The crystal structure of the herpes simplex virus 1 ssDNA-binding protein suggests the structural basis for flexible, cooperative single-stranded DNA binding.

All organisms including animal viruses use specific proteins to bind single-stranded DNA rapidly in a non-sequence-specific, flexible, and cooperative manner during the DNA replication process. The crystal structure of a 60-residue C-terminal deletion construct of ICP8, the major single-stranded DNA-binding protein from herpes simplex virus-1, was determined at 3.0 A resolution. The structure reveals a novel fold, consisting of a large N-terminal domain (residues 9-1038) and a small C-terminal domain (residues 1049-1129). On the basis of the structure and the nearest neighbor interactions in the crystal, we have presented a model describing the site of single-stranded DNA binding and explaining the basis for cooperative binding. This model agrees with the beaded morphology observed in electron micrographs.

Viruses of the Herpesviridae family infect almost all vertebrates, including man, causing a variety of diseases. Of the seven viruses identified as human infectious agents, herpes simplex virus-1 (HSV-1) 1 is the prototype of the ␣herpesvirus subfamily and of the family as a whole. The HSV-1 singlestranded DNA (ssDNA)-binding protein (SSB), ICP8, is a nuclear protein that, along with the six other HSV replication proteins (the viral polymerase (UL30) and its accessory factor (UL42), the trimeric helicase-primase complex (UL5-UL8-UL52), and the origin-binding protein (OBP), coded by the gene ul9), is required for viral DNA replication (1) during lytic infection. Replication has been thought to proceed by a rolling circle mechanism (2) partly because the replication product is a concatamer, although the observation of highly branched replication intermediates could be explained by other mechanisms that would link recombination and replication. ICP8 is a 128-kDa multifunctional zinc metalloprotein (3) encoded by the ul29 gene. It preferentially binds ssDNA over double-stranded DNA in a non-sequence-specific and cooperative manner (4). ICP8 has been reported to interact either directly or indirectly with several other viral proteins. There is evidence that it binds to the C terminus of the OBP and stimulates its helicase activity (5,6), that it promotes the helicase activity of the viral helicase-primase complex (UL5-UL8-UL52) (7), and that it modulates the processivity of the viral polymerase (UL30) (8). Before viral DNA replication commences, these proteins are thought to be co-localized with ICP8 at small punctuate foci called prereplicative sites. With the onset of viral genome amplification, these proteins become redistributed into a larger globular replication compartment (9) whose location is defined by the preexisting host cell nuclear architecture, most probably at the periphery of the nuclear matrix-associated ND10 domains where the viral transactivator ICP0 and the viral input genome are believed to migrate in the early stages of infection (10). ICP8 is also involved in several other events of the DNA metabolism. It can promote DNA strand transfer (11), catalyze strand invasion in an ATP-independent manner (12), and renature complementary DNA strands (13), which indicates that ICP8 plays an important role in HSV genome recombination. The replication of HSV-1 DNA is also associated with a high degree of homologous recombination. Recently it was shown that ICP8 works together with alkaline nuclease (UL12), which is a 5Ј-3Ј-exonuclease, to effect strand exchange (14). In addition to its role in DNA synthesis, ICP8 has been shown to regulate viral gene expression by repressing transcription from the parental genome (15) and stimulating late gene expression from progeny genomes (16).
Genetic and biochemical analyses have failed to identify functionally independent domains within ICP8. Even the extent of the minimal DNA binding region has remained unclear. It has been placed in the C-terminal half of the protein (17) or in regions spanning residues 564 -1110 (18) or 300 -849 (19). The C-terminal 60 amino acid residues were shown to account for most of the cooperative behavior in ssDNA binding (20), possibly modulated by the two cysteines 254 and 455 (21). It has also been shown that the C-terminal 28 amino acids contain a nuclear localization signal (22), that the residues between 499 and 512 host a zinc binding motif (3), and that the residues from 1082-1169 are also important for the stimulation of late gene expression (18).
Here we have reported the first crystal structure of an ssDNAbinding protein of the Herpesviridae, a 60-amino acid C-terminal deletion mutant of ICP8, at 3.0 Å resolution. The structure consists of an unexpectedly large N-terminal folding unit and a small C-terminal ␣-helical domain, both with novel folds. In addition, it * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. has provided insight into the likely mechanism of cooperative ssDNA binding and tempted us to speculate about the possible interaction with the origin-binding protein.

