Crystal Structures of Yellowtail Ascites Virus VP4 Protease

Background: YAV VP4 is a Ser/Lys dyad protease that processes the birnavirus polyprotein: pVP2-VP4-VP3. Results: An acyl-enzyme is formed between Ser633 (nucleophile) and Ala716, confirming the existence of an internal cleavage site. Conclusion: VP4 terminating at Ala716 is active, and Ser633 serves as the nucleophile. Significance: Structural insights into three stages of the VP4 reaction will promote anti-birnavirus compound development. Yellowtail ascites virus (YAV) is an aquabirnavirus that causes ascites in yellowtail, a fish often used in sushi. Segment A of the YAV genome codes for a polyprotein (pVP2-VP4-VP3), where processing by its own VP4 protease yields the capsid protein precursor pVP2, the ribonucleoprotein-forming VP3, and free VP4. VP4 protease utilizes the rarely observed serine-lysine catalytic dyad mechanism. Here we have confirmed the existence of an internal cleavage site, preceding the VP4/VP3 cleavage site. The resulting C-terminally truncated enzyme (ending at Ala716) is active, as shown by a trans full-length VP4 cleavage assay and a fluorometric peptide cleavage assay. We present a crystal structure of a native active site YAV VP4 with the internal cleavage site trapped as trans product complexes and trans acyl-enzyme complexes. The acyl-enzyme complexes confirm directly the role of Ser633 as the nucleophile. A crystal structure of the lysine general base mutant (K674A) reveals the acyl-enzyme and empty binding site states of VP4, which allows for the observation of structural changes upon substrate or product binding. These snapshots of three different stages in the VP4 protease reaction mechanism will aid in the design of anti-birnavirus compounds, provide insight into previous site-directed mutagenesis results, and contribute to understanding of the serine-lysine dyad protease mechanism. In addition, we have discovered that this protease contains a channel that leads from the enzyme surface (adjacent to the substrate binding groove) to the active site and the deacylating water.

among all VP4s in the family Birnaviridae. Sequence alignment analysis predicts Ser 633 to be the nucleophile and Lys 674 to be the general base in the YAV VP4 mechanism. Results from previous site-directed mutagenesis studies support the prediction of Lys 674 acting as the general base but fail to confirm Ser 633 as the nucleophile (24).
Crystal structures of VP4 protease from blotched snakehead virus (25), infectious pancreatic necrosis virus (26), and tellina virus 1 (27) have been elucidated previously. The blotched snakehead virus VP4 structure revealed a native active site with an empty binding groove. The crystal structures of infectious pancreatic necrosis virus VP4 showed an empty binding groove, intermolecular acyl-enzyme complexes, and productbound complexes in mutant active sites. The tellina virus 1 structure presented an intramolecular acyl-enzyme complex in a native active site but with a sulfate bound near the lysine general base.
Here we have demonstrated that YAV contains an internal cleavage site within the VP4 region of the polyprotein. Cleavage at this site produces a truncated yet proteolytically active enzyme that ends at Ala 716 . In one of the YAV VP4 crystal structures presented here, the C-terminal region (residues 711-716) of each molecule in the asymmetric unit is bound in the active site of an adjacent molecule, either in the form of a trans product complex or a trans acyl-enzyme complex. The acyl-enzyme complex clearly demonstrates that Ser 633 acts as the nucleophile in the YAV VP4 hydrolytic reaction. This is the first VP4 structure to show an intermolecular acyl-enzyme complex and product complex in the presence of a native active site and native substrate. The second crystal structure presented here is that of the YAV VP4 active site mutant (K674A). There are two molecules in the asymmetric unit, with one forming an intermolecular acyl-enzyme complex and the other showing an empty active site. This structure directly reveals changes that occur within the enzyme upon substrate binding. These crystal structures of YAV VP4 will be of value in the design of anti-birnavirus compounds and provide insights into the serine-lysine dyad catalytic mechanism. These structures also provide insights into previous site-directed mutagenesis studies. In addition, these structures reveal a channel that runs from the surface of VP4 to the proposed deacylating water.

