Inhibited Bovine Trypsin, and Bovine Trypsinogen : the Refined Structure and Mechanism of a Not-Too-Serene Protease Citation Chambers, John L. (1977) Inhibited Bovine Trypsin, and Bovine Trypsinogen : the Refined Structure and Mechanism of a Not-Too-Serene Protease. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/69pr-0q16. https://resolver.caltech.edu/CaltechTHESIS:12022025-193957018 Abstract Refined three-dimensional structures of diisopropylphosphoryl (DIP)-inhibited bovine trypsin and of bovine trypsinogen are described, as determined by x-ray crystallography. A constrained difference Fourier method for rapid, economical protein structure refinement was developed and applied to these structures. The refinement system, which has been written to operate on a minicomputer, has reduced the standard crystallographic R-factor for DIP-trypsin from 47.2% at 2.7 Å resolution to 21.5% at 1.5 Å resolution, making it one of the most highly and inexpensively refined protein structures to date. For trypsinogen, R is 31% at 1.9 Å resolution, and refinement of both structures is nearing completion. The refined structures, the mechanism of serine proteases, and the mechanism of activation of their zymogens are discussed. During data collection for the studies described, new, more efficient techniques of correction for background radiation in protein data measured by x-ray diffractometry were developed. Also presented are the structures of trypsin inhibited by the metal ions Ag + and Cu2 + . These ions, which are powerful trypsin and chymotrypsin inhibitors, bind tightly and specifically at the active site of the enzyme, between Asp 102 O δ1 and His 57 N δ1 . They may thus be used as specific probes of the active site of serine proteases. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: (Chemistry) Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemistry Thesis Availability: Public (worldwide access) Research Advisor(s): Stroud, Robert M. Thesis Committee: Unknown, Unknown Defense Date: 4 May 1977 Funders: Funding Agency Grant Number National Institutes of Health (NIH) UNSPECIFIED Record Number: CaltechTHESIS:12022025-193957018 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:12022025-193957018 DOI: 10.7907/69pr-0q16 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 17780 Collection: CaltechTHESIS Deposited By: Benjamin Perez Deposited On: 02 Dec 2025 22:04 Last Modified: 02 Dec 2025 22:05 Thesis Files PDF - Final Version See Usage Policy. 59MB Repository Staff Only: item control page

INHIBITED BOVINE TRYPSIN, AND BOVINE TRYPSINOGEN: 1HE REFINED STRUCTURE AND MEo--IANISM OF A NOT-TOO-SERENE PRaTEASE Thesis by John L. Chambers In Partial Fulfillment of the Requirements for the ~gree of Doctor of Philosophy California Institute of Technology Pasadena, California 1977 (Submitted 4 May 1977) ii TO MY GRANDFATIIER iii PREFACE This thesis is in large part an introduction to and summary of the work described in detail in the appendices . The central theme of this work has been the elucidation of the detailed, high-resolution three-dimensional structures of inhibited derivatives of bovine trypsin and of bovine trypsinogen , and the implications of these studies for protease mechanism, and for structure and function of enzymes in general. Chapter I is a historical sketch concerning protease structure and mechanism, and introduces the present work on diisopropylphos• phory1 (DIP) - , Ag + - , and eu2+ -1·nh·b· 1 1ted trypsins. This work is described more fully in Appendices 1 and 6, and in Chapter IV. Chapter II and Appendices 4 and 5 deal with trypsinogen and the structural basis for the inactivity of this precursor of trypsin, as well as its mechanism of activation. In order for these structures to be interpreted reliably, extension of their resolution and structure refinement were necessary . A major portion of this thesis work has thus been development of techniques for routine, inexpensive refinement of macromolecular structures . These techniques and their application in the case of DIP- trypsin are presented in Chapter III and Appendix 6. The reliability of even the most highly refined structure is, however, affected by the accuracy of the data. Improved methods of correcting protein data for background radiation, with a corresponding improvement in iv accuracy, were developed in the course of this thesis work, and are presented in Appendix 2. Portions of the DIP-trypsin structure at its current level of refinement (R = 21.5% at 1.5 A resolution) are shown in Appendix 7, and protease mechanism is discussed in detail in Appendix 3. V ACKNOWLEDGMENTS I would like to thank my fellow graduate students and colleagues, Tony Kossiakoff, Monty Krieger, Steve Spencer, Mike Ross, Roger Koeppe, Dave Agard, Mel Jones, Gary Christoph and Steve Skedzeleski, in addition to those identified in the appendices, for their help and suggestions during the course of this work. I am grateful for the assistance of Dee Barr in typing the manuscripts of this thesis and of the papers appearing in the appendices (these latter ones several times each, as I remember). Her patience and advice during my stay at Caltech have been very helpful. My deep appreciation goes to C.L. Borders, Jr., S.E. Anderson, L.W. Haynes, T.R. Williams, R.H. Bromlllld, D.L. Powell, W.F. Kieffer, J.W. Chittum and J.D. Reinheimer. Their instruction and encourage- ment during my lllldergraduate years have proven invaluable. I thank the National Institutes of Health for financial support during my graduate career. My parents, grandparents, and the other members of my family gave me their support and encouragement throughout this work, for which I am extremely grateful. Special thanks go to my brother, Scott, for staying in Ohio during the last five years. Most of all I would like to thank my advisor, Robert Stroud, who taught me to ask whether it is better "to suffer the slings and arrows of outrageous torsions, or by opposing bend them?" His vi guidance has been most valuable, and working in his laboratory has been a llllique experience as well as a very rewarding one. vii ABSTRACT Refined three-dimensional structures of diisopropylphosphoryl (DIP)-inhibited bovine trypsin and of bovine trypsinogen are described, as detennined by x-ray crystallography. A constrained difference Fourier method for rapid, economical protein structure refinement was developed and applied to these structures . The refinement system, which has been written to operate on a minicomputer, has reduced the standard crystallographic __R-factor for DIP-trypsin from 0 0 47.2% at 2.7 A resolution to 21.5% at 1.5 A resolution, making it one of the most highly and inexpensively refined protein structures to 0 date. For trypsinogen, R is 31% at 1.9 A resolution, and refinement of both structures is nearing completion. The refined structures, the mechanism of serine proteases, and the mechanism of activation of their zymogens are discussed . During data collection for the studies described, new, more efficient techniques of correction for background radiation in protein data measured by x-ray diffractometry were developed. Also presented are the structures of trypsin inhibited by the metal ions Ag+ and eu2+. These ions, which are powerful trypsin and chymotrypsin inhibitors, bind tightly and specifically at the active site of the enzyme, between Asp 102 0 01 and His 57 N01 . They may thus be used as specific probes of the active site of serine proteases. viii TABLE OF CONTENTS Dedication ii Preface iii Acknowledgments V .Abstract . . . . vii Table of Contents vi ii Chapter I: Chapter II: Chapter III: Chapter IV: Serine Protease Structure and Mechanism: Introduction and Backgrotmd . . . . . . . 1 The Structure and Activation Mechanism of Bovine Trypsinogen . . . . . . . . . . . . 10 Protein Structure Refinement: The Refined High-Resolution Structures of DIP-Trypsin and Trypsinogen . . . . . . . . . . . . . 13 A Crystallographic Study of Trypsin Inhibition by eu2+ Appendix 1: . . . . . . . . . . . . . . . . . . . . 20 Silver Ion Inhibition of Serine Proteases : Crystallographic Study of Silver-Trypsin . 27 Data Collection in Protein Crystallography: Capillary Effects and Backgrotmd Corrections 33 Structure-Ftmction Relationships in the Serine Proteases . . . . . . . . . 43 Structure of Bovine Trypsinogen at 1. 9 A Resolution . . . . . . . . . 64 Appendix 5: Mechanisms of Zymogen Activation. 76 Appendix 6: Difference Fourier Refinement of the Structure of DIP-Trypsin at 1.5 A With a Minicomputer Technique . . . . . . . . . . . . . . . . . 118 One Picture is Worth a Thousand Words: A Trypsin Pictorial Assay . . . . . . 172 Appendix 2: Appendix 3: Appendix 4: Appendix 7: 1 Chapter I Serine Protease Structure and Mechanism: Introduction and Background 2 The wide occurrence of the family of peptide-bond splitting enzymes, the serine proteases, along with their participati on in a variety of important physiological pr ocesses , has placed them among the most extensively studied proteins . The digestive enzymes ~ trypsin, chymotrypsin, and elastase , have been especially well studied. Other members of the serine protease family include thrombin, which is important in blood clotting (l); enzymes of the complement system (2); acrosomal protease, important in fertilization (3); and plasmin , which is involved i~ dissolution of blood clots (4) . Increased plasmin activity has also been shown to be responsible for a ntnnber of morphological changes in the transfonnation of cancer cells (5,6,7). Study of the serine proteases has resulted in not only a better understanding of these physiological processes, but also development of general principles of enzyme catalysis and specificity . The reaction catalyzed by these enzymes is schematized below : + where E, S, P1 and P2 are respectively free enzyme , substrate, the amine (or alcohol in ester hydrolysis) product, and carboxylic acid product. ES and EP 2 are Michaelis complexes, EST and EPI are tetra- hedral intermediates, and EA is the acyl-enzyme intermediate (8-14). 