EXPERIMENTAL PROCEDURES
The preparation and crystallization of the ICP8 protein missing the last 60 amino acids of the C terminus and with the mutations C254S and C455S (ICP8⌬Ccc) have been described previously (23).
Protein Expression and Purification of Selenomethionine-ICP8⌬Ccc-Selenomethionine (SeMet)-enriched ICP8⌬Ccc was expressed in High 5 insect cells grown as monolayer culture. Confluent cells were infected with the same recombinant virus used for expression of the native protein in a methionine-containing Sf-900II SFM medium (Invitrogen), supplemented with 10 g/ml penicillin-streptomycin and 2% fetal calf serum. After 30 h, cells were washed with phosphate-buffered saline solution; methionine-free IPL-41 medium (Applichem) was added for starvation. The methionine-free medium was renewed after 4 h with SeMet-containing (50 mg/l) IPL-41 medium. Cells were incubated for a further 26 h before harvesting. The purification protocol was the same used for the native protein (20), with the only exception that all buffers were flushed with N 2 and supplemented with 10 mM reducing agent (dithiothreitol or ␤-mercaptoethanol) to overcome a more pronounced tendency of the selenomethionine-containing protein to oxidize and aggregate.
Crystallization-SeMet-ICP8⌬Ccc crystals were grown at 22°C in hanging drops by equilibration of 5 mg/ml protein in 10 mM Tris-HCl (pH 8.0), 300 mM NaBr, 20% glycerol, 10 mM dithiothreitol against 12-14% polyethylene glycol 3000 and 100 mM sodium-potassium phosphate, pH 6.3. This crystallization condition is similar to that for the native crystal growth. Within about a week, fragile, plate-like crystals (ϳ0.2 ϫ 0.2 ϫ 0.05 mm 3 ) grew by salting in. For derivatization with methyl mercury acetate (MMA), the protein was first dialyzed against a dithiothreitol-free buffer, then against a 5 mM MMA-containing buffer at pH 7.5, and was subsequently used to set crystallization drops. Plate-like crystals appeared in a week in conditions similar to the ones used for native crystal growth. Crystals formed in space group P2 1 2 1 2 1 with two molecules in the asymmetric unit.
Data Collection, Structure Determination, and Refinement-Both Se-Met-and MMA-containing ICP8⌬Ccc crystals were cryoprotected by brief soaks in 20% glycerol buffered at pH 6.3 before cryocooling in liquid nitrogen. Multiwavelength anomalous diffraction data from crystals of SeMet-ICP8⌬Ccc were collected at 100 K using synchrotron radiation at the 17-ID IMCA-CAT beamline of the Advanced Photon Source (Argonne) at three/two different wavelengths around the selenium absorption edge. A full diffraction data set was collected for the MMA derivative at 100 K, using the BW7B beam line of the European Molecular Biology Laboratory Hamburg Outstation. The diffraction data were processed using the HKL program package (24). Data collection statistics are shown in Table I.
The structure was solved by the MAD method (25). Initially from a first SeMet-containing crystal (CRYST-1), F A values were obtained using XPREP (Bruker-AXS Inc.) to 4.0 Å, enabling the selenium substructure to be solved (50 of 56 seleniums) using the program SHELXD (26). Phases were then obtained to 4.0 Å from the two wavelength MAD data. The phases were extended to 3.2 Å by using density modification procedures and 2-fold non-crystallographic symmetry averaging (27). 55% of the model was built using a semiautomatic procedure with the programs MAID (28), RESOLVE (29), and O (30). Later, phases were extended to 3.0 Å using data from another crystal (CRYST-3, see Table  I) by applying multiple crystal averaging (31). The resultant phases allowed the Se substructure of CRYST-2 to be determined using an anomalous difference Fourier at 4.0 Å. Then single isomorphous replacement with anomalous scattering was used to calculate phases, and phase combination was performed to 4.0 Å with the phases generated from multiple crystal averaging. Finally, phases were extended to 3.0 Å using density modification and 2-fold non-crystallographic symmetry averaging. At this stage, the quality of the map improved significantly. Model building was continued in a similar manner to that described above, and 70% of the model could be built. Refinement of the structure was performed using simulated annealing, followed by positional and restrained B-factor refinement as implemented in CNS (32). As the model became more complete, a new mask was calculated and used in the multiple crystal averaging and phase combination. Density modification and 2-fold non-crystallographic symmetry averaging were repeated, followed by the semiautomatic procedure for model building. The model produced in this way was nearly complete except for some missing loops, and there was interpretable density for 90% of the residues. In the final stage, refinement was continued using non-crys-tallographic symmetry restraints and a bulk solvent correction in the program CNS (32). The refinement was monitored using the free Rfactor calculated with 10% of observed reflections. The refinement statistics for CRYST-3 (which, although a mercury derivative, were the best 3.0 Å data) are shown in Table I. Of 1136 residues, 107 residues in chain A and 105 in chain B are not visible in the electron density and are probably disordered. The major disordered loops are located at the interface of the neck and head. The rest of the disordered loops are situated at different parts of the shoulder region and are shown as dotted lines in Fig. 1.
