J. Mol. BioE. (1991)

221,

309-325

Refined Crystal Structure of the Complex of Subtilisin BPN’ and Streptomyces Subtilisin Inhibitor at l-8 A Resolution Yasuo Takeuchi Pharmaceutical Research Center, Me@ Seika Kaisha, Morooka, Yokohama 222, Japan

Yoshinori Faculty

Ltd,

Satow

of Pharmaceutical Sciences, University Hongo, Tokyo 113, Japan

of Tokyo

Kazuo T. Nakamura and Yukio Mitsui f Faculty

of Engineering, Nagaoka University of Technology Nagaoka, Niigata 940-21, Japan

(Received The

crystal

3 January

1991; accepted 24 April

1991)

structure

of subtilisin BPN’ complexed with a proteinaceous inhibitor SSI inhibitor) was refined at 1.8 A resolution to an R-factor of @177 with a root-mean-square deviation from ideal bond lengths of O-014 A. The work finally established that the SSI-subtilisin complex is a Michaelis complex with a distance between the Oy of active Ser221 and the carbonyl carbon of the scissile peptide bond being an intermediate value between a covalent bond and a van der Waals’ contact, 2.7 8. This feature, as well as the geometry of the catalytic triad and the oxyanion hole, is coincident with that found in other highly refined crystal structures of the complex of subtilisin Novo, subtilisin Carlsberg, bovine trypsin or Streptomyces griseus protease B with their proteinaceous inhibitors. The enzyme-inhibitor p-sheet interaction is composed of two separate parts: that between the Pi-P, residues of SSI and the 125-127 chain segment (the “S1_3 site”) of subtilisin and that between the P,-P, residues of SSI and the 102-104 chain segment (the “S4_6 site”) of subtilisin. The latter b-interaction is unique to subtilisin. In contrast, the /?-sheet interaction previously found in the complex of subtilisin Novo and chymotrypsin inhibitor 2 or in the complex of subtilisin Carlsberg and Eglin C is distinct from the present complex in that the two types of fl-interactions are not separate. As for the flexibility of the molecules comprising the present complex, the following observations were made by comparing the B-factors for free and complexed SSI and comparing those for free and complexed subtilisin BPN’. The rigidification of the component molecules upon complex formation occurs in a very localized region: in SSI, the “primary” and “secondary” contact regions and the flanking region; in subtilisin BPN’, the S,_, and S,., sites and the flanking region.

(Streptomyces subtilisin

Keywords:

crystal

structure;

subtilisin;

SSI; enzyme

structure;

protein-protein

interaction

1. Introduction $ Abbreviations used: SSI, Streptomyces subtilisin inhibitor; BPTI, bovine pancreatic trypsin inhibitor; CI-2, chymotrypsin inhibitor 2; 0MTKY3, ovomucoid inhibitor from turkey, domain 3; PSTI, pancreatic secretory trypsin inhibitor; SGPB, Streptomyces griseus protease B; n.m.r., nuclear magnetic resonance; r.m.s.

SSIS is one of the few well-characterized microbial protein protein&se inhibitors and represents a distinct family of protein protease inhibitors 7 Author to whom all correspondence should be

root-mean-square;

addressed. 0022-2836/91/17030~17

$03.00/O

309

p.p.m.,

parts 0

per million. 1991 Academic Press Limited

Y. Takeuchi et al.

310

(Laskowski & Kato, 1980). SSI is a stable dimer (I,) composed of two identical subunits, each of molecular weight 11,500 (Murao & Sato, 1972). It strongly inhibits a microbial serine proteinase, subtilisin BPN’(E), forming an E,I, complex of molecular weight 79,ooO. There are many reasons for believing that the structures of the complexes of serine proteinases with their proteinaceous inhibitors are only subtly different from those of the complexes wlt,h true substrates. A proteinaceous inhibitor bound on the target enzyme can be regarded as a substrate trapped in a potential energy minimum. The structure and function of the SSI-subtilisin system have been extensively studied, as reviewed by Mitsui et al. (1985) and in a recent monograph (Hiromi et aZ., 1985). The crystal structure of free SSI crystallized in a trigonal space group P3121 has been solved (Mitsui et a,l., 1977. 1979a,b) and now refined at’ 2.05 a resolution (Y. Takeuchi, K. (1 A = wl nm) Ishikawa & Y. Mitsui, unpublished results). On the other hand, the crystal structure of SSI-subtilisin complex crystallized in an orthorhombic space group I222 was initially solved at 4.3 A resolution (Hirono et al.; 1979) and the unrefined structure of the complex analyzed at, 2.6 A resolution has been described (Hirono et al., 1984). For the latter structure analysis, the known struct’ure of free subtilisin BPN’ (or equivalently Novo) (Wright’ et al.. 1969; Drenth et aE., 1972) was utilized. The structure of subtilisin BPN’ has now been fully refined at 1.8 A resolution (Bott et al., 1988) and the result is utilized in the present structure analysis. We now report’ the structure of SSI-subtilisin BPN’ complex refined against the 1% A synchrotron data. Thanks to the unusually abundant crystallographic data (co-ordinates and B-factors) accumulated for both the enzyme subtilisin BPN’ and a substrate analog SSI in both the free and complexed states, we were able to examine the effect of protein-protein interaction for two component proteins with respect to

Table 1 Experimental

conditions

Crystal

1

Size of crystal

(mm)

Rotation axis Beam current (mA) Collimator aperture Crystal-to-film distance Oscillation angle c0 (deg.)/filma w-Rotation/film movement (deg./mm) Number of oscillations/ film w-scan speed (deg./s) Number of films Total exposure time (s) ’ Actually

of data collection

Fuji imaging

2

0.35 (a) x 0.4 (a) x 065 (b) x 075 (b) x 0.1 (c) 0.12 (c) b 172-167 26%21;9 *3 mm x 0.3 mm 430 mm X.0

9.0

4.0

1..i

5 2.0 13 520

3 2.0 13 351

plates were used as films.

both the static and dynamic nature of the relevant molecules. Furthermore, by comparing the present results with those obt’ained for the other two cryst,al structures of the complex of subtilisin Novo OI subtilisin Carlsberg with their proteinaceous inhibitors (McPhalen et aE., 1985a,b; Bode et aE., 1986: McPhalen & James, 1988) and with those obtained for crystal st’ructures of the complex of trypsin. trypsinogen or Streptomyces proteinases with their proteinaceous inhibitors (Huber &r Bode, 1978: et al., 1982: Bolognesi d al.. 1982: Fujinaga Marquart et al.. 1983; Read et al., 1983; Chen Br Bode. 1983; Tsunogae d aE.. 1986). we were able to classify various st’ructural characteristics found in the present st,udv into (1) what is unique to the present SST--sub&isin complex. (2) what is uniquck to subtilisin complexes. (3) what is unique to the complexes involving serine proteinases and (4) what is perhaps common among protein interaction systems in general.

2. Experimental

Methods

(a) Data collrction The crystallization and crystal structure analysis at 2.6 A resolution by multiple isomorphous replacement methods, using a conventional 4-circle diffractometer. has been described (Hirono et a,Z.; 1979, 1984; Mitsui et al.. 1979a). The space group is 1222 with a unit cell having the dimension of a, = 77.2, b = 1859. c = 695 A and containing half the complex molecule (EI) of 40.250 dalt,on per asymmetric unit. For high resolution data collection, a macromolecule-oriented Weissenberg camera devised by Professor N. Sakabe (Sakabe, 1983) installed at Beam Line 6A of the Synchrotron Radiation Source at the National Laboratory for High Energy Physics (Tsukuba) was utilized. The wavelength was set to 1.04 A and the crystals were cooled at about 15°C. Two crystals were used, one with the b-axis and the other with the c-axis aligned as the rotation axis. A pack of Fuji Imaging Plate (Miyahara et al., 1986). which is about 10 to 20 times as sensitive as conventional X-ray films and yet exhibits lower fog level. was used as a Z-dimensional detector. The conditions of data collection are summarized in Table 1. The X-ray diffraction image stored in the imaging plates was digitized and stored in magnetic tape with a Fuji BAlOO readout system. The data sets were processed using the program WEIS (Higashi. 1989). which is an extensive modification of M. G. Rossmann’s oscillation film processing package (Rossmann, 1979). The scaling between imaging plates was carried out using the SCALE module contained in the same package. No absorption correction was applied. The R,,,(Z) was 4.72’!, and 475% within the data sets taken around the b- and c-axes. respectively. The merging R(Z) factor between the 2 data sets:

where h and i st.and for Miller index and data set number. was 556% for 97,766 observations (up to 1.8 A resoluCon, I > 20(I)). which were reduced to 28,118 independent reflections. In adopting an entirely new method of data collection. we naturally felt it necessary to compare the new data with the old data (Hirono et al.. 1984) which

Rejhed Structure of SSI-Subtilisin

(8)

d II

Table 2 Final rejnement parameters and results

2.8 1

5 I

?? No. of cycles R factor Resolution range (A) No. of reflections No. of protein atoms No. of solvent atoms’ No. of variable parameters r.m.s. deviations from ideal valuesb Distance restraints (A) Bond distance Angle distance Planar 1-4 distance Plane restraint (A) Chiral-center restraint (A3) Conformational torsion angle Restraint planar (“)

IO 0

??

