Active human cytomegalovirus protease is a dimer.

The quaternary state of the human cytomegalovirus (hCMV) protease has been analyzed in relation to its catalysis of peptide hydrolysis. Based on results obtained from steady state kinetics, size exclusion chromatography, and velocity sedimentation, the hCMV protease exists in a monomer-dimer equilibrium. Dimerization of the protease is enhanced by the presence of glycerol and high concentrations of enzyme. Isolation of monomeric and dimeric species eluted from a size exclusion column, followed by immediate assay, identifies the dimer as the active species. Activity measurements conducted with a range of enzyme concentrations are also consistent with a kinetic model in which only the dimeric hCMV protease is active. Using this model, the dissociation constant of the protease is 6.6 microM in 10% glycerol and 0.55 microM in 20% glycerol at 30 degrees C and pH 7.5.

The quaternary state of the human cytomegalovirus (hCMV) protease has been analyzed in relation to its catalysis of peptide hydrolysis. Based on results obtained from steady state kinetics, size exclusion chromatography, and velocity sedimentation, the hCMV protease exists in a monomer-dimer equilibrium. Dimerization of the protease is enhanced by the presence of glycerol and high concentrations of enzyme. Isolation of monomeric and dimeric species eluted from a size exclusion column, followed by immediate assay, identifies the dimer as the active species. Activity measurements conducted with a range of enzyme concentrations are also consistent with a kinetic model in which only the dimeric hCMV protease is active. Using this model, the dissociation constant of the protease is 6.6 M in 10% glycerol and 0.55 M in 20% glycerol at 30°C and pH 7.5.
Viruses of the herpes family, including the human cytomegalovirus (hCMV) 1 and herpes simplex virus, encode a protease essential for viral capsid formation and viral replication (1)(2)(3). The herpesvirus proteases are synthesized as precursor proteins that undergo autoproteolytic processing during viral assembly. One of the natural substrates is the viral assembly protein involved in the construction of intermediate viral capsids within the infected cell nucleus. The other natural substrate is the protease precursor protein. The protease catalytic domain is localized in the N terminus of the precursor, which in the case of hCMV encompasses the N-terminal 256 amino acids of the 708-amino acid precursor protein (4,5).
It has been suggested that the hCMV protease is a serine protease based on its chemical reactivity toward classical serine protease inhibitors (6), and recent site-directed mutagenesis data (7) have implicated the catalytic triad of hCMV protease to be Ser-132, His-63, and Glu-122. However, the catalytic efficiency of the hCMV protease is orders of magnitude less than that expected of a classical serine protease, and no amino acid sequence homology has been found between this enzyme and the well characterized serine proteases (7,8).
This report identifies the dimerization of mature hCMV protease as a critical property governing its catalytic activity. Our data are consistent with a dimer dissociation constant (K d ) in the low micromolar range with the dimeric protease being the active species. Given the proposed identity of the catalytic triad and the classification of this enzyme as a serine protease, our finding marks activation by dimerization as a unique feature of this new member of the serine protease group.

MATERIALS AND METHODS
Enzyme Expression and Purification-Both the wild-type and a mutant form (V141G and V207G mutations) of the hCMV protease were expressed in Escherichia coli as described previously (9). The mutations render the hCMV protease resistant to autoproteolysis. 2 All purification steps were performed at 0 -4°C. For the mutant, lysis of cells was performed with a microfluidizer (Microfluidics Corp., Newton, MA) in 50 mM Tris-HCl, pH 8.0 buffer containing 10% glycerol (v/v throughout), 25 mM NaCl, 1 mM EDTA, 1 mM DTT, and 1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride. The lysate was centrifuged, and the pellet was washed with lysis buffer plus 0.1% Nonidet P-40 and recentrifuged. Inclusion bodies were dissolved with 7 M urea, 50 mM Tris-HCl, 5 mM DTT, pH 8.0, followed by centrifugation and chromatography on a MonoQ column (Pharmacia Biotech Inc.). Elution was performed using a sodium chloride gradient in the urea-containing buffer. Protein folding was accomplished by dilution of protease-containing fractions to 0.2 mg/ml into 25 mM Tris-HCl (pH 7.5), 10% glycerol, 1 mM EDTA, 5 mM DTT, and 1 M guanidine HCl, followed by dialysis in the same buffer without guanidine HCl for 24 h. The resulting protein solution was chromatographed on a MonoQ column and eluted with a sodium chloride gradient in 25 mM Tris-HCl (pH 7.5), 10% glycerol, 1 mM EDTA, and 1 mM DTT to yield the purified enzyme. The wild-type protease was purified similarly. To avoid self-proteolysis, the protein obtained from the first MonoQ step was folded and dialyzed as above, but in the absence of glycerol. The sample was then acidified to pH 5.5 with MES, applied to a Merck Fractogel SO 3 column in 50 mM MES, 1 mM EDTA, 1 mM DTT, pH 5.5, and eluted with a sodium chloride gradient. The wild-type enzyme was stable at pH 5.5 and returned to full activity following dilution to pH 7.5. Enzyme preparations were greater than 95% pure by SDS-PAGE and gave the expected amino acid analysis. The N-terminal 5 residues of both enzymes were MTMDE, showing retention but deformylation of the initial N-formyl methionine. Electrospray mass spectrometry indicated a single species within 10 atomic mass units of the expected mass. The concentrations of stock enzyme solutions were determined by quantitative amino acid analysis. The data presented here were collected with use of the stable V141G/V207G mutant enzyme. The wild-type enzyme was employed to confirm all kinetic phenomena as characteristic of the hCMV protease.