Overall geometric quality of the model was assessed using PRO-CHECK (33). 86% of the amino acid residues of ICP8 were found in the most favorable region of the Ramachandran plot, with the remaining residues (apart from Thr 908 ) in the additional and generously allowed regions. All figures were produced using MOLSCRIPT (34), PyMol (35), and RASTER3D (36).
Modeling of ssDNA-The structure of the EcoSSB-ssDNA complex, where two monomers of EcoSSB cover 26 nucleotides, (37) was used for the modeling of ssDNA onto the neck region of ICP8. One of the two monomers was superimposed with a monomer of human mitochondrial SSB (HsmtSSB) (38) with a root mean square deviation of 1.6 Å on the 97-C␣ target pair, and then the monomer of HsmtSSB was overlaid on the neck region of ICP8. In this way the relative orientation of ssDNA was modeled onto the complete structure of ICP8 with the ssDNA covering the neck region of ICP8.
The two independent molecules that form the protein chain have different spatial arrangements of the C-and N-terminal domains; however, the distance over the disordered region (1038 -1048) is approximately the same (16.1 and 19.2 Å). The relative orientation of ssDNA was created for each monomer of the proposed protein chain formed by applying non-crystallographic symmetry. Visual inspection using computer graphics allowed the nucleotides between two non-crystallographic symmetry-related monomers (chains A and B) and between two crystallographically related monomers (chain B and symmetry mate of chain A) to be added while avoiding clashes with the protein chain. Again the crystallographic symmetry was applied to the newly built nucleotides. In this way, it was possible to join the ssDNA in a continuous chain. In the model the continuous chain of ssDNA contains 98 nucleotides covered by 7 monomers and the distance between 5Ј-and 3Ј-ends is 350 Å. The coordinates of the model are available from the authors upon request, and a more detailed illustration is included in the supplemental material.

RESULTS
Crystallization and Structure Determination-The structure and function of a number of other prokaryotic and eukaryotic SSBs have been described (39), but ICP8 is much larger (128 kDa) relative to other SSBs. For example, the monomers of bacterial SSBs are typically ϳ20 kDa, and although the heterotrimeric eukaryotic SSB, RepA, is ϳ116 kDa, this is believed to contain more than one DNA binding region. Crystallization studies of ICP8 have been reported earlier (23), and it was shown that crystals of full-length ICP8 diffracted too poorly to be useful. A mutant, ICP8⌬Ccc, with the C-terminal 60 residues deleted and two point mutations (C254S and C455S) has been crystallized under similar conditions to the full-length ICP8 and diffracted to at least 3 Å resolution. The mutant has been shown to bind ssDNA with much reduced cooperativity (22).
The structure was solved by MAD and single isomorphous replacement with anomalous scattering methods using SeMetsubstituted ICP8. The SeMet crystals formed in space group P2 1 2 1 2 1 with two molecules in the asymmetric unit and diffracted to 3.2 Å resolution. The model (residues 9 -1129 with the disordered regions described below) was refined to a crystallographic R value of 23.5% (R free ϭ 28.6%) using data from 20.0 to 3.0 Å resolution (Table I).