3. These enzymes work optimally near pH 8 (15). While the kinetics of peptide or ester hydrolysis are very similar for different members of the family, each member exhibits its own substrate specificity. For example, trypsin, chymotrypsin, and elastase preferentially hydrolyze peptide bonds to the carboxyl side of residues with long positivelycharged side chains, large hydrophobic side chains, and small side chains, respectively (16). The side chains of residues Ser 195 and His 57 (using the chymotrypsinogen numbering system), conser~ed throughout the serine protease family, were demonstrated to be essential for enzymatic activity by chemical modification (17,18). Solution of the three0 dimensional structure of tosyl-a-chymotrypsin at pH 4.2 and 2 A resolution by Blow and coworkers showed that these two residues were in fact adjacent to one another in the catalytic site, with Ser 195 Oy hydrogen-bonded to His 57 Ni::, and indicated the importance of a third residue, Asp 102, whose carboxyl group was hydrogen bonded to the other side of the His 57 imidazole ring (19,20). In 1966, Oppenheimer et al. (9) had shown the importance of the amino terminus in the reactivity of chymotrypsin. Matthews et al. (19) noted that the amino terminus formed a salt bridge to Asp 194, stabilizing the conformation of the protein around the active-site serine. Henderson (21) and Robertus et al. (22) later suggested the importance of hydrogen bonds from the N--H groups of Gly 193 and Ser 195 to the substrate carbonyl oxygen in stabilization of reaction intermediates. 4. Subsequently the structure of diisopropylphosphoryl (DIP)-trypsin 0 at pH 7 was solved at 2.7 A resolution by Stroud, Kay, and Dickerson (23). Diisopropylfluorophosphate is a very powerful general inhibitor of serine proteases. The DIP group becomes covalently attached to the active-site serine, and probably resembles a tetrahedral intermediate in the proteolytic reaction (24). In spite of a sequence identity between trypsin and chymotrypsin in only 40% of the residues, their overall three-dimensional folding was shown to be remarkably similar (23,24). These authors also noted that the carboxyl group of Asp 189 (serine in chymotrypsin) is in an ideal position to form a hydrogen bond to the positively charged side chain of a specific substrate . Krieger, Kay and Stroud showed that this interaction does indeed give trypsin its specificity, when they solved the structure of trypsin reversibly inhibited by the specific side chain analog, benzamidine (25). 0 Elastase at pH 5, solved at 3.5 A resolution by Shotton and Watson (26), is also remarkably similar in structure to trypsin and chymotrypsin. Residue 216, which is Gly in these latter two enzymes , is valine in elastase. This valine side chain blocks access to the lower portion of the specific binding pocket, thus giving elastase its specificity for small side chains. The pH-rate profiles for these enzymes were shown in early experiments to depend upon deprotonation of a group with a pKa between 6 and 7 (27,28). Since this is the normal range for ionization of a histidine imidazole, His 57 was generally asst.nned responsible for this 5. limb of the profile (11,1.8,29,30). However, in the mechanism sug- gested by Blow and coworkers Asp 102, in its tmusual environment , acts as a base rather than simply to stabilize a positive charge on the His 57 imidazole, or to maintain structure, and accepts a proton f r om the imidazole while the imidazole in turn receives a proton from Ser 195 Oy (31). The explanation of the pH-rate profile given such a mechanism and the assumed normal pKa ' s of Asp 102 and His 57, however, was then tmclear . In 1973, Htmkapiller et al. (32), from 13 C nrnr studies of a-lytic protease, presented strong evidence that His 57 remained neutral throughout the pH range 4 - 9, and suggested that the pKa of 6.8 observed in the rate profile arose from Asp 102. Koeppe and Stroud, using difference infrared titration, were able to assign directly a pKa of 6.8 to the carboxyl group of Asp 102 (33). One of the studies presented in this thesis is the crystallographic determination of the inhibitory site for Ag+ and eu2+ in trypsin. These ions are strong inhibitors of trypsin and chymotrypsin (KI ~10- 4 M), which impair catalysis without hindering substrate binding (34,35) . Using a difference Fourier technique, we found that the major binding site for these ions is between Asp 102 Oo 1 and His 57 No 1 , with approximately linear coordination of the metal ion . The resulting disruption of the charge-relay system greatly diminishes the activity of the enzyme. Not only did an understanding of the inhibi- tion mechanism result from this study, but also the idea of using Ag+ and eu 2+ as highly specific reversible probes of the catalytic site of serine proteases. These ions were subsequently used by Koeppe and 6. Stroud to assign the peaks corresponding to titration of Asp 102 in the difference infrared spectra (33). The Ag+-trypsin study is presented in more detail in Appendix 1 of this thesis. The cu 2+-trypsin project is reported in Chapter IV. During the collection of data from silver trypsin, we developed new techniques for accurate approximation of background measurements. Use of these methods greatly reduces the time spent measuring backgrotn1ds, increasing the useful x-ray exposure life of the crystal, and is particularly advantageous in protein crystallography, where the crystals often decay rapidly in the x-ray beam. The methods are sub- stantial improvements over earlier techniques used for proteins which assumed that the backgrotn1d depends only on scattering angle. This work is reported in Appendix 2. The mechanism of serine proteases at our current level of tn1derstanding is described in .Appendix 3. References 1. Magnusson, S. 321. In The Enzymes (P.D. Boyer, Ed.), Vol. 3, pp. 277- Academic Press, New York, 1971. 886 pages. 2. Muller-Eberhard, H.J. (1969), Ann. Rev. Biochem. 38, 389-414. 3. Stambaugh, R., Brackett, B., and Mastroianni, L. (1969), Biol. Reprod. !_, 223-227. 4. Goodnight, S.H. In Cold Spring Harbor Symposium on Proteases and Biological Control (E. Reich, D.B. Rifkin, and E. Shaw, Eds.), 7. pp. 191-197. Cold Spring Harbor Laboratory, New York, 1975 . 1021 pages. 5. Quigley, J.P. Ossowski, L., and Reich, E. (1974), J. Biol. Chem. 249, 4306-4311. 6. Quigley, J.P., Ossowski, L., and Reich, E. (1974), J. Biol . Chem. 249, 4312-4320. 7. Tobler, J., Krieger, M., and Stroud, R.M., unpublished results . 8. Zerner, B., Bond, R.P., and Bender, M.L. (1964), J . Amer. Chem. Soc. 86, 3674-3679. 9. Oppenheimer, H.L., Labouesse, B., and Hess, G.P. (1966), J. Biol . Chem. 241, 2720-2730. 10. Caplow, M. (1969), J. Amer. Chem. Soc. 91, 3639-3645. 11. Hess, G.P., Mcconn, J., Ku, E., and Mcconkey, G. (1970), Phil. Trans. Roy. Soc. (Lond.) B257, 89-104. 12. Fersht, A.R. and Requena, Y. (1971), J. Amer. Chem. Soc. 93, 7079-7087. 13. Fastrez, J. and Fersht, A.R. (1973), Biochemistry 12, 2025-2034 . 14. Fersht, A.R. and Renard, M. (1974), Biochemistry 13, 1416-1426. 15. Northrop, J.H. and Kunitz, M. (1932), J. Gen. Physiol. 16, 295311. 16. Stroud, R.M. (1974), Scientific .American ~31, 74-88 . 17. Jansen, E.F., Nutting, M.D.F., and Balls, A.K. (1949), J. Biol. Chem. 179, 201-204. 18. Schoellman, G. and Shaw, E. (1963), Biochemistry~, 252-255. 8. 19. :Matthews, B.W., Sigler, P.B., Henderson, R., and Blow, D.M. (1967), Nature 214, 652-656. 20. Steitz, T.A., Henderson, R., and Blow, D.M. (1969), J. Mol. Biol. 46, 337-348. 21. Henderson, R. (1970), J. Mol. Biol. 54, 341-354. 22. Robertus, J .D., Kraut, J., Alden, R.A., and Birktoft, J .J. (1972), Biochemistry 11, 4293-4303. 23. Stroud, R.M., Kay, L.M., and Dickerson, R.E. (1971), Cold Spring Harbor Sympositnn on Quant. Biol. 36, 125-140. 24. Stroud, R.M., Kay, L.M., and Dickerson, R.E. (1974), J. Mol. Biol. 83, 185-208. 25. Krieger, M., Kay, L.M., and Stroud, R.M. (1974), J. Mol. Biol. 83, 209-230. 26. Shotton, D.M. and Watson, H.C. (1970), Phil. Trans. Roy. Soc. (Lond.) B257, 111-118. 27. Hartley, B.S. and Kilby, B.A. (1954), Biochern. J. 56, 288-297. 28. Hammond, B.R. and Gutfreund, H. (1955), Biochern. J. 61, 187-189 . 29. Weil, L., James, S., and Buchert, A.R. (1953), Arch. Biochern. Biophys. 46, 266-278. 30. Bender, M.L. and Kezdy, F.J. (1964), J. Arner. Chern. Soc . 86, 3704-3714. 31 . Blow, D.M., Birktoft, J.J., and Hartley, B.S. (1969), Nature 221, 337-340. 32. Hunkapiller, M.W., Srnallcombe, S.H., Whitaker, D.E., and Richards, J.H. (1973), Biochemistry 12, 4732-4743. 9. 33. Koeppe, R.E. II and Stroud, R.M. (1976), Biochemistry 15, 34503458. 34. Martinek, K., Vill, Kh., Streltsova, Z.A., and Berezin, I.V . (1969), Molek. Biol. I, 554-565. 35. Martinek, K., Savin, Y.V., and Berezin, I.V. (1971), Biokhimiya 36, 806-815. 