Kinetic Assays and Equilibrium Constants-Peptide substrates used were either the fluorophore-labeled (Dabcyl)-RGVVNASSRLA-(Edans) (Bachem Biosciences, Philadelphia, PA; Ref. 10) or Ac-RWGVVNAS.Abu.RLATR-amide (Midwest Biotech, Indianapolis, IN). Products were quantified on high pressure liquid chromatography with fluorescence monitoring of the SSRLA-(Edans) product (350 nm excitation, 500 nm emission) or the Ac-RWGVVNA product (280 nm excitation, 350 nm emission). Reactions were performed for 1 min in a pH 7.5 buffer containing 52 mM MES, 52 mM TAPSO and 100 mM diethanolamine, 1 mM EDTA, 1 mM DTT, 0.05% BSA, and various concentrations of glycerol, and were quenched with the addition of urea to a final concentration of 3 M. For all of the kinetic assays conducted, hydrolysis consumed less than 15% of the substrate, and activities shown in the figures are averages of two determinations. With the exception of the * 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. 1 The abbreviations used are: hCMV, human cytomegalovirus; Abu, L-␣-aminobutyric acid; Dabcyl, 4-(4Ј-dimethylaminophenazo)benzoic acid; Edans where v d is the velocity for dimers, [E] t is the total concentration of enzyme (in monomer equivalents), and [M] is the concentration of monomers. The value of [M] is given by and is derived from the equilibrium condition.
Sedimentation Velocity-hCMV protease was equilibrated in 100 mM sodium HEPES, pH 7.5, 1 mM EDTA, 1 mM DTT, either with or without 20% glycerol, using Bio-Rad Biospin 6 spin columns. Samples (425 l) were sedimented at 20°C, 45,000 rpm in a Beckman XLA analytical ultracentrifuge. Scans were recorded at 280 nm using 0.003-cm point spacing every 8 min for total run time of 4 h. The data were fit to an approximate solution of the Lamm equation using the program SVED-BERG (11). The solvent densities were measured to be 1.0063 in the absence and 1.0619 g/ml in the presence of 20% glycerol using an Anton Paar DMA48 density meter. The partial specific volume of hCMV protease, v , was calculated to be 0.727 at 20°C using the method of Cohn and Edsall (12). Glycerol induces preferential hydration of proteins and causes v to increase. Using the data of Gekko and Timasheff, a correction factor for was found to be ⌬v /⌬ (% volume glycerol) ϭ 3.33 Ϯ 0.38 ϫ 10 Ϫ4 (average of 4 proteins) (13).
