Functional domains of bacteriophage P22 scaffolding protein1

Article No. mb981917

J. Mol. Biol. (1998) 281, 69±79

Functional Domains of Bacteriophage P22 Scaffolding Protein Matthew H. Parker1, Sherwood Casjens2 and Peter E. Prevelige Jr1* 1

Department of Microbiology University of Alabama at Birmingham, Birmingham AL 35294, USA 2

Department of Oncological Science, University of Utah Health Science Center Salt Lake City, UT 84132, USA

Assembly of the bacteriophage P22 requires a 303 amino acid residue scaffolding protein. Two scaffolding protein deletion mutants, consisting of residues 141 to 303 and 141 to 292, have been described. We report here that the 141-303 fragment, but not the 141-292 fragment, promoted procapsid assembly in vitro, bound to preformed shells of coat protein, and bound to a coat protein af®nity column. These ®ndings suggest that the carboxyl-terminal half of the scaffolding protein is suf®cient for promoting assembly, and that the 11 amino acid residues at the extreme carboxyl terminus are required for binding to the coat protein. Analysis of the products of in vitro assembly reactions suggests that the maximum amount of scaffolding protein that can pack into a procapsid is dictated by the internal volume of the procapsid rather than by a ®nite number of binding sites. However, when the amount of scaffolding protein was reduced to limiting values, both the wild-type protein and the 141-303 fragment assembled procapsids with the same number, rather than the same mass, of scaffolding protein molecules. When the 141-292 fragment was added to a mixture of coat and scaffolding proteins, the initial phase of procapsid assembly was inhibited, but the ®nal yield and composition of the procapsids were not affected. Assembly by a covalent dimeric mutant scaffolding protein (R74C/L177I) was not inhibited by the 141-292 fragment, which suggests that the inhibition is due to the formation of inactive heterodimers between the 141-292 fragment and the monomeric scaffolding protein. The 141-303 fragment, which has less tendency to self-associate than the wild-type protein, formed aberrant species as well as normal procapsid-like particles when the rate of assembly was high, suggesting that scaffolding protein dimerization may play a role in ensuring ®delity of assembly. Alternatively, residues 1 to 140 may play a direct structural role in preventing inappropriate scaffolding/ coat protein interactions. # 1998 Academic Press

*Corresponding author

Keywords: bacteriophage P22; virus assembly; scaffolding protein; self-assembling systems; protein-protein interactions

Introduction The assembly pathway of the Salmonella typhimurium bacteriophage P22 has been characterized in considerable detail (reviewed by Prevelige & King, 1993). Like other double-stranded DNA (dsDNA) phage (Hendrix, 1985; Casjens & Hendrix, 1988), as well as herpesviruses (Rixon, 1993; Donaghy & Jupp, 1995; Hong et al., 1996; Abbreviations used: dsDNA, double-stranded DNA; GuHCl, guanidinium hydrochloride; HSV-1, herpes simplex virus type 1; CMV, cytomegalovirus. E-mail address of the corresponding author: [email protected] 0022±2836/98/310069±11 $30.00/0

Newcomb et al., 1996; Trus et al., 1996; Wood et al., 1997) and adenoviruses (D'Halluin et al., 1978), P22 initially assembles a metastable structure called a procapsid. The P22 procapsid has T ˆ 7 icosahedral symmetry and consists of a shell of 420 coat protein subunits, a dodecamer of portal protein located at one 5-fold vertex (Bazinet et al., 1988), and 12 to 20 copies each of three minor proteins (King et al., 1973). All of these proteins are present in the mature phage. The P22 procapsid contains an inner core of 200 to 300 molecules of a scaffolding protein, which exit before or during DNA packaging (King & Casjens, 1974; Casjens & Hendrix, 1988). The precise roles of scaffolding proteins in directing virus assembly are not well # 1998 Academic Press

70 understood. However, assembly in vivo and in vitro in the absence of the scaffolding protein is inef®cient, and results primarily in the formation of aberrant, spiral-shaped particles (King et al., 1973; Earnshaw & King, 1978; Prevelige & King, 1993). Similar results have been obtained with the related virus herpes simplex type 1 (HSV-1; Desai et al., 1994; Thomsen et al., 1994). It is clear that, for many icosahedral viruses, a scaffolding protein is required for ef®cient assembly of properly dimensioned procapsids. The P22 scaffolding protein appears to consist of several segmented helical domains and has little or no globular core (Tuma et al., 1996). The scaffolding protein forms a mixture of monomers, dimers and tetramers in solution (Parker et al., 1997b). P22 scaffolding and coat proteins can be puri®ed, and when mixed together in vitro assemble particles of size and shape nearly identical with those formed in vivo (Prevelige et al., 1988). The rate of in vitro assembly can be determined by following increases in turbidity, and the composition of the assembled particles can be assessed by sucrose gradient sedimentation or electron microscopy (Prevelige et al., 1988). Because conditions for ef®cient in vitro assembly have been developed, P22 provides an excellent system for studying the effects of scaffolding protein mutations (Greene & King, 1996; Parker et al., 1997b) and deletions (Parker et al., 1997a; Tuma et al., 1998) on the rate and ®delity of procapsid assembly. These data can give information about the functions of various domains of the scaffolding protein. Toward this end, we have constructed two deletion mutants (Parker et al., 1997a; Tuma et al., 1998), in which 140 amino acid residues were removed from the amino-terminal end of the 303 residue scaffolding protein. In one of these proteins, an additional 11 residues were removed from the carboxyl terminus. Both of these scaffolding protein fragments form monomers and dimers, but do not associate into tetramers; other aspects of the structures of these mutants have been elucidated by Raman spectroscopy, circular dichroism and NMR (Parker et al., 1997a; Tuma et al., 1998). We report here on the effects of these amino and carboxyl-terminal truncations on the in vitro assembly activity of the scaffolding protein. These ®ndings reveal roles of the various domains of the P22 scaffolding protein in directing the assembly of a T ˆ 7 procapsid lattice.