10. Chapter II The Structure and Activation Mechanism of Bovine Trypsinogen 11. Physiological control over hydrolysis by the digestive serine proteases, and in other systems as well, is maintained by presence of specific inhibitors and by synthesis of these enzymes as inactive precursors, called zymogens or proenzymes. The digestive serine pro- tease zymogens are activated by a single tryptic split, between Lys 15 and a hydrophobic residue, 16 (1,2). Loss of these few residues from the N-terminus results in conformational changes which make the enzyme catalytically active (3). The three-dimensional structure of chymotrypsinogen was solved by Freer et al. (4) and was further interpreted by Wright (5,6). Al- though the structure of the zymogen, including orientation of the catalytic site residues Asp 102, His 57, and Ser 195, was very similar to the chymotrypsin structure of Birktoft and Blow (7), the specific binding pocket (residues 187-194 and 214-220) was somewhat less open and accessible to a substrate side chain than in the active enzyme . Lack in the zymogen of the Ilu 16 - Asp 194 salt bridge resulted in a very different conformation for residues 191-194, according to the study. This chapter of the thesis is concerned with the detennination of the high resolution structure of bovine trypsinogen, carried out primarily by Tony Kossiakoff, and by Bob Stroud and the author. This study indicates that trypsin and trypsinogen are structurally significantly more alike than chymotrypsin and chymotrypsinogen. This may be partly due to the existence of chymotrypsin as a dimer in the pH 4.2 crystals (8), and to the fact that those structures are in a lower 12. state of refinement that the trypsin/trypsinogen pair. This work is discussed in detail in Appendices 4 and 5. itrJ. additional observation made from this study is that relatively small structural changes (on the order of an angstrom) can be very important to the function of a protein. Moreover, in low-resolution electron density maps the possibility of errors in interpretation is high. Refinement of protein structures can greatly improve the accuracy in determination of such changes, and is the subject of the next chapter of this thesis. References 1. Neurath, H. and Dixon, G.H. (1957), Fed. Proc. 16, 791-801. 2. Northrop, J.H., Kunitz, M., and Herriot, R.M. Enzymes, p. 96. Crystalline Columbia Univ. Press, New York, 1948. 352 pages. 3. Neurath, H., Rupley, J.A., and Dreyer, W.J. (1956), Arch . Biochem. Biophys. 65, 243-259 . 4. Freer, S. T. , Kraut , J. , Robertus , J. D. , Wright, H. T. , and Xuong, Ng. H. (1970), Biochemistry g_, 1997-2009. 5. Wright, H.T. (1973), J. Mol. Biol. 79, 1-11. 6. Wright, H.T. (1973), J. Mol. Biol. 79, 13-23. 7. Birktoft, J.J. and Blow, D.M. (1972), J. Mol. Biol. 68, 187-240. 8. Sigler, P.B., Jeffery, B.A., Matthews, B.W., and Blow, D.M. (1966), J. Mol. Biol. 15, 175-192. 13. Chapter III Protein Structure Refinement: The Refined High-Resolution Structures of DIP-Trypsin and Trypsinogen 14. Ever since the method of multiple isomorphous replacement (MIR) enabled the first protein structure, that of hemoglobin, to be solved by Perutz and coworkers (1), MIR has been the traditional method for solution of new macromolecular structures. The technique has enabled protein crystallography in a short time to become one of the most powerful methods for elucidation of biochemical problems. However, the errors in such a structure determination, arising from interpretation of the electron density map, and from difficulties in building and measuring atomic coordinates from a skeletal model of the protein with reference to such a map, can be substantial (2). The MIR elec- tron density map unfortunately provides no feedback concerning the correctness of interpretation. It has become apparent that small positional changes (on the order of an angstrom or less) can be highly significant mechanistically, for example in evaluating differences between oxidized and reduced forms of a protein, between an enzyme and its zyrnogen, between members of an enzyme family, or in comparing homologous proteins from different species. Structure refinement is now performed routinely on structures of small molecules, and has recently been applied to some protein structures (3-8), although the size of the problem makes most of the methods normally applied to small molecules unfeasible or prohibitively expensive for large structures (6,9). Refinement greatly improves the accuracy with which the types of differences mentioned above, as well as other important structural features, can be described. It is moreover particularly valuable in the case of proteins, 15. • where these differences may lie well within the resolution of the data, in obtaining the maxiIIIl.Ull amount of information from the limi ted diffraction patterns typically available from protein crystals. Generally, in structure refinement the disagreement (e . g., a weighted sum of the squared differences) between the observed structure factor amplitudes (F0 ) and those calculated from the model of the structure (Fe) is minimized with respect to the positional and thermal parameters for each atom. The most conunonly quoted index of t his dis- agreement is the standard crystallographic R-factor (R = EJF 0 -F C J/EF0 ). For a small molecule R is often in the range 30% - 40% prior to refinement, and can fall to around 4% for a wellrefined structure. In the case of a protein, R usually depends strongly on the resolution and overall temperature factor of the data , but is typically about 45% - 50% for a medium-resolution unrefined structure, falling to less than 25% for most of the refined st ruct ures to date. The most spectacular refinement of a protein structure curr ently 0 is that of rubredoxin (3) for which R is 11% at 1.2 A resoluti on (11) . However, the structure-factor least squares procedure used for t his small protein (molecular weight 6000), which forms unusually well ordered crystals, is expensive (9), and would be prohi biti vely so for a large structure . Presently, the most widely used technique for protein structures is real-space refinement, in which use is made of the programs written by Diamond (12,13) for fitting a model with standard bond lengths and 16. angles to a measured coordinate set (model-building), and for optimizing the fit of a set of model coordinates to an electron density map (real-space refinement). These programs are, however, very complex and require very large amotmts of storage and time on a large computer . While these requirements do not increase as rapidly for large structures as is the case with least squares refinement, many more cycles are required to produce a refined structure, and the method is still extremely costly. Refinement by difference Fourier methods is poten- ially much more economical. The method has previously been used effectively for proteins in conjunction with Diamond's programs (5-8, 14,15). In the present study the three-dimensional structures of DIPtrypsin and trypsinogen were refined with a constrained difference Fourier technique. The project was undertaken for several reasons. It was felt that the changes between these two proteins, leading to a model for the relative inactivity of the zymogen, could be much more accurately described with reference to high-resolution, refined structures, and this is in fact the case. In addition, the DIP- trypsin structure was an excellent candidate for revealing fine details of general protein structure upon refinement. The crystals are well ordered, diffract to at least 1.1 A resolution, and can be grown to relatively large size (about 0.75 x 0.75 x 1.5 nnn) . It was expected that features such as nonplanar amide groups and praline rings, distortions in helices and sheet structures, solvent structure, and unusual values of bond lengths and angles, could be accurately 17. identified in the refined structure. There were, moreover, several regions in the MIR-phased map where interpretation was ambiguous, particularly near the surface of the molecule and in some of the tight loops in the structure. Extension of the resolution of the map and refinement of the structure were necessary to insure correct interpretation in these regions, and to remove the errors arising from construction and measurement of the wire model. Finally , because of the dramatic improvements in description of a structure which can be gained through refinement, development of a system for efficient, inexpensive, and routine refinement of macromolecular structures was highly desirable. The author has written a series of programs for routine refinement of protein structures, which can operate entirely on a minicomputer (with the current exception of computation of density maps) . Use of this system has enabled the DIP-trypsin refinement to proceed at an extremely low cost from a starting R of 47% at 2.7 A resolution 0 to 21.5% at 1.5 A resolution (the best statistics to date for a refined enzyme structure). This work is described in Appendix 6. Some additional stereo figures of points of interest in the refined structure, generated with this system, are presented in Appendix 7. This series of programs has also been used for refinement of the 0 trypsinogen structure to an R of 31% at 1.9 A resolution. The results and comparison with the trypsin structure are discussed in Appendices 4 and 5. As part of this work, I developed a similar system for pro- tein refinement for use on the CDC-7600 computer at Brookhaven 18. National Laboratory, and refinement of the trypsinogen structure is being continued there by Dr. A. Kossiakoff . Parts of the system have also been used recently in optimi zi ng the fit of coordinates for oxidized and reduced forms of tuna heart cytochrome£ to the MIR electron density (16). Bob Almassy i s cur- rently using the syst em with the structure of cytochrome £ss 1 (17) . References 1. Green, D.W., Ingram, V.M. and Perutz, M.F. (1954), Proc . Roy . Soc. A225, 287-307. 2. 01.ambers, J.L. and Stroud, R.M. (1977), Acta Crystallogr . - - in press (Appendix 6 of this thesis). 3. Watenpaugh, K.D., Sieker, L.C., Herriot, J . R., and Jensen, L.H . (1973), Acta Crystallogr. B29, 943-956. 4. Diamond, R. (1974), J. Mol. Biol. 82, 371-391. 5. Moews, P. C. and Kretsinger, R.H. (1975), J . Mol . Biol . 91 , 201 228. 6. Freer, S.T., Alden, R.A., Carter, C.W. Jr., and Kr aut, J . (1975) , J. Biol . Chem. 250, 46-54 . 7. Deisenhofer, J. and Steigemann, W. (1975), Acta Crystallogr . B31 , 238 - 250 . 8. Bode , W. and Schwager, P. (1975), J. Mol. Biol. 98, 693-717 . 9. Konnert, J.H. (1976), Acta Crystallogr. A32 , 614-617 . 19 . 10. Stout, G.H. and Jensen, L.H. X-ray Structure Detennination: Practical Guide, pp. 395-397. Macmillan, New York, 1968. A 466 pages. 