Crystal Structure of the Gamma-2 Herpesvirus LANA DNA Binding Domain Identifies Charged Surface Residues Which Impact Viral Latency

Latency-associated nuclear antigen (LANA) mediates γ2-herpesvirus genome persistence and regulates transcription. We describe the crystal structure of the murine gammaherpesvirus-68 LANA C-terminal domain at 2.2 Å resolution. The structure reveals an alpha-beta fold that assembles as a dimer, reminiscent of Epstein-Barr virus EBNA1. A predicted DNA binding surface is present and opposite this interface is a positive electrostatic patch. Targeted DNA recognition substitutions eliminated DNA binding, while certain charged patch mutations reduced bromodomain protein, BRD4, binding. Virus containing LANA abolished for DNA binding was incapable of viable latent infection in mice. Virus with mutations at the charged patch periphery exhibited substantial deficiency in expansion of latent infection, while central region substitutions had little effect. This deficiency was independent of BRD4. These results elucidate the LANA DNA binding domain structure and reveal a unique charged region that exerts a critical role in viral latent infection, likely acting through a host cell protein(s).


Introduction
Herpesviruses are ubiquitous viruses, which infect many mammalian species and are a leading cause of human viral disease. There are two known human viruses from the gamma herpesvirus sub-family. The Epstein-Barr virus (EBV), a gamma-1 herpesvirus (Lymphocryptovirus), and the Kaposi's sarcomaassociated herpesvirus (KSHV), a gamma-2 herpesvirus (Rhadinovirus). KSHV has an etiologic role in Kaposi's Sarcoma, the leading AIDS malignancy, as well as primary effusion lymphoma and multicentric Castleman's disease. EBV is also associated with human cancers. These human viruses do not readily infect small laboratory animals. The murine gammaherpesvirus 68 (MHV-68 or murid herpesvirus 4), a rhadinovirus that was isolated from naturally infected rodents, is structurally and functionally related to human gamma-herpesviruses and readily infects mice, thus providing a mouse model for the investigation of gammaherpesvirus pathogenesis [1][2][3].
A key feature of herpesvirus infection is their lifelong persistence in the host in the form of latency. During latent infection, viral genomes persist as multi-copy, circularized, extrachromosomal episomes (plasmids). Only a small subset of viral genes is expressed during latency. In the case of gamma-herpesvirus sub-family members, latent infection is predominantly established in B lymphocytes. Intransal infection of mice with MHV-68 causes an acute self-limiting respiratory infection, followed by the establishment of splenic latency, which peaks 14 days after infection [4,5]. Splenic germinal centre (GC) B cells have a key role both in the establishment and maintenance of viral latency [6]. Like EBV [7], MHV-68 establishes latency predominantly in activated germinal centre B cells. This strategy favours physiological access of the virus to the memory B-cell pool, the major reservoir of long-term latency.
KSHV latency-associated nuclear antigen (kLANA), a 1,162 amino acid protein encoded by ORF73 of the viral genome, is the predominant gene expressed in latent infection. LANA mediates persistence of episomal DNA and thus is required for long-term maintenance of viral genomes in dividing cells, a role central to viral latency. LANA promotes episomal replication and segregation to progeny nuclei after mitosis. The C-terminal region of kLANA harbours a DNA-binding domain (DBD) that selfassociates and binds to terminal repeat (TR) DNA sequences of the viral episome [8][9][10][11][12][13][14]. LANA cooperatively binds to two sites within the TR DNA, LBS1 and LBS2. LBS1 is a high affinity site capable of facilitating the cooperative binding of LANA to LBS2. Both sites contribute to the ability of LANA to suppress transcription and to facilitate DNA replication [10]. An Nterminal region of kLANA is able to attach to chromosomes through binding to histones H2A and H2B [15]. Together, these binding properties allow kLANA to tether the viral genome to chromosomes during mitosis thus ensuring segregation of the episome to daughter nuclei [16,17]. Association of viral genomes with host mitotic chromosomes via a virus-encoded protein is a strategy employed by a number of different latent DNA viruses. Similar to the KSHV LANA, EBV EBNA1 and papillomavirus E2 proteins play a role in viral genome maintenance [18][19][20][21][22].
All gamma-2 herpesviruses encode homologs of KSHV LANA. The ORF73 from MHV-68 encodes a much smaller, 314 amino acid, 50 kDa nuclear protein (termed mLANA thereafter) which lacks the extensive internal acidic and glutamine-rich repeat region of kLANA. The C-terminal region of mLANA, comprising amino acid residues 140 to 263, has amino acid similarity to the kLANA DNA-binding domain. Similar to kLANA, mLANA was recently shown to act on TR elements of the MHV-68 genome to mediate episome maintenance and to associate with mitotic chromosomes [23], and mLANA DNA-binding sites have been identified within MHV-68 TR DNA [24]. In a mouse model of infection, latencyassociated nuclear antigen (LANA) protein is selectively expressed in infected GC B cells [25]. Recombinant viruses that do not express mLANA [26,27], or bear disruptive mutations in the predicted DNA-binding domain fail to establish viable latency [24]. mLANA therefore performs functions equivalent to those of kLANA regarding episomal maintenance and is essential for establishment of viable latency in vivo.
