A mutation at the interface between domains causes rearrangement of

peng$$0103

Protein Engineering vol.10 no.1 pp.45–52, 1997

A mutation at the interface between domains causes rearrangement of domains in 3-isopropylmalate dehydrogenase

Chunxu Qu, Satoshi Akanuma, Hideaki Moriyama, Nobuo Tanaka1 and Tairo Oshima2 Department of Life Science, Faculty of Bioscience and Biotechnology, Tokyo Institute of Technology, Nagatsuta 4259, Yokohama 226 and 2Department of Molecular Biology, Tokyo University of Pharmacy and Life Science, Horinouchi 1432-1, Hachioji, Tokyo 192-03, Japan 1To

whom correspondence should be addressed

The structure of a thermostable Ala172Leu mutant, designated A172L, of 3-isopropylmalate dehydrogenase from Thermus thermophilus was determined. The crystal belongs to space group P21, with cell parameters a J 55.5 Å, b J 88.1 Å, c J 72.0 Å and β J 100.9°. There is one dimer in each asymmetric unit. The final R factor is 17.8% with 69 water molecules at 2.35 Å resolution. The mutation is located at the interface between domains and the Cα trace of the mutant structure deviates from that of the native structure by as much as 1.7 Å, while the structure of each domain barely changes. The mutant enzyme has a more closed conformation compared with the wild-type enzyme as a result of the replacement of Ala with Leu at residue 172. These structural variations were found independent of the crystal packing, because the structure of wild type was the same in crystals obtained in different precipitants. The hinge regions for the movement of domains are located around the active cleft of the enzyme, an observation that implies that the mobility of domains around the hinge is indispensable for the activity of the enzyme. The larger side chain at the mutated site contributed to the thermostability of the mutant protein by enhancing the local packing of side chains, and also by shifting the backbone of the opposing domain. Keywords: closed conformation/domain hinge-motion/domain interface mutation/3-isopropylmalate dehydrogenase

Introduction 3-Isopropylmalate dehydrogenase (IPMDH; threo-D-3-isopropylmalate NAD1 oxidoreductase; EC 1.1.1.85) is an enzyme that catalyzes a reaction in the leucine biosynthetic pathway as follows: 3-isopropylmalate 1 NAD1 → 2-ketoisocaproate 1 NADH 1 CO2 The structure of the enzyme (designated 10T hereafter) from Thermus thermophilus has already been determined (Imada et al., 1991). The enzyme is composed of two identical subunits. Each polypeptide chain of 345 amino acids is folded into two domains. The first domain consists of N-terminal (residues 1–99) and C-terminal (residues 252–345) amino acids, while the second domain consists of the middle part (residues 100–251), as shown in Figure 1. The second domains make contact with each other to form a functional dimer. The temperature for 50% inactivation is one parameter used © Oxford University Press

to characterize the thermal denaturation of enzymes. The difference in this value between IPMDH from T.thermophilus (76°C) and that from Bacillus subtilis (45°C) is a remarkable 38°C (Numata et al., 1995). In various efforts to gain some insight into the mechanism involved in the stabilization of IPMDH from T.thermophilus, various mutant enzymes (Miyazaki et al., 1994; Sakurai et al., 1995) and chimeric enzymes (Onodera et al., 1994; Numata et al., 1995) have been created and investigated. The construction of the native A172L mutant enzyme was based on the results of an experiment in which a search was made for thermostable mutants of chimeric enzyme 2T2M6T by a molecular evolutionary method (Akanuma, 1995). The chimeric enzyme 2T2M6T was produced by replacing the amino acid sequence of the enzyme from T. thermophilus, from the 75th to the 133rd residues by the corresponding part of the enzyme from B.subtilis. A mutant of this chimeric enzyme, A172L–2T2M6T, was less sensitive to increases in temperature than the 2T2M6T chimeric enzyme itself. When the same mutation was generated in the enzyme designated 10T, the melting temperature of the mutant protein was found to be 3.0°C higher than that of the original 10T. Although all the previously reported mutants have been crystallized isomorphously with 10T IPMDH, the 10T mutant A172L (Ala 172 → Leu) was crystallized non-isomorphously with the native enzyme. This mutant enzyme has two independent subunits per asymmetric unit, both of which have a closed conformation compared with the native enzyme. Changes in the conformation of the enzyme in response to binding of NAD, Mg21 and 3-isopropylmalate were found through a small-angle X-ray scattering experiment and it was proposed that the enzyme adopts a closed conformation in its active state (Kadono et al., 1995). However, this mutant has activity similar to that of the native enzyme. This paper describes the structural analysis of the A172L mutant of 10T which was undertaken to elucidate the reason for the enhanced thermostability of the mutant enzyme. Moreover, to clarify the mechanism of domain movement, the structure of the native enzyme, crystallized under similar conditions as the mutant, was also analyzed. Materials and methods Construction and purification of the mutant enzyme Construction of the mutant enzyme and its purification were performed as described by Akanuma et al. (1995). Assays of thermostability and activity The thermostability of each enzyme was determined by using changes in circular dichroism to monitor the unfolding as a function of temperature. The melting temperature (Tm) of each enzyme was measured at pH 10.9 in a solution of NaHCO3, at an enzyme concentration of 0.2 mg/ml. The values for A172L and the wild-type enzyme were 67.5 and 64.5°C, respectively. Activities were determined at 60°C in 100 mM potassium phosphate (pH 7.6) that contained 1 M KCl, 20 45