Results The carboxyl terminus of the scaffolding protein is required for assembly In order to map functional domains of the 303 residue P22 scaffolding protein, two proteins corresponding to residues 141 to 292 and 141 to 303 were tested for their ability to promote the assembly of puri®ed coat protein. The assembly was followed by monitoring the time-dependent increase

P22 Scaffolding Protein Domains

in turbidity at 250 nm due to light-scattering (Figure 1(a)). The reactions were followed for four hours, at which point the assembly was judged to be essentially complete. When added to the coat protein, the amino acid 141-303 fragment caused a

Figure 1. The wild-type scaffolding protein and the 141-303 fragment assemble procapsid-like particles in vitro; the 141-292 fragment does not. (a) The wild-type scaffolding protein (triangles), the 141-303 fragment (circles) or the 141-292 fragment (crosses) at ®nal concentrations of 10 mM were mixed with 22 mM coat protein to initiate procapsid assembly. The assembly was followed at 20 C by monitoring the turbidity at 250 nm. Data points were taken at ten second intervals for 30 minutes, followed by 220 minutes at 75 second intervals; only every third point is shown for the 141-292 fragment. Turbidity is reported as the amount of change in the absorbance at 250 nm over the background levels observed before the addition of scaffolding proteins. (b) The assembly reactions with the scaffolding protein or the 141-303 fragment were allowed to proceed for four hours and then sedimented through 5 ml linear gradients of 5 to 20% (w/w) sucrose in buffer B as described in the text. The gradients were separated into 15 fractions and analyzed by SDS-PAGE. The bottom and top of the gradients are fractions 1 and 15, respectively. Upper panel, wild-type scaffolding protein; lower panel, the 141-303 fragment.

71

P22 Scaffolding Protein Domains

rapid increase in turbidity with a half-time of approximately one to two minutes. In contrast, the 141-292 fragment did not cause any signi®cant increase in turbidity, even after long reaction times. The slight increase in turbidity observed with the 141-292 fragment was similar in magnitude to that observed for the coat protein in the absence of the scaffolding protein in control experiments (data not shown). The effect of adding the wild-type scaffolding protein is shown for comparison. In accord with previous observations (Prevelige et al., 1988, 1993), the wild-type protein caused an increase in turbidity with a half-time of approximately four to ®ve minutes. At all times, the assembly reaction containing the 141-303 fragment displayed a substantially greater amount of turbidity than did the parallel reaction with the wild-type protein. This may re¯ect either a greater yield of procapsid-like particles or the polymerization into aberrant forms with larger light-scattering signatures (see below). In order to investigate the types of particles formed with the 141-303 fragment, the products of the assembly reactions were analyzed by sucrose gradient sedimentation and SDS-PAGE. Conditions were chosen such that procapsids would be expected to sediment approximately two-thirds of the way through the gradient. The gradients were separated into fractions, which were analyzed by SDS-PAGE (Figure 1(b)). Under these reaction conditions (i.e. a twofold molar excess of the coat protein over the scaffolding protein or the 141-303 fragment), approximately 30% of the coat protein was polymerized into procapsid-like particles. Electron microscopy con®rmed that both the wildtype scaffolding protein and the 141-303 fragment assembled procapsid-like particles (see Figure 7). The 141-303 fragment has been observed to assemble procapsid-like particles in an in vivo expression system (S.C., L. Sampson & P. Weigele, unpublished results). The mass ratio of the wild-type scaffolding protein to the coat protein in the procapsid-like particles was estimated (by scanning the SDS-PAGE gels of the sucrose gradient fractions) to be 0.40, corresponding to about 230 molecules of the scaffolding protein per particle. The mass ratio of the 141-303 fragment to the coat protein in the procapsid-like particles was estimated to be 0.43, corresponding to about 470 molecules of the 141-303 fragment per particle. As described above, the 141-303 fragment appeared to polymerize coat protein more rapidly than the wild-type scaffolding protein, as indicated by a faster increase in turbidity (Figure 1(a)). However, the ®delity of assembly was affected by the loss of the amino-terminal domain, causing rapidly sedimenting aberrant species to be formed as well as procapsid-like particles (Figure 1(b)). The contribution of these aberrant species to the total turbidity may account for some or all of the apparent increase in the rate of assembly compared to the wild-type scaffolding protein. The relative amounts

of aberrant species formed increased as the 141-303 fragment concentration was raised, up to a point where the aberrant species contained approximately 60% of the total coat protein (as opposed to only 25% in the form of procapsid-like particles). This observation is explored in more detail below. Assembly reactions with the 141-292 fragment were analyzed in parallel. Sucrose gradient sedimentation studies showed no increase in coat protein polymerization above background levels (data not shown), which along with the lack of an increase in turbidity (Figure 1(a)) indicated that the 141-292 fragment does not promote the assembly of procapsid-like particles. Similar results were obtained using trypsin-digested or chymotrypsindigested scaffolding protein (B. Greene & J. King, personal communication). These results suggest that residues 1 to 140 of the P22 scaffolding protein are not required for procapsid assembly, but that this region of the scaffolding protein may play a role in ensuring proper form determination in the procapsids. In contrast, the 11 residues at the carboxyl terminal of the scaffolding protein are absolutely required for assembly. When present in limiting concentrations, the scaffolding protein and the 141-303 fragment package similar numbers of molecules into procapsids Prevelige et al. (1988) have demonstrated that, when the ratio of the scaffolding protein to the coat protein in in vitro assembly reactions becomes limiting, the number of scaffolding protein molecules that are found in the assembled particles decreases to a limiting value of approximately 140. This suggests that a minimum set of coat-scaffolding binding interactions is required for assembly, and that the additional 100 to 150 scaffolding protein subunits normally found in a procapsid are not absolutely required. In order to determine if truncation of the scaffolding protein altered the minimum number of molecules required for assembly, in vitro assembly reactions were performed at various ratios of 141-303 to coat protein and allowed to proceed for four hours. The products were then sedimented through sucrose gradients, fractionated, and analyzed by SDS-PAGE. The ratio of scaffolding to coat protein molecules in the assembled procapsids was determined by measuring the relative staining intensities of the coat protein and either the wildtype scaffolding protein or the 141-303 fragment. The results are plotted in Figure 2 for the wildtype scaffolding protein (triangles) and the 141-303 fragment (circles). As the amount of either scaffolding protein in the reaction was reduced, the assembled procapsids contained decreasing numbers of scaffolding protein molecules. In agreement with previous results (Prevelige et al., 1988), this number extrapolated to an estimated limiting value of approximately 100 molecules per procap-