LANA proteins are also modulators of transcription from a variety of cellular and viral promoters and affect cellular growth [28,29]. Both kLANA and mLANA are able to regulate transcription through E3-ubiquitin ligase activity [30][31][32]. mLANA promotes the polyubiquitination and subsequent proteasomal-dependent nuclear degradation of host nuclear factor-kappa B (NF-kB) [31]. The mechanism involves the assembly of an Elongin C/Cullin5/SOCS (suppressors of cytokine signalling)-like complex, mediated by an unconventional viral SOCS-box motif (residues 199-206), homologous to the kLANA Cul5 box, present in the C-terminal domain of mLANA [30,31]. Another motif also within the C-terminal region of mLANA (QAKKLK motif, residues 226-231) has been shown to bind to several members of the BET (Bromodomain and Extra Terminal domain) family of proteins, including BRD2, BRD3 and BRD4 [33,34], that interact with acetylated histones. mLANA interaction with BET proteins leads to activation of cell cycle promoters.
Secondary structure predictions have suggested that the Cterminal regions of both mLANA and kLANA have similarity to the X-ray structure of the DNA-binding domain of the EBV nuclear antigen 1 protein (EBNA1) [35]. As a first step towards understanding the mechanism by which the C-terminal region of mLANA exerts its functions, we have determined the crystal structure of this mLANA domain (amino acids 140-272). mLANA 140-272 forms a b-barrel induced dimer and has an overall fold with an a-helix arrangement similar to that of the EBNA1 and E2 latency protein structures [36,37]. However, the structure reveals unique features that exert important effects on viral persistence in an animal model.

Results
X-ray crystal structure of C-terminal mLANA reveals an alpha+beta fold that assembles to form a dimer The structure of the C-terminal domain of mLANA was determined at 2.2 Å resolution. Data collection and refinement statistics are reported in Table 1. The initial structure was solved using the coordinates of the EBV EBNA1 dimer, which has both functional and predicted secondary structure homology, as a molecular replacement model [36]. The crystallized domain contained residues 140-272, designated mLANA 140-272 ( Figure  S1). The refined electron density map was readily interpretable with exceptions only at the termini and loop regions. The tertiary structure of mLANA 140-272 exhibits an a+b ferredoxin-like fold ( Figure 1A) which assembles to form a dimeric eight-stranded antiparallel b-barrel that is central to its functional architecture. This dimer can be viewed from opposite sides of the b-barrel core which for clarity is termed the ventral (top) and the dorsal sides (bottom) ( Figure 1B). Each monomer contributes an anti-parallel 4-stranded b-sheet ''half-barrel'' from which helix a2 and a3 pack onto the plaited sheet with their axis parallel to the strands of the sheet. The core of the barrel is occupied by large hydrophobic surface ( Figure  S2) that contributes to the stability of mLANA ( Figure 1C). Size exclusion analysis confirms that mLANA 140-272 is also a dimer in solution ( Figure S2B and C). When the dimer interface was disrupted by removal of residues 254-261, which lie below the b2-b3 loop of the adjacent monomer, the melting temperature of mLANA 140-253 decreased drastically (T m 43uC), indicating the importance of these residues for stability ( Figure 1D). The flanking helix a1 caps this helical arrangement and lies perpendicular to the central b-barrel. Helix a2, situated on the ventral side, is postulated to be the DNA recognition helix (see below).
LANA cooperatively binds to two adjacent sites within TR DNA, similar to the EBV DS element required for EBNA1 binding [38], suggesting a dimer-dimer interaction [39]. Analysis of the mLANA crystal packing shows that dimers pack so that the DNA-binding flanking helix a1 in each dimer is in position to interact with both the flanking helix a1 and helix a3 from the adjacent dimer ( Figure 1E). Two phosphate ions on the ventral surface contribute and coordinate the binding of the a1-a39

Author Summary
Herpesviruses establish life-long latent infections. During latency, gammaherpesviruses, such as Kaposi's sarcomaassociated herpesvirus (KSHV), persist as multicopy, circularized genomes in the cell nucleus and express a small subset of viral genes. KSHV latency-associated nuclear antigen (LANA) is the predominant gene expressed during latent infection. C-terminal LANA binds KSHV terminal repeat (TR) DNA to mediate DNA replication. TR DNA binding also allows tethering of the viral genome to mitotic chromosomes to mediate DNA segregation to daughter nuclei. We describe here the crystal structure of the murine gammaherpesvirus 68 LANA DNA binding domain, which is homologous to that of KSHV LANA. The structure revealed a dimer and we identified residues involved in the interaction with viral DNA. Mutation of these residues abolished DNA binding and viable latency establishment in a mouse model of infection. We also identified a positively charged patch on the dimer surface opposite to the DNA binding region and found this patch exerts an important role in the virus's ability to expand latent infection in vivo. This work elucidates the structure of the LANA DNA binding domain and identifies a novel surface feature that is critical for viral latent infection, likely by acting through a host cell protein.
dimer-dimer interactions and suggest at least 2 or 3 nucleotides may span the spacer region between the LBS-1 and LBS-2 sites ( Figure 1F). Consistent with this finding, size exclusion chromatography demonstrates that mLANA 140-314 forms tetramers in solution ( Figure S2B). In addition, mLANA 140-272 dimer forms a tetramer to cooperatively bind adjacent mTR DNA binding sites (described below).