8-

‘;; 2

6-

0

.

h 4-

2

??

I

??

??

I

0.02

I

I

0.06

O-10 sin

I

o-14

311

Complex

I

0.18

146 0.177 61.8 26,760 2702 260 11,989

0014 0.034 0043 9011 0.110

(9017) (0027) (0037) (9020) (9100)

48 (25)

“Including 2 Ca2+ ions in the total. bThe values of CT,in parentheses, are the input estimated standard deviations that determine the relative weights of the corresponding restraints (see Hendrickson, 1985).

8/X

Figure 1. Plot of the relative difference in structure factor (RF = ClF(new) - F(old)@F(new)) between the 1.8 A synchrotron data (F(new)) and the 2.6 A diffractometer data (F(old)) as a function of resolution range.

was collected up to 2.6 A resolution by a 4-circle diffractometer Rigaku AFC-4 on a sealed-tube generator (45 kV, 24 mA). The relative difference in structure factor, RF = (ZlF(new) - F(old)l/ZF(new)) were 609%, 7.11% and 602% when the reference data were the b-axis data set, the c-axis data set and the 2-axes-merged data set, respectively. The resolution breakdown of these statistics is shown in Fig. 1. It will be noticed that the statistics, especially up to medium resolution, are very good considering the completely different geometry and detection system. (b) Starting models An electron density map at 2.6 A resolution was calculated using previously obtained multiple isomorphous replacement phases. The models of SSI were constructed as described (Hirono et al., 1984). For the interpretation of the enzyme part, refined co-ordinates of subtilisin BPN kindly provided by R. Bott (Bott et al., 1988) were fitted to the electron density map by a least-squares procedure (Urata et al., 1981). Several parts of the model, however, had to be rebuilt as clearly indicated by the (F,, - F,) difference Fourier map. For this procedure, the program TOM (Jones, 1978) running on a Silicon Graphics IRIS 3030 display was used. It turned out that, in the old interpretation, there was an error of 5 degrees in the orientation of the enzyme molecule. A global least-squares superposition of the a-carbon chains of old model and this model in the enzyme using locally developed program LSFIT (Urata et al., 1981) gave an r.m.s. deviation of 088 A. However, since the enzyme model had been placed in the crystal lattice taking the enzyme-inhibitor interface as a pivot axis, the present rebuilding does not affect the previous description (Hirono et al., 1984) of major features of the enzyme-inhibitor interface.

(c) Refinement The restrained-parameter least-squares refinement program of Hendrickson (1985), modified locally by H. Akagawa for compatibility with a HITAC S-820 supercomputer, was used to refine the structure of the complex against 26,760 independent reflections at 1.8 A resolution (F 2 a(F)). The percentage against theoretically possible reflections were 97.8, 950, 81.6, 62.6 and 355% for 5 shells divided at co, 4, 3, 2.6, 22 and 1.8 A. The stereochemical restraints were initially relatively weak, to prevent the model from falling into a local minimum. As the refinement progressed, and the model became more accurate and complete, the restraints were tightened to impose better geometry. However, no intermolecular distance restraints were imposed. 110 cycles of refinement, interrupted by 3 sessions of manual rebuilding, reduced the R-factor to 6211 (for 50 to 1.8 A resolution). A total of 262 solvent molecules were located. Two strong solvent densities were redefined as Ca*+ ions on the grounds of strong residual electron density in the (F, - FJ maps and proper distances to possible ligands co-ordinated to it. After 37 more cycles, refinement converged at R = 0177 (6.0 to 1.8 A resolution). Thirty-seven of the solvent molecules had B > 70 A* and were omitted from the atoms list. Because occupancies and temperature factors are strongly correlated, the former were fixed at 1.0 for all solvent molecules and for all protein atoms. See Table 2 for a summary of the refinement results. The co-ordinates have been deposited with the Protein Data Bank (identity code 2SIC).

3. Results and Discussion Because of the significant error in the orientation of subtilisin BPN’ in the previous model (see Experimental Methods, section (b)), an updated version of the a-carbon chain drawing of the SSI-subtilisin BPN’ complex (as an E,I, complex) is shown in Figure 2. To facilitate further discussion, notation of the secondary structures and functional notation for the reactive site amino acid residues in SST are schematically shown in Figure 3.

312

Y. Takeuchi

et’al.

Figure 2. cd‘arbon chains of a complex of MI dimer (bold bonds) with :! subtilisin bar represents 10 a. This is an updated version of Fig. 3 of Mitsui et nZ. (1979,).

(a) The SSI-subtilisin

BPN’ interface

The region of contact in the final refined model is shown in Figure 4 and the nature of intermolecular contacts are summarized in Table 3. There is a total of 154 contacts of less than 4.0 A between the inhibitor and the enzyme. The hydrogen-bond scheme between the polypeptide main-chains of SSI and the enzyme is shown in Figure 5 together with the appropriate distances. As described in the previous paper (Hirono et al., 1984), there are two separate antiparallel P-sheet interactions, one between the P,-P, residues of SSI and the Ser125-Gly127 chain segment of the enzyme and the other between the P,-P, residues of SSI and the Gly102-Tyr104 chain segment of the enzyme. Referring to the Schechter & Berger (1967) notation for the binding site of enzymes, the former and latter chain segments of subtilisin BPN’ were named “S,., site” and “S,., site”, respect’ively (Hirono et al., 1984). Since our previous publication, two highly refined structures of subtilisin complexed with proteinaceous inhibitor have appeared: (1) subtilisin Novo complexed with chymotrypsin inhibitor 2 (hereafter referred to as CT-subtilisin complex) (McPhalen et al., 1985b: McPhalen & James, 1988) and (2) subtilisin Carlsberg complexed with Eglin C (hereafter referred to as Eglin-subtilisin complex) (McPhalen et al., 1985a; Bode et al., 1986; McPhalen & James, 1988). Note that subtilisin Novo is known to be chemically identical with subtilisin BPN’ used in the present study. Between subtilisin BPN’ (or Novo) and subtilisin Carlsberg, there are 82 amino acid changes and one deletion; thus they are 70% identical. As for the S1_3 and S,., sites mentioned above, there is only one change in amino acid residues; Gln103 (in BPN’) versus SerlO3

KF’X’ rr~oI~~~~~lw (thin bonds).

SC&

(in Carlsberg). McPhalen dt James (I 988) stated that there are less t,han I A displacement between t’he r-carbons of subtilisins BPN’ and (la&berg in these regions.

u

SSI Subunit

I N-teml.

Figure 3. Schematic drawing of one SSI subunit showing the /?-strands (arrow plates), a-helices (filled bonds), p-turn (shaded bonds), S-S bridges (zigzag line). scissile peptide bond (triangle). N terminus and (’ terminus. Notations (P, P,’ and so on) after Schechter h Berger (1967) are indicated for some of the amino acid residues along the reactive site loop (or the primary contact region (Mitsui et al., 1979a)). Similar notation (Q,,. Q1’ and so on) after Hirono et al. (1984) are also indicated for some of the amino acid residues along the secondary contact region (Mitsui et al.. 1979a) having some direct contacts with the surface of subtilisin BPN’ (see Table 3).