Overall Structure-The structure of ICP8 (9 -1129) ( Fig. 1) is composed of a large N-terminal domain (9 -1038) and a smaller ␣-helical C-terminal domain (1049 -1129). The first 8 residues and the last 7 residues of the construct are not visible in the electron density and are presumed to be disordered. The Nterminal domain can be described as consisting of head, neck, and shoulder regions. The head consists of the eight helices ␣14, ␣15, ␣16, ␣21, ␣22, ␣23, ␣24, and ␣25 (Fig. 1B). The front side of the neck region consists of a five-stranded ␤-sheet (␤16, ␤17, ␤23, ␤26, and ␤27) and two helices (␣17 and ␣27), whereas the back side is a three-stranded ␤-sheet (␤24, ␤25, and ␤28) (Fig. 1). The shoulder part of the N-terminal domain contains an ␣-helical and ␤-sheet region. The head, neck, and shoulders are interconnected in such a way that their individual structural folds are not formed by contiguous polypeptide chains. From the N terminus, the polypeptide chain forms a first helical region in the head and then one of the two ␤-sheet regions  1. A, structure of ICP8. Overall view of the ICP8 structure. Dotted lines represent disordered regions with blue and red balls signifying the N-and C-terminal ends of the disordered regions. (Sequence information using the same color code is given Fig. 4). The shoulder region is colored blue; the zinc binding region is green; the part of the polypeptide chain linking the neck and shoulders as a single folding unit is orange. The neck is colored yellow (front) and gray (back). The head is red, and the C-terminal helical domain is purple. B, the structure rotated 60°along x-axis relative to Fig. 1A. The blue to red color gradient follows from the N to the C terminus. In this orientation, the C-terminal domain is behind the neck.
belonging to the neck. The strands ␤16 and ␤17 in the neck lead to strands ␤18-␤22 in the shoulders before returning to strands ␤23-␤26 in the neck. The strands ␤18-␤22 are involved in interaction with residues in other strands from the N terminus (see Fig. 1A). This explains why limited proteolysis experiments have never yielded either soluble or functionally active fragments (20) and why so many mutant proteins have proven to be insoluble (see, for example, Ref. 18).
DNA Binding Region-No structurally related protein can be retrieved from the DALI (40) or SSM (www.ebi.ac.uk/msd-srv/ ssm/cgi-bin/ssmserver) servers using the whole ICP8 molecule or the individual subdomains as search models. Although no structural homology is detectable for any of the ICP8 regions ( Fig. 1), the front side of the neck region shows some structural resemblance to the oligonucleotide/oligosaccharide binding (OB) fold (41), which is responsible for ssDNA binding in all SSBs so far described with the exception of the adenoviral SSB (42). The topology is different (39), but the principle is the same, namely a crossed ␤-sheet with disordered connecting loops containing conserved basic and aromatic residues. The direction of each ␤-strand of the neck region that resembles the OB-fold is similar to that of HsmtSSB (38) (Fig. 3A). The proposed DNA binding region on the front side of the neck (Fig.  3A) contains elements of the sequence between amino acids 530 and 1028, similar to the boundaries suggested by Gao and Knipe (18). Limited proteolytic analysis studies had suggested that the putative boundaries of the minimal DNA binding region are between residues 300 and 849 (19). More recent evidence, based on ICP8 photo-affinity labeling with oligonucleotides, indicated a slightly different region, namely between residues 386 and 902 (43). There are a number of aromatic and positively charged residues from the front side of the neck that are exposed to the surface or lie in the disordered loops that are relatively well conserved across the Herpesviridae. These are Tyr 543 , Asn 551 , Arg 772 , Lys 774 , Arg 776 , Tyr 988 , Phe 998 , and Asn 1002 (Fig. 4), which we believe are involved in ssDNA binding either by base stacking or electrostatic interactions.