Size Exclusion Chromatography-Either a Pharmacia Superdex 75 column (300 ϫ 10 mm) or two Bio-Rad BioSelect 125 columns (300 ϫ 7.4 mm) connected in series were used immersed in a controlled-temperature water bath. Chromatography with the Superdex column was at 20°C in buffer (0.1 M sodium HEPES, pH 7.5, 1 mM EDTA, 1 mM DTT) containing either 0% or 20% glycerol, and sample volumes injected were 200 l. Chromatography with the BioSelect columns was at 10°C except where indicated, and in pH 7.5 buffer containing 52 mM MES, 52 mM TAPSO, 100 mM diethanolamine, 1 mM EDTA, and 20% glycerol (assay buffer without BSA). For K d determinations, samples were maintained at 30°C in a controlled temperature sample compartment prior to loading onto columns equilibrated in 20% glycerol-containing buffer at 10°C. Detection of eluting hCMV protease was by simultaneous monitoring of absorbance at 280 nm and fluorescence (excitation 280 nm, emission 350 nm). Comparison of integrated peak areas using the two methods showed a consistent 1.26-fold higher fluorescence response for monomers than for dimers, so that calculations of total protein in the monomer fluorescence peaks were adjusted accordingly. The absorbance extinction coefficients at 280 nm for monomers and dimers were assumed to be equal. The peak shapes and intensities obtained were unchanged for samples preincubated at 30°C from 1.5 to 8 h. Column buffers were degassed with and maintained under helium. Proteins used for standardization were BSA (64 kDa), ovalbumin (43 kDa), chymotrypsinogen (23.2 kDa), myoglobin (16.9 kDa), and bovine pancreatic RNase (12.6 kDa). Complete sets of standards were run at the beginning, the end, and in the middle of each solvent study, with little variation in elution volumes.

Dependence of Protease Specific Activity on Enzyme and
Glycerol Concentrations-Dilution of a concentrated solution of hCMV protease produces a time-dependent change in activity to a new, lower level. The rate and extent of the change in protease activity is a function of temperature. At 37°C, the decrease in protease activity occurs within minutes, while at 0°C, no significant change is observable for a period of hours. The change in hCMV protease activity at four temperatures upon dilution as a function of time is shown in Fig. 1. At temperatures above 20°C, the change in activity upon enzyme dilution is rapid enough (t 1 ⁄2 Յ 30 min) that an accurate assess-ment of activity requires a short assay (ϳ1 min).
The specific activity of the hCMV protease increases at higher enzyme and glycerol concentrations. These effects at 30°C are shown in Fig. 2. The specific activities of the protease measured after enzyme dilution and incubations of 1.5 and 3.5 h prior to reaction are shown in Fig. 2A. The data reveal that the hCMV protease specific activity tends toward zero as its concentration is lowered. The negligible difference between the determinations at 1.5 and 3.5 h suggests that the active form of the enzyme has reached equilibrium within 1.5 h, for both the 10% and 20% glycerol samples. The activity of the protease incubated in 10% glycerol (v/v) is lower than that incubated in 20% glycerol (v/v). It can be shifted back to the higher activity seen for 20% glycerol by addition of an equal volume of buffer containing 30% glycerol to produce a solution containing 20% glycerol, followed by further incubation. Thus, the dependence of hCMV protease activity on enzyme and glycerol concentrations is reversible. The reversal effect is shown in Fig. 2B.
The kinetic parameters V max and K m for the substrate (Dabcyl)-RGVVNASSRLA-(Edans) were determined with different hCMV protease concentrations preincubated in 20% glycerol using assay conditions as in Fig. 2. The V max and K m values obtained with 0.5 M enzyme are 360 nmol min Ϫ1 mg Ϫ1 and 92 M, respectively, while for 2.0 M enzyme they are 600 nmol min Ϫ1 mg Ϫ1 and 97 M. Thus it is the apparent turnover rate of the enzyme that varies with enzyme concentration.
Analytical Centrifugation-The quaternary state of the hCMV protease was characterized by sedimentation velocity measurements. Table I shows the results of sedimentation analyses of samples at a concentration of 20 M in the presence or absence of 20% glycerol (v/v). An approximate solution of the Lamm equation can be fit directly to the data to obtain the sedimentation coefficient, s 20,w , the diffusion coefficient, D 20,w , and by use of the Svedberg relation, the molecular weight (11). For both samples good fits are found for a single sedimenting species. In the absence of glycerol the molecular weight obtained is 29.7 ϫ 10 3 , which is close to the molecular weight of a monomeric hCMV protease. In 20% glycerol, 48.5 ϫ 10 3 is obtained, indicating that the enzyme is in a predominantly   FIG. 1. Activity of the hCMV protease as a function of time  after dilution at 0, 25, 30, and 37°C. A 2.5 M hCMV protease sample (monomer equivalents) at 0°C was diluted 50-fold to 50 nM into a buffer (52 mM MES, 52 mM TAPSO, 100 mM diethanolamine, 1 mM EDTA, 1 mM DTT, 20% glycerol, 0.05% BSA, pH 7.5) at the temperature indicated, and at the times shown an aliquot was withdrawn and assayed with the substrate Ac-GVVNAS.Abu.RLATR-amide at the same temperature. Assays were performed for a period of 1 min as described under "Materials and Methods." dimeric form. 3 The increase in s 20,w and the decrease in D 20,w shown in Table I are both consistent with the existence of a stable dimeric protease in 20% glycerol.