Rejined Structure of SSI-Subtilisin

Complex

313

Figure 4. The region of contact between SSI (bold bonds; residue numbers for SSI are shown with the suffix I) and subtilisin RPN’ (thin bonds). The view direction is similar to Fig. 5. All residues related to the intermolecular short contacts of less than 4 A (see Table 3) are shown. This is an updated version of Fig. 4 of Mitsui et al. (1979a) although the direction of view is completely different.

The /?-sheet interaction between the Pi-P, residues of the inhibitor and the S1_3 site of the enzyme found in SSI-subtilisin complex is a common feature with the above two subtilisin complexes, the CI-subtilisin complex and Eglin-subtilisin complex. Moreover, the same type of interaction has been found in all the so far examined complexes of the serine proteinases of the trypsin-chymotrypsin family with their proteinaceous inhibitors although, in these cases, the residue numbers of the S1_a site are 214, 215 and 216 reflecting the completely different chain folding of these enzymes from subtilisin-type enzymes (trypsin-BPTI complex (Huber et al., 1974; Huber & Bode, 1978), SGPB-OMTKYS complex (Fujinaga et al., 1982; Read et al., 1983), trypsinogen-PST1 complex (Bolognesi et al., 1982), Kallikrein A-BPTI complex (Chen & Bode, 1983), trypsin complexed with a Bowman-Birk type protease inhibitor (Tsunogae et al., 1986), cc-chymotrypsin complexed with mucous proteinase inhibitor (Griitter et al., 1988)). On the other hand, another /l-sheet interaction, between the P,-P, residues of the inhibitor and the S4_6 site of the enzyme, found in SSI-subtilisin complex seems to be unique to subtilisin. However, there is a considerable variation in the details of the b-interaction among the three subtilisin complexes. Thus, unlike the case of SSI-subtilisin complex (see Fig. 5), in the case of CI-subtilisin and Eglin-subtilisin complexes, there are no hydrogenbond connection between the peptide NH and

carbonyl oxygen of the P, residue (Gly54 in CI-2 or Gly40 in Eglin C; the numbering scheme is after Fig. l(b) of McPhalen & James (1988)) and peptide NH of the S, sites (Tyr104 in both subtilisins Carlsberg and BPN’ (or Novo); McPhalen & James. 1988). Instead, there is a strong hydrogen bond between the peptide NH of the P, residue (Thr58 in CI-2 or Thr44 in Eglin C) and the carbonyl oxygen of the S, site (GlylOO in both subtilisins Carlsberg and BPN’). Thus the P-sheet interaction found in the Eglinor CI-subtilisin complexes has been described as a three-stranded antiparallel /J-sheet with the inhibitor polypeptide segment as the central strand (McPhalen & James, 1988). In contrast, in the case of SSI-subtilisin complex, it seems more appropriate to describe it as consisting of two separate two-stranded antiparallel b-sheets, one involving the S,_, site and the other involving the S4_6 site. The absence in the enzymes of the trypsin-chymotrypsin family of potential binding sites similar to the S,, site of subtilisin has been noted based on the quantitative comparison of the active site surfaces between the two kinds of enzyme (Hirono et al., 1984). In the interface of SSI-subtilisin complex, there is another region of contact involving a chain segment of SSI which is closely located to the reactive site loop (Pr to P, residues and the flanking region) thanks to the Cys71 (P,)-CyslOl disulfide bridge. This region was named “secondary contact region” (Mitsui et al., 1979a) to discriminate it from the reactive site loop (the “primary contact region”).

Y. Takeuchi et al.

314

I

.

the Q1 residue (CyslOlI) (the suffix I denotes inhibitor amino acid residue and will be attached only when necessary to avoid confusion) and the Q4 residue (Ser981) are in direct contact with subtilisin BPN’. Especially the Oy of Ser981 forms a strong hydrogen bond (2.55 d) with 0” of Glu156. Thus in the case of SST-subtilisin complex, the secondary contact region contributes to intermolecular interaction in addition to its obvious role as a supporting device for the primary contact region. In contrast. in the case of 0MTKY3 and PSTI, which also have secondary contact regions evolutionarily (Mitsui it al., 1979a) and structurally (Hirono et al.. 1984) closely related to the corresponding region of SSI. the secondary contact regions are never in direct contact with their target enzymes (SGPB in the case of OMTKYS, trypsinogen in the case of PSTI). Tn the proteinaceous inhibitor Eglin C and U-2 there are no chain segments similar to the secondary cont,act region of SSI. Thus in the Eglin-subtilisin and (IT-subtilisin complexes the intermolecular cont,acts are exclusively through the reactive site loop (or the primary contact, region) of the inhibitor.

rj,,____3~‘P___~=~

I

(b) The catalytic environment

Figure 5. Schematic drawing showing the hydrogenbond scheme and distances in the active site. Amino acid residues of SSI are doubly squared with the scissile peptide bonds indicated by a wedge. Secondary contact which is connected with the primary contact region, region through the 71-101 disulfide bridge. is also shown. See the text.

So-called Q-notation (see Figs 3 and 5) was proposed to represent the amino acid residues in this region (Hirono et al., 1984). As shown in Table 3,

As shown in Figure 6, there is a very good hydrogen-bonded system of the catalytic triad in the SSI-subtilisin complex. Thus the distance between the Oy of the act,ive serine (Ser221) and NE1 of His64 is 2.7 .A and that between Nd2 of His64 and a carbonyl oxygen of Asp32 is 2% A. indicating a strong hydrogen bond in each case. As shown in Table 4, this geometry is almost coincident within the experimental error among the highly refined enzyme structures bound with proteinaceous inhihitors. Thus at least when bound with proteinaceous inhibitors, the original scheme noted by Blow rt al. (1969) as to the hydrogen-bonded network of the

Table 3 Contacts of less than 4.0 d between subtilisin

Residue

Site

No. of contacts

Primary contact region Srg651 P, Glu671 p, Asp681 PC Va169I P, Met701 P‘l

13 10 9 24

cys7 1I Pro721 Met731

P, P, p,

7 13 32

Va1741

P,’

14

Tyr751

P,’

Secondary contact region Ser981 Q4 CyslOlI

QI

6

RPN’ and KY’/

(‘omments

(Ilose contacts of side-chain with Gly127. Glyl28, Pro129 side-chain Close contacts with side-chain of Tyr104, Serl30 B-Bridge with TyrlO4, close contacts with Gln103. Tyr104 Close contacts with Gly102, Tyr104 B-Bridge with 01~102; hydrophobic. with S4 binding pocket (SerlOl. (:l,vlO%. T,vr104 side-chain, Leu126 side-chain, Gly127, Leu135 side-chain) B-Bridge with Gly127, close contacts with GlylOO, Leul26 Hydrophobic, with 52 binding pocket (His64 side-chain, GlylOO. Ser12.5) Carbonyl 0 in oxyanion hole (Ser221. Asn155); 3.1 A from NH to Serl25 CO: hydrophobic, with Sl binding cleft (Leu126. Gly127. Gly154. Asnl55); close contacts His64, Asn155, Gly219, Thr220. Rer221 (!lose contacts with His64 side-chain, Asn155 side-chain. Asn21X. Ser221 side-chain. Met222 side-chain B-Bridge with Asn218; hydrophobic, with Phe189: close contacts with Asnl55 side-chain, Asn218 Close contacts of side-chain with Asnl55. 01~156 side-chain. Hydrogen bond of 256 A from Oy to Glu156 OE’ Close contacts of side-chain with Asp99. SerlOl

Rejined Stvucture of SSI-Subtilisin

Complex

315

Figure 6. The catalytic environment in the SSI-subtilisin interface. The P,(Pro72), P,(Met73), Pi’(Va174) and P,’ (Tyr75) residues flanking the scissile peptide bond are shown with bold bonds (residue numbers for SSI are shown with the suffix I), while the amino acid residues of subtilisin constituting the catalytic triad and oxyanion hole are shown with thin bonds.

catalytic triad is strictly realized in the target serine proteinases. As for the state of the scissile peptide bond and its location against the Oy of active Ser221, the tentative conclusion presented in the previous paper (Hirono et al., 1984) is finally borne out. Thus the refined distance between the Oy atom and the carbonyl carbon of the scissile peptide bond between Met731 and Va1741 is 2.7 A (no distance restraints were imposed for this pair of atoms). The value is an intermediate between the covalent distance and the mere sum of the van der Waals’ radii and, within the experimental error, coincides with the corresponding values found in other inhibitors with proteinaceous (see complexes Table 4). The notion that the SSI-subtilisin complex is a Michaelis complex seems most directly sustained by observation (Kainosho et al., 1982; an n.m.r. Kainosho & Miyake, 1988) made for the isotopeenriched sample of SSI in which the peptide N and carbonyl C of the scissile peptide bond were specifically enriched by i5N and 13C isotopes, respectively. Thus Kainosho and his colleagues were able to observe the splitting due to the bicinal coupling constant 1J(‘5N-‘3C) in their 13C n.m.r. spectra.