11. Watenpaugh, K.D., Sieker, L.C., and Jensen, L.H. (1977), American Crystallographic Association Abstracts, Series 2, ~(1), 17. 12. Diamond, R. (1966), Acta Crystallogr. 21, 253- 266. 13. Diamond, R. (1971), Acta Crystallogr. A27, 436-452. 14. Huber, R. , Kukla, D. , Bode, W. , Schwager, P. , Bartels, K. , Deisenhofer, J., and Steigemann, W. (1974), J. Mol. Biol. 89, 73101. 15. Epp, 0., Lattrnan, E.E., Schiffer, M., Huber, R., and Palm, W. (1975), Biochemistry 14, 4943-4952. 16. Mandel, N., Mandel, G., Trus, B.L., Carlson, G., and Dickerson, R.E. (1977), J. Biol. Chem.--in press. 17. Almassy, R.J., personal communication. 20. Chapter IV A Crystallographic Study of Trypsin Inhibition by cu 2+ 21. Introduction The metal ions Ag+ and Cu 2+ are strong inhibitors of trypsin and chymotrypsin (1,2). These ions compete with one another for binding to the enzyme in an inhibitory fashion. As described in Chapter I and Appendix 1 of this thesis, Ag+ inhibits these enzymes by disrupting the catalytic site, binding between the Asp 102 and His 57 side chains. In the difference infrared titration experiment of Koeppe and Stroud (3), the peaks corresponding to titration of the Asp 102 carboxyl with a pKa of 6.8 were assigned with the use of cu2+ ions, since the experimental difficulties with Ag+ are considerably greater. We wished to determine directly the site of eu 2+ binding to trypsin, since the fact that Cu2+ and Ag+ compete in the kinetic experiments does not necessarily mean that they bind at the same site in the enzyme. The determination was made using x-ray crystallographic techniques by Roger Koeppe, Lois Kay, and the author. In addition, Mel Jones and the author are currently attempting to locate the Cu 2+ binding site in benzarnidine-inhibited trypsin. Experimental DIP-trypsin was crystallized as described by Stroud et al. (4). The eu2+ derivative was prepared by placing the DIP-trypsin crystals in a pH 6.9, 0.05 M cacodylate buffer containing 17 mM Cu(NO 3) 2 (Merck reagent) and 0.66 M Mgso4 . 22. Reflection intensities were measured with a Syntex Pi diffractometer using graphite-monochromatized CuKa radiation, and equipped with a helium tube between the crystal and detector. current were 40 KV and 20 ma, respectively. Tube voltage and The step-scan method of Wyckoff et al. (5) was employed to record intensities from a single 0 crystal to 2.53 A (28 = 35.5°). Background corrections were made with the anisotropic interpolation method of Krieger et al. (6). Further data reduction and scaling were performed according to Stroud et al. (4). The intensities of five standard reflections were monitored throughout the data collection as an indication of crystal decay arising from x-ray exposure. At the termination of data collection none of these intensities had fallen by more than 15% from their initial value. Of the 7704 independent reflections in the 2.53 A sphere, those 6905 with structure factors greater than twice their standard deviation were included in the final data set. A 2.7 A difference electron density map was computed with coi<P efficients m (FCuDIP - FDIP) e , where FCUDIP and FDIP are respectively the structure factor amplitude for eu 2+-DIP-trypsin and DIPtrypsin, and¢ and mare the 'best' phase and figure of merit (7) determined by Stroud et al. (4) 23. Results and Discussion The section of the difference density map obtained in this study, shown in Figure 1, is very similar to that obtained for the Ag+ derivative (8). These _a re the only significant peaks in the map, the larger one arising from a copper ion bound between Asp 102 06 1 and His 57 No 1 , and the smaller one from movement of the His 57 imidazole to accommodate the copper ion. This detennination therefore indicates that the site of eu 2+ inhibition of trypsin is the same as for Ag+, confinning the method of assignment of difference infrared peaks by Koeppe and Stroud (3). References 1. Martinek, K., Vill, Kh., Streltsova, Z.A., and Berezin, I.V. (1969), Molek. Biol. l, 554-565. 2. Martinek, K., Savin, Y.V., and Berezin, I.V. (1971), Biokhimiya 36, 806-815. 3. Koeppe, R.E. II, and Stroud, R.M. (1976), Biochemistry 15, 34503458. 4. Stroud, R.M., Kay, L.M., and Dickerson, R.E. (1974), J. Mol. Biol. 83, 185-208. 5. Wyckoff, H.W., Tsernoglou, D., Hanson, A.W., Knox, J.R., Lee, B., and Richards, F.M. (1970), J. Biol. Chem. 245, 305-328. 24. 6. Krieger, M., Chambers, J.L., Christoph, G.G., Stroud, R.M., and Trus, B.L. (1974), Acta Crystallogr. A30, 740-748 (Appendix 2 of this thesis). 7. Dickerson, R.E., Kendrew, J.C . , and Strandberg, B.E. (1961), Acta Crystallogr. 14, 1188-1195. 8. Chambers, J.L., Christoph, G.G., Krieger, M., Kay, L., and Stroud, R.M. (1974), Bioch. Bioph. Res . Conmrun. 59, 70-74 (Appendix 1 of this thesis) . 9. Henderson, R. and Moffat; J.K. (1971), Acta Crystallogr. B27, 1414-1420. 25. Figure Legend FIGURE 1. A view down the r_-axis of the 2.7 A FCuDIP-FDIP dif- ference Fourier map described in the text. Four sections of the map, 0 giving a total thickness of about 5 A, are represented. Positive density is indicated by solid contours, negative density by dotted contours. Contour levels begin at ±0.2 electrons/A3 with intervals of ±0.1 e/A3 . The RMS error in this map (9) is 0.08 e/A3 . The bars at the lower right indicate distances of 2 A. The only significant peaks in the map are shown, the larger corresponding to a cu 2+ ion positioned between H57No and Dl02Oc , and 1 1 the smaller one to the movement of the H57 irnidazole out toward the solvent to acconunodate the copper ion. The positions of the Ag+ and H57 movement peaks in the Ag+-DIP-trypsin difference map (8) are marked with asterisks (*). 26. Figure 1 I ·Cl) ::> u 0 27 . Appendix 1 Silver Ion Inhibition of Serine Proteases: Crystallographic Study of Silver-Trypsin 28. Vol. 59, No. 1, 1974 BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS SILVER ION INHIBITION OF SERINE PROTEASES: CRYSTALLOGRAPHIC STUDY OF SILVER-TRYPSIN J. L. Chambers, G. G. Christoph, M. Krieger, L. Kay, and R. M. Stroud Norman W. Church Laboratory of Chemical Biology California Institute of Technology Pasadena, California 91109 U.S. A. Received May 3,1974 Summary Silver ion is a potent inhibitor of trypsin and chymotrypsin, with Kr's of 4 x 1 o- 5 M. and 3 x 1 o- 5 , r ·e spectively. A crys_tallographic study shows that the primary silver ion binding site on trypsin is at the active center between the carboxyl group of Asp 102 and the 6-nitrogen of His 57. This result is correlated with the fact that Ag+ interferes primarily with the acylation rate constant, k 2 , and does not significantly affect the binding constant, Ks. The location of this site explains the potent inhibitory effect of silver (I) ions on trypsin activity: The imidazole ring of His 57 is repositioned 1. 8 A further out into the solvent to accommodate the silver ion, preventing its normal interaction with the hydroxyl group of Ser 195. Consequently, His 57 cannot directly assist the proton transfer in the catalyzed reaction. Since silver ion binds to the catalytic site in this highly specific manner, silver may be used as a specific probe of the active site of serine proteases. This communication reports the 2. 7 A resolution structure of the isomorphous silver derivative of bovine trypsin inhibited by diisopropyl-fluoro. 1, 2 phosphate (DIP). Martinek et al. found that silver ion is a potent trypsin (or chymotrypsin) inhibitor with a K 1 4 x 1 o- 5 M. They concluded, first, that silver ion prevents the acylation of the enzyme while not appreciably interfering with substrate binding. Second, silver ions compete with pro- tons for the binding site and the silver binding depends on a group with an apparent pKa of 7. 1, which they suggested was the imidazole of His 57. In light of more recent evidence, however, the pKa reflected in those experi3 4 ments is more likely that of Asp 102 • . EXPERIMENT AL To prepare the silver (I) derivative crystals of DIP-trypsin 5 were soaked in solutions containing O. 012 M AgNO 3 for periods of four to eight days. Three-dimensional, 2. 7 A data sets for both the native DIP-trypsin 29. Vol. 59, No . 1, 1974 BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS and the silver (I) derivative were collected using a Syntex Pl automated diffractometer. Data reduction and scaling were accomplished 1,1sing 5 standard techniques. RESULTS AND DISCUSSION From our data and the phases previously determined for DIP-trypsin, a difference Fourier map was obtained. of the active site is shown in Fig. 1. 5 The region of this map in the area The large peak "A" corresponds to the position of the fully occupied, primary silver ion binding site. In addi- tion, there is a smaller region of positive density, "B" , above and behind the primary site. This peak results from a movement of the imidazole ring of the catalytic site residue His 57 into the solvent by approximately 1. 8 :t (S 0. 2) A . A secondary silver ion binding site of 35% refined occu- pancy was found elsewhere on the surface of the molecule, in the vicinity of His 40. Fig. 2 shows an ORTEP silver (I) trypsin derivative. 6 representation of the catalytic site in the The silver ion is coordinated in an approxi- mately linear fashion between the lower carboxyl oxygen of Asp 102 and the 6-nitrogen of His 57. Bond distances are: 061 (Asp 102)-Ag+, 2. 3:t0.2A; Ag+-N6 (His 57), 2. 3 :t0. 2 A . This suggests that the silver ion is in a two 7 coordi~ate sigma-bonded complex characteristic of silver (1) . This con- figuration is also structurally very similar to complexes of silver (I) 8 • hf ree amino • wit ac1' d s. The silver DIP-trypsin structure provides a model for the mechanism of the silver (I) inhibition of trypsin. The distance between the E - nitrogen of His 57 and the position of the y-oxygen of Ser 195 found previously for benzamidine-trypsin (where the serine oxygen was hydrogen bonded to the EN of His 57) is 4. 