction coefficient of 17,880 calculated using ProtParam (https://web.expasy.org/protparam/). The typical yield was~1.5 mg per L of TB culture. Due to different isoelectric points for numerous EBV NEC constructs, pH of the buffers was adjusted accordingly (see S9 Table). Lysis buffer always contained 50 mM buffer, 500 mM NaCl, 0.5 mM TCEP, 10% glycerol and gel-filtration buffer always contained the final concentration of 20 mM buffer, 100 mM NaCl and 0.5 mM TCEP.
HSV-1 NEC220 was expressed and purified with the same techniques as EBV NEC215-N31S BFLF2 . The buffers can be found in S9 Table. Crystallization and data collection Crystals of native EBV NEC195Δ65 were grown by vapor diffusion at 20˚C in hanging drops containing 2 μL protein in gel filtration buffer (20 mM HEPES pH 8.0, 100 mM NaCl, 0.5 mM TCEP) equilibrated over 500 μL reservoir solution (25% PEG 3350, 0.1 M Tris-HCl pH 8.5, 0.2 M Li 2 SO 4 ). No reservoir solution was added to the drops. The crystals were identified using an automated microscope [59,60]. Crystals would not grow in a 1:1 hanging drop of protein:reservoir solution. Tetragonal crystals appeared after 1 day and grew to their final size after 2 days. Crystals were cryopreserved using ethylene glycol in a step-wise cryoprotectant transfer protocol [61]. Briefly, crystals were harvested with a loop and transferred into a 2 μL drop containing the harvesting buffer (1.5x gel filtration buffer and no ethylene glycol). Then, 2 μL of the harvesting buffer plus 5% ethylene glycol was added on one side and then 2 μL was removed from the other side, taking care to leave the crystal in the drop. This process was repeated sequentially with the harvesting buffer plus 10%, 15%, 20%, 25%, 30%, 35%, 40% ethylene glycol. The crystal was then flash frozen in liquid nitrogen.
Initially, native diffraction data were collected at the Advanced Photon Source at Argonne National Labs at beamline 24ID-E at the wavelength of 0.979180 Å at 100 K and processed to 3.93 Å using XDS [62] as implemented in the Northeastern Collaborative Access Team (NE-CAT) rapid automated processing of data (RAPD) software (https://rapd.nec.aps.anl. gov). Crystals took the tetragonal space group P4 1/3 2 1 2 with unit cell dimensions a = b = 238.698 Å, c = 138.074 Å and α = β = γ = 90˚. Molecular replacement as implemented in Phaser [63] using the crystal structures of NEC homologs from HSV-1, PRV, or HCMV (RCSB PDB 4zxs, 4z3u, or 5d5n, respectively) did not yield any solution. Only the structure of the EBV NEC "base" (RCBS PDB 6t3z), composed of EBV BFRF1 1-192 bound to the "hook" fragment of BFLF2 59-87, yielded a correct solution for four copies of the EBV NEC base in space group P4 3 2 1 2 (LLG = 929). However, no connected density was observed for the globular domain of BFLF2.
All known NEC structures contain a Zn ion coordinated by four strictly conserved residues within the UL31 homologs, three cysteines and one histidine. Therefore, we hypothesized that the EBV NEC also contained a bound Zn that could be used for structure determination by single-wavelength anomalous dispersion (SAD). Fluorescence scans on an EBV NEC crystal at the APS beamline 24ID-C confirmed the presence of Zn. SAD data were collected at the wavelength of 1.28215 Å (Zn) at 100 K, and two data sets from different crystals were merged using RAPD and processed together to 3.97 Å. Four high-occupancy Zn sites were located using the SHELX macromolecular substructure solution program [64], and the Phenix AutoSol program [65] was used to calculate experimental phases and generate a preliminary electron density map, within the RAPD SAD pipeline. However, the electron density did not permit either autotracing in Autosol or manual tracing.
Therefore, the structure of the EBV NEC was ultimately solved by a combination of molecular replacement and Zn SAD. First, phases calculated from the molecular replacement solution obtained using the structure of the EBV NEC base (RCBS PDB 6t3z) were used to locate the four Zn sites in the SAD dataset, in Autosol [65]. Each Zn site was located near a base where it would be predicted to be based on the expected BFLF2 location. A density modified map revealed the density for the three copies of BFLF2. Due to the poor density, autotracing was not feasible. Therefore, the structure of the globular core of HCMV UL53 (a homolog of EBV BFLF2 and HSV1-UL31), from RSCB PDB 5d5n, was manually placed into the density modified map and underwent extensive rebuilding in Coot [66]. The initial model was refined against the SAD data in phenix.refine [65] to 3.97 Å. Prior to refinement, 5% of all data were set aside for cross-validation. After several cycles of refinement and model rebuilding, the density for the 4 th copy of BFLF2 became interpretable, which allowed its manual placement. Surprisingly, after additional rounds of refinement and rebuilding, the density for the 5 th copy of EBV NEC became evident, and the 5 th copy was built into it.