This observation clearly intermediate. Moreover

excludes the case of acyl the case of tetrahedral

intermediate was safeiy excluded on the basis of the observed 13C chemical shift of the carbonyl carbon of the P, residue (Met73), which would have been at least 196 p.p.m. upfield shift if the carbon is in the sp3 hybrid state (Kainosho et al., 1982). Similar observation was also made for the 13C-labeled ST1 (Hunkapiller et aZ., 1979; also see Laskowski & Kato, 1980). Pyramidalization of the carbonyl carbon atom of the scissile peptide bond should indicate how far an enzyme-bound inhibitor has moved along the reaction pathway toward a tetrahedral intermediate. A review of many serine proteinase-inhibitor complexes (Read & James, 1986) concludes that although small distortions of this bond are observed, the size of the distortion is comparable to the errors in co-ordinates bfthe structures.However, earlier examination by Marquart et al. (1983) of 16 crystal structures of the complex of trypsin and trypsinogen with various proteinaceous inhibitors concluded that the peptide groups may be substantially non-planar and this non-planarity can be reliably analyzed under the conditions of their study. In the present study, the following computa-

316

Y. Takeuchi

et al

Table 4 Catalytic

geometries observed in the crystals of serine proteinases

complexed

with their proteinaceous

inhibitors

Inhibitor

BPTI”

PSTIb

SW

Eglin-c

0MTKY3d

Enzyme

Trypsin

Trypsinogen

Subtilisin BPN’ h

Hubtilisin Carlsberg

SGPB’

Reference Distance (A) Pi carbonyl C to Ser221 Oyk Hydrogen-bond distances (A) Asp-His Hisser Oxyanion hole

f

B

26

2.68

2.7

2%

2.7

2.7 27 NH (Ser195) NH (Gly193)

2.79 2.66 NH (Ser195) NH (Gly193)

2% 27 NH (Ser221) NH, (Asn155)

2.7 2.7 NH (Ser221) NH, (Asn155)

255 NH (Ser195) NH (Gly193)

J

a Bovine pancreatic trypsin inhibitor (Kunitz). bPancreatic secretory trypsin inhibitor. ’Streptcvmycessubtilisin inhibitor. dOvomucoid inhibitor from Turkey. domain 3. e Streptomyces griseus protease B. ‘Huber & Bode (1978); Marquart et al. (1983). g Bolognesi et al. (1982). “Present study. ‘McPhalen rt al. (1988). j Fujinaga et al. (1982); Read et al. (1983). ‘The equivalent residue from the subtilisins to trypsin and SGPB is Ser221 + Ser195.

tional facts were observed. After the completion of usual restrained crystallographic refinement (see Experimental Methods), which imposed planar restraints on all the peptide bonds, the planar restraints were removed and several cycles of refinement were carried out. The g4 angle (out-of-plane bending of the carbonyl oxygen with respect to C”. C’ and N) for the I’1 residue of SST (Met73) was (b8”. while the r.m.s. deviation among all residues was 2.6”. In fact, many residues showed larger e4 angle than that of Met73, the maximum deviation from 0” being 9” for Lys89. Thus we could not find any strong evidence for the specific pyramidal distortion of the scissile peptide bond. However, we should be careful about interpreting the above observation since the relaxation on planarit,y restraints was applied only after the long processes of refinement where the planarity restraints had been constantly imposed. Again the measurement of the bicinal coupling constant 1J(15N-‘3C) in the r3C n.m.r. spectra mentioned above (Kainosho et al., 1982; Kainosho & Miyake, 1988) should provide the clearest evidence. According to Kainosho (private communication), the preliminary estimated value of ‘J(‘sN-‘3C) (about 10 Hz) is consistent with only a fairly small value of e4 if any. We feel that the best way in assessing this problem is to wait for final estimation from the n.m.r. experiment. (c) Comparison of free SSI and SSI in the present complex Crystal structure of free SSI in a trigonal crystal was initially determined at 2,6 A resolution by

isomorphous replacement methods (Mitsui et al., 19796). Since then the native data were recollected using a four-circle diffractometer mounted on a Rigaku RU-200 rotating-anode generator operated at 50 kV and 200 mA. The structure was refined against the new data (Y. Takeuchi, K. Ishikawa & Y. Mitsui. unpublished results) using a restrained crystallographic refinement program PROLSQ (Konnert & Hendrickson. 1980: Hendrickson. 1985) and simulated annealing program XPLOR (Briinger et al.. 1987). The final &value was 0.194 for 4998 reflections in the 6 to 2.05 A sphere with the rootmean-square deviation from ideal bond lengths being 0.019 A. A global least-squares superposition of the a-carbon chains of free and complexed SSI using the locally developed program LSFIT (C-rata et al., 1981) gave an r.m.s. deviation of 0.63 A. Nine c’” atoms differed in position by more than I.0 A: residues 7 to 8, 36, 65 to 66, 70, 107 and 112 to 113. In both free and complexed SSIs the N-terminal six residues are not clearly visible in the electron density map. most probably due to proteolytic degradation in t’he majority of the molecules (as detected by the Edman degradation analysis: Ikenaka et al., 1974). Thus the residues Tyr7 to Ala8 are the actual terminal residues and liable to conformation change. Ala36 is in t,he loop between a,-strand and /I,-strand (see Fig. 3) and located on the surface of the inhibitor. Arg65 is the P, residue in the reactive site loop and in direct contact with the enzyme surface (see Table 3). Met70 is the P, residue with its main-chain firmly accommodated in the S, pocket of subtilisin BPN’ (see Table 3 and Fig. 4). Gly107 is in no direct contact with the

-

A ,.‘:;:~.‘.::.::::!LALAL2Z

?:::‘:‘:::;::::‘:,‘:~.~

:::*z ’ I: ::

‘cd

+i

m

t

-

4

6 -

5: 10+3rl,-_8

c

318

Y. Takeuchi et al.

enzyme but it has the highest B-factor (251 A2 as the average over all main-chain atoms) in free SSI. Phel13 is the C-terminal residue of SSI. The B-factor versus residue number curve for the SSI subunit in the SSI-subtilisin complex is shown in Figure 7 for main-chain and side-chain atoms, respectively. The average B-values are 38 A2 for both main-chain and side-chain atoms. These values are higher than the corresponding values for the molecule of subtilisin BPN’ $iven in section (d), below) by approximately 24 A . This is a reflection of generally much weaker electron density found in the SSI region of the present complex. The latter fact, in turn, is understandable considering that the intermolecular contacts in the present crystal structure are mainly found between the symmetryrelated molecules of subtilisin BPN’. Lower B-values for the B1- and /I,-strands are in agreement with the general tendency in many protein structures. In free SSI (Y. Takeuchi, K. Ishikawa & Y. Mitsui, unpublished results), the same tendency is observed for the p4- and b,-strands as well. In the complexed SSI, however, the latter features are obscured due to large change in B-factor upon complex formation. In the AB-factor curves shown in Figure 7(c) and (d) (for main-chain and side-chain atoms, respectively), the base line is not necessarily meaningful since it is the difference between the two B-factors, which are derived from completely different crystal forms. It should be recalled that the B-factors reflect not only the mobility of the atoms but also the crystal lattice disorder, which can be largely dependent on the particular crystal form. Most probably it will be reasonable to assume that the AB value (about 12 A2) at the core of the PI-strand (around residue number 17) is the realistic base line, since the position is one of the hardest core of the SSI subunit and is sufficiently remote from the contact region with the enzyme. Examining Figure 7(c) and (d), it is clear that the most marked decrease in B-factors (or the rigidification of the molecule) upon complex formation occur in the following regions: (1) the primary contact region (P,(Glu67) through P,‘(Tyr75)) involving the scissile peptide bond between Met73 and Va174; (2) the secondary contact region (Q,(Ser98) through Qi(CyslOl), see Fig. 3 for the Q-notation); and (3) the polypeptide chain segments flanking the region (1) and (2) above, especially the former half of the P,-strand, the latter half of the P,-strand and the cr,-helix. The above observation shows that the rigidification of SSI upon complex formation with the target enzyme occur highly locally: the largest degree of rigidification occurs in the region directly in contact with the target enzyme and lower degrees of rigidification occur in the regions flanking (along the polypeptide chain) the above direct contact region in decreasing degrees as they deviate from the direct contact region. As described in section (d), below, a similar effect was also observed for subtilisin BPN’ molecules and it seems that the rigidilication of protein molecules upon complex formation with their ligands generally occur in this