Despite functional conservation, mLANA and EBNA1 proteins do not exhibit amino acid similarity ( Figure 2A). However, a PHYRE2 program [40] search predicts mLANA secondary structure homology between the C-terminal domains of mLANA and EBNA1 (not shown), similar to other predictions [35]. A superposition of both monomer structures is shown in Figure 2C.
Protruding perpendicular to the ventral face of mLANA lies a partially disordered b2-b3 loop (199-215) ( Figure 2B) that harbours a motif required for the interaction of mLANA with Elongin B/C (EloBC)-cullin-SOCS box (ECS)-type E3 ubiquitin ligase, resulting in the polyubiquitination of NFkB [31]. Mutations at the base of this loop ( 199 VSCLPLVP 206 , underlined residues mutated to Alanine) disrupt binding of mLANA to Elongin C and consequently mLANA's E3-ubiquitin ligase activity. A recombinant virus containing these mutations fails to establish latency in vivo [31].
Ventral face of mLANA -phosphate binding traces the site of DNA interaction The structural basis for DNA recognition can be inferred from the observed phosphate binding pattern. In total, seven phosphates ions were located at the surface of mLANA. Five of these are situated on the ventral face ( Figure 2D-F) while the remaining two cap the dorsal side of helix a1. Therefore, we consider the ventral face as the putative DNA binding site of mLANA. We modelled mLANA-DNA on the EBNA1-DNA structure (PDB ID: 1B3T) and the mLANA structure accommodated the DNA well ( Figure 2G and Figure S3).
Mutation of predicted DNA phosphate contacts disrupts DNA binding to TR DNA To directly assess mLANA DNA binding, we used an electrophoretic mobility shift assay (EMSA) to assess the ability of mLANA to bind DNA. DNAse I footprinting analysis with recombinant mLANA has shown three adjacent protected regions within the mTR sequence [24], ( Figure 3A, green and blue regions). Two overlapping probes, mLBS1 or mLBS2, each containing a portion of the protected sequence, or control, overlapping adjacent sequence were used to assess mLANA binding ( Figure 3B). After incubation with mLANA, a shifted complex was only observed after incubation with mLBS1, and not with mLBS2 or control adjacent sequence ( Figure 3B). The mLBS1 complex was supershifted with anti-FLAG antibody and was effectively competed with excess, unlabelled mLBS1 or mLBS1-2 competitor (containing the entire DNAse I protected region), but not unlabelled mLBS2 ( Figure 3B). Despite the inability of mLANA to shift mLBS2 probe, size exclusion chromatography analysis demonstrated that mLANA 140-272 bound adjacent mLBS1 and mLBS2 sites in mLBS1-2 oligonucleotide as a tetramer ( Figure S6), consistent with cooperative binding to both sites. Therefore, similar to KSHV LANA DNA binding [10], mLANA cooperatively binds to adjacent high and low affinity TR DNA binding sites.
We then constructed mutations in the predicted mLANA recognition helix intended to decrease interactions with its cognate DNA. mLANA H186D/K187E contains substitutions predicted to disrupt phosphate interactions and abolish DNA binding, and this mutant was unable to complex with mLBS1 ( Figure 3C). In contrast, mLANA K224A/K228A/K229A, mLANA K224E/K228E/K229E, mLANA K251A/K253A, and mLANA K251E/K253E, each contain substitutions within a dorsal positively charged patch (discussed below) opposite the predicted DNA interface, and each mLANA gel shifted the mLBS1 probe ( Figure 3C . Therefore, as predicted by the mLANA DNA binding domain structure, mLANA H186D/ K187E substitutions abolished DNA binding, but DNA binding was preserved with dorsal mutations. The finding that dorsal alanine substitutions of lysine reduced DNA binding, while glutamate substitutions did not, suggests that substituting non charged residues may have subtle effects on DNA binding domain folding. We characterised mLANA TR DNA binding by surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) to determine the dissociation constant (K D ) for mLBS1. The K D for wild type and H186D/K187E were determined by   Figure S1). (B) A representation of the mLANA 140-272 quaternary structure and its dimeric conformation viewed along the non-crystallographic axis and illustrates the a-helical arrangement around the b-barrel core. The two chains are coloured in blue (chain A) and green (chain B). (C) Cross section of the dimer interface with coloured surface properties. A large area (,1440 Å 2 ) constituting the hydrophobic core is displayed in yellow with the hydrogen bonding residues displayed in blue. Deleting residues P254-P259 in red titrating mLANA 124-272 against immobilized mLBS1 ( Figure 3E and F). The K D 's were 169 nM and 960 nM for wild type and mLANA H186D/K187E, respectively ( Table 2). Thermodynamics binding parameters in solution measured by ITC using the more soluble mLANA 124-314 fitted well to a one-site model, indicating one distinct mLANA binding site ( Figure 3G), and provided a calculated K D of 66 nM ( Table 3). The discrepancy between the SPR and ITC dissociation constants was due to the inaccessibility of more than 50 percent of mLANA 124-314 for DNA binding as observed from the calculated N value from ITC (0.2760.03 dimer per TR-DNA)(Supporting Information, Text S1).