C.Qu et al.

Table I. Data collection and structural refinement statistics for A172L and 10T Item

Fig. 1. Ribbon diagram showing domain structures of and the site of the mutation in 3-isopropylmalate dehydrogenase. The side chains of A172 and H300 are represented by ball-and-stick diagrams. The N- and C-termini are indicated by N and C. This figure was drawn using MOLSCRIPT (Kraulis, 1991) and the coordinates of Imada et al. (1991).

mM MnCl2, 800 µM NAD and 800 µM D,L-IPM (Akanuma et al., 1995). Crystallization and collection of data The hanging drop vapor diffusion method was applied using the sparse matrix method (Jancaric and Kim, 1991) at 25°C since crystals suitable for X-ray analysis were not obtained under the crystallization conditions used for the analysis of the native enzyme (Sakurai et al., 1992). Trials indicated that the mutant enzyme could give crystals of a suitable size near its isoelectric point of 4.6 with polyethylene glycol (PEG) 4000 as the precipitant. The protein was dissolved in 20 mM potassium phosphate buffer, pH 7.6. The reservoir solution was 6.5% PEG 4000 dissolved in 0.1 M sodium acetate buffer, pH 5.2. A 5 µl aliquot of a solution of 10 mg/ml protein was mixed with 5 µl of reservoir solution on a cover-slip. Rodlike crystals of 0.230.430.4 mm3 were obtained after a few days. Intensity data were collected up to 2.35 Å resolution with a Rigaku R-Axis IIc diffractometer (Sato et al., 1992). A set of image data was converted into an indexed intensity data set by the data processing package Process (Rigaku). The full data set was obtained by merging and scaling the indexed data set by the method of Fox and Holmes (1966). The crystal was found to belong to space group P21, with cell parameters a 5 55.5 Å, b 5 88.1 Å, c 5 72.0 Å and β 5 100.9°. There were two subunits per asymmetric unit. In order to clarify the effect of the precipitant, the native enzyme was crystallized under conditions similar to those described above. Crystals grew from a well in which the reservoir solution was 3% PEG 4000, pH 4.8. A crystal of 0.230.430.4 mm3 was mounted on the R-Axis IIc diffractometer. Intensity data were collected up to 2.4 Å resolution. However, data beyond 3 Å resolution suffered from a high Rmerge value. The space group and cell parameters were determined from data that had been processed up to 3 Å resolution. The crystal was found to belong to space group P21, with cell parameters a 5 82.5 Å, b 5 86.7 Å, c 5 121.1 Å and β 5 90.7°. There were two dimers per asymmetric unit. A summary of the data collection and statistics for both crystals is given in Table I. 46

A172L

Native (PEG 4000)

Data collection: Space group P21 a (Å) 55.5 b (Å) 88.1 c (Å) 72.0 β (°) 100.9 No. of molecules per asymmetric unit 2 2.4 Vm (Å3/Da) Completeness 93.8% Rmergea 5.96%

P21 82.5 86.7 121.1 90.7 4 2.7 92.7% 5.43%

Structural refinement: Resolution No. of reflections used R factorb R.m.s.d. of bond length R.m.s.d. of bond angle