way. Various spectroscopic observations made for protein-protein interaction in solution also seem to support the above general notion. For example, for the system of free and trypsin-complexed BPTI, Pershina & Hvidt (1974) observed that the number of (peptide) NH protons with slowed exchange rates in the complexed inhibitors exceeds the number of amino acid residues at the active site. For the same system, Simon et al. (1984), based on the ‘H n.m.r. spectrum, observed that trypsin binding has a highly localized effect on the NH exchange rate of BPTT. Thus the exchange rate for NH of Tyr35 is slowed by a factor of > lo3 in the complex, while the other NH values measured (Glu31, Phe33 and Phe45) are slowed by a factor of only 3 to 15. Tyr35 is the Q4 residue, which, unlike the other residues mentioned above, is close to the Qi’(Tyr39) through Q4’(Cys42) chain segment (the secondary contact region), which is in direct contact with trypsin (see Table 81(b) of Mitsui (1985), which is based on t,he crystal structure of Riihlmann et al., 1973). For the same SSI-subtilisin BPN’ system as used in the present crystal structure analysis, Kainosho & Miyake (1988) observed the change in 13C n.m.r. chemical shift upon complex formation and found that the amino acid residue for which the carbonyl i3C chemical shift undergoes significant carbon (larger than 605 p.p.m.) change is highly localized. In the bottom strip of Figure 7(d), the positions of such residues are marked. As seen. their distribution is almost perfectly coincident with the regions (1). (2) and (3) mentioned above where marked decrease in B-factors were observed upon complex formation. Thus the crystallographic observation and spectroobservation made on solution exhibit scopic striking similarity. An apparent paradox is t’hat. while the change in crystallographic B-factors obviously reflects the change in mobility of the atoms, the change in 13C chemical shift is believed to reflect mainly the change in electronic environ ment’ of the 13C nucleus. This point will be further investigated in co-operation with Dr Kainosho and his colleagues.

(d) Comparison of free sub&sin BPN’ and subtilisin BPN’ in the present complex Crystal structure of free subtilisin BPN’ has been refined at 1.8 A resolution to an R-factor of 0.14 by Bott et al. (1988). A global least-squares superposition of the a-carbon chains of free and complexed subtilisin BPN’ using the program LSFIT (Urata et aZ., 1981) gave an r.m.s. deviation of 660 A (Fig. 8). Nine C” atoms differ in position by more than 2.0 A: residues 3 to 4, 47 to 51 and 92. Residues 3 to 4 are N-terminal on the surface of subtilisin BPN’. Residues 47 to 51 are on the /I,-strand, the outermost strand of the P-barrel (McPhalen & James, 1988). The jl,-strand appreciably moves inward upon complex formation. Tn the complex crystal, the carbonyl carbon of Gly47 makes a hydrogen bond of 3.0 A with Gln271 0” in a

ReJined Structure of SSI-Subtilisin

Cmplex

319

Figure 8. Superposition of the a-carbon chains of free subtilisin BPN’ (Bott et al., 1988) in thin bonds and the same enzyme in the SSI-subtilisin complex (present study) in bold bonds.

symmetry-related molecule. Thus the inward moving is most probably due to crystal contacts to a symmetry-related molecule in this complexed subtilisin. Ala92 is on the j,-strand, which is located next to the outermost /?,-strand. The &-strand also moves inward upon complex formation. This moving is most probably coupled with the &strand movement mentioned above. The B-factor versus residue number curves for subtilisin BPN’ in the SSI-subtilisin complex is shown in Figure 9(a) and (b) for main-chain and The average side-chain atoms, respectively. B-values are 13 A2 and 14 A2 for main-chain and side-chain atoms, respectively. As expected, the B-factors are generally small for the central core consisting of seven parallel B-strands (8, through 8, strands, marked in the lower strip of Fig. 9(a)) and generally higher for the nine external a-helices (helix A through I, marked in the upper strip of Fig. 9(a)) except for helix F, which is almost buried (the nomenclature for the secondary structural elements is after McPhalen & James, 1988). In the AB-factor curves shown in Figure 9(c) and (d), we see that the dramatic decrease in B-factors (or the rigidification of the molecule) occurs only in the S,., (residues 125 through 127) and the S,., site (residues 102 through 104), which form two antiparallel B-sheets with the P,-P, residues of the inhibitor (see section (a), above). This is another example that the rigidification of protein molecules upon complex formation with ligands is highly localized. Note that the S1_3 site and the flanking region (residues 124 through 130) are the barrier between the S, and S, binding pockets of subtilisin. Thus the rigidification here is due to three factors: (1) interaction of the P, residue with the S1 pocket, (2) interaction of the P, residue with the S4 pocket, and (3) hydrogen-bonded interaction between the

main-chain of P,-P, residues of the inhibitor and that of S,-S, residues of the enzyme. As for the catalytic triad, we see that the sidechains of Ser221 and His64, but not that of Asp32, are markedly rigidified. Whether or not this kind of rigidification occurs when the ligand is a true substrate rather than an inhibitor, needs careful consideration (see section (f), below). The decrease in B-factor of the side-chain atoms of Asn240 is most probably due to crystal contacts to a symmetryrelated molecule. The detailed description of the nature of P,-S, and P,-S, interaction, especially the induced-fit movement of Tyr104 comprising the S4 pocket will be given in a separate paper (Y. Takeuchi, et al., unpublished results). (e) Solvent structure The final refined model includes 297 solvent molecules; 295 of these have been refined as water molecules and the remaining two as Ca2+ ions. During the course of the structure refinement, two solvent sites in subtilisin BPN’ refined rapidly to occupancies of 1.0 and low B-factors when their scattering contributions were included as water oxygen atoms. Subsequently, these two ion sites were refined as Ca2+ ions with no restraints on the Ca’+-to-oxygen ligand distances. The final B-factors of these two ions are 7 A2 and 22 A2, respectively. Subtilisin BPN’ is known to be stabilized by Ca2+ (Matsubara et al., 1958). Many other extracellular proteinases are stabilized by Ca2+ ions as a protection against autolytic degradation (Kretsinger, 1976; Bode & Schwager, 1975). Two calcium ions found in the present subtilisin BPN’ molecule were presumably copurified with the enzyme, since no calcium ion was added to the crystal growth solutions. The ligand to Ca2+ ion distances and the co-

40

40

60

60

80

8”

120

I40

140

180 200

160 180 200

I$0

Residue number

I60

100 Ii0

240

260

220 240 260

220

-lO20

20

40

40

60

I

60

80

80

140

120

.I 140

lS’-3

120

160

160

Residue number

100

I.1

100

180 200

180 200

220

220

240

240

260

260

Figure 9. H-factor versus residue number curves for subtilisin in the SST-subtilisin complex ((a) and (b)) and A&factor (= B(complexed subtilisin) - &free subtilisin)) versus residue number curves ((c) and (d)). The data for free subtilisin were kindly provided by T)r Et. Hott (Genentech). In (a) and (c), the B-factors were averaged over all main-chain atoms, while in (b) and (d), they were averaged over all side-chain atoms. Tn the bottom of (a), the position of the secondary structures are marked as 2 strips, the upper one showing the cr-helices A through I and the lower one showing the b-strands 1 through 7 (the nomenclature for the secondary structures is after McPhalen & James, 1988). The strips in the bottom of (c) and (d) indicate all the amino acid residues participating in close (less than 4.0 A) contacts with 881 (Table 3). The S,., site and the S,., site are labeled with the active Ser221 indicated by an arrow.