mLANA exhibits a novel positive patch on the dorsal face and interacts with BRD4
Electrostatic surface analysis of the mLANA protein reveals an extensive positive patch on the dorsal side which includes residues Arg-156, Lys-224, 225, 228, 229, 231, 251, 253, and Arg-232 ( Figure 4A and B and S5B). Both Arg-156 and Lys-229 interact with a phosphate ion which is located on the dorsal face of helix a1. Lys-251 and 253 lie on the 2-fold non-crystallographic symmetry axis of the dimer interface ( Figure 4B). The positive electrostatic surface also encompasses the identified 226 QAKKLK 231 motif that is involved in binding to BET proteins [34]. Both mLANA and kLANA have been described to bind to the ET domain of BET proteins and the cluster of positive charges is energetically organized to establish charge-charge interactions with a negatively charged patch on the helical ET domain of the BET proteins BRD2 and BRD4. Four key Lysine residues are conserved in both mLANA (Lys-228, Lys-229, Lys-251, and Lys-253) and kLANA (Lys-1113, Lys-1114, Lys-1138 and Lys-1140) ( Figure 2A).
Since mLANA residues 226-231 were previously shown to be important for BRD4 binding [34], and since C-terminal mLANA bound the region upstream of the ET domain, we assessed the ability of mLANA, or mLANA mutated at positively charged residues, including within 226-331, to bind BRD4 471-594.  Figure 4G). The finding that mLANA 124-272 H186D/K187E had reduced BRD4 binding suggests a potential role for these residues in interacting with Brd4. Therefore, substitution of negatively charged, but not neutral, residues for positively charged amino acids on the dorsal positive patch of C-terminal mLANA reduces BRD4 binding.

Loss of DNA binding abolishes virus persistence in mice
We assessed the role of mLANA binding to viral DNA on virus persistence in mice. A recombinant virus was generated bearing the mLANA H186D/K187E mutations in the predicted mLANA DNA recognition helix (designated vmLANA H186D/K187E ; Figure 5A). We first compared the kinetics of lytic viral replication in vitro and during the acute phase of infection in lungs. Lytic replication kinetics was essentially preserved, though changes in the pattern of lytic infection could be identified ( Figure 5B and C; Supporting Information, Text S1). This result is consistent with mLANA having a predominant role in latency. We next proceeded to investigate the impact of the introduced mutations in the ability of MHV-68 to expand latent infection in GC B cells and establish persistent infection in mice. vmLANA H186D/K187E showed a marked defect in in vitro reactivation from latency ( Figure 5D), and viable virus was barely detectable beyond the limit of detection of the assay as early as 14 dpi. This viral mutant also showed a clear defect in the frequency of viral-DNA positive splenocytes ( Figure 5E and Table 4). Consistent with this deficit, the frequency of vmLANA H186D/K187E infection in GC B cells at the peak of virus latency (14 dpi) was three orders of magnitude lower ( Figure 5F and Table 4). This result was confirmed by visualisation of latently infected cells within GCs in splenic sections ( Figure 5G, a-b). Notably, analysis of vmLANA H186D/K187E at 21 and 50 dpi showed that loss of mLANA DNA binding severely diminished virus persistence in mice ( Figure 5D and E). These results demonstrate that mLANA binding to its cognate DNA is essential for the expansion of latent infection in GC B cells and is critical for virus persistence. from the C-terminus destabilizes the dimer, as it contributes to the hydrophobic stability of the dimer interface (see also Figure S2). (D) Fluorescencebased thermal shift analysis of mLANA truncations (arrows display the midpoint of the unfolding transitions). The midpoint melting temperatures of the protein-unfolding transition (T m ) for mLANA DBD truncations are shown for the optimal buffers. (E) mLANA tetramer assembly presumed for cooperative TR DNA binding as observed in the crystal packing with the symmetry equivalent dimer. Phosphate interactions coordinate the dimerdimer interface. Two dimers form a tetramer stabilized by an interface with a surface area (,535 Å 2 ) per monomer that is approximately 12% of that of the total surface of a single subunit (,7,500 Å 2 ). (F) A ribbon representation of the dimer-dimer interface view from the ventral side. In the interface Arg-148 from dimer-2 stacks against Tyr-241, and makes a weak hydrogen bond with Tyr-237 from dimer-1. The arginine residue is conserved in both mLANA and kLANA proteins which has the ability to intercalate between hydrophobic aromatic amino acids on the second dimer. doi:10.1371/journal.ppat.1003673.g001 Electrostatic surface residues on the dorsal face exert a role in expansion of latent infection in GC B cells We investigated the role of the positive electrostatic surface residues on the pathogenesis of MHV-68 infection taking into account their charge and relative position on the dorsal face. Recombinant viruses were engineered to contain neutral or negatively charged mutations at the centre or periphery of the positive patch ( Figure 5A). This set of mutations did not affect viral replication in vitro or affect lytic infection in lungs of mice ( Figure 5B and C). However, when the latent phase of the infection was analysed a clear phenotype could be ascribed to mutations located at the periphery of the positively charged dorsal motif. Whereas vmLANA K251A/K253A and vmLANA K251E/K253E showed levels of latency comparable to wild type virus ( Figure 5D, F and Table 4), viruses vmLANA K224A/K228A/K229A and vmLA-NA K224E/K228E/K229E showed a significant (p,0.0001) latency deficit at 14 dpi ( Figure 5D). This deficit was confirmed by quantification of the frequency of viral DNA-positive cells in total splenocytes ( Figure 5F and Table 4). This phenotype is unlikely to reflect inadvertant mutations introduced during the mutagenesis process since the equivalent phenotype was observed in both independently derived mutant viruses. Moreover, revertant viruses in which the mutated K224/K228/K229 residues were restored to wild type status did not show any defects in latent infection ( Figure 5E), demonstrating that phenotypic changes observed were intrinsic to this locus and not a consequence of mutations elsewhere in the viral genome. Quantification of the frequency of infection in purified GC B cells ( Figure 5G and Table 4) and visualisation of latently infected cells by in situ hybridisation ( Figure 5H, c-d) revealed that replacing a positively charged lysine for the oppositely charged glutamate accentuated the attenuated phenotype when compared to replacement with a non-polar alanine. Notably, the attenuated latency phenotype ascribed to Lys-224, 228, 229 did not correlate with the activity of mLANA to bind BRD4. That conclusion is based on the result that BRD4 binding was reduced for mLANA K251E/K253E ( Figure 4F and 4G), however vmLANA K251E/K253E showed no latent infection phenotype (Figure 5D, F, G and H, e-f). Further, mLANA K224A/K228A/K229A bound BRD4 normally ( Figure 4F and  4G), yet vmLANA K224A/K228A/K229A was deficient for latent infection. Taken together these data demonstrate that Lys residues 224, 228 and 229 located at the periphery of the positively charged dorsal motif are required for the efficient expansion of infected GC B cells, likely acting through a host cell protein other than BRD4.