2 :t0. 2 A in the silver derivative. This long distance, coupled with the unfavorable directionality between these two atoms, prevents proton transfer from the hydroxyl group of Ser 195 to the His 57 imidazole. The effect of Ag + on k , the acylation rate con2 stant, thus gives an indication of the contribution of His 57 and Asp 102 to enhancement of the catalytic rate of serine proteases. As well as defining the mechanism of the silver ion inhibition, the silver derivative data have been incorporated into the phase refinement .. -~ , ,_ ~ .... -' i, \., -.\ _ " I ,. ! . ' () , ... c \ ' - ....., I . \,, 1• -· ; :. J, .. --~~ - ,-~ i :~· ·,,.,_,,:, '-1)_: • +i~-=- \ \ ~J-~n ( C: ! ~ ~ ~,.' .~-(: C·. - \._.,/:, ~\'0,Jlj (, ; , J '< , _, .> • ~ -D , -· l\ ) "' ( ... , ' .• (~} \ ~r \'-:,) " jJ ,, . . ~ -, Figure 2. An ORTEP drawing of the catalytic site of silver DIP-trypsin, viewed approximately down the crystallographic b-axis. The DIP group has been omitted for clarity. The 'l-oxygen of Ser 195 is in the position found for DIP-trypsin, close to that found in tetrahedral intermediates or in acyl enzymes. ~ Figure 1. A view parallel to the crystalloiraphic a-axis of the active site region of a difference Fourier map calculated using coefficients !Fsilverl-!Fnativel . The view shown is a projection through 12 sections, each 1. 1 A apart. Solid contours enclose regions of positive density ; the broken lines, regions of negative density. Peak "A" corresponds to the location of the main silver ion binding site. Peak "B" results from the movement of the His 57 imidazole ring. Contours begin at :t: 2. 5 standard deviations and the contour interval is one standard deviation in the difference electron density. 2A f ,,. ,- 1-·l',,~,., ) , '"' ' t·• g. z CJ) 6 z ► -I () C i 0 () I XI ("\ )> CJ) m m ::<, ► ,-- ri CJ) -< I 0"'ti O> 0 z )> ► ,-- ri ~ :::r m 0 () 0:, .i,.. 'l -<) z0 _-<) u, ~ 0 31. Vol. 59, No. 1, 1974 BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS of the 2. 7 A structure of DIP-trypsin. The overall mean figure of merit of the 2. 7 A phases increased to 0. 83 during this refinement. The details of this refinement will be discussed elsewhere. One of the key problems in the assignment of microscopic pKa's to residues at the active center of serine proteases lies in the requirement that one should be examining the "native" enzyme rather than a modified derivative of it. Two familiar problems are thus apparent: 1) presence of a covalent label near the active site perturbs the system under study in an unknown fashion; 2) spectroscopic techniques which study the native enzyme often have problems of assignment to particular residues. Silver ion can be used to as sign peaks to the active center residues by perturbation of the native enzyme spectrum, primarily affecting peaks due to Asp 102 and His 57. We are currently applying this technique to assign peaks associated with carboxylic acid groups in the difference 9 infrared titration spectra of native serine proteases. ACKNOWLEDGEMENT This work is contribution no. 4850 from the Norman W. Church Laboratory of Chemical Biology of the California Institute of Technology. This work has been carried out with the support of the United States Public Health Service Grants GM-19984 and GM-12121, whose help is gratefully acknowledged. One of us (MK) is the recipient of a Danforth Graduate Fellowship, another (RMS) is the recipient of a National Institutes of Health Career Award (GM-70469), and another (GGC) is a National Institutes of Health Postdoctoral Fellow. REFERENCES 1. 2. 3. Martinek, K. , Savin, Y. V. , and Berezin, I. V. , Biokhimi ya ~. 806-815 (1971). Martinek, K., Vill, K., Strel'tsova, Z. A., and Berezin, I. V., Molekulyarnaya Biol. 3, 554-565 (1969). Krieger, M., Kay, L. M:-; and Stroud, R. M., J. Mol. Biol. 83, 209-230 (1974). 32. Vol. 59, No. 1, 1974 4. 5. 6. 7. 8. 9. BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS Hunkapiller, M. W., Smallcombe, S. H., Whitaker, D.R., and Richards, J. H., Biochemistry 12, 4732-4743 (1973). Stroud, R. M., Kay, L. H., and Dickerson, R. E., J. Mal. Biol. 83, 185-208 (1974). Johnson, C. K., "OR TEP: A Fortran Thermal Ellipsoid Plot Program for Crystal Structure Illustrations", Oak Ridge National Laboratory, Oak Ridge, Tenn., Publ. No . ORNL-3794, revised 1965. Cotton, F. A. and Wilkinson, G. "Advanced Inorganic Chemistry", Interscience, New York, 2nd ed. , 1966, p. 1040. Acland, C. B. and Freeman, H.C., J. Chem. Soc. Dl971, 1016-1017. Koeppe, R. and Stroud, R. M. --in preparation. 33 . Appendix 2 Data Collection in Protein Crystallography: Capillary Effects and Background Corrections 34. Reprinted from Acta Crystallographica, Vol. A 30, Part 6, November 1974 rRINlT.D IN OBNAMRK A,·ta Cry.'1. (1974). A30, 740 Data Collection in Protein Crystallography: Capillary Effects and Background Corrections* BY M . KRIEGERt, J . L.CHAMHERS, G.G.CIIRISTOPllf, R. M.STROUn§ Cr<'llin laboratory of Chemistry AND B. L. TRUS' Norma11 W . Church laboratory ,if Clwmiral Biology. California fmlifute of TechnnlogJ•, Pa.,adena, Culifomla 91109, U.S.A . (R,•ceived 16 January 1974; accepted24 May 1974) In protein cryslallography, observed diffraction intensities must be corrected for background radiation due lo scatter from air and scatter and absorption by Cllpillary, crystal and molher liquor. A systematic study shows I hat a major contrihution lo background intensity is air scatter arising from the air intercepted by the dirccl X-ray beam as 's.:en' hy lhc receiving-counter upcrlure. As a result lhcre is a firstorder dependence of background on the Wangle. The second-order varialions in this function are principally due to absorption of the direct beam or air-scattered radiation by the Cllpillary and to diffraction hy the glass in the direct beam. To reduce dais collection time and crystal e,posure, individual background measurements may be appm,imated by interpolation from empirical background curves or, alternatively, by collecting background inlensities for short times and titling lhese data wilh a multidimensional funclion. If isotropic interpolation is used, I.e., if background is considered to be a function of 211 alone, syslematic errors of up to aboul 30% can be introduced into the interpolated backgrounds. Mel hods or accounling for the anisotropy in the background are derived and shown lo reduce Ihis error to I 2 '%. lntro,tuclion Because protein crystals arc quite susceptible to radiation damage, crystallographers have been e,ploring different methods for accurate data collection which minimize the X-ray e•posure of the <:rystal. With automated diffractometers, one widely used method of data collection involves counting the background on one or both sides of each measured reflection. There arc several methods of reducing the time of data collection and thereby increasing the number of reflections collected per crystal (,•.g., Wyckoff, Tsernoglou, Hanson. Know_ Lee & Richards. 1970; Watson, Sh,>tton, Cox & Muirhead, 1970). In one method, the observation of individual backgrounds is omitted and background corr~-clions for the measured intensities are calculated from an empirical curve of background 11rr.ms sin 11/). measured li>r each cryslal. If the backgrounds do not vary signilicantly during the time of collec1ion, almost all of the crystal exposure time can be devoted to intensity data collection, and empirical background curves can be measured after the intensity data have been collected . Empirical background curves have generally been obtained from background values measured with long counting times for points along one lattice row, and applied using the • Contribulion No. 4726. Supported by U.S. Public Health Service Grants GM-t9984 and GM-12121. t Danforth Foundation fellow. i National lnstilutcs or Health Postdoctoral Fellow. § National lnstilutcs of Health Career Development Awar.. dee, U. S. Public Health Service Grant No. GM-70469. • Jane Coffin Childs Memorial Fellow. approximation that the bal'kl(mund depends only on the Bragg angle and not ,111 the other setting :mglcs (Matthews, Levine & Argos, 1972; Jensen, 1972; Sulemme, Freer, Xuong, Alden & Kraut, 1973; Wyckoff. Doscher, Tsernoglou, Jnagami, Johnson, Hardman, Allewell, Kelly & Richards. 1967). This appro•imation can .give rise systematic errors of about 30% hetwL-en interpolated and observed backgrounds. Hill & Banaszak (1973) have reported observing an additional, 20-invariant q, dependence. Our experiments show that the background radiation can vary systematically with q, and x in addition to 211. and furthermore that such rp and x dependences are functions or 211. These dependences can he accounted for by additional components in a simple interpolative procedure. In contrast to Hill & Banaszak, we conclude that the variation with q, arises primarily from capillary absorption and scattering rather than ab- · sorption of background radiation hy the protein crystal and its mounting to the capillary. There is substantial improvemcnl in interpolated backgrounds if the rp and x-depcndent variations are accounted for. A second approach to streamlining data collection involves observing very short backgrounds with each reflection, and fitting, by least squares, a function of the diffractometer setting angles to all of the data. Such a function must include cross terms between x and 20, and rp and 20, and can· vary in complexity depending upon the range of 20 and the experimental conditions. By pooling the data in this manner. the resulting backgrounds are more accurate than the individual short measurements from which the background function was constructed. 