6.0–3.0 Å 24 207 14.9% 0.007 Å 1.36°

6.0–2.35 Å 24 963 17.8% 0.007 Å 1.36°

aReliability

factor among symmetry-equivalent reflections: 5 (Σ |Ii/Gi – ,Ii.|)/Σ,Ii. where Ii is one of the intensities of reflection that can be related to a symmetry operation, Gi is a scale factor and ,Ii. is the average of Ii/Gi. bR factor is defined as Σ | |F | – |F obs calc| |/Σ|Fobs|. Rmerge

Table II. Comparison of structures between two mutant molecules and the native molecule Molecule

Fitting part

subunit 1

Entire molecule Domain 1 Domain 2

subunit 2

Entire molecule Domain 1 Domain 2

Twist (°)

8.9

1.9

Shift (Å)

R.m.s.d. of Cα atoms (Å)

14.8

1.71 0.44 0.37

10.7

1.03 0.46 0.28

Determination and refinement of structures The molecular replacement method was applied to the analysis of A172L, with the native dimer as a searching model. The program XPLOR (Bru¨nger et al., 1989, 1990) was used for calculations. After the molecules had been located in the cell, the difference Fourier map, 2(Fo – Fc)Φc, was calculated and the structure modification corresponding to Ala to Leu mutation at position 172 was achieved with the Frodo program (Jones, 1985). Further refinement was continued with 24 963 reflections within the resolution range 6.0–2.35 Å. Two types of difference Fourier maps [2(Fo – Fc)Φc and (Fo – Fc)Φc] were calculated and used for several cycles of modelling. The final R factor was 17.8% with 69 water molecules. The root mean square deviations (r.m.s.d.) of the bond lengths and bond angles from the ideal values were 0.007 Å and 1.36°, respectively. In the analysis of the native enzyme that had been crystallized from PEG 4000, the monomer structure, determined from crystals obtained in ammonium sulfate (Imada et al., 1991), was used as a searching model. Four peaks were observed after PC refinement in the molecular replacement calculation. When the model molecule was put into the corresponding positions of the asymmetric unit by rotation and translation, an R factor of 26% was obtained with rigid body refinement at a resolution range 6.0–4.0 Å. Further refinement was performed in the resolution range 6.0–3.0 Å using 24 207

Mutation at the interface between domains

Fig. 2. Stereo view of four molecules in an asymmetric unit of the native crystal, crystallized with PEG 4000 as precipitant. Subunits 1 and 3 form one functional dimer and subunits 2 and 4 form the other. The two dimers are related by a twofold axis which is parallel to the z-axis.

reflections. Positional analysis showed that the four subunits formed two functional dimers, which were related by a twofold axis parallel to the z-axis. As shown in Figure 2, subunits 1 and 3 formed one dimer and subunits 2 and 4 formed the other. As in the structure of the crystal obtained by ammonium sulfate precipitation, there was also a twofold axis that related the two subunits of each dimer. Non-crystallographic symmetry constraints were applied to the two dimers, and the structure was refined by slow cooling protocols. The final R factor was 14.9%, and the r.m.s.d.s of the bond length and bond angle from the ideal were 0.007 Å and 1.36°, respectively. Results Thermostability and activity The A172L mutant enzyme is a more thermostable protein than the wild-type enzyme. Its melting temperature is 3°C higher than that of the native enzyme, while its activity is similar to that of the native enzyme. Structural comparison between the A172L mutant enzyme and the native enzyme crystallized from a solution of ammonium sulfate Overall structure. The crystallographic asymmetric unit of A172L contains a dimer and the two subunits are related by a non-crystallographic twofold axis, with a rotation angle of 177.4°. The two independent subunits, designated subunits 1 and 2, have a different arrangement of domains from the native enzyme. As shown in Table II, although the r.m.s.d. of Cα atoms of the whole molecule are higher, the corresponding values calculated for the separate domains are lower, indicating that the conformation within each domain is very similar. To clarify this point, we superimposed Cα atoms of separate domains of the A172L structure on those of the native structure and then calculated the r.m.s.d. of backbone atoms. As plotted in Figure 3, when we fitted the first domain, the r.m.s.d. were small for the residues within this domain, whereas the r.m.s.d. for the residues within the other domain were very large. The same was true for the second domain. It appears, therefore, that a major structural difference exists between the mutant and the native enzyme that is due to rearrangement of domains. To describe the movement of domains, we calculated the domain twist and shift values for the mutant structure relative to the native one. When two objects are superimposed, two values, a rotation angle and a translation vector, can be determined. Here, the rotational axis was almost parallel when we fitted the separate domains, so the domain twist value was defined as the difference in rotation angles for fitting separate