The Role of Zinc Binding Region-ICP8 is a zinc metalloprotein containing one zinc atom/molecule (3) that, as predicted, is coordinated by three cysteines (Cys 499 , Cys 502 , Cys 510 ) and a histidine (His 512 ). Of these four residues, the cysteines are totally conserved among the Herpesviridae SSBs, but the histidine is only conserved in the Simplexvirus genus (Fig. 4). Thr 513 is, however, fully conserved and stabilizes the zinc loop further by hydrogen bonding to the main chain oxygen of res-idue 507 (Fig. 5A). It has been shown that mutation of Cys 499 and Cys 502 produces a non-functional protein that fails to complement a temperature-sensitive UL29 mutant at the nonpermissive temperature (44) and that zinc-depleted ICP8 molecules transiently retain DNA binding activity (3), suggesting the zinc binding confers structural integrity to the protein. This is confirmed by the crystal structure. The loop containing the zinc finger interacts with two regions of the protein. The first is FIG. 3. Interaction of ICP8 with short stretches and long filaments of ssDNA. A, the neck region (red) is superimposed on the OB fold of HsmtSSB (3ULL) colored in blue. Visual inspection indicates that the structure-based sequence alignment extends over 79 residues with a root mean square deviation of 2.7 Å. The DNA binding cleft is shown as a concave surface. The loop between ␤26 and ␤27 corresponds to L 4-5 and between ␤17 and ␤23 to L 1-2 of HsmtSSB. A 10-nucleotidelong stretch of ssDNA is modeled on the neck region of ICP8 as derived from superimposition with the OB fold of HsmtSSB with reference to the view of Fig. 1A. B, the mode of cooperative ssDNA binding of ICP8. Ribbon diagram showing the arrangement of monomers with the Cterminal domain of one monomer docking into the back of the neck of the next monomer in the protein chain to create an oligomer on the ssDNA filament (here 98 nucleotides long). C, enlarged electron micrograph of ssDNA covered by ICP8 to approximately the same scale taken from Ref. 5.

FIG. 2. Orientation of the C terminus.
A stereopair showing the structural superposition of the N-terminal domains of the two independent molecules, illustrating the difference in relative orientation of their C-terminal domains. These C-terminal domains would dock onto the next molecule in the protein chain as indicated in Fig. 3. The relative orientation of C-and N-terminal domains is probably determined to some degree by the packing of the protein chains in the solid state.  (Roseolovirus, Cytomegalovirus, Muromegalovirus), and two genus from the gammaherpes virus subfamily (Rhadinovirus, Lymphocryptovirus) are used in the sequence alignment. The Swiss-Prot codes of ICP8 orthologues from these sources are P04296, Q89549, Q9E6P0, O56282, P17147, P30672, O36360 and P03227, respectively. Horizontal cylinders above the sequences indicate ␣-helices (labeled ␣1-␣32). Horizontal arrows indicate ␤-strands (labeled ␤1-␤28). The secondary structure elements are colored red for the head, blue and orange for the shoulder, yellow and gray for the neck region of the N-terminal domain (a similar color code is used in Fig. 1A), purple for the C-terminal helical domain, and light green for the zinc binding loop, including two helices (␣12 and ␣13) that are involved in interaction with part of the N and C termini. Three cysteines and a histidine involved in binding to zinc are shown by a bar above the corresponding residues. The dotted lines indicate regions that are disordered in toward the N terminus and includes the region between ␤7 and ␣2. It is primarily a hydrophobic interface formed by residues Leu 112 and Leu 115 packing against Leu 501 , Val 511 , and the hydrophobic residues on one face of helix ␣11, but there is also a hydrogen bond between Asn 111 N␦2 and the carbonyl oxygen of Leu 501 . The second is toward the C terminus and contains the region between ␤24 and ␣26 (Fig. 1, sequence numbering as in Fig. 4). The interaction involves the positioning of His 508 , which hydrogen bonds with the carbonyl oxygen of Ser 934 and is stabilized by a hydrogen bond between Trp 933 N⑀1 and Glu 470 O⑀1. The zinc binding region is remote to the proposed ssDNA binding region and therefore unlikely to be important for DNA binding.
The Existence of a Protein Chain through the Crystal Suggests How Flexible Cooperative ssDNA Binding Is Achieved-In the crystal, the C-terminal domain of each monomer fits loosely into a concave surface of the back side of the neck region of the N-terminal domain that belongs to a non-crystallographic symmetry-related molecule. A continuous chain with a beaded appearance is then formed by a molecule related to the first by translation of one unit cell along a. Although relatively small (ϳ919 Å 2 ), both independent interaction surfaces are similar. However, the spatial arrangement between the two domains differs (Fig. 2), indicating the flexible nature of the protein chain that, in this case, is probably determined by the crystal packing. A continuous chain of molecules through the crystal lattice can also be observed in the case of T4 gene 32 (45) and the adenovirus ssDNA binding proteins (42) as well as in the organization of Escherichia Coli SSB tetramers (37). The arrangement of ICP8 molecules is similar to the beaded morphology observed in negatively stained electron micrographs (5) of ICP8 decorated ssDNA (Fig. 3C). We previously established that the 60-residue C-terminal region of the ICP8 is a principal determinant for cooperative DNA binding (23), at least on shorter oligonucleotides. Although the crystallized construct does not contain this region, we have postulated that the remaining weaker interactions generate the same chain in the crystal that would form on DNA in solution. Ser 254 and Ser 455 (Cys 254 and Cys 455 in the native protein) are located in the loop region between helices ␣6 and ␣7 and the C-cap of the ␣10 the crystal structure. The dashed line indicates that the region was absent in the construct. The triangle above 3 residues in the C-terminal region (last 60 residues) indicates the region encompassing the FNF motif. The star sign above two cysteines shows mutation to serine in the structure presented here. helix, respectively. These residues are neither solvent exposed nor close to the region in the neck that we have proposed to be involved in ssDNA binding. It is therefore unlikely that these two cysteines are involved in cooperativity as has been previously suggested (21).