Model refinement included gradient minimization refinement of xyz coordinates and individual thermal parameters with optimization of X-ray/stereochemistry and X-ray/ADP weights, and secondary structure restraints. Rigid-body refinement was employed during early stages of refinement. Experimental phase restraints were used until the late stages of refinement. Electron density maps were sharpened. The final model encompasses five polypeptide chains of BFRF1 (A, C, E, G, and I), five polypeptide chains of BFLF2 (B, D, F, H, and J), and five Zn ions. Side chains were modeled into the available electron density. When electron density for the side chains was missing, the most common rotamer was used. According to Mol-Probity [67] as implemented in phenix.refine, 92.61% of residues lie in the most favored regions of the Ramachandran plot and 6.29% lie in the additionally allowed regions. The overall fitting of the model to the electron density yielded an R-free of 0.308. The following residues are resolved in each chain: Chain A,

Liposome preparation
Liposomes were prepared as described previously [25]. Briefly, lipids were mixed in a molar ratio of 58% POPC/11% POPE/9% POPA/9% POPS/5% cholesterol/5% DGS-NTA/3% POPE Atto594. 5 μL was then spread on the surface of an ITO-covered slide and vacuum desiccated for 30 minutes. Once dry, a vacuum-greased O-ring was placed around the lipid mixture and the VesiclePrep Pro (Nanion Technologies) was used to produce an AC field (sinusoidal wave function with a frequency of 8Hz and amplitude 2V) before adding 270 μL of lipid swelling buffer (300mM sucrose dissolved in 5 mM Na-HEPES, pH 7.5). A second ITO-covered slide was then placed to cover the lipid/buffer mixture after 3 minutes followed by a 2-hour swell and a 5-minute fall step. GUVs were used immediately and diluted 1/20 with 20 mM Tris pH 8.5, 100 mM NaCl, 0.5 mM TCEP.

GUV budding assay
Fluorescently labeled GUVs were co-incubated with the soluble NEC and the membrane impermeable dye, Cascade Blue Hydrazide (ThermoFisher Scientific) for 3 minutes. The GUVs contained 18% negatively charged lipids (58% POPC/11% POPE/9% POPA/9% POPS/ 5% cholesterol/5% DGS-NTA/3% POPE Atto594), which resembles the INM of uninfected cells [68,69]. The nuclear membrane composition during HSV-1 infection is unknown but may differ from that of uninfected cells. Budding events display as intraluminal vesicles (ILVs) containing Cascade Blue within the GUVs (Fig 6C). 5 μL of the above GUV composition and 2 μL Cascade Blue Hydrazide were mixed with a final concentration of 1.5 μM NEC for a total volume of 100 μL. Each sample was visualized using a Leica SP8 confocal microscope. Background levels of ILVs, counted in the absence of NEC, were subtracted from counts of ILVs in the presence of NEC. Background levels are typically around 10%. Experiments were performed with at least 3 technical replicates and at least 3 biological replicates. The standard error of the mean is reported from at least three individual experiments. Data was plotted using GraphPad Prism 9.0.
BFLF2 (rcsb pdb 7t7i) and homologs from HCMV UL53 (rcsb pdb 5d5n), HSV-1 UL31 (rcsb pdb 4xzs), and PRV UL31 (rcsb pdb 4z3u). Secondary structure assignments were obtained from DSSP [73]. Secondary structures are color coded, α-helices (light blue) and β-strands (dark blue). All images created in PyMol [44]. (TIF) S2 Fig. Sequence alignment of EBV BFRF1 and BFLF1. Alignment of sequences for BFRF1 (A) and BFLF2 (B). Secondary structure from the crystal structure is shown above sequence alignment blocks. Vertical black arrows denote EBV crystal construct boundaries. Unresolved residues are shown in grey boxes. Residues at the arm-hook interaction are colored, BFRF1 (cyan) and BFLF2 (red). Residues at the globular interface are colored, BFRF1 (green) and BFLF2 (yellow). Alignment made with Clustal Omega [47] and annotated using Espript [48]. Interface analysis done with PDBePISA [46]. (TIF) S3 Fig. Sequence alignment of EBV and homologous NECs. Alignment of sequences for BFRF1 and homologs (A) and BFLF2 and homologs (B). Secondary structures from crystal structures are shown above sequence alignment blocks. Vertical black arrows denote EBV crystal construct boundaries. Alignment made with Clustal Omega [47] and annotated using Espript [48]. Interface analysis done with PDBePISA [46]. Residues at the arm-hook interaction are colored, BFRF1 homologs (cyan) and BFLF2 homologs (red). Residues at the globular interface are colored, BFRF1 homologs (green) and BFLF2 