Fig. 3. Root mean square deviations of main-chain atoms showing differences in arrangement of domains relative to native IPMDH (ammonium sulfate crystals) for the two independent molecules of mutant A172L. The r.m.s.d.s were calculated from (a) superimposition of the first domain (segments 1–99 and 252–345) and (b) superimposition of the second domain (segment 100–251).

domains, and the domain shift vector was defined as the difference between translation vectors. The calculated values are shown in Table II. The movement of domains causes the mutant structure to be closer than the native structure, as illustrated in Figure 4. To detect the hinge region associated with movement of domains, we used the method reported by Dixon et al. (1992). Sixteen-residue segments of the mutant enzyme were superimposed on the corresponding parts of the native structure and r.m.s.d. of backbone atoms were calculated. The segments were chosen sequentially from the N-terminus to C-terminus and distributions of r.m.s.d. are shown in Figure 5a. The hinge region is around the linkage between the second domain and C-terminal segment of the first domain. In the case of subunit 1, the hinge extends from residue 244 to residue 262, and this 47

C.Qu et al.

Fig. 4. Cα atoms of the two independent molecules in the mutant crystal (thick lines) superimposed on those of the native structure (thin lines, ammonium sulfate crystal): (a) according to the first domain (segments 1– 99 and 252–345) and (b) according to the second domain (segment 100– 251)

region is a helix (244–248)–loop (249–257)–strand (258–262) structural unit. The average temperature factor of main-chain atoms in this region is similar to the overall value in both structures, as shown in Table III. In the case of subunit 2, the hinge region consists of residues 266–290. Although this part contains mainly a loop region, with a short β-strand (268– 271) and part of a helix (288–290), the average temperature factor of main-chain atoms is also similar to the overall values. It seems, therefore, that these parts are not especially mobile or disordered. Given that main-chain dihedral angles can also reflect the main-chain trace very effectively, we calculated the pseudotorsion angles, defined by four adjacent Cα atoms, for each structure, and analyzed the differences between the mutant structure and the native one. Using seven-point smoothing techniques, we found several peaks appeared in the distribution profile (Figure 5b). As shown in Figure 5b, the two hinge regions described above were also detected, residues 248–256 in subunit 1 and residues 261–283 in subunit 2. However, the average temperature factor of the main-chain atoms in segment 248–256 was higher than the overall value, in particular in the case of subunit 1 (45.3 versus 29.4 Å2, Table III). In the structure of subunit 1, the temperature factor of residues 251– 256 was found to be higher than 40 Å2, showing high thermal mobility, whereas in the case of the native enzyme, only slightly higher temperature factors were found in this region. Thus, the large difference in torsion angle might be due to structural disorder of subunit 1. By contrast, the hinge region of subunit 2 retained a similar average temperature factor to the overall value, an indication that this region is not especially mobile in these two conformations. Several other regions were also detected, as indicated in Figure 5b. There were several common regions: residues 43– 51 in subunit 1 (and 43–57 in subunit 2), residues 65–82 in subunit 1 (and 66–87 in subunit 2) and residues 327–343. In all three structures, these regions gave higher average values 48

Fig. 5. (a) Plot showing the calculated r.m.s.d.s between subunit 1 and the native enzyme (continuous line) and between subunit 2 and the native enzyme (dashed line). The calculations were based on backbone atoms of 16-residue segments, from the N-terminus to the C-terminus. (b) Plot showing absolute differences in main-chain pseudo-dihedral angles, defined by four adjacent Cα atoms, between subunit 1 and the native enzyme (upper) and between subunit 2 and the native enzyme (lower). The lines were smoothed by seven-point averaging.