72

Figure 2. The ratio of scaffolding protein to coat protein in the products of in vitro assembly is a function of the input ratio. Assembly reactions were performed at 20 C at a constant coat protein concentration of 0.9 mg/ml using various concentrations of the scaffolding protein or the 141-303 fragment. After four hours, each sample was sedimented through a 5 ml linear 5% to 20% (w/w) sucrose gradient as described in the text, separated into 20 fractions, and analyzed by SDS-PAGE. Fractions containing procapsids were scanned and the relative masses of scaffolding and coat proteins present were calculated from the relative staining intensities. The numbers of scaffolding protein (triangles) or 141-303 fragment (circles) molecules per procapsid were calculated from the mass ratios and the molecular masses of the proteins.

sid for the wild-type scaffolding protein, and the same result was obtained with the 141-303 fragment. This suggests that similar coat-scaffolding interactions occur during assembly with the 141-303 fragment and the wild-type scaffolding protein. As the amount of the wild-type scaffolding protein supplied in the reactions was increased, the mass ratio of scaffolding to coat protein in the assembled particles increased to a value between 0.5 and 0.6, in agreement with earlier ®ndings (Prevelige et al., 1988). A similar mass ratio was obtained using the 141-303 fragment but, since the fragment has approximately half the mass of the wild-type scaffolding protein, the number of molecules of the fragment found in the particles was more than twice as high. This ratio corresponds to about 300 molecules of the wild-type scaffolding protein or 650 molecules of the 141-303 fragment per particle (Figure 2). These data suggest that the maximum number of scaffolding protein molecules that can be incorporated during assembly is determined by the internal volume of the procapsid and not by a ®nite number of binding sites. The 141-303 fragment, but not the 141-292 fragment, re-enters and binds to empty procapsid ``shells'' The polymerization of coat protein subunits into a procapsid is likely to require at least two steps: binding of the coat protein to the scaffolding pro-

P22 Scaffolding Protein Domains

tein, followed by positioning of the coat protein into the growing T ˆ 7 procapsid. In order to determine whether the 141-292 scaffolding protein fragment was capable of binding to preformed coat protein shells, we used a re-entry assay. The full-length scaffolding protein is capable of reentering and binding to preformed procapsid ``shells'' from which the scaffolding proteins have been extracted (Greene & King, 1994), a process that can be monitored by measuring increases in turbidity. If the 141-292 fragment can bind to coat protein subunits, but cannot promote assembly, it should be able to bind to shells. Conversely, if the 141-292 fragment does not bind to shells, this would suggest that its inability to direct assembly is caused by loss of the coat protein-binding domain. The kinetics of re-entry of the wild-type scaffolding protein and the 141-303 and 141-292 fragments into procapsid shells are illustrated by Figure 3. Both the wild-type scaffolding protein and the 141-303 fragment displayed a time-dependent increase in turbidity, indicating that these proteins re-entered and bound to procapsid shells. In contrast, no increase in turbidity was observed for the 141-292 fragment, indicating that it does not bind to preformed procapsid shells, and suggesting that its inability to assemble procapsid-like particles is caused by loss of the coat protein-binding domain. The 141-303 fragment re-entered and bound to procapsid shells more rapidly than the wild-type protein. Neither kinetic pro®le could be ®t well by a single exponential (data not shown), indicating

Figure 3. The wild-type scaffolding protein and the 141303 fragment re-enter and bind to empty procapsid shells; the 141-292 fragment does not. The wild-type scaffolding protein (triangles), the 141-303 fragment (circles), or the 141-292 fragment (crosses) at ®nal concentrations of 30 mM were added to solutions containing 8.6 mM coat protein in the form of shells. The re-entry and binding of scaffolding proteins were followed at 20 C by monitoring the turbidity at 250 nm (reported as the change in absorbance at 250 nm over the background level observed before addition of scaffolding proteins). Data points are shown at 42 second intervals for the initial 29 minutes and at 45 second intervals afterward; only every third data point is shown for the 141-292 fragment.

P22 Scaffolding Protein Domains

that the re-entry of each of these scaffolding proteins is a multistep process, in agreement with previous ®ndings for the wild-type protein (Greene & King, 1994). In addition, the ®nal level of turbidity was somewhat higher for the 141-303 fragment than for the wild-type protein, suggesting that the 141-303 fragment may be able to pack more densely into a preformed shell. To investigate this, we sedimented the products of the re-entry experiments through sucrose gradients as described above and measured the ratios of each of the scaffolding proteins to the coat protein in the procapsid fractions. For the wild-type protein, the scaffolding protein to coat protein ratio was 0.43(0.02), corresponding to approximately 250 molecules of scaffolding protein per particle. This value is similar to the number found in procapsids assembled in vivo (Casjens & King, 1974; Eppler et al., 1991). For the 141-303 fragment, the ratio was 0.38(0.04), corresponding to about 410 molecules of the 141-303 fragment per particle. Therefore, as described above for the in vitro assembly reactions (Figure 2), roughly equivalent masses of the scaffolding protein or the fragment pack into the procapsid shells, suggesting that the determining factor in packaging is the ®nal scaffolding protein density rather than the absolute number of molecules. The fact that, on average, fewer molecules of either scaffolding protein were found in the products of re-entry compared to the products of in vitro assembly suggests that there may be differences in the accessibility of scaffolding protein binding sites between these two forms. The higher ®nal turbidity obtained with the 141-303 fragment compared to the wild-type scaffolding protein may re¯ect subtle differences in the packing arrangements within the procapsids, which may give rise to different light-scattering signatures.

73 would be indicated by a delay in the elution from the column relative to the control protein. The results of these experiments are depicted in Figure 4. As shown in Figure 4A, the wild-type scaffolding protein (continuous line) eluted from the column with pronounced peak-trailing compared to the negative control (broken line). This suggests that the wild-type protein was retained on the column due to its ability to interact with the coat protein subunits. The experiment was repeated using the 141-292 fragment (Figure 4B). Unlike the wild-type scaffolding protein, this fragment did not display any peak-trailing compared to the control, and thus did not appear to bind to immobilized coat protein. Figure 4C demonstrates that the 141-303 fragment was retained on the column, and displayed a large degree of peak-trailing, suggesting a stronger interaction than was the case for the wild-type scaffolding protein (see below). Therefore, since the wild-type scaffolding protein and the 141-303 fragment could bind to immobilized coat protein subunits, and the 141-292 fragment could not, we conclude that the inability of the 141-292 fragment to promote assembly was caused by loss of the coat protein-binding domain.