35. M. KRIEGER. J. L. CHAMBERS, G. G. CHRISTOPH, R. M. STROUD AND B. L. TRUS It should he noted that systematic variations between real and interpolated values for the background can give rise to systematic errors in phase determination, in AF terms used in calculating difference maps, and in scaling. Phases and AF's free from such errors are crucial to accurate descriptions of molecular shifts from difference maps (Dickerson, Kopka, Varnum & Weinzierl, I967; Henderson & Moffat , I971 ; Krieger, Kay & Stroud, 1973). Furthermore, with protein crystals whose reflection intensities arc small, a reflection which is only twice as inlense as the background may be ten standard deviations ahovc background. Therefore, well determined reflection intensities (,;;Ja) may he in error by as much as JO% if anisotropy in the background is not accounted for. As the background intensity is independent of crystal absorption, lhc badgrounds, estimated or measured, should he sublracted from the observed dit~ fraction intensities before applying the standard absorption and Lp corrections. In this paper we describe two methods for obtaining 741 background corrections of similar quality to individually measured backgrounds, but at significant savings in the time spent on observing hackgrounds. Although it is desirable to minimize the background present (i.e., improve the signal-to-noise characteristics of the experiments), we arc primarily concerned here with developing background correction methods whose use would he applicable to a wide variety of existing fouraxis diffractometers, without modification of the experimental hardware. As a result or this analysis, and the consideration of lhe origins of components in the background intensity, experimental methods of reducing the background, and thereby increasing the signa lto•noise ratio, are suggested . •:xperimental Measurements were made on severnl different diffractometers both with and without a monochromator. Details of each experimental arrangement used arc descrihcd in Table I. E~ch machine will henceforth Table I. Characteri.,tir data/or tliffrurtom,·ter.1· us<',/ i11 thi.1· study In all cuses the h.1<i1C four-circle ma.:hincs have been rehuill ;rnd modified; in all cuscs cxapl A amt I) modificulions have been cxlcnsivc 10 the point uf hHal rcmachining . (,IJ Syutell. l'f (ll) (ii' XRD-4'/0 (/)) (CJ (/iJ llatc>.-G L:: llilgcr Watts• UucrHCr .su1lpcf't (/') s. Samsonf hyhrid Source Target si1.c (mm) Voltage (kV) Current (mAl Radial ion I0x I 40 20 CuKcx 12·5 , 0·k 45 45 18·5 IS Cu Ka Cu K« 3" J' 3· Graphilc None None Graphite Graphite 0" 0·.1 -0·4" ~ 4-0(l·0J ~')·5 (1 ·0) 23-5 H·0(2·0J 9·0 11 ·5 (0·6) 15 ·9 (0·<>J 19·4 4·0 !l·0J 5·5 12·0 (1 ·5) 20·8 (1 ·8 ) 22·5 Geometry Take-off-angle 6" M,mochromator type: <irnphih.: Monochromalor mounting (t?)§ 90" 0 ·3'" Monocl11omator dispersion Soun.:c to crystal Distance from source to : 1st apcrlUrc(cm) (si7.c in mm) Monochromator c1yslal Apcrlure (size) hyhrid 12·S x 0 ·M !Ox I 46 16 C:u K« 8 x0·4 20 12·S x 0·8 45 IK Cu Ka Mo Ko ]3 'I()'" 0·3' ~-5 (2·01 6-K 19·8 (1·0 - 1·5) 21 ·.l (I ·5- 2·0) 27·3 4·5 (l·0) 5·0 8·0 (l ·0J 14·5 (l·HJ 15·5 None 4·5 (1 ·5) 14·3 (I ·M) 15·0 l>istanccs frum i.:rystal In : J\pc.:rturc (size) Aperture (sit.c) Counter 32·5 (2·5) 3H·S (l·0 -2 ·0J 40·5 4·0(1·0 1·5) 12·0 (2·0 2·51 l<,·0 22·2 17·0 (.1·0) 22·5 0·01 25·0 6·0(2·2) 1 l·H (1 ·5--2·0J IJ ·5 8·0(2·11) 11·4 (5·(14) 21·6 Noise count rah! with X-rays olf (per 100 sJ 35 55 25 3H 30 13 AperlUre (size) Crysl:11 None Crystal tu cuunlcr 11 9·2 (l·0J • Standard four-circle machine with no mon0<:hromator. t This machine was essentially redesigned by, and rebuilt under the din.-clion of, R. M. Stroud, and uses Dalcx automation for operation in a ~-slcp scan mode (two-circle Weisscnbcrg geometry). t t--'our-\.'.ircle diffraL·lomctcr a.Cisemblcd locally under the direction of S. Sam so n. Fh.-..:tronics and Alloys goniostat, Varian computer, Datcx rcsovcnlyncs, and Ortcc counter chain. l.! is the .inglc between lhc normals to the plane-; of inc.:idcnce at the monochromalor. 1 • Counter tyre : .1 ll ltilfrnc:1omctcrs use sl"in1illatiun counter. l lcrmctically sc,1lcd Tl · llriftcd Nal ,·rystals arc individually scli.:dcJ and dtci.:ki.·<l IL'~t1larly for decay ;rnd tlamagc . * 36. 742 DATA COLLECTION IN PROTEIN CRYSTALLOGRAPHY be referred to by a letter (A-F) which in turn corresponds to one arrangement in Table I. Each of the figures refers to a specific case, although the discussion considers results obtained from many different experiments. It should be noted that these different arrangements include diffractometers both with and without monochromators, with both of the generally used monochromator settings, with copper and molybdenum Ka radiation, and with or without helium-filled pathways to the counter. Glass capillaries were 0·3-1 ·0 mm in diameter with walls approximately 0·01 mm thick . Crysta ls of trypsin, trypsinogen, cytochrome c and their derivatives with dimensions ranging from O· I x O· I x 0·35 mm to 0·4 x 0·4 x 0·7 mm were used in the crystal-capillary experiments. Trypsin crystals grow with acicular habit, cytochrome crystals are rectangular prisms, and trypsinogen crystals grow as approximately equidimensional trigonal bipyramids. Thus crystals of the most commonly encountered morphological forms have been used. 400 ,b- ~ _ .1 .. .... ---"'-t o 1 .J • · ' -~ L · -• , _ ... 40 28 Fig. 2. The background 20 dependence for : (u) air scatter only (0, S min/point), (b} on empty capillary( <', , 3 min/r>oint), and (c) for a trypsinogen crystal, mounted in the same 0·5 mm diameter capillary ([.J. 3 min/point) . All measure~ mcnts were ma<le at x==90", usin,t diffrac:lometcr A as de• scribed in lhc te,cl. ➔ €1 5 CollifflOIOt" Backgrounds were measured under the following conditions: (I) withoul cryslal or capillary in the beam; (2) with a well centered capillary (mounled coaxially with q,) in the beam; (3) with a capillary miscentered, but mounted parallel to the ,p axis; and (4) with a crystal in a capillary such that the crystal, but not the capillary, is cenlercd in the beam (the standard data-collection condition). Background counting times were usually 3 lo 5 min per point for the empirical backgrou.nd curves, and 10 to 120 s for individual backgrounds used for comparison. Reflection intensities were measured in both thew scan and the Wyckoff step-scan (Wyckoff el al., 1967) modes. Measurements were made over the ranges 0,;20,;46 ·'; 0 !'.:,p~ J60 °; and O:,;x,;: J60 °. Attenuators were used to obtain data for small 20 angles down to O"'. Results Fig. I. The usual diffractometer data-collection arrangement. The 'visible volume• is 1hc volume of air irradiated by the direct hcam that is seen by the receiving-counter aperture. This volume changes only wi1h 20, and governs the 20 dependence of the air-st:altcred part of the background. Por1ions of the •visible volume' am at limes hidden from the X-ray beam and al times from the counter by the intcrposi• tion of the capillary ;,tnd crystal. The overall background will be affected hy changes in the capillary -t.:rysral setting angles. as well as !he amt111nt or capillary glass in the direct beam. Fig. I shows the usual diffractometer data collection arrangement and illustrates the origins of the different variations in the background intensity. Primary ~ff<'cls: 211 dependence Fig. 2 shows the overall background as a function of 20: (a) with no crystal, no capillary; (b) for a well L-entered capillary; and (c) for a capillary containing a trypsinogen crystal (dimensions approximately 0·3 x 37. M . KRIEGER, J. I.. CHAMBERS, G . 0. CHRISTOPH, R. M. STROUD AND ll. I. TRUS 0·25 x 0-35 mm). The intrinsic absorption of a glass capillary was measured and founJ to be generally less than 25 ¾. The absorptive component due to the glass which affects the curv~s of Fig. 2 must consequently be less than about 25 '}., _ It is much less where 'visible volume' dimensions (sec Fig. I) arc greater than the diameter of the capillary which is often the case. It is therefore clear that scattering by nir is a major contributing factor to the background . In the presence of a capillary the background is reduced by absorption by the glass capillary. The hackground intensity increases at higher 21/ angles where scatter from the glass predominates over ils absorption . Variations in the orientation of the capillary lead to f/1• and x-dependcnt secondary perturbations in the background . j ~ 0 E 104 0 1.00 u 096 0 z 09 2 . o ··lio - - - -- - -· · - 1 a 4' Fit,t. 3. The hadgruund If' Jcpcndcn1.:c ohscrvcd al 211 -• 22·(," and .z = 90-' for a DIP-lrypsin cryslal. The average intcn~ity is nonnali,.cd 10 1·0 (f)iffrnctomclcr A). ... 211 108 74.1 S<'c11111/ury ,:(/,•c/s: 'I' depenclcnr<' When ii is necessary to collect data wit h a crystal mounted in a capillary whose diamrl<'r is greater than the radius of the X-ray henm, there can be a significant systematic variation of the background radiation with the diffractometer I/' setting. Because !here is lillle difference between the angular distribution of the background of an empty capillary a m.I that of a capillary containing a crystal and its mother liquor, the <p dependence must he almost wholly due to the eccentricity of the capillary about the I/' axis. Fig. 3 shows typical <p dependence of background intensity for n capillary containing a protein crystal, and Fig. 4 shows this dependence for an eccentrica lly mounted empty capillary al different 211 values. A well centered capillary shows no I/' dependence. while with eccentrically.mounted capillaries the background variation is an approximately sinusoidal function of 2,p as the capillary rotates in and out of the hcam. At low 211 values the background, primarily due to air scatter, is modulated hy absorption of the eccentric capillary; therefore, the background is greatest when the capillary blocks the direct be:1m least (f/1 = 0 and IKO • in Fig. 5). At higher 21/ angles the capillary scatters more radiation than it absorbs, and backgrounds are highest al I/' = 90 and 210 · in Fig. 5. Al intermediate ~ti values, near the crossing point of Fig. 2, the absorption and scattering of the capill:ary are nearly eq1wl and the backgrounds arc essentially 1pindependent. The extent of the b:tckground variation with <p will depend on the eccentricity and diameter of the capillary and the value of 20. The curves in Fig. 6 depict the background as a function of 2// measured at two values of I/'; one when the low-angle backgrou,uf is a minimum (curve A ~0 . ,..,(211)] and the other when the low-angle background is a maxi mum (curve /J~,. ,,,.pll)). The magnitude of the differences between the lwo curves is the a mplitude of the f/1• dependent variation. There may be circumstances under which these curves would not cross. although we have not observed this. In such a c:rsc the amplitude of the f/l•dc1>cndent variation with 20 would not change sign. 22-r- I 27•6" 360 4> Fig. 4. The f' dependence of the b;.1ckground for an CCl.'Cnlrically mounted empty capillary Bt various 20 values Cr = 90" in alt cases}. While the amplitude and sign of 1he varialions change with 20. the pha~c is con~tant. The curves were measured at 20 = 27 ·ff·. 