Table III. Comparison of average temperature factors (Å2) of main atoms in hinge regions Subunit 1 vs native

Subunit 2 vs native

Region

Native

Subunit 1

Region

Native

Subunit 2

Overall 244–262 6–11 43–51 65–82 96–100 141–149 248–256 327–334

24.9 24.4 20.4 28.9 35.3 21.5 28.0 28.5 31.5

29.4 33.2 22.8 33.3 44.3 24.6 33.2 45.3 38.6

Overall 266–290 43–57 66–87 103–109 117–122 132–139 261–283 327–334

24.9 22.8 29.4 36.4 21.4 23.8 21.9 22.3 31.5

27.1 26.1 35.2 33.0 22.9 29.3 24.8 26.3 34.6

of the temperature factor (Table III), indicating that these regions have higher thermal mobility and the difference in main-chain dihedral angles might be caused by structural disorder. However, some regions around residue 100, another domain connector, and the arm region (residues 134–158;

Mutation at the interface between domains

Fig. 6. Stereo view showing adjacent residues within a sphere of radius 5 Å, with the site of mutation at its center. This figure was drawn using MOLSCRIPT (Kraulis, 1991) and the coordinates of Imada et al. (1991).

Fig. 7. Cavities around the site of the mutation detected with the QUANTA program, using a probe of radius 1.4 Å. Cavity 1 is formed by the hydrophobic residues around the Cβ atom of A172, and cavity 2 is connected to the domain cleft channel.

Imada et al., 1991) had average temperature factors similar to the overall values. In this analysis, the method involving Cα pseudo-dihedral angles gave much more detailed information than the method of Dixon et al. (1992). The hinges for domain movement were located around the domain linkage and arm region, that is, near the active cleft (Kadono et al., 1995). The high mobility of some of these regions reflects the intrinsic flexibility of the protein, which is indispensable for its activity. Local structure around the site of the mutation. As shown in Figure 1, A172 is located at the domain interface. The mutation of Ala to Leu does not cause any local deformation of structure (Figure 4b). Figure 6 depicts all of the residues adjacent to the site of mutation. Residues L103, L129, V131, V168, A169, E171, A173 and V232 belong to the second domain, whereas H300 belongs to the first. All these residues, including residue A172, are located within either α-helices or β-strands, and they form a hydrophobic cavity (Figure 7) around the Cβ atom of residue 172. When the alanine residue is replaced by leucine, the cavity is not a suitable site for leucine’s side chain. In the native structure, the distance between the Cβ atom of A172 and the Nδ2 atom of H300 is 4.4 Å. When the side chain of

residue 172 is enlarged, it is too close to residue H300 and has a tendency to repel it. Because these two residues are located in the rigid parts of different domains, the interaction causes the two domains to move. As shown in Figure 4, the cleft with the mutation is more open, while the opposite cleft, the substrate-binding cleft (Kadono et al.,1995), is more closed. The distances between the Cα atoms of residues 172 and 300 are 7.56 Å in the native enzyme, 9.68 Å in subunit 1 and 9.17 Å in subunit 2. Therefore, the driving force for the movement of the domains is the van der Waals interactions among interdomain residues that are too close together. Comparison of the structures of the native protein crystallized from solutions of different precipitants The native crystal was grown in a solution of ammonium sulfate, pH 7.6, and the mutant crystal was grown in PEG 4000, pH 5.2. It is well known that the precipitant has an effect on protein conformation. Motion of domains around a hinge has already been characterized in the T4 lysozyme, and in this case, the bending is an intrinsic property of the lysozyme molecule and not an artifact due to mutation (Zhang et al., 1995). Moreover, a small-angle X-ray scattering experiment indicated that the present enzyme has several conformations in solutions with different additive (Kadono et al., 1995). In an effort to understand the mechanism of domain movement in the mutant structure, we analyzed the native structure using crystals obtained in PEG 4000. As shown in Table IV, domain movement of small amplitude was also found in the 3 Å structure of the native protein that had been crystallized under similar conditions to the mutant protein. The largest twist angle was about 2° among the four subunits in one asymmetric unit. This result indicates that this enzyme has the capacity for movement of domains around a hinge, which is a prerequisite for the folding of the A172L mutant. As also shown in Table IV, the r.m.s.d. of Cα atoms did not change significantly even when the entire molecules were compared. This result implies that the precipitant does not affect the conformation of the enzyme very much and that PEG 4000, as the precipitant, does not trap a closed conformation (Kadono et al., 1995). Thus the different arrangement of domains in A172L is caused by the mutation. Discussion Stability Although only one mutation was introduced to produce A172L, this mutation generates a new interface between domains. 49

C.Qu et al.