20

(b)

20

ReJined Structure of SSI-Subtilisin ordination angles are given in Table 5 and stereoviews of some of the ion-binding sites are shown in Figure 10. The low B-factor and the mean ion-tooxygen ligand distance of 2.3 A strongly suggests that the ion site 1 is occupied by a Ca2+ ion. Site 2 has a relatively high B-factor and a mean ligand distance of 2.9 A. The co-ordination geometry matches that found at Ca2+ binding sites in several subtilisin structures (Drenth et al., 1972; McPhalen & James, 1988; Bott et al., 1988). (f) Why is the protein SSI an inhibitor rather than a substrate? Like other proteinaceous inhibitors such as BPTI and STI, SSI is a very strong inhibitor of subtilisin BPN’ with the Ki value of approximately lo-” M (Tonomura et al., 1985). The unusually stable nature of the SSI-subtilisin BPN’ complex enabled the crystallographic and spectroscopic works mentioned above, both of which require very long life-time of the complex. Yet the catalytic geometry as elucidated by the highly refined crystal structure (see Fig. 6) is surprisingly similar to what is expected for the complex of the enzyme with a substrate. It is true that the distance of -2.7 A between the Oy of the active Ser221 and the carbonyl carbon of the scissile peptide bond, which has been repeatedly observed in all the highly refined structure (see Table 4), is definitely longer by - 1.3 A than the potential Oy-C’ covalent bonds having a distance of - 1.4 A. We used to think that such a deviation from the ideal catalytic geometry is enough to spoil the enzymatic function, since the quantum-mechanical processes involving the transfer of electrons are known to be very sensitive to the geometrical parameters (see, for example, Biirgi & Dunitz, 1983). However, the recently accumulated kinetic data for the modified enzyme in which the catalytic residues are replaced with other (homologous) residues through gene manipulation tend to show that the geometrical condition required for the emergence of catalytic activity is not so strict as has been conceived so far, although the modified geometry does affect the catalytic efficiency. For example, in the case of ribonuclease T,, the replacement of a catalytic residue Glu58 with Asp58 which is a -2 A reduction in the length of the side-chain, decreased the k,,,/K,,, value by a factor of 10 but it was still definitely active (Nishikawa et al., 1986). Moreover, molecular dynamics simulation of proteins shows that the r.m.s. fluctuation of atoms in proteins can be as much as 0.4 A for main-chain atoms and 1.5 A for atoms of longer side-chains (Karplus & McCammon, 1983). Then it will be natural to interpret the above site-directed mutagenesis experiments as resulting from the fact that, despite the distortion in the catalytic geometry, ideal geometry was still realized during a fraction of the reaction time as a result of dynamic fluctuation of the catalytic groups. There is also a possibility of an induced-fit rearrangement of the catalytic devices so as to approach the ideal geometry.

Complex

321

Table 5 Ligand-ion distances and co-ordination angles for Ca’+ distance (A)

Site

Atom

la

Gln2 0” Asp41 06’ Asp41 06’ Leu75 0 Am77 Odl Ile79 0 Va181 0 Mean

2.3 23 24 2.3 25 24 22 23

2s

Gly169 0 Tyr171 0 Va1174 0 Glu195 0 Asol 0” w;tt725 Wat726 Mean Atom 1’

Atom 2

2.9 32 2.9 31 30 2.0 30 29 Angle (“)

Gln2 Oel Gln2 0” Gln2 0” Gln2 gC’ Gln2 0” Gln2 0” Asn41 Odl Asp41 Odl Asp41 Odl Asp41 Od’ Asp41 Odl Asp41 06’ Asp41 Odz Asp41 0” Asp41 Odz Leu75 0 Leu75 0 LeLl75 0 Am77 Odl Asn77 Odl Ile79 0 Gly169 0 Gly169 0 Gly169 0 Gly169 0 Gly169 0 Gly169 0 Tyrl71 0 Tyr171 0 Tyr171 0 Tyrl71 0 Tyr171 0 Va1174 0 Va1174 0 Va1174 0 Va1174 0 Glu195 0 Glu195 0 Glu195 0 Asp197 0” Asp197 Odz Wat725

Asp41 Odl Asp41 0” Leu75 0 Asn77 Odl Ile79 0 Va181 0 Asp41 0” Leu75 0 Am77 Odl Ile79 0 Va181 0 Leu75 0 Asn77 Odl Ile79 0 Va181 0 Asn77 Odl Ile79 0 Va181 0 Ills79 0 Va181 0 Va181 0 Tyrl71 0 Va1174 0 Glu195 0 Asp197 Odz Wat725 Wat726 Va1174 0 Glu195 0 Asp197 0” Wat725 Wat726 Glu195 0 Asp197 0” wat725 Wat726 Asp197 0” Wat725 Wat726 Wat725 Wat726 Wat726

Site 1

2

154 154 80 80 90 81 52 86 77 97 121 109 124 87 74 86 160 93 75 161 103 79 99 91 108 156 113 69 157 136 126 91 134 68 106 138 66 66 74 67 102 57

‘The B factor for CaZC is 7 A’. ‘The B factor for Ca2+ is 22 A*. c In each case, the apex of the angle (defined by atoms 1 and 2) is the Ca2+ ion.

In any case, considering the experimental and theoretical facts as mentioned above, it is clear that the mere fact that the Oy-c’ distance is -2.7 A

322

Y. Takeuchi

et al.

(b)

Figure 10. Stereo representation of the environment of the Ca'+ Relevant ligand-to-ion distances and angles are given in Table 5. (a) CaZC binding site 1 in subtilisin. (b) Ca’+ binding site 2 in subtilisin.

cannot be taken as the crucial cause of SSI or other proteinaceous inhibitors being a potent inhibitor. Here we would like to put forward the following working hypothesis: (1) the real cause of a protein being an inhibitor is the absence of such extent of flexibility in the enzyme-inhibitor interface as would be expected in the interface between enzymes and true substrates and (2) such absence of flexibility leads to inability of the system to go through such dynamic fluctuations and induced-fit movements as are required for the catalytic processes. At first sight, the rigidification of the SSI-subtilisin

interface upon complex formation as manifested in the decrease in B-factors (see Fig. 9) may appear to directly support the above working hypothesis. However, such rigidification upon intermolecular interaction is perhaps a general tendency in protein-protein interaction. The real question is whether or not the extent of rigidification realized in the enzyme-inhibitor interface is in fact larger than that realized in the enzyme-substrate interface. It is known that the interaction between the enzyme and a substrate generally occur with the Km values higher than - 10e6 M, which is by far weaker than

Refined Structure of SSI-Subtilisin the interaction between an enzyme and a proteinaceous inhibitor typically characterized by the Ki values around lo-” M. While the relation between the extent of (mechanical) rigidification and the extent of free energy decrease, which is directly or indirectly reflected in the Ki or Km values, is not straightforward, the free energy observation mentioned above points to one pitfall for the experimental attempt at addressing the above working hypothesis by studying the complex with substrates (perhaps more practically poor substrates) in a way similar to the complex with inhibitors: it is extremely difficult to study such a complex as is characterized by a Km value higher than N 10e6 M by means of X-ray or n.m.r. methods since these methodologies require the presence of a stable complex enduring at least for a few hours. Despite the difficulty with direct experimental approach, we feel that there is plenty of circumstantial evidence supporting the above working hypothesis as described in the following items (1) and (2). (1) Amino acid replacements in SSI most likely leading to increased flexibility of SSI (and thus most, probably to increased flexibility of the SSI-subtilisin interface) changes SSI into a “temporary inhibitor”, the inhibitor which is gradually degraded after complex formation and thus reactivates the target enzyme (Laskowski & Sealock, 1971). It is clear that the temporary inhibitor is equivalent to a very slowly degraded substrate. Indeed, some of the well-known natural trypsin inhibitors such as PST1 are known to be intrinsic temporary inhibitors (Laskowski & Sealock, 1971). In the case of SSI, the elimination of the 71-101 disulfide bridge (Fig. 3) by replacing both Cys71 and CyslOl with Ser exhibited such effects (S. Kojima et al., unpublished results). Also, the removal of a strong hydrogen-bonded interaction between the terminal carboxylate at Phell3 and the guanidine group of Arg29 (Mitsui et al., 1979b) by replacing Arg29 with Met caused a similar effect of temporary inhibitor formation. The presence of the above hydrogen-bond seems to be important for correctly positioning the C-terminal polypeptide segment, which is attached to the hydrophobic core of the SSI subunit mainly through the side-chains of Phelll and Phel13. The correct positioning of the C-terminal segment, in turn, seems to be important for maintaining the “secondary contact region”, which is directly in contact with the enzyme surface (see Table 3) and is connected to the primary contact region through the 71-101 disulfide bridge. (2) When the target enzyme is a-chymotrypsin, the scissile peptide bond of SSI is easily cleaved (Omichi et al., 1980; Tonomura et al., 1985). In this case, the relevant Ki value is much larger, around lOA M (Inouye et al., 1975). Thus the protein SSI can be a substrate when the target enzyme is appropriately chosen. Most dramatically, Estell & Laskowski (1980) found that such well-known trypsin inhibitors as BPTI or STI, which are the very strong inhibitors of mammalian pancreatic trypsin, behave as very good substrates when the target