22·7·· , 17-k", 16·1 ". 14·5 .. , 6·4", and 4·H". The trat.·cs. cKccpl for 21} .::, 14·5", represcnl the hcst least· M,tuares lit of 24 roitlls mea."iurc<l .it 15" intcrvuls lU lhc fund ion : Normalized intCO"iil)" .:..:. I ·0 I a cos 2(q, •·· '°") . 'I he average intcnsily al c.u:h 20 is normalit.ct1 to I ·0. The ...:urvc for 20 = 14·5"' is the best slrnight line through the datu nl this 20 value. Mcasun:men1-. were made on c.liffrnc1ontt:lcr A. For a DI !'-trypsin crystal in a 0·6 mm capillary and a I nun hcam, the background at 23 '' in 20 va ried 11 % in intensity with rotation about I/'· The average reflection intensity in the shell 22 < 211< 26 "' for this crystal was seven times the background level. For a reflection with this 'average' intensity. using interpolated backgrounds uncorrected for the <p dependence would produce a systematic error up to I ·6 % in its net intensity. Weaker reflections, which are still important for heavy-atom refinement and difference maps, are subject to substantially larger erro rs: a reflection with raw intensity twice background, althoui:h at /ea.H 10 '7 (.1·ta111/ard dt•1>ialillm) ah,.,,,. ha.-kgrouml, would he in error by as much as 11 % . These errors arc systematic and must be corn.."\.'lcJ for, if .u.:curalc plmscs arc lo he calculated. 38. DATA COLLECTION IN PROTEIN CRYSTALLOGRAPHY 744 The following scheme was tested for generating interpolated backgrounds corrected for f{J dependence: (J) Measure the f{J dependence at a low 211 value (~6 ") with a q, scan al x=90°. (2) Measure the 20 dependence, A...,., 00(2/1), at the rp value ('{Jo) for which the background is a minimum on the rp scan. (3) Measure the 211 dependence, B.,. , 00 (2/J), at the rp value (f{J 1 =rp0 +90 '" ) for which the background is a maximum on the f{J scan . (4) Tabulate the difference, D.(20) = B, ,.,90(2//)A,0 _,90(2/}). (N ormally, it is convenient to interpolale the values between the observed backgrounds on the A...,.,,.(2/)) and tJ.,.,.,,(211) curves at intervals in sin (0)/.t] (5) Calculale the backwounds, BG(20,f{J)- For background correclions (just as for absorption corrcclions) the rp setting angle, f{J, must he modified to account for geometrical contributions lo rotalion with respect to the incident beam from x and,,,_• BG(20,f{J,l= A,0 .,,.,121/H- D.(20) sin' (q,, -'{)0 ) x dependence The background x dependence is analogous to the q, dependenc~. but a rigorous definition µf the problem is more difficult as u result or more complex geometry. However, the x dependence may be understood qualitatively. As X varies from O to 90", the amount of glass in the equatorial plane increases approximately as the secant of x- This, in turn, increases the ahsorption of air-scattered radiation. The exact nature of this angular backl\round variation with x depends strongly on the experimental conditions.? t We have ohscrveJ lha l each difTraclomctcr has a characlerislic x - 20 dependence, as shown in Figs. 7 and 8. Subtle differences in the geometry of the difTraclometcrs (,•. g .• collimator-to--crystal distam·c) arc apparently 1csronliihlc for the difference in thc:sc pallerns. 1 hi: correction scheme outlined below has proven ctTcc1i\'e for datu coll'-'\;lt:d on different diffrac1omc1crs with difTerenl x- 20 patlcrns. (I) where f{J,=q,, - tan · ' (cos x tan m). (II) For 1500 .reflections ohservcd over a wide range or W, f{J , and x for a DIP-trypsin crystal (0·25 x 0·3 x 0·4 mm) mounted in a I mm capillary on diffractomeler A, the. agreement between observed and interpolated backgrounds typically improved by 40 % when q,-depcndent anisolropic, rather than isotropic, interpolation was used . This scheme assumes that the q, dependence is sin usoidalt .. nd that the ampliludc of the variation is the difference (D.(211)] hctwc-en the A••· , 90(20) and B,,. ,.,,,(211) curves or Fig. 6. If possible, crystals should be mounted only in capillaries whose diameters are smaller than the radius of the X-ray beam. Under these conditions there is no observable ,p dependence. • Any addiliC'lnal co rrcuion for a 20 component in ,p, (due to ro1a1ion with rc'ipcd tt1 the ruuntcr) is much less significant in general where the visihk vo lume (Fig. 1) is larger lhan the irradialclt portion of the capill;1ry. The corro.:lion for x anJ ,,, deals with the sourer of ~1tlering directly, while the 20 com:L:tion deals with o nly ,, small cornpom:111 of the scallcn·d ratlia1ion. In any case an .1d1litinnal 2/J com.·dion wnnld he· ""'ymmclric amt comrk.\ a1HI i" uM1ally u11111."\:,:e:"l.sary. .• t We have always observe\\ ,1 cerlain asymmclry in the ,p scans which we presume to he.• Jue 10 the difference in posi1ion of the eccentrically mounlcJ uapillary with respect to the counter at ,p and ,p+ l~O'-. In order lo account for this asym• mclry, we have calc..:ul.,tcd backgrounds using the expression: . ll<,(20,q,) = A••. ,,.<211>+ v.c211, [R~ -R~l F(,p,)- ,--<.;., where, for example, F(q,) = 11(x - :c0)1+b(y·· Yu)l +c'(x - xu) (y- Yo) : X ::-: sin rp, y::. cos <p, and x 11 , y 0 , a, h, and r arc rdinabk parameters. Alh.:r• natively, one could use ;111 interpolated ,p curve for flip). Howe ve r, we have ohtaincd the hcsl rf'sult s usins the sine fundion in equation (I). Fig. 5. View parallel to tile ,p Blllis of a capillary and crystal. l'lhowing the amount or direcl•hcum raJialion they inler• cept af vnrious If' sc1 lin gs. The capillary diameter is grcah:r lhan the r;1tliu~ of the X•ray he.am. ,·~ ·····lV\!\ i ~ l'. ,: / ~ \. '"jl.... 100 jr.· ,\ . ..~ ;.- . . io • • • • I .,j Fig. 6. C'urvcs A•o,x90 ant.I ll411 ,x90 • the background 20 dcpcnJenu observed at two ,p se ttings, Q() apart, for a capillary• mounted n11•.1rypsi n crystal. Inse t. u ,p sca n for this crystal taken at 20 -~ 6", sh,1wing the posil:ons of rhe fl'o and , 1 settings f I )iffraL·lomclcr :I). 0 39. M. KRIEGER, J. I.. l'HAMIIERS, G. G . CllRISTOPU, R. M. STROUD AND U. L. TRUS Figs. 7 and 8 show the 20 dependence or the background radiation for 1)1 P-lrypsin and cytochrome c crystals :it x=O" [(',,0 . , 0 (211)] and x, ')0'' IA, 0 .,,,0 (20J]. and the variation of the x dependence for several values or 20. Just as with tht· ,p dependence, the amplitude or the x variation depends upon the difference betwet·n the backgrounds at the c,trcmcs. x = 0 and x ~ 9() " . Depending upon the experimental conditions. it is 0 possible to observe a (.'.rossovcr of the A.,u. ,Q0 (20) and 745 c.,,. ,0(21/) curves analogous to that for the 'fJ curves. In such cases, the high-angle background increases as x goes from 0 to 90", and the shape or the curve t'an vary from that observed at low 211 (Fig. X). The x-depcndcnt background variation can represent a significant fraction of the total background. We have routinely observed 15 % differences in backgrounds between x = O and x = 90 ", and differences as high as 25% arc not uncommon. The following interpolation scheme, analogous lo that used for ,p, was used 10 correct for the x dependence: (I) Measure the 2// dependence, C,0 .,0 (20), at x = 0 • (2) Measure the 211 dependence, A, 0 ., 90(2/1), at x=90". (3) Tabulate the differences, l>,(211) = C, 0 _,.(2U) A. 0 .,.,,poJ. (4) Measure the x dependence at a 2// = 21/m for which the difference, D,(211), is large. If possible, 20m should be representative of the W's in the data set. (5) Cilculalc the backgrounds from: 8(' 211 ,( 21 ,X)= A ••. ,.,.( I) t IJ 211 [ 1-'(X) - F('JO'J] ,( ) F(O'') -- l-'(90") 21/m (Ill) where F(x) is either an empirical function representing the x dependence or an interpol,ited x curve at 2/lm. Our best results have been obtained with: Fig. 7. Curves A._, 0 ,x90 ;tnd C.,0 .x0 : the hackground 20 dependence ohservcd al l.. .:. 90 and x:.: ()" on <liffractomcter A for a DI P-1rypsin cry:-.lal mounted in a I mm capillary. f..adt dala point was counted for 4 min. Inset: lhc x dcJ)enJcnl·c observed al scvc..·rnl 20 values. Each poinl W<.tS mcasurc-.1 for S min. ,oo F(x)=uexp{ -- h/(cosx+c)J BG(W,91,x)= A... , ..(211)+ D.(211) sin' (q,,-l(Jo) F( x)-- F(90")] + D,(20) [ F(O") - F(90") 20•• § 30 (IV) where the coeflicients o,b,c arc determined by a leastsquares lit. If the background shown in Fig. 7. for example, is not corrected for x dependence, i.e., if isotropic interpolated backgrounds (depending only on 211) arc used. systematic errors of up to 30% of the background can result. x-dependent (anisotropic) interpolation can be improved upon by subdividing the data into smaller 211 ranges and applying appropriate x curves lo the data in each range. When both 'fJ and x dependences arc present, backgrounds may be approximated by combining the 'fJ and x corrections: (V) 0 u 200 .. 100 , J. _ I. .. 