Table IV. Comparison of structures between the four molecules in PEG crystals and that in ammonium sulfate crystals of the native protein

Table VI. Solvent accessible areaa lost as a result of formation of the domain interface

Molecule

Molecule

Subunit 1

subunit 2

subunit 3

subunit 4

Fitting part entire molecule domain 1 domain 2 entire molecule domain 1 domain 2 entire molecule domain 1 domain 2 entire molecule domain 1 domain 2

Twist (°)

2.0

0.1

0.2

–0.8

Shift (Å)

R.m.s.d. of Cα atoms (Å)

2.8

0.46 0.37 0.32

2.5

0.46 0.37 0.32

1.4

0.39 0.43 0.29

1.0

0.38 0.42 0.29

10T A172L-subunit1 A172L-subunit2

Domain interface ASA ( Å2) Hydrophobicb

Hydrophilic

Total

1472 (71.8%) 1402 (73.7%) 1371 (72.7%)

577 499 506

2049 1902 1887

aSolvent-accessible

area (ASA) was calculated with the Quanta program, using a 1.4 Å probe rolling on the protein’s surface. bEach number in parentheses represents the percentage of the hydrophobic area taken from the total interface area.

Table V. Hydrogen bonds at domain interfaces in 10T and A172L 10T Conserved

Non-conserved

100 101 103 252 260 262 102 102 309

N...262 O N...262 O N...260 O N...249 O N...103 O N...101 O Nδ2...260 O Nδ2...261 Oγ Nη2...175 O

A172L-subunit 1

A172L-subunit 2

100 101 103 252 260 262

100 101 103 252 260 262 102 102

N...262 N...262 N...260 N...249 N...103 N...101

O O O O O O

N...262 O N...262 O N...260 O N...249 O N...103 O N...101 O Nδ2...260 O Nδ2...261 Oγ

107 Nζ...312 Oε2 164 Nη2...98 Oδ2

Three thermostability-related factors (Chan et al., 1995) that could involve the interface between domains, namely salt bridges, hydrogen bonds and solvent accessibility, were examined. There are no salt bridges across the domain clefts in either the native or the mutant structure. As shown in Table V, the main-chain hydrogen bonds at the interface are conserved, whereas the side-chain hydrogen bonds are changed. In one subunit, the total number of hydrogen bonds decreases, but in the other subunit in the asymmetric unit it increases. Therefore, it is unclear whether hydrogen bonds at the interface play an important role in the thermostability of the mutant protein. The area of the interface between domains can reflect the packing of two domains. If the domain interface is large, and its hydrophobic region is more extensive, inter-domain interactions will make a larger contribution to structural stability. Therefore, we calculated the solvent-accessible area that was lost by formation of a domain interface (Table VI). In the mutant, both molecules have a slightly smaller domain interface, as a result of the repulsive effect between L172 (domain 2) and H300 (domain 1). However, the ratio of the hydrophobic region taken from the interface in the mutant was slightly larger, being favorable for a domain interface. From the analysis of the three factors, it was difficult to conclude which arrangement of domains has a lower conformational energy. A small-angle X-ray scattering experiment indicated that the enzyme adopts a closed conformation when it has trapped its substrate (Kadono et al., 1995). In terms of the activity of the enzyme, domain closure must be 50

Fig. 8. Backbone drawing showing how crystal packing affects the domain closure in the mutant crystal. The labeled regions indicate the interacting parts of subunits. The first domain (segments 1–99 and 252–345) of subunit 1 (left) in one dimer interacts with both domains of subunit 2 (right; the open circle marks the site of the mutation) in the other dimer. Not all of the unit cell is shown.

rapid and the transition between open and closed forms cannot involve high-energy barriers (Gerstein et al., 1994). Therefore, we suggest that the arrangement of domains in the mutant has similar energy to that in the native enzyme. For the region adjacent to the site of the mutation, we calculated the solvent accessibility of various residues and their Voronoi volumes (Richards, 1985) using the VOLUME program (Richards, 1974). As listed in Table VII, the solvent accessibility of four residues (underlined in Table 7), including the mutated residue, increased, in particular that of residues A172L and H300, indicating that the local cavity is not suitable for leucine. One residue, 171E, gave the opposite result. This residue is on the salt bridge net formed by K178–D208–R174– E171. Thus, it is expected that a decrease in solvent accessibility to this residue will increase the strength of the salt bridge and make a positive contribution to the thermostability of the mutant structure. The reduction in Voronoi volumes of all these residues is also presented in Table VII, and it indicates that these residues are packed more tightly. Leucine has three more carbon atoms than alanine, so it adds more hydrophobic interactions to the local structure and stabilizes the mutant structure. As mentioned in the Results section, the side chain of leucine is too long to be allocated in the cavity near A172 and it pushes the opposite domain away. In domain 2, the hydrophobic side chain of leucine fills the cavity to some extent, and in domain 1, the interface residues which lost some interactions from domain 2 readjust themselves and acquire a tighter packing intensity as indicated by the smaller Voronoi