EXPERIMENTAL PROCEDURES
YAV VP4 Constructs-The cDNA for segment A of YAV was generously provided by Dr. Syunichirou Oshima. The fulllength YAV VP4 construct (residues 509 -734, from the pVP2/ VP4 cleavage site to the VP4/VP3 cleavage site) was PCR-amplified using a forward primer with the sequence 5Ј-GGA CTC CAT GGC CAG CGG CAC AGA CAC TGG G-3Ј and a reverse primer with the sequence 5Ј-CTG GCC TCG AGT GCA GTT GTT CTC ATT AGT TCC CC-3Ј. Vent DNA polymerase (New England BioLabs) was used. The amplicon was cloned into the restriction sites NcoI and XhoI of plasmid pET28b ϩ (Novagen) using T4 DNA ligase (Fermentas). The ligation mix was transformed into Escherichia coli strain NovaBlue (Novagen) for plasmid isolation. This creates a full-length VP4 construct that contains residues 509 -734 from segment A of YAV, with two additional residues at the N terminus (Met-Ala) and eight addi-tional residues at the C terminus (Leu-Glu and a His 6 affinity tag) (Fig. 1A). The truncated YAV VP4 construct was amplified using the same forward primer as mentioned above along with a reverse primer with sequence 5Ј-CTG GCC TCG AGT GCT TTC TGC ACT GGT AGT GCT CC-3Ј. This DNA region was amplified using the polymerase chain reaction with Pfu DNA polymerase (Fermentas). The amplicon was cloned into the restriction sites NcoI and XhoI of plasmid pET28b ϩ (Novagen) using T4 DNA ligase (Fermentas). The ligation mix was transformed into E. coli strain NovaBlue (Novagen) for plasmid isolation. This creates a VP4 that contains residues 509 -716 from segment A of YAV, with the same additional residues at the termini as described for the full-length construct. To generate the active site mutant (K674A) constructs, site-directed mutagenesis reactions using Pfu DNA polymerase (Fermentas) were performed using primers with the following sequences: 5Ј-TGC GGT GTA GAC ATC GCA GCC ATC GCC GCC CAT-3Ј and 5Ј-ATG GGC GGC GAT GGC TGC GAT GTC TAC ACC GCA-3Ј. The NCBI reference sequence for segment A of YAV indicates an asparagine at position 616 (a surface residue), but the sequencing result indicates an aspartic acid residue at this position. We used site-directed mutagenesis to change the sequence to match the reference sequence (Asn 616 ). PfuUltra TM polymerase (Stratagene) along with the primers of sequences 5Ј-AAA GAG ATC AAG AAG AAC GGA AAC ATC GTG GTG-3Ј and 5Ј-CAC CAC GAT GTT TCC GTT CTT CTT GAT CTC TTT-3Ј were used in the reaction. The sequences of all VP4 constructs were verified by DNA sequencing (GenBank TM accession number NC_004168, UniProt accession number P89521).
Protein Expression and Purification-The expression vectors containing the YAV VP4 constructs were transformed into E. coli strain Tuner (DE3), followed by selection on Luria-Bertani (LB) agar plates supplemented with 50 g/ml kanamycin. Six liters of cultures were grown for each batch of protein purified. A 10-ml aliquot of overnight culture was inoculated into each liter of LB/kanamycin medium. The cultures were incubated at 37°C while shaking for 4 h before induction with 0.5 ml of 1 M isopropyl ␤-D-1-thiogalactopyranoside. The induced cultures grew overnight at 25°C. Cells were harvested by centrifugation at 9,110 ϫ g for 7 min. To facilitate cell lysis, the cell pellet was stored at Ϫ80°C for 15 min. The frozen cell pellets were then completely resuspended in lysis buffer (50 mM Tris-HCl buffer, pH 8.0, 10% glycerol, 1 mM dithiothreitol (DTT), 7 mM magnesium acetate, 0.1% Triton X-100, and 1 unit/ml benzonase). The cells were sonicated three times at 30% amplitude for 5 s with a 10-s rest between each of the pulses (Fisher sonic dismembrator model 500). The sonicated cells were then lysed using an Avestin Emulsiflex-3C cell homogenizer (1,250 bars for 3 min). Cell debris was removed by centrifugation at 28,964 ϫ g for 20 min. The resulting supernatant was applied to a 5-ml nickel-nitrilotriacetic acid metal affinity column (Qiagen). The column was equilibrated with 5 column volumes of standard buffer (20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 10% glycerol, and 1 mM DTT). A step gradient containing 100, 300, and 600 mM imidazole in standard buffer was used to elute the His 6 -tagged protein. Fractions positive for VP4 were pooled, concentrated to 5 ml, and loaded onto a size exclusion chroma-tography column (HiPrep 26/60 Sephacryl S-100 HR) equilibrated with crystallization buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 10% glycerol, and 1% ␤-mercaptoethanol). The flow rate of the buffer was controlled by an ÄKTA Prime TM system (GE Biosciences) set at 0.7 ml/min. Fractions with pure VP4 were pooled and concentrated to 30 mg/ml using a Millipore centrifugal filter with a molecular mass cut-off of 10 kDa. Protein concentration was measured with a Nanodrop UV spectrophotometer (Thermo Scientific) using an extinction coefficient of 9,970 M Ϫ1 cm Ϫ1 as calculated by the ProtParam server based on the amino acid sequence (28).
YAV VP4 Full-length Self-cleavage Assay-A 1.5-l aliquot of truncated VP4 (YAV VP4 509 -716 or YAV VP4 509 -716 K674A, 30 mg/ml) was added to 120 l of the reaction buffer (20 mM MES, pH 6.5; the same buffer and pH as the crystallization conditions). A 1.5-l aliquot of full-length YAV VP4 (residues 509 -734, 30 mg/ml) with a mutant active site (K674A) was added to this mixture as a substrate. The reaction was carried out at 23°C and 20-l aliquots were taken at times 0, 4, and 7 h and overnight. The samples were mixed with an equal volume of 2ϫ sample buffer and loaded onto a 15% SDS-polyacrylamide gel. The gel was then stained with Coomassie Blue for visualization.