Size Exclusion Chromatography-The hCMV protease preincubated in the absence of glycerol and applied to a size exclusion column elutes with an apparent molecular weight of 37 ϫ 10 3 or 56 ϫ 10 3 , depending the loading concentration of the protease. With 20% glycerol present during preincubation, the protease elutes as a single 56 ϫ 10 3 species except with relatively low protease loading concentrations (Ͻ10 M), where two fractions emerge with apparent weights of 55 ϫ 10 3 and 33 ϫ 10 3 , as summarized in Table II. Since the molecular weight of the hCMV protease calculated from amino acid sequence is 28 ϫ 10 3 , these results suggest that the protease exists in a monomer-dimer equilibrium. Eluted enzyme samples corresponding to a dimeric protease show no evidence of covalent (disulfide) linkages as demonstrated by SDS-PAGE under nonreducing conditions. The elution profiles obtained from samples equilibrated in 20% glycerol at 30°C, analyzed on the BioSelect columns run at 10°C in 20% glycerol, are shown in Fig. 3A.
Assignment of the early and late elution peaks in Fig. 3A as dimer and monomer, respectively, allows the estimation of a K d for dimerization. Using the ratios of the areas under-the-peak of the early and late peaks, a K d value of 0.54 M is found for 20% glycerol as shown in Fig. 3B. The maximum fraction of dimeric protease extrapolated from Fig. 3B is 0.92. In separate experiments using protease at a loading concentration of 100 M, the maximum fraction of dimeric protease is Ն0.95. When activity assays are conducted at 0°C, no detectable activity is found in the eluted peak corresponding to the monomer while hydrolytic activity (Ͼ50-fold of detectable level) is found for the dimer peak. The same analysis applied to enzyme pre-equilibrated in 10% glycerol produces a K d of 5.5 M. The activity data sets shown in Fig. 2 can now be justifiably treated with a model wherein an inactive monomeric hCMV protease exists in equilibrium with an active dimeric hCMV protease. Fits of this model (see "Materials and Methods") to the kinetic data (Fig. 2) give average dissociation constants (K d ) for hCMV protease of 6.6 M in 10% glycerol and 0.55 M in 20% glycerol.
Changes in sample loading volume, column temperature, and chromatography time have been made to affirm that equilibrium exchange between protease monomers and dimers is negligible during size exclusion chromatography at 10°C in 20% glycerol. No significant variation in dimer-monomer peak ratios occurs when injection volumes of 5, 10, 20, or 30 l (15-90 pmol) of protease sample are made. Column tempera- 3 Similar results (J. Cole, unpublished data) were obtained from equilibrium sedimentation measurements. At protein concentrations below 1 M, we find that in the absence of glycerol the molecular weight of the hCMV protease is within 5% of the monomer value. In the presence of 20% glycerol, the apparent weight-average molecular weights increase at intermediate protein concentrations and then decrease at higher concentrations; this behavior is typical of a dimerizing protein which exhibits thermodynamic nonideality (14). The sedimentation equilibrium data obtained in the presence of 20% glycerol at a protease concentration of 50 M fit to a nonideal dimer model giving an apparent molecular weight of 52,100. A thorough sedimentation equilibrium study will be presented separately. a At low concentrations of enzyme, two 280 nm absorbing peaks were resolved in the presence of glycerol in the chromatography buffer. The apparent molecular weight is listed for each peak. tures of 5, 10, and 18°C produce essentially identical results as well. Some coalescence of elution peaks toward the monomeric form is observed at 25°C, and complete peak merging occurs at 30°C to yield mostly monomer. With the chromatography temperature at 10°C, as in the analyses presented here, the dimermonomer ratio has also been compared for the use of one versus two sizing columns. While the resolution with two columns in tandem (shown in Fig. 2A) is slightly better than with one column alone, the dimer-monomer ratio observed is identical, despite the fact that chromatography runs are completed in one-half the time with the single column. We conclude that despite an approximate 100-fold dilution of sample during the chromatography at 10°C, the aggregation state of the sample upon injection is well approximated by the elution patterns observed, due to the slow monomer-dimer equilibration at low temperatures.