We believe that, because of the nature of the ICP8 domain connection, ssDNA is covered in a flexible manner while keeping the ssDNA in an extended form (Fig. 3, B and C) that prevents formation of secondary structures. On the basis of the ICP8-ssDNA model ( Fig. 3 and supplemental material), it seems that ϳ14 oligonucleotides would be covered per ICP8 molecule, and this is in good agreement with biochemical data (23).
Interactions of the C Terminus-There is evidence (20) that the deletion of the C terminus seriously reduces cooperative binding for ICP8, suggesting that there is an additional protein-protein interaction involving this region. An F(N/D)F motif (amino acids 1142-1144) in the C terminus is identifiable in the Alphaherpes viruses and possibly in the Roseolovirus genus of the Betaherpesvirus subfamily (Fig. 4). We believe that at least one of the phenylalanines is involved in a hydrophobic interaction with a hydrophobic region formed by Phe 827 , Phe 843 , Trp 844 , Leu 857 , and Ile 865 (Fig. 6) of the head. This region of ICP8 is also very well conserved among Alphaherpes viruses (Fig. 4). The C-terminal residue (Glu 1129 ) of the model is in the vicinity of the head of the N-terminal domain of another molecule. Modeling a continuation of the C terminus would allow it to pass around the head and dock part of the F(N/D)F motif into the hydrophobic region mentioned above (Fig. 6). A similar interaction is important for the formation of the protein chain in the adenovirus SSB (42). The last C-terminal 28 residues that contain the nuclear localization signal must presumably remain free to facilitate nuclear import.
The deletion of 27 residues at the C terminus of the HSV1 origin-binding protein has been shown to reduce its specific affinity for ICP8 (46), suggesting that there may be an interaction between this region of OBP and ICP8 (46). However, there is no biochemical evidence that would help locate the corresponding interaction region of ICP8. Because a number of hydrophobic residues in the 27-residue C-terminal region of the protein are conserved among those Herpesviridae for which an OBP is present and among these is a V(N/D)F sequence, we have tentatively suggested that the V(N/D)F motif of OBP interacts with the same hydrophobic patch described above. Thus we speculated that the C terminus of ICP8 and OBP could compete for the same site, depending upon the nature of the protein-protein and, importantly, the protein-DNA interaction. This is consistent with a model of initiation of DNA replication in which the ICP8-OBP interaction is required to complete origin unwinding but is "replaced" by ICP8-ICP8 intermolecular interactions upon replication onset when the OBP is released from DNA and processive replication ensues.
Potential Protein-Protein Interaction Sites-We have identified the function of head, neck, and the C-terminal helical region of ICP8; however, a large part of the shoulder region is not, according to our model, involved in cooperative ssDNA binding. It is unlikely that such a large part of the N-terminal domain would have no functional role, because the necessity of packaging the viral genome should tend to enhance the evolution of multifunctional proteins and reduce the likelihood of producing non-functional coding regions. Regulation of late gene expression could involve the ssDNA binding region, but presumably nuclear positioning involves the exposed part of the protein chain. Recent work (47) has identified, by immunocoprecipitation, a number of cellular proteins that co-localize with ICP8. Some of these co-localizations are not dependent on mediation by DNA and are involved in (cellular) DNA replication, repair, and recombination. Structurally there are two regions that are likely to be involved in some of these interactions. The first is a deep pocket (Fig. 5B) with conserved aromatic residues (Tyr 20 , Phe 61 , Tyr 90 ) at the base and capped by the hydrogen bonding interaction between Glu 58 and Arg 193 . The second is a cleft (Fig. 5C) containing the fully conserved Cys 116 and Arg 120 , both on ␣2, at the side.