Fluorometric Peptide Cleavage Assays-The fluorometric peptide substrate benzyloxycarbonyl-PVQKA-4-methylcoumaryl-7-amide was synthesized by LifeTein. The peptide was dissolved in 100% DMSO at a 50 mM stock concentration. Every reaction had a final DMSO concentration of 1% (v/v). Each 100-l reaction contained 100 M VP4 protease (YAV VP4 509 -716 or YAV VP4 509 -716 K674A) and 500 M fluorometric peptide in reaction buffer (20 mM MES, pH 6.5). The control reaction contained the fluorometric peptide in reaction buffer. The assay was performed in triplicate on a 96-well Clear black plate from Greiner Bio-One. The fluorescence (in relative fluorescence units) was monitored at 37°C using a SpectraMax M5 multimode microplate reader (Molecular Devices) with an excitation wavelength at 380 nm and emission wavelength at 460 nm. Data reduction was performed to overlay the graphs from three different conditions.
The YAV VP4 fluorometric peptide cleavage assay with and without DTT was performed with the substrate N,N-Dimethyl-p-aminobenzeneazobenzoic acid-KALPVQKAQG-ASE-5-[(2-aminoethyl)amino]naphthalene-1-sulfonic acid that was synthesized by ChinaPeptides. It contains residues P7-P4Ј (residues 710 -720) of the VP4 internal cleavage site. The peptide was dissolved in 100% DMSO at a 50 mM stock concentration, and every reaction had a final DMSO concentration of 1% (v/v). Each 100-l reaction contained 67 M VP4 protease (YAV VP4 509 -716) and 500 M fluorometric peptide in reaction buffer (20 mM MES, pH 6.5) with or without 10 mM DTT. In the control reaction, the native active site enzyme was replaced with the inactive mutant (YAV VP4 509 -716, K674A). The assay was run in triplicate on a 96-well Clear black plate from Greiner Bio-One. The fluorescence (in relative fluorescence units) was monitored at 37°C using a SpectraMax M5 multimode microplate reader (Molecular Devices) with an excitation wavelength at 340 nm and emission wavelength at 540 nm.
Crystallization-All crystallization trials were carried out at room temperature (ϳ296 K). The crystals used in the diffraction studies were grown using the sitting-drop vapor diffusion method. For the native active site VP4, a 1-l protein sample (30 mg/ml) was mixed with 1 l of reservoir reagent and allowed to reach vapor equilibrium with 1 ml of reservoir reagent in a tape-sealed chamber. The initial crystallization condition was obtained from the Hampton Research Crystal Screen I condition 23 (30% PEG 400, 0.1 M HEPES, pH 7.5, and 0.2 M magnesium chloride). The optimized condition gave rodshaped crystals that grew out of precipitation in 35% PEG 2000, 0.1 M MES, pH 6.5, and 0.3 M magnesium chloride. These monoclinic crystals belonged to space group P12 1 1 and had unit cell dimensions of 41.6 ϫ 64.3 ϫ 187.7 Å (␤ ϭ 95.8°) with five molecules in the asymmetric unit (45% solvent, Matthews coefficient of 2.24). For the active site mutant, 1 l of the native VP4 (30 mg/ml) was added to a 30-l aliquot of the active site mutant (30 mg/ml) immediately prior to crystal plating. Three microliters of this VP4 mixture was mixed with 3 l of the mother liquor and plated as described for the native active site VP4. The active site mutant crystals used for diffraction data collection grew out of 25% PEG 2000, 0.1 M MES, pH 6.5, and 0.45 M magnesium chloride. These cubic crystals had unit cell dimensions of 273.5 ϫ 273.5 ϫ 273.5 Å with two molecules in the asymmetric unit (solvent content 74%, Matthews coefficient of 4.79) and belonged to space group F4 1 32.
Data Collection-Although the two crystal forms grew out of very similar crystallization conditions, the cryo-solvent conditions for the two crystals were significantly different. The cryosolution for the native active site VP4 crystals was 30% 2-methyl-2,4-pentanediol in mother liquor (35% PEG 2000, 0.1 M MES, pH 6.5, and 0.3 M magnesium chloride), whereas 20% glycerol in the mother liquor (25% PEG 2000, 0.1 M MES, pH 6.5, and 0.45 M magnesium chloride) was used for the active site mutant crystals. The crystal of the native active site VP4 was transferred into cryo-solution immediately prior to mounting in a loop with a copper base. Diffraction data were collected at the macromolecular x-ray diffraction data collection facility in the Molecular Biology and Biochemistry Department of Simon Fraser University, which includes a MicroMax-007 Microfocus x-ray generator, osmic confocal VariMax high flux optics, an R-AXIS IVϩϩ image plate, and an X-stream 2000 cryosystem. The program CrystalClear was used for data collection. The crystal-to-detector distance was 220 mm. The crystal was exposed for 3 min with an oscillation angle of 0.5°and an x-ray wavelength of 1.5418 Å. The crystal of the active site mutant was transferred into cryo-solution and immediately mounted in a loop with a copper base and then flash-cooled with liquid nitrogen prior to transport to the Canadian Light Source for data collection at beamline 08ID-1. The crystal-to-detector distance was 310 mm. The crystal was exposed for 1 s with an oscillation angle of 0.4°and an x-ray wavelength of 0.97949 Å.