The scaffolding protein and the 141-303 fragment bind to immobilized coat protein; the 141-292 fragment does not The 141-292 fragment was unable to assemble procapsid-like particles in vitro (Figure 1(a)) or in vivo (L. Sampson & S. C., unpublished results). This fragment was also unable to enter and bind to empty procapsid shells (Figure 3). Taken together, these ®ndings suggested that the 11 residues at the carboxyl terminus of the scaffolding protein (residues 293 to 303) form a critical part of the coat protein-binding domain. To test this directly, we employed a similar approach to that utilized by Teschke & Fong (1996) for the analysis of coat protein mutants. The coat protein was immobilized on a solid support, the wild-type scaffolding protein and the 141-292 and 141-303 fragments were applied to this column, and the elution of the scaffolding proteins from the column was followed by monitoring the absorbance at 280 nm. Bovine serum albumin was used as a negative control. Interactions between scaffolding protein in the mobile phase and the immobilized coat protein

Figure 4. The wild-type scaffolding protein and the 141-303 fragment bind to immobilized coat protein; the 141-292 fragment does not. A 0.7 cm  6 cm coat protein af®nity column was prepared as described in the text. Samples (50 ml) of the wild-type scaffolding protein A, the 141-292 fragment B, and the 141-303 fragment C, at 0.7 mg/ml were applied to the column and developed at 4 C with buffer B at a ¯ow-rate of 0.5 ml/min. Continuous lines indicate the absorbance at 280 nm for the scaffolding proteins in a 0.5 cm path-length cell. In a separate experiment, BSA (broken lines) was applied to the column as a negative control. The peak of the BSA absorbance peak corresponds to 0.01 absorbance unit.

74 The higher degree of peak-trailing in the elution pro®le of the 141-303 fragment compared to the full-length scaffolding protein suggests that this fragment interacts with the coat protein in a fundamentally different manner, perhaps binding more tightly. This may account for the increased rate of re-entry described above (Figure 3), and may explain the decreased ®delity of assembly observed for the 141-303 fragment (see below; and Figures 1(b), 7). The ®nding that removal of nearly half of the scaffolding protein molecule apparently increases its af®nity for the coat protein was unexpected.

P22 Scaffolding Protein Domains

to a system constrained to a dimer/tetramer equilibrium. This supports a model in which the formation of heterodimers containing the scaffolding

The 141-292 fragment inhibits assembly of procapsid-like particles Scaffolding protein monomers exist in a rapidly reversible equilibrium with dimers and tetramers, and dimerization of the scaffolding protein appears to be part of the mechanism of procapsid assembly (Parker et al., 1997b). Both the 141-292 and 141-303 fragments are capable of dimerizing (but do not form tetramers; Tuma et al., 1998). We reasoned that the inactive 141-292 fragment might inhibit assembly by forming heterodimers with the scaffolding protein. If this were true, then we would expect that a covalent dimeric R74C/L177I mutant scaffolding protein (Parker et al., 1997b), which cannot dissociate into monomers, would be inhibited to a lesser degree. We therefore examined the effect of including the 141-292 fragment in in vitro assembly reactions with the wild-type and R74C/ L177I scaffolding proteins. Figure 5(a) shows the initial phase of an assembly reaction with the wild-type scaffolding protein in the presence and absence of the 141-292 fragment. Inclusion of the 141-292 fragment in the assembly reaction decreased the overall rate of assembly. In Figure 5(b), the results of a similar experiment using the covalent dimeric R74C/L177I mutant scaffolding protein are shown. Assembly with the R74C/L177I mutant was not inhibited by the 141-292 fragment. The R74C/L177I dimer was then preincubated with DTT in order to reduce the disul®de bonds that form the covalent crosslinks between monomers. The reduced form of the R74C/L177I scaffolding protein was inhibited by the 141-292 fragment (Figure 5(c)), suggesting that the lack of inhibition portrayed in Figure 5(b) was due to the presence of covalent dimers. DTT pretreatment had no effect on the degree of inhibition of assembly with the wild-type scaffolding protein (data not shown). The active 141-303 fragment did not inhibit procapsid assembly when added in equimolar amounts to the wild-type scaffolding protein; the initial rate of assembly was midway between the rates observed with either protein alone (data not shown). The assembly-inactive 141-292 fragment was capable of inhibiting procapsid assembly when added to a scaffolding protein monomer/dimer/ tetramer equilibrium system, but not when added

Figure 5. The 141-292 fragment inhibits in vitro procapsid assembly with the wild-type scaffolding protein, but not with a covalent dimeric scaffolding protein mutant. The 141-292 fragment was included in assembly reactions containing the coat protein and either the wildtype (a) or the dimeric R74C/L177I mutant (b) scaffolding protein. The reactions in which the 141-292 fragment was included are indicated by broken lines. (c) The effect of reducing the disul®de bond of the R74C/L177I scaffolding protein by pretreatment with 10 mM DTT. The assembly was followed at 20 C by monitoring the turbidity at 250 nm (reported as the change in absorbance at 250 nm over the background level observed before addition of the scaffolding proteins); data points were taken at 7.5 second intervals.

75

P22 Scaffolding Protein Domains

protein and the 141-292 fragment results in inhibition of assembly. These heterodimers may be either inactive or partially active. The formation of heterodimers would involve a reversible equilibrium process, while procapsid assembly is essentially irreversible. Since heterodimers would thus eventually dissociate, active dimers of the scaffolding protein would then be able to form and participate in assembly. We therefore anticipated that the yield of procapsid-like particles assembled in the presence of the 141-292 fragment might, after assembly has proceeded to completion, be similar to that observed in the absence of the fragment. This was supported by the observation that, after the initial few minutes of procapsid assembly, the turbidities of reactions conducted in the presence and absence of the 141-292 fragment slowly converged (data not shown). To determine whether the inhibitory effect of the 141-292 fragment is indeed transient, assembly reactions were carried out in the presence and absence of the 141-292 fragment, allowed to proceed to completion, and then sedimented through sucrose gradients. Analysis of the fractions by SDSPAGE (Figure 6) shows that the presence of the 141-292 fragment made no difference to either the yield of procapsid-like particles or the ratio of the scaffolding to coat protein in the particles. In addition, no 141-292 fragment was detected in the procapsid fractions. Taken together, these results suggest that if heterodimers of the 141-292 fragment and the wild-type scaffolding protein do form, they are completely inactive, and not simply reduced in activity. This model is supported by in vivo assembly experiments, which demonstrated that the 141-292 fragment is not incorporated into procapsids when it is coexpressed with the wildtype scaffolding protein (L. Sampson & S.C., unpublished results).