DISCUSSION
This report describes our findings that the hCMV protease activity is dependent upon protein dimerization. That the activity of the protease is a function of time after dilution (Fig. 1) and of enzyme concentration (Fig. 2) suggests the existence of a protomer-oligomer equilibrium with the oligomeric enzyme being the active species. Physical data from sedimentation velocity and size exclusion chromatography experiments are evidence that the multimeric form of the protease is a dimer. Sedimentation velocity data (Table I) obtained in the absence of glycerol gives 29.7 ϫ 10 3 as the molecular weight of the hCMV protease, as expected for a monomeric enzyme species. In the presence of 20% glycerol, molecular weights obtained from sedimentation velocity (Table I) and sedimentation equilibrium 3 runs are consistent with a dimeric protease. The molecular weights estimated from size exclusion chromatography for the two enzyme species (Table II) are relative values that depend upon the calibrating proteins used, but they are in agreement with sedimentation results. Most importantly, the conversion of the protease from dimers to monomers is quantitatively reversible (Fig. 2B). The isolated monomer is found to be inactive, when tested at 0°C to prevent re-equilibration, while its dimeric counterpart is active (Fig. 3). These results, when considered together, are consistent with a simple monomer-dimer equilibrium model in which only the dimeric protease is the active form (Scheme 1).
Although only the dimer form of the protease is depicted to bind substrates in Scheme 1, 4 it is not possible for us to state unequivocally that the monomeric form of the enzyme does not bind substrate or does not possess a very low catalytic activity. Also unclear at present is the stoichiometry of substrate binding to the enzyme. Our efforts continue in clarifying these issues.
Scheme 1 further defines the kinetic parameters of the hCMV protease based on the results reported here. Determinations of the K d can be complicated during substrate turnover by the interchange of enzyme forms on the minute time scale (Fig. 1), so a short reaction time is required for accurate estimates. Our kinetic results (Fig. 2) indicate that the affinity of the monomeric protease for itself at 30°C is weak with a K d in the low micromolar range: 6.6 and 0.54 M in the presence of 10% and 20% glycerol, respectively. Essentially the same K d values, 5.5 and 0.55 M, respectively, are obtained from size exclusion measurements (Fig. 3B). Kinetic estimates of K d with the alternate substrate Ac-RWGVVNA-S.Abu.RLATR-amide are in complete agreement with these values (data not shown).
Given Scheme 1 and the rate of activity relaxation at 30°C shown in Fig. 1, the 1-min assay measurements should approximate the dimer catalytic activity prior to subunit dissociation by dilution and prior to substrate perturbation of the monomerdimer equilibrium. The V max measurements then allow calculation of a k cat for fully dimeric enzyme using the assumption of a single active site per subunit. The V max values in 20% glycerol of 360 and 600 nmol min Ϫ1 mg Ϫ1 for 0.5 and 2.0 M total enzyme, respectively, combined with the corresponding fractions of enzyme present as dimer calculated from the K d , 0.48 and 0.69, give similar k cat values of 21.2 and 24.2 min Ϫ1 . The nm. Injection volumes were adjusted to give the same total protein injected (30 pmol, monomer equivalents). B, fraction of total protein appearing in the dimer peak (26.2 min) as a function of protein concentration. The solid line corresponds to a fit of the monomer-dimer equilibrium function (see "Materials and Methods") to the data, yielding a K d of 0.54 M, with a maximum dimer fraction of 0.92. In separate experiments using a loading concentration of 100 M protease, the maximum fraction of dimeric protease is Ն0.95, suggesting that all the hCMV protease in our purified sample is fully capable of dimerization.
velocity of dimeric enzyme, v d , expected to be found in the enzyme titration of Fig. 2 can be calculated from the average of these k cat values, 22.7 min Ϫ1 , taking consideration of substrate concentration used (92 M) relative to the average K m under these conditions (94 M) to yield v d equal to 401 nmol min Ϫ1 mg Ϫ1 . In fact, v d from the curve fit of data in Fig. 2 yields 400 nmol min Ϫ1 mg Ϫ1 in 20% glycerol. The true k cat /K m value for this dimeric enzyme with (Dabcyl)-RGVVNASSRLA-(Edans) at 30°C in 20% glycerol is therefore 4000 M Ϫ1 s Ϫ1 , several times higher than previously reported for this and related substrates (8 -10).
In summary, the data presented here reveal that the hCMV protease exists as a dimeric species under conditions that enhance the specific activity of the enzyme, such as high enzyme concentration or the presence of glycerol. The activating effect of glycerol has been noted previously (8,15), but not the underlying mechanism of activation. Activity of the protease observed in assays containing no glycerol is likely due to a small concentration of dimers in equilibrium with monomers. Although unusual for a serine protease, this mode of activation may provide an appropriate temporal trigger for proteolytic activity during the assembly of hCMV capsids, a process wherein the protein components become highly concentrated.