Structure Solution and Refinement-The indexing and integration for both data sets were carried out using MOSFLM (29), and scaling steps were performed with SCALA (30) from the CCP4i suite (31). The program POINTLESS was used to confirm the space group assignment (30). Molecular replacement was employed to obtain phase estimates. For the native active site VP4, the solution was found using the program MOLREP (32) utilizing the infectious pancreatic necrosis virus VP4 structure (chain A of PDB entry 2PNM) as the search model. Rigid body and restrained refinement were performed using REFMAC5 (33)(34)(35). For the active site mutant, the coordinates from the native active site YAV VP4 were used as a molecular replacement search model in the program PHASER (36). Rigid body and restrained refinements were performed using REFMAC5 (33)(34)(35). Editing and fitting of atomic coordinates were performed using the program Coot (37).

RESULTS AND DISCUSSION
Self-cleavage at the YAV VP4 Internal Cleavage Site-Based on sequence similarity with infectious pancreatic necrosis virus VP4, it was previously proposed that YAV VP4 (Fig. 1, A and B) contains an internal cleavage site (between residues 716 and 717) preceding the VP4/VP3 cleavage site, which is between residues 734 and 735 (24). Consistent with this proposal, we have observed that a truncated YAV VP4 construct (residues 509 -716) with a C-terminal His 6 tag self-cleaved the tag upon incubation.
We were unable to isolate the full-length wild-type YAV VP4 enzymes (residues 509 -734) due to a low expression level. However, the same construct with the active site mutation K674A was purified with a high yield. Therefore, we were able to test the ability of a truncated YAV VP4 with a native active site to cleave the full-length YAV VP4 protein at the internal cleavage site in a trans fashion. We found that the truncated YAV VP4 (residues 509 -716) was active and fully processed the full-length YAV VP4 (residues 509 -734, K674A) after an overnight incubation (Fig. 1C). In contrast, no processing was observed when the active site mutant (K674A) version of YAV VP4 509 -716 was used in the reaction (Fig. 1C). In addition, we used a fluorometric peptide assay to show that the YAV VP4 truncated at the internal cleavage site (residues 509 -716) is active. The peptide sequence used (PVQKA) within the substrate corresponds to residues P5-P1 of the internal cleavage site (residues 712-716 of the YAV polyprotein). This peptide is modified at the C terminus with a 4-methylcoumaryl-7-amide group. Upon cleavage of the amide bond between the 4-methylcoumaryl-7-amide group and C-terminal residue, the fluorophore 7-amino-4-methylcoumain is released, which results in a fluorescence emission signal at 460 nm when excited at 380 nm (50). We found that the truncated YAV VP4 (residues 509 -716) was active because it recognized and cleaved at the internal cleavage site sequence (Fig. 1D). In contrast, no activity was observed when the active enzyme was substituted with either the truncated active site mutant (residues 509 -716, K674A) or the reaction buffer control.
Self-cleavage at the YAV VP4 Internal Cleavage Site Promotes Crystallization-As described above, the native active site version of YAV VP4 with a C-terminal affinity tag slowly cleaved itself after Ala 716 , the P1 residue for the internal cleavage site. The resulting processed protein produced monoclinic crystals FIGURE 1. YAV VP4 protease truncated at the internal cleavage site is active. A, the genomic arrangement of segment A in YAV is shown. The major sites of cleavage are denoted by arrowheads with the P1 and P1Ј residue numbers listed. B, a list of known major YAV VP4 protease cleavage sites. The residues preceding P1Ј in the internal cleavage site are visible in the crystal structures and are boxed in red. C, to demonstrate processing at the internal cleavage site (after residue 716), truncated YAV VP4 (residues 509 -716) was incubated with full-length YAV VP4 with an active site mutation (residues 509 -734, K674A) for 0, 4, and 7 h and overnight at 23°C. As a control, the same experiment was repeated with an enzyme with an active site mutation (residues 509 -716, K674A). A 20-l aliquot of the reaction was run on an SDS-polyacrylamide gel and visualized by Coomassie stain. D, a fluorometric peptide (benzyloxycarbonyl-PVQKA-4-methylcoumaryl-7-amide (Z-PVQKA-MCA)) corresponding to the sequence for residues P5-P1 of the YAV VP4 internal cleavage site (residues 712-716 of the YAV polyprotein) was incubated with truncated YAV VP4 (E; residues 509 -716), truncated YAV VP4 active site mutant (Ⅺ; residues 509 -716, K674A) and reaction buffer (‚; 20 mM MES, pH 6.5) alone. The relative fluorescence units (RFU; reduced values) are plotted over time.
with five molecules in the asymmetric unit that diffracted to 2.5 Å resolution. An active site mutant version (K674A) of YAV VP4 produced cubic crystals with two molecules in the asymmetric unit that diffracted to 2.3 Å resolution. Interestingly, crystals of the active site mutant only appeared when a small amount of native VP4 was added to promote cleavage at Ala 716 . The refined crystal structures display electron density for residues 515-716 in each molecule of the asymmetric unit. The crystallographic statistics and validation are listed in Table 1.