Discussion
C-terminal LANA is the most highly conserved LANA domain, and its ability to bind DNA is essential for viral DNA replication and episome persistence [12]. Here, we solve the X-ray crystal structure of the mLANA DNA binding domain. The overall fold comprises a rigid and thermostable b-barrel dimer flanked on either side by an a-helix arrangement. The structural features of mLANA are common to EBV EBNA1 and HPV E2 proteins, indicative of their conserved biological functions in episome maintenance and transcriptional activation.
This work identifies ventral C-terminal mLANA as the DNA binding interface and demonstrates its essential role in latent viral infection. The similarity in the protein folds of mLANA with EBNA1, the sites of mLANA phosphate ion binding, and the modelled superposition of the EBNA1-DNA complex all support that the ventral side interfaces with DNA. Furthermore, a high affinity TR DNA LANA binding site (mLBS1) was demonstrated by EMSA and shown by ITC to bind mLBS1 with nanomolar affinity, 66 nM, with enthalpy as the major driving force for binding. Similar to kLANA, mLANA cooperatively binds adjacent high and low affinity TR DNA binding sites. Disruption of targeted residues on the ventral surface abolished DNA binding (Figure 3), providing direct evidence that this surface is the DNA binding interface. Moreover, engineering the same disruptive mutations into the genome of MHV-68 abolished the ability of the virus to expand latent infection in GC B cells and eliminated persistence of virus that was capable of reactivation. This result was consistent with recent findings that virus containing mLANA incapable of binding DNA was highly deficient for virus ex vivo reactivation from splenocytes [24]. This work establishes that LANA binding to TR DNA is critical for viral persistence in proliferating GC B cells. Most importantly, it supports a model in which a strategy of expansion of virus infection in GC B cells to access persistence in long-lived memory B cells is vital for host colonisation. This establishes a parallel with EBV where the same strategy for virus persistence has been proposed based upon clinical investigation [7].
Prior genetic and functional analyses of the LANA DBD can now be interpreted in light of the solved mLANA structure. Consistent with their predicted contacts with DNA, mutations within a helices 1 and 2 affect DNA binding. mLANA Pro179Thr or Leu150Pro resulted in loss of the ability of mLANA to protect its binding site from DNase I cleavage [24,39,42]. Pro-179 is found   Table 3. doi:10.1371/journal.ppat.1003673.g003 at the N-terminus turn of the a2-helix, and restricts the dihedral angles of the preceding amino acid His-178, diminishing the flexibility of His-178, a residue expected to be important for DNA interaction ( Figure 2F). The Leu150Pro substitution in helix a1 causes helix disruption as proline is expected to destabilize the ahelix structure, causing a kink. Multiple mutations within a helices 1 and 2 of the kLANA DBD also disrupted DNA binding [24,39,42]. As some kLANA mutations located in regions other than a helix 1 and 2 also disrupted binding, it is likely that these disturb secondary structure and folding of the DBD. Consistent with the large surface area interface of the dimer, only larger deletions of ,10-15 amino acids were able to disrupt kLANA selfassociation [12,39,43].
Inhibition of NF-kB signalling is essential for MHV-68 latency [31]. mLANA recruits EC 5 S E3 ubiquitin ligase complex to target NFkB for degradation by the ubiquitin-proteasome pathway by way of an unconventional SOCS box motif ( 199 VSCLPLVP 206 ). SOCS box containing proteins mediate interactions with elongins B and C to target substrates for degradation. The mLANA structure shows that this motif forms a loop perpendicular to the DNA binding interface, likely allowing accessibility for elongin binding without disrupting DNA binding. A strictly conserved leucine residue [44,45] is part of an essential hydrophobic interface required for Elongin C binding. This feature may be fulfilled by residues Leu-202, Pro-203 and/or Leu-204 that are positioned on the b2-b3 loop to accommodate an ElonginBC-Cul5 complex. A conserved cysteine is critical for SOCS-box interaction with Elongin C [44], and the LANA SOCS-box motif includes Cys-201 conserved in both mLANA and kLANA. The mLANA SOCS loop is an atypical structural motif since a typical SOCS-Elongin C interaction forms a ''folded-leaf'' four helix cluster. Since LANA does not display a standard SOCS box consensus sequence, it is likely that LANA has acquired a unique strategy to recruit an EC 5 S ubiquitin ligase complex.