Figure 6. The 141-292 fragment does not affect the yield or composition of procapsid-like particles after the assembly has proceeded to completion. The scaffolding protein (12 mM) was mixed with the coat protein (22 mM) in the absence or presence of 36 mM of the 141-292 fragment (upper and lower panels, respectively). After four hours at 20 C, the mixtures were sedimented through sucrose gradients as described in the text, fractionated, and analyzed by SDS-PAGE.

Deletion of 140 residues from the amino terminus of scaffolding protein causes decreased fidelity of assembly We have established that the carboxyl-terminal 11 residues of the scaffolding protein are required for assembly, and that the amino-terminal 140 residues are dispensable. What role, then, does the amino-terminal domain play? At low concentrations of the 141-303 fragment, sucrose gradient sedimentation indicated that procapsid-like particles were the most abundant product of assembly (data not shown). However, as the concentration of the 141-303 fragment was raised, increasing amounts of aberrant species began to assemble as well. Conversely, even at high concentrations, the wild-type scaffolding protein did not assemble signi®cant amounts of aberrant species (Figure 1(b)). Figure 7 shows electron micrographs of the products of assembly reactions conducted using 2.5 mM wild-type scaffolding

Figure 7. Electron microscopy demonstrates that the 141-303 fragment assembles procapsid-like particles and aberrant structures in vitro. The coat protein (19 mM) was mixed with the scaffolding protein or the 141-303 fragment and incubated for four hours at 20 C. Aliquots were applied to Formvar/carbon-coated copper grids and negative-stained with 2% uranyl acetate. The scale bar represents 50 nm; the magni®cation is 150,000: A, 2.5 mM scaffolding protein; B, 2.5 mM 141-303 fragment; C, 40 mM 141-303 fragment.

76 protein (A), 2.5 mM 141-303 fragment (B), and 40 mM 141-303 fragment (C). At the higher concentration of the 141-303 fragment, 46% of the particles produced were aberrant. At the lower concentration, the percentage of aberrants decreased to approximately 8%. The percentage of aberrants produced by the wild-type scaffolding protein was approximately 2% over the entire range of concentrations. There appeared to be proteinaceous cores in the interiors of many of the aberrant structures formed with the higher concentration of the 141-303 fragment, suggesting that some of the fragment remained bound. These results suggest that the amino-terminal domain of the P22 scaffolding protein, while not strictly required for procapsid assembly, plays a role in form determination. This role apparently becomes more critical at higher overall rates of assembly.

Discussion By making deletions in the bacteriophage P22 scaffolding protein, we have demonstrated that the carboxyl-terminal 11 residues of this protein are required for procapsid assembly and that these same residues are involved in binding to the coat protein. Spectroscopic measurements have indicated that this domain (residues 293 to 303) of the P22 scaffolding protein is highly a-helical, and is stabilized by contacts with the 141-292 domain (Tuma et al., 1998). The nucleotide sequence of the scaffolding protein gene has been determined (Eppler et al., 1991). The 292-303 domain has a heptad repeat pattern of hydrophobic residues, which suggests that they form an amphipathic a-helix, with a hydrophobic stripe on one face consisting of Tyr292, Leu295 and Leu299. Although the active form of the scaffolding protein appears to be a dimer (Parker et al., 1997b; Tuma et al., 1998), we believe it is unlikely that the carboxyl-terminal region contributes to the dimer interface. Six of the 12 residues closest to the carboxyl terminus are basic (and none is acidic); therefore, dimerization of this region would require overcoming substantial electrostatic repulsion. On the basis of analytical ultracentrifugation and spectroscopic data (Tuma et al., 1998), we favor a model in which the carboxyl-terminal helical region is stabilized by folding back and interacting with a nearby helical region within the same subunit. Sequence analysis suggests that this region may encompass approximately residues 274 to 286, with Ile276, Met280, and Ala284 capable of forming hydrophobic interactions with Leu299, Leu295, and Tyr292 on the carboxyl-terminal helix. Preliminary NMR data support this model (Y. Sun, M.H.P., P.E.P. & N. Krishna, unpublished results). We propose that the P22 scaffolding proteins bind via their carboxyl-terminal domains to coat protein subunits, possibly via electrostatic interactions involving the cluster of basic amino acid

P22 Scaffolding Protein Domains

residues at the scaffolding protein carboxyl terminus, and activate the coat proteins for assembly by dimerizing. The scaffolding protein might serve two functions: it could use the free energy of dimerization to overcome the activation energy barrier to coat protein self-association, thereby acting as an ``entropy sink.`` In addition, it seems likely that dimerization of the scaffolding protein assists in the conformational switching of the coat protein subunits into their appropriate quasiequivalent conformations (Johnson, 1996; Johnson & Speir, 1997). A similar mechanism may be involved in the assembly of herpesviruses. It has been shown that 15 to 25 amino acid residues at the carboxyl termini of the scaffolding proteins of the herpesviruses cytomegalovirus (CMV; Beaudet-Miller et al., 1996) and HSV-1 (Matusick-Kumar et al., 1995; Hong et al., 1996; Oien et al., 1997) interact with their respective major capsid proteins. As in the P22 scaffolding protein, the critical carboxyl-terminal domains in the herpesvirus scaffolding proteins have hydrophobic heptad patterns suggestive of amphipathic a-helices, and helicity in this region has been directly demonstrated to be important for binding to the herpesvirus major capsid proteins. Like the P22 scaffolding protein, the herpesvirus scaffolding proteins have been shown to selfassociate (Desai & Person, 1996; Pelletier et al., 1997; Wood et al., 1997), and oligomerization appears to be required for binding to the major capsid proteins. We report here that an inactive, truncated scaffolding protein (residues 141 to 292) that is capable of dimerizing (Tuma et al., 1998) can inhibit procapsid assembly, and our results suggest that the mechanism of inhibition may be the formation of heterodimers with the full-length scaffolding protein. Although we do not know whether this putative heterodimer is inactive or simply partially active, the 141-292 fragment was not carried into assembling procapsids in vitro or in vivo, suggesting that a dimer containing two intact carboxyl termini is required for assembly. When molar excesses of either the wild-type scaffolding protein or the assembly-active 141-303 fragment were added to coat protein monomers or to empty procapsid shells, the masses of either protein that bound within the procapsids were similar. Since the fragment is approximately half the size of the wild-type protein, this indicated that the upper limit to the number of scaffolding protein molecules that can pack into a procapsid is dictated by the internal volume of the procapsid. There are 420 coat protein subunits in a P22 procapsid, but as many as 650 molecules of the 141-303 scaffolding fragment can be incorporated during assembly. (The number of scaffolding protein subunits in the HSV-1 procapsid also exceeds the number of major capsid proteins; Newcomb et al., 1993.) It is possible that more than one molecule of the 141-303 fragment can bind to each coat protein subunit. Alternatively, some of the molecules of the 141-303