323

Cmplex

enzyme is the trypsin from a species of star fish, Dermasterim imbricata. Again, in this example, the Ki valuk is estimated to be about 3 x lo-’ M. The hypothetical complex of SSI and a-chymotrypsin was constructed using computer graphics by a technique of least-squares superposition described by Hirono et al. (1979) (Y. Mitsui, unpublished results). Considerable amount of interatomic collision was observed between the side-chains of P, through P6 residues of SSI and the enzyme surface. Thus we speculate that the SSI-a-chymotrypsin interface would have much more flexibility than the SSI-subtilisin complex because of the bad complementarity between the surfaces of SSI and a-chymotrypsin. We speculate that a similar situation occurs in the interface of BPTI or ST1 with the starfish trypsin mentioned above. The authors thank Professor W. Hendrickson for advice in the use of PROLSQ and Dr R. Bott of Genentech for providing us with the refined co-ordinates of subtilisin BPN’. Data collection at the synchrotron radiation source, Tsukuba was made possible through generous assistance by Professor N. Sakabe and Dr A. Nakagawa. We also thank Professors M. Laskowski Jr, K. Miura, B. Tonomura and M. Kainosho, Drs S. Kojima and Y. Migake for discussion, Professors D. M. Blow, S. Murao, k. Hiromi, M. Miwa, S. Ishii, K. Hamaguchi and A. Wada for encouragement.

References Blow, D. M., Birktoft, J. J. & Hartley, B. S. (1969). Role of a buried acid group in the mechanism of action of chymotrypsin. Nature (London), 221, 337-340. Bode, W. & Schwager, P. (1975). The refined crystal structure of bovine /?-trypsin at 1% A resolution. II. Crystallographic refinement, calcium binding site, benzamidine binding site and active site at pH 7.0. J. Mol. Biol. 98, 693-717. Bode, W., Papamokos, E., Musil, D., Seemueller, U. & Fritz, H. (1986). Refined 1.2 b crystal structure of the complex formed between subtilisin Carlsberg and the inhibitor Eglin C, molecular structure of Eglin and its detailed interaction with subtilisin. EMBO J. 5, 813-818. Bolognesi, M., Gatti, G., Menegatti, E., Guarneri, M., Marquart, M., Papamokos, E. & Huber, R. (1982). Three-dimensional structure of the complex between pancreatic secretory trypsin inhibitor (Kazal type) and trypsinogen at 1.8 A resolution. J. Mol. Biol. 162, 83%868. Bott, R., Ultsch, M., Hossiakoff, A., Graycar, T., Katz, B. 6 Power, S. (1988). The three-dimensional structure of Bacillus amyloliquefaciens subtilisin at 1.8 A and an analysis of the structural consequences of peroxide inactivation. J. Biol. Chem. 263, 7895-7906, Briinger, A. T., Kuriyan, K. & Karplus, M. (1987). Crystallographic R-factor refinement by molecular dynamics. Science, 235, 458461. Biirgi, H. B. & Dunitz, J. D. (1983). From crystal statistics to chemical dynamics. Ads Chem. Res. 16, 153-161. Chen, Z. & Bode, W. (1983). Refined 2.5 A X-ray crystal structure of the complex formed by porcine kallikrein

Y. Takeuchi et al.

324

A and the bovine pancreatic trypsin inhibitor. J. Mol. Biol. 164, 283-311. Drenth, J., HOI, W. G. J., Jansonius, J. & Koekoek, R. (1972). Subtilisin Novo, the three-dimensional structure and its comparison with subtilisin BPN’. Eur. J. Biochem. 26, 177-181. Estell, D. A. & Laskowski, M., Jr (1980). Dermasterias imbricata trypsin 1: an enzyme which rapidly hydrolyzes the reactive-site peptide bonds of protein trypsin inhibit)ors. Biochemistry, 19, 124-131. Fujinaga, M., Read, R. J., Sielecki, A. R., Ardelt, W., Laskowski, M., Jr t James, M. N. G. (1982). Refined crystal structure of the molecular complex of Streptomyces griseus protease B, a serine protease, with the third domain of the ovomucoid inhibitor from turkey. Proc. Nat. Aead. Sci.. U.S.A. 79, 48684872. Griitter, M. G., Fendrick, G., Huber, R. & Bode, W. (1988). The 2.5 A X-ray crystal structure of the acidstable proteinase inhibitor from human mucous secretions analyzed in its complex with bovine EMBO J. 7, 345-351. u-chymotrypsin. Hendrickson, W. A. (1985). Stereochemically restrained refinement of macromolelcular structures. In Methods in Enzymology (Wyckoff, H. W., Hirs, C. H. W. & Timasheff, S. N., eds), pp. 252-270, Academic Press, New York. Higashi, T. (1989). The processing of diffraction data taken on a screenless Weissenberg camera for macromolecular crystallography. J. Appl. Crystallogr. 22. 9-18. Hiromi, K.. Akasaka, K., Mitsui, Y., Tonomura, B. & Murao. S. (1985). Protein Protease Inhibitor ~ The Case of Streptomyces Subtilisin Inhibitor (SSI) , pp. 1481, Elsevier; Amsterdam. Hirono, S., Nakamura, K. T.. Iitaka, Y. & Mitsui, Y. (1979). Crystal structure of the complex of subtilisin subtiBPN’ with its protein inhibitor Streptomyces lisin inhibitor: the structure at 43 A resolution. J. Mol. Biol. 131, 855-869. Hirono, S., Akagawa, H., Mitsui, Y. & Iitaka, Y. (1984). Crystal structure at 2.6 A resolution of the complex of subtilisin BPN’ with Streptomyces subtilisin inhibitor (SSI). J. Mol. Biol. 178, 38!%413. Huber, R. & Bode, W. (1978). Structural basis of the activation and action of trypsin. Accts Chem. Res. 11, 114-122. Huber, R., Kukla, D., Bode, W.. Schwager, P., Bartels, K., Deisenhofer, J. & Steigemann, W. (1974). Structure of the complex formed by bovine trypsin and bovine pancreatic trypsin inhibitor. II. Crystallographic refinement at l-9 A resolution. J. Mol. Biol. 89, 73-101. Hunkapiller, M. W., Forgac, M. D., Yu, E. H. & Richards, J. H. (1979). 13C NMR studies of the binding of soybean trypsin inhibitor to trypsin. Biochem. Biophys. Res. Commun. 87, 25-31. Ikenaka, T., Odani, S., Sakai. M., Nabeshima, Y., Sato, S. & Murao, S. (1974). Amino acid sequence of an alkaline proteinase inhibitor (Streptomyces subtilisin inhibitor) from Streptomyces albogriseolus S-3253. J. Biochem. 76, 1191-1209. Inouye, K., Tonomura, B. & Hiromi, K. (1975). Interactions between Streptomyces subtilisin inhibitor (SSI) and cc-chymotrypsin. Agric. Biol. Chem. 39, 1154-1161.