Mutation at the interface between domains

Table VII. Solvent accessibility and Voronoi volumes of adjacent residues (radius 5 5 Å) around the site of the mutation No.

172 103 129 131 168 169 171 173 232 300

Res.

A/L L L V V A E A V H

10T

Subunit1

Subunit2

Acc. (%)

Vol. (Å3)

Acc. (%)

Vol. (Å3)

Acc. (%)

Vol. (Å3)

0 0 2 0 12 0 48 0 0 3

115.4 196.5 197.5 166.5 160.9 100.0 151.3 88.1 159.3 195.9

7 0 3 0 15 0 37 0 0 16

197.0 195.7 187.5 170.4 154.4 101.4 144.1 87.8 161.7 176.9

6 0 4 0 15 0 38 0 0 18

187.1 189.4 189.6 168.4 146.4 104.8 135.4 90.6 162.0 184.2

volume in Table VII. Obviously, leucine influences the local packing of side chains to a greater extent than would a cavityfilling mutant. Activity The mutant enzyme has a closed conformation. However, its activity is similar to that of the native enzyme. As suggested by Hurley and Dean (1994), the width of the activity cleft is measured by the distance between Cα atoms of G255 and I279. In the complex of IPMDH with NAD, this distance is 12.4 Å (Hurley and Dean, 1994). In the free enzyme, the distances are 13.6, 13.9 and 13.3 Å for 10T, the crystallographically independent subunit 1 and subunit 2 of the mutant enzyme, respectively. Thus, closing of the domain cleft only favors interactions of side chains around the site of the mutation and does not affect activity. Conformation of the native enzyme Structural comparisons of crystals of the native enzyme crystallized from solutions of different precipitants showed that the overall conformations were almost the same. The precipitant had little effect on the conformation of the enzyme. However, small-amplitude movement of domains was observed, suggesting that this enzyme has the capacity for domain motion around a hinge. This motion is the prerequisite for the folding of A172L. These observations confirmed that the domain movement in A172L is caused by mutation. The mutation at the domain interface amplifies the capacity for motion around the hinge of the enzyme The mutation is located at a site suitable for an effect on interdomain interactions. The van der Waals interactions between the side chains of residues at positions 172 and 300 that are too close together drive the two domains apart. The hinge regions of the structure are near the active cleft, implying that motion around the hinge is needed for activity. In general, when stability of proteins is examined by sitedirected mutagenesis, amino acid substitutions fall into two basic categories (Matthews, 1993), those on the surface and those in the rigid part of the protein. Substitutions at the surface of a protein have little effect on stability (Hecht et al., 1983; Reidhaar-Olson and Sauer, 1988), while substitutions in the rigid part of the protein can change its stability significantly (Alber et al., 1987). The replacement of Ala172 by Leu can be regarded as a member of the second category. More appropriately, however, this mutation should be regarded as a new kind of mutation, namely a domain interface mutation. If the protein has the ability for movement of domains around a hinge, mutations of various types can be introduced. Some