77

P22 Scaffolding Protein Domains

fragment may be held in place solely by scaffolding-scaffolding interactions within the procapsid. Procapsids can be assembled with as few as approximately 100 molecules of the wild-type scaffolding protein per particle, and we have shown that a similar number of molecules of the 141-303 fragment is suf®cient for assembly. The role of the remaining scaffolding protein molecules that are normally found inside a procapsid is not clear. However, it has been proposed that one role of the scaffolding protein may be to sterically block cellular proteins from being trapped inside the procapsid as it assembles (Earnshaw & Casjens, 1980). The multiphasic kinetics observed in procapsid reentry experiments using both the wild-type scaffolding protein (Greene & King, 1994) and the 141-303 fragment support the idea that there is more than one class of binding interactions within the procapsid. This suggests that the mechanism by which the scaffolding protein promotes assembly is to activate coat protein subunits at certain critical positions in the procapsid lattice. The other coat protein subunits may be able to bind correctly without the assistance of the scaffolding protein. These critical scaffolding-coat interactions are required throughout the assembly process and not simply in the initial stages of assembly (Prevelige et al., 1993). The 141-303 fragment, which displays decreased ®delity of assembly, has an association constant for dimerization that is approximately sevenfold lower than for that of the wild-type scaffolding protein and, unlike the wild-type protein, does not form tetramers (Tuma et al., 1998). We believe that this decreased ability to dimerize (and possibly also to tetramerize) may destabilize the alignment of the coat protein subunits as they bind to the growing procapsid and lead to the formation of aberrant species. Another possibility is that correct binding of coat protein subunits to the growing procapsid lattice is a thermodynamically favorable process, but that kinetic ``traps'' involving incorrect binding orientations can occur as the rate of assembly is raised. The amino-terminal domain of the scaffolding protein might serve to prevent these incorrect interactions from occurring, perhaps by sterically blocking inappropriate coat-scaffolding interactions. The precise size of the proposed aminoterminal ®delity-enhancing domain is not clear from these initial experiments; truncation at residue 141 was chosen as a convenient starting point based on technical considerations. However, several more extensively truncated scaffolding proteins have recently been cloned (P. Weigele & S.C., unpublished results). Scaffolding protein fragments containing as few as 66 residues from the C terminus are active in assembly, albeit with a very high incidence of aberrant assembly (M.H.P. & P.E.P., unpublished results). The relationships among scaffolding protein length, degree of aberrant assembly and self-association will shed further light on the results presented here.

Materials and Methods Preparation of P22 procapsids, shells, coat protein and scaffolding protein Procapsids were prepared as described (Prevelige et al., 1988). Brie¯y, S. typhimurium strain DB7136 was infected with a strain of bacteriophage P22 (2ÿamH200/ 13ÿamH101) that contains mutations in genes responsible for DNA packaging and cell lysis. These mutations prevent full maturation of the phage, and thus cause procapsids to accumulate in the host. After growth at 37 C for three hours, the cells were lysed by repeated freeze/ thawing. Cell debris was removed by low-speed centrifugation, and the procapsids were subsequently harvested by high-speed centrifugation, puri®ed by sizeexclusion chromatography, and stored at 4 C. All puri®cation steps were carried out in buffer B (50 mM TrisHCl (pH 7.6), 25 mM NaCl, 2 mM EDTA) at 4 C unless otherwise noted. The wild-type scaffolding protein was extracted from procapsids by several cycles of treatment with 0.5 M guanidinium hydrochloride (GuHCl). This treatment solubilized the scaffolding protein and the other minor proteins present in procapsids, leaving empty ``shells'' of coat protein, which were harvested by high-speed centrifugation. The scaffolding protein was then puri®ed using ion-exchange chromatography as described (Parker et al., 1997b) and stored in buffer B at ÿ20 C. The procapsid shells were stored at 4 C. Coat protein monomers were prepared for assembly experiments by dissociating the shells with ®ve to ten volumes of 6 M GuHCl, followed by exhaustive dialysis into buffer B. Aggregates were removed by centrifugation in a Beckman TLA-100.3 rotor at 55,000 revs/minute for 30 minutes at 4 C. The coat protein monomers were used within one day of preparation. Scaffolding protein deletion mutants Construction of the plasmids coding for the scaffolding protein amino acid 141-303 and 141-292 fragments, and details of their expression and puri®cation, have been described (Parker et al., 1997a; Tuma et al., 1998). The proteins were overexpressed in Escherichia coli, puri®ed by ion-exchange chromatography, and stored at ÿ20 C in buffer B. Procapsid in vitro assembly reactions The procapsid assembly reactions depicted in Figure 1 were initiated by adding the wild-type scaffolding protein or the 141-303 or 141-292 fragment (®nal concentration 10 mM) to a solution of 22 mM coat protein in a thermostatically controlled cuvette at 20 C. To keep the amount of data reasonable, but provide suf®cient resolution at early times, the turbidity at 250 nm was measured at ten second intervals for 30 minutes, followed by a further 220 minutes at 75-second intervals. Aliquots (200 ml) were then sedimented through sucrose gradients essentially as described (Prevelige et al., 1988; Parker et al., 1997b). For Figures 1(b) and 6, each 5 ml gradient (5 to 20% (w/w) sucrose in buffer B, atop a 150 ml cushion of 60% (w/v) CsCl in 20% sucrose) was centrifuged at 20 C for 35 minutes at 33,000 revs/minute in a Beckman SW-55Ti rotor. The gradients were separated into 15 fractions, which were subsequently analyzed by SDS-PAGE on 15% polyacrylamide gels. For Figure 2,