The most striking feature in the mLANA structure is the positive electrostatic charge on the dorsal face, which is absent in EBNA1 and E2. Residues 226 QAKKLK 231 , within this charged region, were previously identified as binding the BET family of proteins, BRD2 and BRD4 by using BRD4 as a probe against an mLANA peptide array and these findings were confirmed in coimmunoprecipitation experiments with mLANA mutated at these residues [34]. Consistent with these results, we found that substitution of positively charged for negative residues in mLANA K224E/K228E/K229E reduced BRD4 binding, although substitution with neutral residues in mLANA K224A/K228A/ K229A did not. Similarly, mLANA K251E/K253E, but not mLANA K251A/K251A, was reduced for BRD4 binding. These findings suggest possible electrostatic inhibition when positively charged residues were substituted for ones with a negative charge. Our results differ from other work [34] since the region immediately upstream of the BRD4 ET domain, rather than the ET domain itself, mediated C-terminal mLANA binding. Interestingly, in contrast to mLANA, and as previously reported [41], kLANA bound the ET domain rather than the upstream BRD4 region, suggesting independent evolution of this LANA binding. We also found that N-terminal mLANA binds to distinct BRD4 regions compared to C-terminal mLANA, although the Cterminal domain has the more prominent role in BRD4 binding (Figure 4). Interestingly EBNA1 and E2 each bind BRD4 through non C-terminal regions. EBNA1 binds BRD4 through Nterminal residues 61-83 [46], while the N-terminal E2 transactivation domain binds BRD2/4 [47]. Most importantly, our mouse infections with mutated viruses showed that this novel electrostatic charge patch on the mLANA dorsal face is required for efficient expansion of latently infected GC B cells. The charged patch could be divided into two components with different functional effects, consisting of a central region and a peripheral region. Unexpectedly, the attenuated latency phenotype ascribed to mutations in Lys-224, 228, 229 could not be attributed to BRD4. This finding suggests that this region is likely to be acting through a host cell protein other than BRD4, possibly a different BET protein.
The resolution of the crystal structure of the DBD of mLANA provides advantages for pathogenesis studies in a mouse model. In this study, solving the mLANA DBD structure enabled the rational design of mutations in the DNA binding interface that resulted in loss of DNA binding and eliminated virus associated GC cell proliferation and persistence of viable virus in the host. In addition, the mLANA quaternary structure revealed a novel structural motif composed of a patch of positively charged lysine residues on the dorsal face. Results showed that this feature exerts a key role in the expansion of latently infected GC B cells. Thus, combining structural, in vitro and cellular studies with an animal model of infection offers a unique opportunity to investigate viral pathogenesis.

Cloning, expression and purification of mLANA
A series of C-terminal truncations of mLANA DBD were cloned into the pET-49b vector (Novagen) and the proteins expressed with an N-terminal GST-His-tag ( Figure S1) in Escherichia coli BL21 Star (DE3) cells (Invitrogen) containing the pRARE2 plasmid (Novagen). Proteins were purified on an Ä KTA Explorer (GE Healthcare) FPLC system. Point mutations within mLANA  were engineered using the QuikChange multisite-directed mutagenesis kit (Stratagene). Truncations of mLANA ( Figure S1A) were PCR cloned into pET-49b as BamHI-NotI fragments using the appropriate primers ( Figure S1B). All constructs were confirmed by DNA sequencing. The vector carries an N-terminal GST-tag and His-tag coding sequences followed by a recognition site for the human rhinovirus (HRV) 3C protease (LEVLFQ/GP). Cultures of BL21 Star (DE3) cells transformed with mLANA plasmids were grown until the mid-log phase (A600 nm = 1.8) in Terrific Broth medium supplemented with kanamycin (50 mg/ml) and chloramphenicol (30 mg/ml), induced with 0.5 mM isopropyl-1-thio-b-Dgalactopyranoside, and grown for an additional 16 h at 18uC. Cells were harvested by centrifugation, resuspended in a buffer containing 50 mM Na/K phosphate (pH 7.0), 500 mM NaCl, 1 mM Tris (2-carboxyethyl) phosphine (TCEP), 10 mM imidazole, EDTA-free protease inhibitor cocktail (Roche) and 5 mg/ml OmniCleave, (Epicentre), and stored at 280uC. The cell pellet was disrupted using an APV-2000 continuous homogenizer (APV, Madrid, Spain) at 6000 psi. After centrifugation, the supernatant was loaded onto a HisTrap HP (GE Healthcare) column, preequilibrated in 50 mM Na/K phosphate (pH 7.0), 500 mM NaCl, 1 mM TCEP, 10 mM imidazole, and proteins eluted with a linear gradient from 10-500 mM imidazole in the same buffer using a Ä KTA Explorer (GE Healthcare) FPLC system ( Figure S1C). The GST-Histidine tag was removed by overnight incubation with 3C protease at 4uC and the cleaved protein passed over a GSTPrep FF (GE Healthcare) column (to remove any uncleaved protein as well as the GST-tagged 3C protease) connected end to end to a HiPrep Heparin FF (GE Healthcare) column. The columns were disconnected and the heparin bound mLANA proteins eluted with a linear gradient from 0.3-2 M NaCl. Following 3C cleavage the recombinant protein contained the vector-derived sequence GPGYQKDP at the NH2-terminus. Finally, the protein was applied to a Superdex S75 16/60 filtration column (GE Healthcare) in 10 mM MES buffer, pH 6.5, 300 mM NaCl and 1 mM TCEP and stored at 280uC.