homologs (yellow). Secondary structures from HCMV (rcsb pdb 5d5n), HSV-1 (rcsb pdb 4xzs), and PRV (rcsb pdb 4z3u). (TIF) S1 Table. Resolved residues for each individual EBV BFRF1 (top) and BFLF2 (bottom) crystal structure chains. For each EBV chain, the boundaries of resolved residues and % resolved residues are listed. (DOCX) S2 Table. Structural alignments of the five EBV NEC heterodimers in the asymmetric unit. Heterodimers (NEC) and individual BFRF1 or BFLF2 chains were aligned. For each EBV NEC or chain, RMSD (Å) is listed followed by the number of aligned residues in parentheses. In all cases, "SSM superpose" in WinCoot [66] was used to carry out the structure alignments and calculate RMSDs, except for those denoted by � in which case "LSQ Superpose" command was used. (DOCX) S3 Table. Sequence conservation among homologous herpesvirus NEC components. Clustal Omega [47] was used for sequence alignment and calculations of % identity. The following UniProtKB IDs were used: HSV-1 UL34 (P10218), HSV-1 UL31 (P10215), PRV UL34 (G3G8R3), PRV UL31 (G3G955), HCMV UL50 (P16791), HCMV UL53 (P16794), EBV BFRF1 (P03185), and EBV BFLF2 (P0CK47). (DOCX) S4 Table. Structural alignments of the EBV NEC to its homologs in HSV-1, PRV, and HCMV. For each EBV NEC or chain, RMSD (Å) is listed followed by the number of residues aligned in parentheses. In all cases, "SSM Superpose" in WinCoot [66] was used to carry out the structure alignments and calculate RMSDs, except for those denoted by � in which case "LSQ Superpose" command was used. RMSD and residues aligned comparisons for BFLF2 chains B-J are not available for EBV due to the presence of only 32 residues of BFLF2 in rcsb pdb 6t3z [42]. Crystal structures of EBV BFRF1 (rcsb pdb 6t3z) and NEC homologs from HSV-1 (rcsb pdb 4xzs), PRV (rcsb pdb 4z3u and 5e8c), and HCMV (rcsb pdb 5d5n and 5dob) were used. (DOCX) S5 Table. Interface areas for the globular, hook-in-groove, and total heterodimeric interfaces for the EBV BFLF2:BFRF1 interaction. PDBePISA [46] analysis of the interface area between the globular domains, hook-in-groove and total heterodimeric interface. The globular interface area was obtained by deleting the arm-hook portion of BFLF2/HSV-1 UL31/PRV UL31/HCMV UL53 from files before analysis ending at V111/K88/K55/G88, respectively. The arm-hook interaction was obtained by subtracting the globular interface area from the total heterodimeric interaction. Total heterodimeric interface area was obtained by inputting the entire crystal structure. Crystal structures the NECs homologs from HSV-1 (rcsb pdb 4xzs), PRV (rcsb pdb 4z3u and 5e8c), and HCMV (rcsb pdb 5d5n and 5dob) were used. (DOCX) S6 Table. Interface areas for the oligomeric interfaces of EBV NEC and all applicable homologous NEC crystal structures. PDBePISA [46] analysis of the solvent accessible area buried at the oligomeric interface. Crystal structures of the NEC homologs from HSV-1 (rcsb pdb 4xzs) and HCMV (rcsb pdb 5d5n) were used. (DOCX) S7 Table. Conservation of residues at the oligomeric interfaces. Interface residues in the two EBV NEC dimers in the crystal structure were mapped using PDBePISA [46] analysis (Fig 4). Interface residues that are identical across in BFLF2 and its homologs or BFRF1 and its homologs from EBV, HSV-1, and HCMV (Identical interface residues) were divided by the total number of interface residues in the respective dimer (Total interface residues) to yield % Identical Interface Residues. % Total Sequence Conservation is the % identical residues in BFLF2 or BFRF1.  Table. Protein purification buffers. Due to different isoelectric points (PIs) of EBV NEC constructs, pH of buffers used for purification was adjusted to be~1 pH unit away from the pI. Cells were lysed in lysis buffer (50 mM indicated buffer, 500 mM NaCl, 0.5 mM TCEP, 10% glycerol). Affinity tag removal changed the PIs, so buffer with a different pH was used during and after PreScission protease cleavage. After affinity tag removal, protein was further purified in gel filtration buffer (20 mM indicated buffer, 100 mM NaCl, 0.5 mM TCEP). Salt concentration varied for ion exchange chromatography (see Materials and Methods). (DOCX) Martin Hunter (University of Massachusetts College of Engineering) for assistance with fluorescence microscopy experiments. We thank Andrew Bohm (Tufts University School of Medicine) for use of his automated microscope for protein crystal screening. We also thank the staff at the NE-CAT (Advanced Photon Source) for help with collecting X-ray diffraction data,