78 the gradients were centrifuged for 35 minutes at 37,500 revs/minute and separated into 20 fractions. The gels were stained with Coomassie blue dye and scanned with an Alpha Imager 2000 densitometer (Alpha Innotech Corp., Hayward, CA). In control experiments, the staining intensities of the wild-type scaffolding protein and the 141-303 fragment were compared. Both proteins stained equally well, and staining for both proteins was linear over the range of intensities used in these experiments. For reactions in which inhibition by the 141-292 fragment was measured (Figure 5), similar conditions were used, except that the data points were collected at 7.5 second intervals. The coat protein concentration was 19 mM, and the following scaffolding protein or 141-292 fragment concentrations were used: A, 12 mM wild-type scaffolding protein and 29 mM 141-292 fragment; B, 7.5 mM R74C/L177I mutant scaffolding protein and 37 mM 141-292 fragment; C, 12 mM R74C/L177I mutant and 29 mM 141-292 fragment. For A and C, the wild-type or mutant scaffolding protein was preincubated in 10 mM DTT overnight at 4 C. This treatment reduces the disul®de crosslink in the R74C/L177I mutant.

Procapsid shell re-entry experiments Procapsid shells at a ®nal concentration of 0.40 mg/ ml were placed into a thermostatically controlled cuvette at 20 C, and the wild-type scaffolding protein, the 141-303 fragment or the 141-292 fragment was added to initiate the reaction. The kinetics of scaffolding protein re-entry were measured by following the turbidity at 250 nm. Data points were collected at six second intervals for 29 minutes, followed by a further 220 minutes at 45 second intervals. The products of the re-entry experiments were then sedimented through sucrose gradients as described above. The fractions that contained procapsid-like particles were identi®ed by SDS-PAGE. These fractions were subsequently analyzed in triplicate by SDS-PAGE and scanned to determine the relative amounts of the proteins present in the particles.

Coat protein affinity column Monomeric coat protein, prepared as described above, was dialyzed into buffer containing 0.1 M sodium bicarbonate and 0.5 M NaCl at pH 8.3. Approximately 15 mg of coat protein was mixed with 1.5 ml of Sepharose 4BCNBr (Sigma Chemical Co., St. Louis, MO) prepared according to the manufacturer's instructions. After overnight reaction at 4 C, coupling was quenched by the addition of 1 M Tris base (pH 9), followed by exhaustive washing. Approximately 70% of the coat protein was conjugated as determined spectrophotometrically. In order to reduce the average density of the coat protein, the Sepharose 4B/coat protein beads were diluted by the addition of an equal volume of Sepharose 4B. A 0.7 cm  6 cm column was prepared and equilibrated in buffer B. Samples (50 ml) of the various scaffolding proteins or bovine serum albumin (BSA) at 0.7 mg/ml initial concentration were applied to the column. The column was developed at 4 C with buffer B at a ¯ow-rate of 0.5 ml/minute. The elution pro®le was determined by continuous monitoring at 280 nm. The BSA standard peak height in Figure 4 corresponds to an absorbance of 0.01 in a 0.5 cm path-length ¯ow cell.

P22 Scaffolding Protein Domains Electron microscopy Aliquots (10 ml) of the products of assembly reactions were applied to Formvar/carbon-coated copper grids and negative-stained with 2% (w/v) uranyl acetate (Huxley & Zubay, 1960). Electron microscopy was carried out using a Hitachi H-7000 microscope operating at 75 kV.

Acknowledgments This work was supported by grants from the National Institutes of Health (GM47980, to P.E.P., and training grant T32-AI07150 for support of M.H.P.) and the National Science Foundation (MCB-9600574, to S.C.). We thank Kenneth W. French for technical assistance.

References Bazinet, C., Benbasat, J., King, J., Carazo, J. M. & Carrascosa, J. L. (1988). Puri®cation and organization of the gene 1 portal protein required for phage P22 DNA packaging. Biochemistry, 27, 1849± 1856. Beaudet-Miller, M., Zhang, R., Durkin, J., Gibson, W., Kwong, A. & Hong, Z. (1996). Virus-speci®c interaction between the human cytomegalovirus major capsid protein and the C terminus of the assembly protein precursor. J. Virol. 70, 8081± 8088. Casjens, S. & Hendrix, R. (1988). Control mechanisms in dsDNA bacteriophage assembly. In The Bacteriophages (Calendar, R., ed.), vol. 1, pp. 15 ± 91, Plenum Press, New York. Casjens, S. & King, J. (1974). P22 morphogenesis I: catalytic scaffolding protein in capsid assembly. J. Supramol. Struct. 2, 202± 224. D'Halluin, J. C., Martin, G. R., Torpier, G. & Boulanger, P. A. (1978). Adenovirus type 2 assembly analyzed by reversible cross-linking of labile intermediates. J. Virol. 26, 357± 363. Desai, P. & Person, S. (1996). Molecular interactions between the HSV-1 capsid proteins as measured by the yeast two-hybrid system. Virology, 220, 516± 521. Desai, P., Watkins, S. C. & Person, S. (1994). The size and symmetry of B capsids of herpes simplex virus type 1 are determined by the gene products of the UL26 open reading frame. J. Virol. 68, 5365± 5374. Donaghy, G. & Jupp, R. (1995). Characterization of the Epstein-Barr virus proteinase and comparison with the human cytomegalovirus proteinase. J. Virol. 69, 1265± 1270. Earnshaw, W. C. & Casjens, S. R. (1980). DNA packaging by the double-stranded DNA bacteriophages. Cell, 21, 319± 331. Earnshaw, W. & King, J. (1978). Structure of phage P22 coat protein aggregates formed in the absence of the scaffolding protein. J. Mol. Biol. 126, 721± 747. Eppler, K., Wyckoff, E., Goates, J., Parr, R. & Casjens, S. (1991). Nucleotide sequence of the bacteriophage P22 genes required for DNA packaging. Virology, 183, 519± 538. Greene, B. & King, J. (1994). Binding of scaffolding subunit within the P22 procapsid lattice. Virology, 205, 188± 197.