Jones, T. A. (1978). A graphics model building for macromolecules. ment system Crystallogr. 11, 268-272.

and refineJ. Appl.

Kainosho, M. & Miyake, Y. (1988). Structural study of SSI-subtilisin complexes by means of stable-isotopeaided NMR spectroscopy. Nogeikagaku, Kaishi, 62. 1822-1827. (in Japanese) Kainosho, M., Tsuji, T., Akagawa, H. & Mitsui. Y. (1982). Structure and function of serine proteinases as inhibitors. viewed their proteinaceous from Tanpakushitsu Kakusan Koso, 27, 12&133. (in Japanese) Karplus, M. & McCammon. ,I. A. (1983). Dynamics of proteins: elements and function. Annu. Rev. Biochpnt. 53. 263-300. Konnert, ,J. H. $ Hendrickson, W. A. (1980). A restrained-parameter thermal-factor refinement proredure. Acta Crystallogr. Sect. A, 36, 344-350. Kretsinger, R. H. (1976). Calcium-binding prot,eins. Annu. Rev. Biochem. 45. 239-266. Laskowski, M., Jr & Kato, T. (1980). Protein inhibitors of prot,einases. Annu. Rev. Biochem. 149, 593426. Laskowski, M.. *Jr & Sealock, R. W. (1971). Protein proteinase inhibitors --molecular aspects. In 7% Enzymes (Bayer. P. D., ed.), 3rd edit. vol. 3, pp. 375 474, Academic Press, New York. Marquart, M., Walter, J., Deisenhofer. J., Bode, \V. B Huber. R. (1983). The geometry of the reactive site and of the peptide groups in trypsin. trypsinogen and its complexes with inhibitors. Acta, Crystallogr. .sect. R, 39. 48&490. Matsubara, H., Hagihara, B.. Nakai, M.. Komaki, T.. Yonetani, T. & Okunuki. K. (1958). Crystalline bacaterial proteinase II. General properties of crystalline proteinase of Bacillus aubtilis N’. .I. Biochewc. 45. 251~255. McPhalen. (1. A. & tJames. M. S. G. (1988). Structural comparison of two serine proteinase-protein inhibitor complexes: Eglin-Gsubtilisin Carlsberg and CI-2-subtilisin Novo. Biochemi&ry, 27, 6582-6598. McPhalen. c’. A.. Schnebli, H. P. & James. M. N. (4. (1985a). Crystal and molecular structure of the inhibitor Eglin from leeches in complex with subtilisin Carlsberg. FEBS Letters, 188, 55-58. McPhalen. (‘. A., Svendsen, T.. Jonassen. I. & James. M. N’. G. (19856). Crystal and molecular structure of chymotrypsin inhibitor 2 from barley seeds irl complex with subtilisin Novo. Proc. Yat. Acad. IS’&.. G.S.A. 82. 7242-7246. Mitsui, Y. (1985). Three-dimensional structure of the complex of SST with subtilisin BPN’. In Protein l’rotease Inhibitor The Case of Streptomyces Subtilisin Inhibitor (Hiromi, K., Akasaka. K.. Mitsui. Y., Tonomura, B. & Murao. S.. eds). chapt. 8. pp. 265.-290, Elsevier. Amsterdam. Mitsui, Y.. Akagawa, H.. Onishi. H., Itoh. S. B Nakamura, K. T. (1985). Crystal structure of subtilisin complexed with its trapped substrate Streptomyces subtilisin inhibitor - Protein--protein interaction and evolution of serine proteinases and their proteinaceous inhibitors. J. Hiosci. (India). 8. 48lm-489. Mitsui. Y., Satow. I’. & Sakamaki, T. (1977). Crystal structure of a protein protease inhibitor, Streptomyces subtilisin inhibitor, at 2.3 A resolution. J. Biochem. (Tokyo), 82, 295-298. Mitsui, Y., Satow, Y., Watanabe, Y., Hirono, S. & Iitaka. Y. (1979a). Crystal structures of Streptomyces subtilisin inhibitor and its complex with subtilisin BPN’. Natwe (London), 277, 447V452. Mitsui, Y., Satow. Y.. Watanabe, Y. & Iitaka, Y. (19796). C:rystal structure of a bacterial proteinase inhibitor

Refined Structure of SSI-Subtilisin

(Streptomyces subtilisin inhibitor) at 2.6 A resolution. J. Mol. Biol. 131, 697-724. Miyahara, J., Takahashi, K., Amemiya, Y., Kamiya, N. & Satow, Y. (1986). A new type of X-ray area detector utilizing laser stimulated luminescence. NucE. Instrum. Methods, A246, 572-578. Murao, S. & Sato, S. (1972). S-SI, a new alkaline protease inhibitor from Streptomyces albogriseolus S-3253. Agric. Biol Chem. 36, 166163. Nishikawa, S., Morioka, H., Fuchimura, K., Tanaka, T., Uesugi, S., Ohtsuka, E. & Ikehara, M. (1986). Modification of Glu58, an amino acid of the active center of ribonuclease T, to Gln and Asp. Biochem. Biophys. Res. Commun. 138, 789794. Omichi, K., Nagura, N. & Ikenaka, T. (1980). On the reactive site of Streptomyces subtilisin inhibitor. J. Biochem. (Tokyo), 87, 483489. Pershina, L. & Hvidt, A. (1974). A study by the hydrogen-exchange method of the complex formed between the basic pancreatic trypsin inhibitor and trypsin. Eur. J. Biochem. 48, 339344. Read, R. J. t James, M. N. G. (1986). Introduction to the protein inhibitors: X-ray crystallography. In Proteinase Inhibitors (Barrett, A. J. & Salvesen, G., eds), pp. 301-336, Elsevier, Amsterdam. Read, R. J.. Fujinaga, M., Sielecki, A. R. & James, M. N. G. (1983). Structure of the complex of Streptomyces griseus protease B and the third domain of the turkey ovomucoid inhibitor. Biochemistry, 22, 44204433. Rossmann, M. G. (1979). Processing oscillation diffraction data for very large unit cells with an automatic convolution technique and profile fitting. J. Appt. Crystallogr. 12, 225238. Riihlmann, A., Kukla, D., Schwager, P., Bartels, K. &

Complex

325

Huber, R. (1973). Structure of the complex formed by bovine trypsin and bovine pancreatic trypsin inhibitor. J. Mol. Biol. 77, 417436. Sakabe, N. (1983). A focusing Weissenberg camera with multi-layer-line screens for macromolecular crystallography. J. Appl. Crystallogr. 16, 542-545. Schechter, I. t Berger, A. (1967). On the size of the active site in proteases. I. Papain. Biochem. Biophys. Res. Commun. 27, 157-162. Simon, I., Tiichsen, E. & Woodward, C. (1984). Effect of trypsin binding on the hydrogen exchange kinetics of bovine pancreatic trypsin inhibitor p-sheet NH’s Biochemistry, 23, 2064-2068. Tonomura, B., Senda, M., Tsuru, D., Komiyama, T., Miwa, M., Akasaka, K., Kainosho, M., Tsuji, T., Takahashi, K., Hiromi, K., Ikenaka, T. & Murao, S. (1985). Interaction of SSI and proteases in solution. In Protein Protease Inhibitor - The Case of Streptomyces Subtilisin Inhibitor (Hiromi, K., Akasaka, K., Mitsui, Y., Tonomura, B. & Murao, S., eds), chapt. 9, pp. 291-362, Elsevier, Amsterdam. Tsunogae, Y., Tanaka, I., Yamane, T., Kikukawa, J., Ashida, T., Ishikawa, C., Watanabe, K., Nakamura, S. & Takahashi, K. (1986). Structure of the trypsinbinding domain of Bowman-Birk type protease inhibitor and its interaction with trypsin. J. Biochem. 106, 1637-1646. Urata, Y., Mitsui, Y. t Nakamura, K. T. (1981). A leastsquares minimization procedure to attain optimum spatial fit between homologous (molecular) structures and its application to studying protein structural homology. Chem. Pharm. Bull. 29, 2737-2742. Wright, C. S., Alden, R. A. & Kraut, J. (1969). Structure of subtilisin BPN’ at 2.5 A resolution. Nature (London), 221, 235242.

Edited by B. W. Matthews