Overall Architecture and Active Site of YAV VP4 Protease-Consistent with the theoretical isoelectric point of 4.6, the surface of YAV VP4 is predominantly negatively charged. The protease has an ␣/␤ protein fold consisting of 13 ␤-strands, four ␣-helices, and two 3 10 -helices ( Fig. 2A). ␤-Strands 1-7 form a platform on which the binding pockets for substrate recognition are constructed. Most of these ␤-strands interact in an anti-parallel fashion. Two 3 10 -helices (1 and 2) reside on a loop between ␤3 and ␤4. Strands ␤8, ␤11, and ␤12 form a parallel ␤-sheet flanked by three ␣-helices (␣2, ␣3, and ␣4) and a ␤-hairpin (␤9 and ␤10). This region hosts the catalytic dyad with the serine nucleophile (Ser 633 ) found immediately preceding ␣-helix 2 (␣2) and the lysine general base (Lys 674 ) arriving from ␣-helix 3 (␣3). There is clear electron density for the Lys 674 side chain in all five molecules in the asymmetric unit of the native active site structure. The N for Lys 674 has an average B-factor of 21.3 Å 2 . The N for Lys 674 is partially buried and has an average accessible surface area of 6.6 Å 2 . The average acces-sible surface area for the nine other lysine N atoms in chain A is 50.6 Å 2 . The N of Lys 674 is highly coordinated by the O␥1 of Thr 655 , which resides between ␤-strands 8 and 9, the O␥ of is the intensity of an individual reflection, and ͗I(hkl)͘ is the mean intensity of that reflection.
where F o and F c are the observed and calculated structure-factor amplitude, respectively. e R free is calculated using 5% of the reflections randomly excluded from refinement. f Wilson plot estimated B-factor was calculated using program Truncate from the CCP4 program suite (31,61). g Correlation coefficient was calculated using Refmac5 from the CCP4 program suite (31,35). h Ramachandran plot was calculated using Procheck from the CCP4 program suite (41). Cysteine Residues-YAV VP4 contains five cysteine residues at positions 588, 604, 634, 669, and 688. VP4 from other Birnavirus species have on average just two cysteine residues. Residues 588 and 604 form a disulfide bridge (Cys 588 -Cys 604 ) that links ␤-strands 5 and 6 ( Fig. 2A). This is the first disulfide bond observed within a VP4 protease. Despite being purified and crystallized in the presence of reducing agents, some of the molecules in the asymmetric unit for both structures show electron density off the thiol group of Cys 634 and Cys 669 , which is consistent with them being oxidized to S-hydroxycysteine (CSO), a sulfenic acid. Conversely, the electron density for Cys 688 S␥ always appears unmodified. The modification state for each cysteine is summarized in Table 2.
Different C-terminal Conformations-In the native active site structure, each protein chain in the asymmetric unit has a C terminus that is in an extended conformation ( Fig. 2A). In contrast, the active site mutant (K674A) structure reveals two sep-FIGURE 3. The YAV VP4 protease intermolecular (trans) acyl-enzyme complex, product complex, and empty active site. YAV VP4 protease is able to cleave at an internal cleavage site (after Ala 716 ) near its own C terminus (after Ala 734 ), the VP4/VP3 junction. Two crystal structures of YAV VP4 (with and without an active site mutation) reveal the internal cleavage site bound within the active site of a neighboring VP4 molecule. Analysis of the electron density within each molecule of the asymmetric unit reveals three different enzyme states of the protease reaction cycle (empty active site, acyl-enzyme, and product complex). The packing of YAV VP4 protease in the asymmetric unit for each structure is shown along with representative 2F o Ϫ F c electron density maps (contoured at 1) at the active site for representative types of complexes. A, in the monoclinic crystal of native active site YAV VP4, five molecules of VP4 are in the asymmetric unit with the C terminus of each molecule bound in the active site of its neighbor. The C-terminal carbonyl carbons of molecules D and E form an intermolecular acyl-enzyme complex (AE) with the active site nucleophilic serine O␥ of molecules C and D, respectively. An enzyme-product (EP) complex is formed between molecule pairs E/A, A/B, and B/C. B, in the cubic crystal of mutant YAV VP4 (Lys 674 general base mutated to alanine), two molecules are in the asymmetric unit. The C-terminal carbonyl carbon of molecule A forms an intermolecular acyl-enzyme complex (AE) with the nucleophilic serine O␥ of molecule B. The active site of molecule A remains in the unbound state (E) because the C terminus of molecule B (shown as red spheres) folds back onto itself instead of being in an extended conformation. MAY 3, 2013 • VOLUME 288 • NUMBER 18

JOURNAL OF BIOLOGICAL CHEMISTRY 13073
arate C-terminal conformations, one that is extended and another where the C terminus packs against the surface of the enzyme (Fig. 2B).