Differential scanning fluorimetry
Assays were performed on an iCycle IQ5 Real Time PCR Detection System (BioRad) with excitation and emission wavelengths of 490 and 575 nm, respectively. Reaction volumes of 20 ml contained 1 mg of protein and 106 of SYPRO orange (Invitrogen) previously diluted in 20 mM HEPES pH 7.5 from an initial 50006 stock. The reaction mix was prepared by adding 2.0 ml of protein-dye mixture solution to 18.0 ml of each buffer from the JBS Solubility kit (Jena Biosciences). For the thermal denaturation the plates were heated at a rate of 1uC/min from 20 to 90uC with a 10 s hold step for every point and fluorescence was measured in 1uC increments. Experiments were carried out in duplicate, and T m values were calculated for each well using nonlinear regression analysis in the curve-fitting program GraphPad Prism. The T m values were incorporated into a Microsoft Excel script (ftp://ftp.sgc.ox.ac.uk/pub/biophysics) for comparison of melting curves and thermal shifts. While crystal optimization by sitting-drop vapour diffusion technique failed for mLANA  . Diffraction-quality crystals for mLANA 140-272 were obtained by exchanging the protein buffer to the optimal buffer system suggested by DSF assays (50 mM Na/ K phosphate pH 7.0). mLANA 140-272 crystals were further optimized by serial streak-seeding in decreasing precipitant concentrations, with optimal growth to 0.0560.0560.15 mm in drops formed by mixing an equal volume of the protein solution with a reservoir solution that contained 0.1 M Na/K phosphate pH 7.0, 0.1 M lithium sulphate, 22% w/v PEG 3350 and 4% v/v 1,4 dioxane. Crystals were transferred to mother liquor supplemented with 20% v/v glycerol, and cryocooled in liquid nitrogen.

Crystallization and structure determination
A diffraction data set was collected from EMBL beamline P14 at PETRA III (DESY, Germany), indexed and integrated with XDS [48], then merged and converted with Pointless [49] to mtz format for scaling in SCALA [49]. The structure was solved by molecular replacement using Phaser [50] with the EBNA1 dimer (PDB ID: 1VHI) as a search model. An initial model of the structure was autobuilt in PHENIX [51] with manual rebuilding of the model performed in Coot [52] and refined in PHENIX. The crystallographic data are summarized in Table 1. The structure was validated using MOLPROBITY [53] as implemented in PHENIX. The analysis of the dimer interface used the Protein interfaces, surfaces and assemblies service PISA at European Bioinformatics Institute [54]. Energy minimization for 3D structures of mLANA-DNA models was performed using YASARA [55]. Structural illustrations were prepared with PyMOL [56] and CCP4MG [57]. unlabeled mLBS1, mLBS2, or mLBS1-2 were included in the incubation. To generate mLBS1-2 oligonucleotide, TATAATG-GATCCggcgccatgcgcccgcgcctggggcgccacgcggcg and TTATATG-GATccccgggtctccagggcgcgc-cgcgtggcgccccag primers were annealed and extended using Prime-It II kit (capital letters indicate additional non TR sequence). The resulting oligonucleotide was ligated into the EcoRV site of pBluescript. mLBS1-2 was then released after digestion with HindIII/XbaI and gel purified (Qiagen #28406). For the supershift, 1 mg anti-FLAG antibody (Sigma #F1804) was added to the reaction. Loading buffer (95% Formamide, 10 mM EDTA, 0.1% xylene cyanol and 0.1% bromophenol blue) was added to incubations prior to loading on a 3.5%/8% non-denaturing TBE-polyacrylamide gel. The gel was run in TBE buffer for 1 hour at 300 V and dried on Whatman paper before exposing to Kodak biomax MS film (Kodak #829 4985). In vitro translated mLANA wt and mutants were detected by  western blot. mLANA was resolved by 10% SDS-PAGE and after transfer to nitrocellulose membrane, the protein detected using monoclonal anti-mLANA (6A3) followed by HRP conjugated goat anti-mouse secondary antibody (Southern Biotech #1031-05).

SPR and ITC analysis
SPR binding experiments were performed on a BIACORE 6100 instrument (GE Healthcare) at 25uC. Single stranded DNA oligonucleotides, mLBS1 with an addition of five thymine's in the 59 end as a spacer were annealed to its pair before injected over a streptavidin-coated sensor chip (SA chip, GE Healthcare) at 1 nM concentration at 10 ml min 21 up to 590 RU. A flow cell was left blank to allow background signal subtraction.
Purified recombinant mLANA 124-272 protein dilution series from higher to lower concentration were injected at a flow rate of 10-30 mL min 21 for 3 min and the regeneration of the binding surface was achieved by 4M NaCl. All the experiments were performed in duplicates. Data were processed using BiaEvaluation (GE Healthcare) and analyzed using SigmaPlot. The equilibrium dissociation constant, K D , was determined by non-linear fitting of the data using a 1:1 Langmuir isotherm.