P22 Scaffolding Protein Domains Greene, B. & King, J. (1996). Scaffolding mutants identifying domains required for P22 procapsid assembly and maturation. Virology, 224, 82 ± 96. Hendrix, R. (1985). Shape determination in virus assembly: The bacteriophage example. In Virus Structure and Assembly (Casjens, S., ed.), pp. 169± 204, Jones and Bartlett, Boston. Hong, Z., Beaudet-Miller, M., Durkin, J., Zhang, R. & Kwong, A. D. (1996). Identi®cation of a minimal hydrophobic domain in the herpes simplex virus type 1 scaffolding protein which is required for interaction with the major capsid protein. J. Virol. 70, 533± 540. Huxley, H. & Zubay, G. (1960). Electron microscope observations on the structure of microsomal particles from Escherichia coli. J. Mol. Biol. 2, 10 ± 18. Johnson, J. E. (1996). Functional implications of proteinprotein interactions in icosahedral viruses. Proc. Natl Acad. Sci. USA, 93, 27 ± 33. Johnson, J. E. & Speir, J. A. (1997). Quasi-equivalent viruses: a paradigm for protein assemblies. J. Mol. Biol. 269, 665± 675. King, J. & Casjens, S. (1974). Catalytic head assembling protein in virus morphogenesis. Nature, 251, 112± 119. King, J., Lenk, E. V. & Botstein, D. (1973). Mechanism of head assembly and DNA encapsulation in Salmonella phage P22. II. Morphogenetic pathway. J. Mol. Biol, 80, 697± 731. Matusick-Kumar, L., Newcomb, W. W., Brown, J. C., McCann, P. J., III, Hurlburt, W., Weinheimer, S. P. & Gao, M. (1995). The C-terminal 25 amino acids of the protease and its substrate ICP35 of herpes simplex virus type I are involved in the formation of sealed capsids. J. Virol. 69, 4347± 4356. Newcomb, W. W., Trus, B. L., Booy, F. P., Steven, A. C., Wall, J. S. & Brown, J. C. (1993). Structure of the herpes simplex virus capsid. Molecular composition of the pentons and the triplexes. J. Mol. Biol. 232, 499± 511. Newcomb, W. W., Homa, F. L., Thomsen, D. R., Booy, F. P., Trus, B. L., Steven, A. C., Spencer, J. V. & Brown, J. C. (1996). Assembly of the herpes simplex virus capsid: characterization of intermediates observed during cell-free capsid formation. J. Mol. Biol. 263, 432± 446. Oien, N. L., Thomsen, D. R., Wathen, M. W., Newcomb, W. W., Brown, J. C. & Homa, F. L. (1997). Assembly of herpes simplex virus capsids using the human cytomegalovirus scaffold protein: critical role of the C terminus. J. Virol. 71, 1281± 1291. Parker, M. H., Jablonsky, M., Casjens, S., Sampson, L., Krishna, N. R. & Prevelige, P. E., Jr (1997a). Cloning, puri®cation, and preliminary characterization by circular dichroism and NMR of a carboxyl-term-

79 inal domain of the bacteriophage P22 scaffolding protein. Protein Sci. 6, 1583± 1586. Parker, M. H., Stafford, W. F., III & Prevelige, P. E., Jr (1997b). Bacteriophage P22 scaffolding protein forms oligomers in solution. J. Mol. Biol. 268, 655± 665. Pelletier, A., Do, F., Brisebois, J. J., Lagace, L. & Cordingley, M. G. (1997). Self-association of herpes simplex virus type 1 ICP35 is via coiled- coil interactions and promotes stable interaction with the major capsid protein. J. Virol. 71, 5197± 5208. Prevelige, P. E., Jr & King, J. (1993). Assembly of bacteriophage P22: a model for dsDNA virus assembly. Prog. Medical Virol. 40, 206± 221. Prevelige, P. E., Jr, Thomas, D. & King, J. (1988). Scaffolding protein regulates the polymerization of P22 coat subunit into icosahedral shells in vitro. J. Mol. Biol. 202, 743± 757. Prevelige, P. E., Jr, Thomas, D. & King, J. (1993). Nucleation and growth phases in the polymerization of coat protein and scaffolding subunits into icosahedral procapsid shells. Biophys. J. 64, 824± 835. Rixon, F. J. (1993). Structure and assembly of herpesviruses. Semin. Virol. 4, 135 ±144. Teschke, C. M. & Fong, D. G. (1996). Interactions between coat and scaffolding proteins of phage P22 are altered in vitro by amino acid substitutions in coat protein that cause a cold-sensitive phenotype. Biochemistry, 35, 14831± 14840. Thomsen, D. R., Roof, L. L. & Homa, F. L. (1994). Assembly of herpes simplex virus (HSV) intermediate capsids in insect cells infected with recombinant baculoviruses expressing HSV capsid proteins. J. Virol. 68, 2442± 2457. Trus, B. L., Booy, F. P., Newcomb, W. W., Brown, J. C., Homa, F. L., Thomsen, D. R. & Steven, A. C. (1996). The herpes simplex virus procapsid: structure, conformational changes upon maturation, and roles of the triplex proteins VP19c and VP23 in assembly. J. Mol. Biol. 263, 447± 462. Tuma, R., Prevelige, P. E., Jr & Thomas, G. J., Jr (1996). Structural transitions in the scaffolding and coat proteins of P22 virus during assembly and disassembly. Biochemistry, 35, 4619 ±4627. Tuma, R., Parker, M. H., Weigele, P., Sampson, L., Sun, Y., Krishna, N. R., Casjens, S., Thomas, G. J., Jr & Prevelige, P. E., Jr (1998). A helical coat protein recognition domain of the bacteriophage P22 scaffolding protein. J. Mol. Biol. 281, 81± 94. Wood, L., Baxter, M., Plafker, S. & Gibson, W. (1997). Human cytomegalovirus capsid assembly protein precursor (pUL80. 5) interacts with itself and with the major capsid protein (pUL86) through two different domains. J. Virol. 71, 179± 190.

Edited by W. Baumeister (Received 14 November 1997; received in revised form 17 April 1998; accepted 27 April 1998)