trans Acyl-Enzyme and Enzyme-Product Complexes in a Native Active Site-In the asymmetric unit of the native active site structure, each substrate binding groove is occupied by the last six residues (residues 711-716) of the adjacent molecule, thus stabilizing the extended conformation for each C terminus ( Figs. 2A, 3A, and 4A). The average buried surface area between the enzyme and the neighboring molecules is 894 Å 2 . Analysis of the electron density reveals that two of the five VP4 molecules in the asymmetric unit form acyl-enzyme complexes with

Chain B (acyl-enzyme)
Cys 634 ϫ ϫ ϫ ϫ Cys 669 ϫ ϫ ϫ ϫ neighboring VP4 molecules (Fig. 3A). The O␥ of the serine nucleophile (Ser 633 ) is covalently linked to the carbonyl carbon of Ala 716 , the P1 residue for the internal cleavage site. The electron density is consistent with a trigonal planar geometry that is expected for an ester bond (Fig. 3A). The P6 -P1 residues (residues 711-716) are stabilized by hydrogen bonding interactions in an anti-parallel fashion with residues 546 -550 on one side of the binding groove and interact in a parallel fashion with residues 627-630 on the other side of the binding groove (Fig. 4A). Five of the hydrogen bonds involved in linking the C terminus of one molecule to the binding groove of the neighboring molecule are conserved in both the native active structure and the mutant active site structure. Several water molecules participate in the hydrogen bonding network within the substrate binding groove. The water that bridges the main chain amide nitrogen of Gln 714 (P3) to the carbonyl oxygen of Pro 628 (Fig. 4) is seen in each of the observed reaction steps. A water that bridges Glu 550 N to Leu 711 O is observed in all of the acyl-enzyme complexes captured. In addition, chain B, which is involved in the product complex also has this water. In some of the acyl-enzyme and enzyme-product complexes, a water molecule bridges the O␥ of Thr 548 to the carbonyl oxygen of Val 546 and the amide nitrogen of Lys 715 (Fig.  4). The most interesting water seen in the enzyme-product complexes is the one that resides within the oxyanion hole (Fig.  4B). This water, which is displaced in the acyl-enzyme complex, makes hydrogen bonding interactions with the carboxylate oxygen of Ala 716 (P1 residue), the O␥ and main chain nitrogen of Ser 633 (nucleophile), and the main chain carbonyl oxygen of Met 630 (Fig. 4B).
The major substrate specificity pockets are the S1 and S3 (Figs. 4 -6). Residues Ile 543 , Pro 544 , Val 545 , Leu 561 , IIe 563 , Ile 592 , Ala 629 , Met 630 , Gly 631 , Pro 632 , Ser 633 , and Cys 634 form a continuous surface (the S1 pocket) for the binding of the P1 residue side chain (Ala 716 ). Residues Val 545 , His 547 , Ser 559 , Leu 560 , Leu 561 , Leu 572 , Gln 576 , Gly 591 , Ala 607 , Pro 609 , Val 624 , Ala 626 , Gly 627 , Pro 628 , and Ala 629 form a continuous surface (the S3 pocket) for the binding of the P3 residue side chain (Gln 714 ). A cleft that is open on one side accommodates the P5 residue (Pro 712 ). The side chain of Leu 711 (the P6 residue) points into a shallow pocket near the end of the substrate binding groove. There are no binding pockets for the P2 and P4 residues because their side chains are pointing toward the solvent, away from the substrate binding groove.
Substrate Binding Groove, Empty versus Bound-The crystal structure of the active site mutant of YAV VP4 (K674A) reveals an acyl-enzyme complex and an empty active site in the same asymmetric unit (Fig. 3B). This gives us the opportunity to observe structural differences within the substrate binding groove during these distinctly different enzyme states. One difference is seen in the loop, between ␤-strands 2 and 3, that resides adjacent to the substrate binding groove (Fig. 2). In the absence of a bound peptide (P6 -P1 from the neighboring molecule), this loop is more disordered, with clear density observed only for the main chain atoms. With bound peptide, this turn becomes ordered and shows clear density for both main chain and side chain atoms. Another clear difference between the bound and unbound states of VP4 is seen within the S3 specificity binding pocket. The side chains of His 547 and Val 545 both show a change in rotamer conformation, which results in a larger S3 binding pocket in the bound state (Fig. 6). The main chain atoms that make up the rim of the S3 pocket have also moved to widen the opening in the bound state.
Aqueous Channel Leads to Active Site-Interestingly, the S1 binding pocket in YAV VP4 is not fully enclosed within the acyl-enzyme or product complex structures; rather it is open to the surface of the enzyme via a channel (Fig. 7). This fenestration leads from the protein surface to an area near the proposed deacylating water within the active site. The channel is made up of atoms from the following residues: Ile 592 , Glu 594 , Asp 595 , Ile 596 , Pro 597 , Ala 629 , Met 630 , Gly 631 , Pro 632 , CSO 634 or Cys 634 , Gln 635 , and Leu 638 . The interior of the channel is filled with hydrogen bond donors and acceptors. In both the native active site and mutant active site structures, ordered water molecules are coordinated at and near the entrance of the channel by Glu 594 , Asp 595 , Ala 629 , Gly 631 , and Gln 635 . In the active site mutant structure, a water molecule is also found near the narrowest region of the channel (bottleneck) in the VP4 with the empty binding groove. However, such a water is absent in the acyl-enzyme containing VP4. Superposition of these molecules reveals no significant change in the position of the residues that form the channel. The average diameter at the narrowest point (bottleneck diameter) in the channel for the five molecules in the native active site structure is 2.1 Å. The residues that form the narrowest point of the channel are Cys 634 or CSO 634 , Ile 592 , Ala 629 , Met 630 , and Gly 631 (Fig. 8). Although the diameter of a water molecule is ϳ2.8 Å, there are examples of water channels with narrow passageways of approximately the same dimensions as seen in the VP4 channel. For example, a selectivity filter of around 2 Å in diameter has been reported in E. coli aquaporin Z, a water channel protein (51,52). It is possible that a deacylating (catalytic) water could utilize such a channel to position itself in the correct trajectory for an attack on the ester bond of the acyl-enzyme.