We further examined mLANA-DNA binding using ITC. Purified mLANA protein 124-314 and mLBS1 DNA (without biotin-TTTTT spacer) were dialyzed against 25 mM Na/K phosphate pH 7.0, 150 mM NaCl, and 5% glycerol. ITC titrations were performed with a MicroCal iTC200 Isothermal Titration Calorimeter (MicroCal) at 25uC. Protein and DNA absorbance were measured after dialysis by NanoDrop (Nano-Drop Technologies) and their concentrations were determined with their respective extinction coefficients. Twenty-five injections of 1.5 ml each of 30 mM mLBS1 DNA were titrated into 8.5 mM protein solution. Data were corrected for nonspecific heats and analysed using MicroCal Origin 7.0 software using a one-site binding model. The experiments were performed in triplicate and showed similar results.
In vitro binding of mLANA and BRD4 GST, GST mLANA or GST BRD4 [41] fusion proteins were expressed in BL21 (DE3) bacteria and collected on Glutathione Sepharose 4B (GE Healthcare) beads. GST fusion binding assays were performed as described [41]. Proteins were resolved by SDS-PAGE using 10% polyacrylamide, and transferred to nitrocellulose membrane. 35 S labelled BRD4 proteins were detected after exposure to film, KSHV LANAD33-888 [58] was detected with anti-T7NTag antibody (Novagen), and mLANA was detected with anti-FLAG M2 antibody (Sigma).

Generation of recombinant viruses
MuHV-4 recombinant viruses were independently generated by mutagenesis of the viral genome cloned as a bacterial artificial chromosome (BAC) [59]. The point mutations were introduced by PCR on the mLANA gene cloned into pCMV-Myc. pCMV-Myc-mLANA was digested with HindIII and PciI or BstEII and PciI to isolate the fragment harbouring the desired mutations, which were inserted into the BamHI-G genomic clone. Recombinant BamHI-G mutant fragments were subcloned into the BamHI site of a pST76K-SR shuttle plasmid and transformed into an Escherichia coli strain (DH10B) containing the wildtype MuHV-4 BAC (pHA3). Following a multistep selection procedure, recombinant BAC clones were identified by DNA sequencing. To generate revertant viruses the wild type BamHI-G pST76K-SR shuttle plasmid was transformed into DH10B cells containing each of the mutant BAC genomes. The integrity of each BAC was confirmed by restriction digestion with BamHI and EcoRI. All viruses were reconstituted by transfection of BAC DNA into BHK-21 cells using X-tremeGENE HP (Roche). The loxPflanked BAC cassette was removed by viral passage through NIH Cre 3T3 cells and limiting dilution to obtain GFP-negative viruses.

Analysis of recombinant viruses
We performed five independent animal infections, which were analyzed by five complementary experiments; ex vivo explant coculture assays to measure virus latency in total splenocytes, flow cytometry coupled to limiting dilution and real-time PCR to quantify the frequency of viral DNA-positive in total splenocytes or to quantify GC B cells, and in situ hybridisation to identify virally infected cells within the spleen. In situ hybridisation for transcripts corresponding to the MHV-68-encoded miRNAs permits the detection of latently infected cells in splenic sections. Quantification of viral infection in GC B cells is highly relevant since mLANA is selectively expressed in proliferating GC B cells [25]. Revertant viruses were analysed at 14 days post-infection by ex vivo explant co-culture assay by three independent animal infections. C57BL/6 mice (Charles River, Spain) with 6-8 weeks of age were intranasally inoculated with 10 4 p.f.u. in 20 ml of PBS under isofluorane anaesthesia. At 3, 7, 14, 21 and 50 days postinfection, lungs or spleens were removed and processed for subsequent analysis. Infectious virus titers in freeze-thawed lung homogenates were determined by suspension assay using BHK-21 cells. Latent viruses were examined using explant co-cultures of single-cell suspension splenocytes with BHK-21 cells. Plates were incubated for 4 (suspension assay) or 5 days (co-culture assay), fixed with 10% formal saline and counterstained with toluidine blue. Viral plaques were counted with a plate microscope. Frequencies of virus-genome-positive cells were determined by limiting dilution combined with real-time PCR, as described earlier [25]. Total splenocytes suspensions were prepared from pools of five spleens. GC B-cell (CD19 + CD95 hi GL7 hi ) populations were purified from pools of five spleens using a BD FACSAria Flow Cytometer (BD Biosciences). The purity of sorted populations was always .95%, as analysed by flow cytometry. Real-time PCR reactions were performed as reported [60]. To determine multistep growth curves, BHK-21 cells were infected with MuHV-4 recombinant viruses, at a low multiplicity of infection (0.01 p.f.u. per cell). After 1 hour of virus adsorption, cells were washed in PBS and at the indicated times post-infection virus titers were determined by plaque assay. p values were determined using ordinary one-way ANOVA (GraphPad Prism). In situ hybridisation was performed on formalin-fixed, paraffin-embedded splenic sessions using digoxigenin-labelled riboprobes, generated by T7 transcription of pEH1.4 [61]. In situ hybridisation for transcripts corresponding to the MHV-68-encoded miRNAs (encoded within pEH1.4) permits the detection of latently infected cells in splenic sessions, specifically within GC reactions. For panels B and C, the A 280 nm signal indicating protein is amplified due to the contribution of DNA in the complex, V e is the sample elution volume and M represents protein standards (Supporting Information, Text S1).

(TIF)
Text S1 Supporting data including a supporting method and supporting results. (DOCX)