mutations will cause rearrangement of domains, rather than the disruption of the structure that is caused by mutations in rigid parts. In the present case, the motion around hinge regions allows the protein to minimize the effects of a potentially deleterious substitution by readjustment of domains to give an alternative structure that is energetically similar to the wild type. Crystal packing Crystal packing complicates the problem. As analyzed above, the conformations of the two molecules in an asymmetric unit of the A172L crystal and that of the four molecules in the asymmetric unit of native enzyme (PEG crystal) are different. In the case of the mutant enzyme, crystal packing has an effect opposite to the mutation effect, as shown in Figure 8. In one crystal cell, the first domain of subunit 1 in one dimer interacts with both domains of subunit 2 in the other dimer. As discussed above, the leucine side chain tends to cause the cleft with the mutation to open up and, therefore, it tends to close the opposite cleft. Because of the inter-subunit interactions, the closing of this cleft is hampered. As shown in Table II, the twist angle of subunit 2 is only about 2°. Therefore, subunit 1 has a more closed conformation than subunit 2. Thus, the driving force for movement of domains in the A172L protein is clearly the mutation. Atomic coordinates and structure factors have been deposited with Protein Data Bank (references: 10SJ and R10SJRF, 10SI and R10SIRF). Acknowledgements This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education Science and Culture of Japan (Nos 05244102 and 07304050). The VOLUME program was supplied by Dr P.J.Fleming, Department of Molecular Biophysics and Biochemistry, Yale University. The authors thank to Professor B.W.Matthews, Research Laboratories, Howard Hughes Medical Institute, for helpful discussions.

References Akanuma,S. (1995) Master’s thesis, Tokyo Institute of Technology, Yokohama. Akanuma,S.,Yamagishi,A. and Oshima,T. (1995) Protein Engng, 8, Suppl., 4. Alber,T., Dao-pin,S., Wilson,K., Wozniak,J.A., Cook,S.P. and Matthews,B.W. (1987) Nature, 330, 41–46. Bru¨nger,A.T., Karplus,M. and Petsuko,G.A. (1989) Acta Crystallogr., Sect. A, 45, 50–61. Bru¨nger,A.T., Krukowski,A. and Erickson,J. (1990) Acta Crystallogr., Sect. A, 46, 585–593. Chan,M.K., Mukund,S., Kletzin,A., Adams,M.W.W. and Rees,D.C. (1995) Science, 267, 1463–1469. Dixon,M.M., Nicholson,H., Shewchuk,L., Baase,W.A. and Matthews,B.W. (1992) J. Mol. Biol., 227, 917–933. Fox,G.C. and Holmes,K.C. (1966) Acta Crystallogr., 20, 886–891. Gerstein,M, Lesk,A.M. and Chothia,C. (1994) Biochemistry, 33, 6739–6749.

51

C.Qu et al. Hecht,M.H., Nelson,H.C.M. and Sauer,R.T. (1983) Proc. Natl Acad. Sci. USA, 80, 2676–2680. Hurley,J.H. and Dean,A.M. (1994) Structure, 2, 1007–1016. Imada,K., Sato,M., Tanaka,N., Katsube,Y., Matsuura,Y. and Oshima,T. (1991) J. Mol. Biol., 222, 725–738. Jancarik,J. and Kim,S.H. (1991), J. Appl. Crystallogr., 24, 409–411. Jones,T.A. (1985) Methods Enzymol., 115, 157–171. Kadono,S., Sakurai,M., Moriyama,H., Sato,M., Hayashi,Y., Oshima,T. and Tanaka,N. (1995) J. Biochem.. 118, 745–752. Kraulis,P.J. (1991) J. Appl. Crystallogr., 24, 946–950. Matthews,B.W. (1993) Annu. Rev. Biochem., 62, 139–160. Miyazaki,K., Kadono,S., Sakurai,M., Moriyama,H., Tanaka,N. and Oshima,T. (1994), Protein Engng, 7, 99–102. Numata,K., Muro,M., Akutsu,N., Nosoh,Y., Yamagishi,A. and Oshima,T. (1995) Protein Engng, 8, 39–43. Onodera,K., Sakurai,M., Moriyama,H., Tanaka,N., Numata,K., Oshima,T., Sato,M. and Katsube,Y. (1994) Protein Engng, 7, 453–459. Reidhaar-Olson,J.F. and Sauer,R.T. (1988) Science, 241, 53–57. Richards,F.M. (1974) J. Mol. Biol., 82, 1–14. Richards,F.M. (1985) Methods Enzymol., 115, 440–464. Sakurai,M. et al. (1992) J. Biochem. 112, 173–174. Sakurai,M. et al. (1995), Protein Engng, 8, 763–767. Sato,M., Yamamoto,M., Imada,K., Katsube,Y., Tanaka,N. and Higashi,T. (1992), J. Appl. Crystallogr., 25, 348–357. Zhang,X.-J., Wozniak,J.A. and Matthews,B.W. (1995) J. Mol. Biol., 250, 527–552. Received July 8, 1996; revised August 26, 1996; accepted August 27, 1996

52