Ser 633 is the YAV VP4 Nucleophile, but Could Thr 655 O␥ Function as a Nucleophile in the Absence of Ser 633 O␥?-The serine/lysine catalytic dyad mechanism is conserved among all birnavirus VP4 proteases studied thus far. Therefore, Ser 633 and Lys 674 were predicted to be the nucleophile and general base, respectively, in YAV VP4 based on sequence alignments (5,24,27). Consistent with these predictions, the processing of the YAV segment A polyprotein was completely abolished when Lys 674 , the proposed general base, was mutated to an aspartic acid in a previous mutagenesis study (24). We show here that mutating Lys 674 to alanine also produces a inactive enzyme. However, previous mutagenesis experiments did not conclusively show that Ser 633 plays the role of the nucleophile. Both pVP2-VP4 and VP4-VP3 fragments from the YAV segment A polyprotein were detected when Ser 633 was mutated to a proline. Interestingly, similar results were also seen in infectious bursal disease virus mutagenesis experiments (53, 54). Our crystal structures of the YAV VP4 trans acyl-enzyme com-plex clearly support the initial hypothesis of Ser 633 being the nucleophile and Lys 674 serving as the general base. The acylenzyme complexes directly reveal electron density linking the O␥ of Ser 633 to the carbonyl carbon of Ala 716 (the P1 residue of the internal cleavage site) (Fig. 3).
Besides the nucleophilic Ser 633 , Thr 655 is the only titratable residue within the vicinity of the general base Lys 674 (Figs. 3 and  4). A serine or threonine is observed in a similar position in all other known proteases utilizing a Ser/Lys catalytic dyad mechanism (55). If it were possible for the hydroxyl of Thr 655 to function as the nucleophile in the absence of a Ser 633 hydroxyl, this would put the attack on the scissile bond in the opposite direction (re-rather than si-). This would probably affect the deacylation step of the reaction, resulting in a single cleavage event, thus providing a possible explanation for the presence of the pVP2-VP4 and VP4-VP3 precursors in the S633P mutant experiment.
Interpreting Non-active Site Residue Mutagenesis Results-Site-directed mutagenesis experiments show that changes to YAV VP4 non-active site residues can affect polyprotein processing (24). A cleavage product that was slightly smaller than the YAV segment A polyprotein was observed when Ile 543 was mutated to a glycine. The side chain of IIe 543 makes up part of the S1 binding pocket (Fig. 9). Replacing IIe 543 with a smaller residue like glycine would enlarge the S1 pocket and allow bulkier side chains to bind. Because S1 and S3 are the major specificity pockets present in the binding groove, such a mutation might significantly alter the substrate specificity. Thus, the enzyme is likely to be functional but with the ability to cleave the polyprotein at more locations. In a V686Q mutant, both the polyprotein and pVP2 were observed (24). This suggested that the cleavage at the VP4/VP3 junction was affected. Residue Val 686 does not lie in the active site, nor does it reside near the substrate specificity pockets. Instead, it is located adjacent to the C-terminal ␣-helix, with its main chain carbonyl oxygen forming a hydrogen bond with the amide nitrogen of Thr 655 , a residue that coordinates the N of the general base (Fig. 9) The main chain interaction at this position is observed in all other VP4 structures solved thus far (25)(26)(27). Because this mutant still generated pVP2 by cleaving at the VP2/VP4 junction, it is unlikely that it affected the catalytic dyad. Amino acid sequence alignment analysis revealed that uncharged residues, such as valine, isoleucine, leucine, and alanine, are found at this position in other birnaviruses. Replacement of Val 686 with a more polar residue like glutamine could affect the packing around the C terminus, the substrate at the VP4/VP3 junction. Moreover, Leu 705 , which resides on the conserved ␣-helix 4 near the C terminus, is located across from Val 686 . The packing interactions of this last ␣-helix are likely to be required for the proper presentation of the VP4/VP3 cleavage site.
cis versus trans Cleavage at the YAV VP4 Internal Cleavage Site-All of the molecules in the crystal structures presented here reveal intermolecular (trans) associations between the cleavage site and active site, but is this what happens in solution? Sequence and structural analysis suggests that in solution, the internal cleavage site processing probably only occurs via an intermolecular (trans) reaction. The previous tellina virus 1 VP4 structure, which does not have an internal cleavage site, reveals that the VP4/VP3 cleavage site (position 734/735 in the YAV polyprotein sequence) is just long enough to form the intramolecular (cis) acyl-enzyme complex (27). Therefore, the  A channel leads to the active site of YAV VP4 protease. A, a channel (red surface), found adjacent to the substrate binding groove, leads from the enzyme surface to a region near the catalytic residues (shown in a stick representation) and the proposed deacylating water. B, the residues surrounding the channel are shown in copper color. The neighboring protein chain forming a product complex is shown in a magenta schematic. C, a close-up view of the channel with the S1 and S3 pockets circled in orange. D, a cross-section view of the channel reveals that it is continuous with the S1 binding pocket (orange circle) and the pocket where the deacylating water resides (pink).
internal cleavage site (position 716/717) in YAV would not be accessible for an intramolecular (cis) cleavage event without significant unfolding of the secondary structure. The intermolecular processing of the internal cleavage site is further supported by our trans-cleavage assay of the full-length YAV VP4 and our observation that the active site mutant enzyme (K674A) would not crystallize until a small amount of the active native active site enzyme was added to cleave the inactive mutant at the internal cleavage site.
How do these intermolecular acyl-enzyme complexes in VP4 protease form, and why are they stable enough to be seen in the crystal structure? With the mutant enzyme, we had to add a small amount of native active site enzyme to cleave at the internal cleavage site before crystals would form. This suggests that the internal cleavage site product was presented to the VP4 active site, driving the reaction in the reverse direction to form the acyl-enzyme. With the general base absent, there is no suitable deprotonated functional group in the vicinity to activate the would-be deacylating water, thus stabilizing the acyl-enzyme complex. It is not so easy to explain how the acyl-enzyme is stabilized in the native active site enzyme. The crystals were formed at pH 6.5, which is probably well below the pK a of the lysine general base; thus, the low pH could be slowing down the deacylation step. It is also possible that the internal cleavage site plays a regulatory role in the VP4 protease. The glutamine found at the P3 position of the internal cleavage site is significantly longer and more polar than the residue seen at the P3 position of the pVP2/VP4 and VP4/VP3 cleavage sites (Ser 506 and Thr 732 , respectively). This allows Gln 714 to make multiple hydrogen bonding interactions with residues at the bottom of the S3 pocket. It is possible that these additional interactions slow down the release of the internal cleavage site from the VP4  binding groove, thereby acting as an internal inhibitor that functions after the main cleavage sites are processed. Interestingly, there have been previous reports of VP4 assembling into tubules (56).
Can Cysteine Modification Have an Effect on YAV VP4 Activity?-The thiol side chain of cysteine can be oxidized by biological oxidants to form CSO, a sulfenic acid. Increasingly, this posttranslational modification reaction is being investigated as a means of regulation (57)(58)(59)(60). Interestingly, despite being purified in the presence of reducing agents, the electron densities for two of the five cysteine residues in YAV VP4 (Cys 634 and Cys 669 ) are consistent with them being oxidized to CSO in some of the molecules in the asymmetric unit (Table 2 and Fig. 10A). This brings up the following question. Could the state of oxidation of these cysteine side chains affect the activity of YAV VP4? To address this question, we incubated VP4 in the presence and absence of an excess of reducing agent and tested its ability to cleave a fluorometric peptide substrate with a sequence matching that of the YAV VP4 internal cleavage site. The reducing environment did not affect the activity (Fig. 10B). Because Cys 634 immediately follows the nucleophile (Ser 633 ), this prompted us to analyze the structures to see if there was a correlation between the oxidation state of Cys 634 and the acylation state of Ser 633 . Two of the five molecules in the native active site structure form an acyl-enzyme complex (molecules C and D). In molecule C, Cys 634 (C) is modified as a CSO, and Ser 633 (C) is covalently bonded to Ala 716 (D), forming an acylenzyme, whereas in molecule D, Cys 634 (D) is not modified, but Ser 633 (D) is again covalently bonded to Ala 716 (E), forming an acyl-enzyme (Table 2). Therefore, the electron density of the crystal structure does not support the suggestion that modification (oxidation to CSO) of Cys 634 affects the activity of YAV VP4. By looking at the native active site structure, we can see that the main chain carbonyl oxygen of Cys 669 is within hydrogen bonding distance to the N of the general base lysine (Lys 674 ). However, no significant change in hydrogen bonding distance is seen between the N of the lysine general base and the O␥ of the serine nucleophile (Ser 633 ) when molecules with different oxidization states at Cys 669 were compared. In addition, no correlation appears to exist between the cysteine modification and the geometry of the YAV VP4 aqueous channel (Fig. 8).

CONCLUSION
In this study, we have shown that YAV VP4 contains an internal cleavage site (position 716/717) near its C terminus (residue 734). We have shown that the resulting truncated VP4 is proteolytically active. We present for the first time the structure of a serine-lysine protease trans acyl-enzyme complex with a native substrate in a native active site. The acyl-enzyme structures directly prove the role of Ser 633 as the nucleophile in the YAV VP4 reaction mechanism. A structure of a YAV VP4 mutant with the lysine general base changed to alanine is also presented. Together, these two structures reveal the substrate binding site for this protease in three separate stages of the protease reaction cycle: empty active site, acyl-enzyme complex, and product complex. Analysis of the two structures has helped us to provide insights into previous mutagenesis results and will be of value in the design of anti-birnavirus compounds and in the basic understanding of the serine-lysine dyad catalytic mechanism. We have discovered that this protease contains a channel that leads from the enzyme surface to the proposed deacylating water.