0 20 30 . ~ . .... - ·' 40 28 Fig. 8. Curves A,.0 ,x~ and C.,0 .x0 : the background 20 dependence observed al x = 90 and x=O" for a cytochrome c crystal using diffrac1omcter B. lnscl: background x dcpenllcncc at several 20 values. All points wen:: measured for 4 min. Non-linear least squares A diftercnt approach to the estimation of backgrounds is the use of non-linear least squares to approximate backgrounds as some function of 20, ({J, and X· This function is determined by titting backgrounds measured for short times in the vicinity of each reflection. The reliahility of such a procedure depends on the choice of a well behaved function with a relatively small number or parameters. Its accuracy also uepcnds on the number of data and time spent 011 measuring the individual backgrounds. 40. 746 DATA COLLECTION IN PROTEIN CRYSTALLOGRAPHY The dependence of background intensity on 211 may be satisfactorily accounted for by a third-order function of 2//: BG(2//) , ath'/'tcT'+dT', (VI) Table 2. Error analysis of backgrm11ul-a1>woxi111atio11 ,,.,chniqut:.\' where T=W and the lower case letters a,b,c, <'IC., are Method ( ,1HG')"'t .vt Rau%§ GOF• R, 1% .. II <, 1·7 ~5·3 10-0 2·22 1·76 Al 41 ·2 30·8 1.-11 1·38 0·88 LS 32 -J 17·1 5·3 1·19 0·87 0 Estimated ha1:kgrounds comparc:d lo ba..:kgrnunds measured for 40 s per poinl. • parameters to bt! dckrmincd. Some improvement can be achieved when the data arc separated into low, middle, and high-angle regions in 21/. Olher terms arc added to BG(2//) to account for the variation of BG (2//) as functions of 1/J and X· In any region of re,·iprocal space (not necessarily over all 211 ranges), it is sufficient to approximate the changing magnitude and sign of the <p correction with a third-order polynomial: l.lB,(,p,21/)=sin' (lf'-,Po) (e+fT+gT2 +hT3 ) (VII) where 1p0 is a parameter corresponding to the 'I' value for minimum or maximum background. If the background docs not have a 91 dependence, AB,(q,,20) is zero. Since the magnitude, curvature, and sometimes the sign of the x correction arc also 21/-depcndent, we have tried various combinations of terms of the form T'x'; i= I to 4;j= I to 4 to dekrmine which terms arc useful to best correct for the x dependence. An expression of the form AB,(x,W)=iTx+JTx' ➔ kTx'+IT'x'+mT'x (VIII) provides the best compromise between the number of refined parameters and the overall quality of fit. The overall expression for least-squares retinement is: BG(2/J,,p,X)=BG(2//)+AB1(,p,2/l)+AB2(X,20). (IX) This function can be used in any 211 range. However, BG(2/I) and A B,(x.21/) can often be simplilied for middle and high 211 ranges with essentially no loss of accuracy: for middle or high 20, BG(20) =a+ bT suffices; for the middle range, tlB,(x,21/)=iTx+jTx'+ kTx'+IT'x'; and for the high range AB,(x.W)=iTx+ • Results arc for 1956 rcllcctions measured 1.ln diffractometer (A) as dcscrihc\t in the le.1tt. The awragc observed background was 501 counls per min. and the rms "(IJG.,t,,) was 27·4 counts per min. The comparisons are all tabulated in ~otmts per min. t ~BG= HGoba ·• JiO~•lc• t s is the error component due to errors arising from the II, Al or LS methods, s' = (llBG')- (u;,,). § Rou=2: lllBGI/ 2: BG.M. • GOF= tllBG/a,,,.)' ]'", the ·goodness-of-fit'. [·I 1956 pts. •• R,· = ~ IFn<iob, - · P1u;u1,.I/ ~ ll-'1tll1>btl• mal distribution. For the II method the distribution of error is far from normal because of the systematic error introduced by ignoring the anisotropy in the background. The error distribution is skewed and distorted regardless of the setting angles used for observing the 20 curve (see Fig. IO). The error distributions for the Al and LS methods, on the other hand, are very nearly Gaussian,• indi• We have shown this in two ways: Jeasl•squarcs fitting a Gaussian to lhe error frc(luency curves in Fig. 9. and by direct compurison to perfectly normal distributions by means of nor• mal probability plots (Abrahams & Kcvc, 196K). Oislribuhon 0 jTx'+kTx'. Discussion The anisotropic interpolation (Al) and non-linear least-squares (LS) techniques have been used routinely to compare and measure buckground intensities for crystals of several proteins on different diffractometers. Data for the comparison to be discussed !irst were taken from a DIP-trypsin crystal using diffractometer A. The data shown in Table 2 and Fig. 9 compare the isotropic interpolation (II), Al and LS methods of estimating background intensity with individual backgrounds, BG 0 . , , each measured for 40 s. Fig. 9 shows the distribution of errors, ABG/110 . , , where ABG=(BG ••, - BG,.,J and a•.,,= Vea ••• ' for each of the three methods and for a perfectly nor- Il· -;: _....::..:.~;.. -6 .5 -4 .\ 0 ~BG/a 0 t, 1 Fig. 9. The distrihulion of errors between imlividuully uhserved (40 s) backgrounds and backgrounds calculated using the three lcchniques described in the text. The statistics plotted are the numbers of points within intervals of .d KG/a.it.. where .dB<J = BG.,"- 8O,::.ic anJ n.,,,,= VBG.,1t~• The total area under each curve is equivalent to 1956 dala points. The curves arc frequency distributions for the errors of the II, or isotropic intcrpola1ed backgrounds, the Al, or anisotropic interpolated backgrounds, and the LS, or least-squares back• grounds. The heavy curve represents an ideal Gaussian distribution of error for the same nu111her of points. (DIPtryr"in crystal on dilfraclometcr A). • 41. M. KRIEGER, J. L. CHAMBERS, G . G . CHRISTOPH, R. M . STROUD AND B. L. TRUS eating that the systematic errors resulting from the JI method have been largely eliminated. The slightly increased breadth of the Al and LS error distributions relative to the perfectly normal distribution is a consequence of the errors introduced hy these methods. We can estimate the extent of these errors,"• using the approximation: .,, ce (ABG') •• (11 2••,) (X) 747 LS corrected · data. Applying absorption corrections to the scan intensities be/or<' subtracting II backgrounds, as recommended by Hill & Banas:w k, increases the R, value to 1·95 %. Clearly the Al and LS methods are substantial improvements over the II method . Conclusion The most significant conclusions to be drawn from where (ADG') is the mean squared .1DG and (u!.. ) these studies of background iniensity are: is the mean squared error in the individually observed (I) The principal 20-dependent component of backbackgrounds. For the data illustrated in Fig. 9 the ground intensity arises from X-rays scattered by the estimated averngc error for backgrounds calculated volume of air illuminated by the direct beam and 'seen' using the Al technique, with empirical background by the receiving-counter aperture. Modulation of this curves observed for 4 min per point, is roughly equiv- air scatter by capillary absorption and scatter gives alent to that whic:h would be obtained by measuring rise to 'I' and x-dependent variations in the background_. each background for 32 s. For LS, s is equivalent to (2) The background intensity is essentially uncorerror expected for backgrounds measured for 91 s each. related with the crystal absorption. Under certain exThe value of -' for Al depends upon the time spent. perimental conditions the two· phenomena may appear measuring each point for the interpolated curves, while to be correlated; for example, where a flat crystal rests that for LS depends on the counting times for indivi- on the side of the capillary such that its long axis lies dually measured backgrounds. In this example, indivi- parallel to the 'I' axis of the diffractometer. dual 40 s backgrounds were used to calculate the LS (3) As a consequence of (2), the background intenfunction. In another case results were essentially iden- sity should always be subtracted from the scan intentical when 10 s backgrounds were. used, and this sity before the absorption corrections are applied, method is clearly most useful when still shorter times contrary to the suggestion of Hill & Banaszak ( 1973). are used. (4) The anisotropy of background scatter as a funcWe have compared the effects of the II, Al and LS tion of 20, 'I', and x can be satisfactorily accounted for methods on the structure factors by calculating the by using a simple interpolative procedure, or hy fitting R values (R,) between the data sets obtained using a suitable function to many backgrounds measured for observed and calculated backgrounds (sec Ta hie I). short times near each reflection. The R,- value for the II' corrected data is approxi( 5) The use of an isotropic background interpolation mately twice as large as that calculated for the Al and is usually unsatisfoctory for accurate data reduction . This method can introduce systematic errors of up to JO% in the estimated background intensities. Errors of this sort lead to systematic errors in phases and increase the noise levels in difference maps. Doth the Al and LS methods substantially redu,e the systematic errors inherent in the II method. As a result, the differences between observed backgrounds and estimated backgrounds closely follow II normal 0 distribution. Both methods represent significant savI ings in time and are improvements over the II methot.l. I The overall accuracy of either method is limited by the extent and duration of the background sampling, and is generally commensurate with that obtained for indi/ vidual hackgrounds. When deciding between the two methods, one must choose between the greater savings in time provided by the A I method, or the somewhat better accuracy of the LS technique at the cost of Fig. 10. The distribulion of errors when diff'en:nt empirical 20 increased crystal exposure and decay. curves arc used in lhc II mclho,t. The frequency distributions Our results identify the principal sources of backare for hack.grounds calculated using: a 20 curve mc.isurcd ground intensity. Consequently, there are several exat 9' = tp,,, x = O" ( •·), a 20 curve measured at ,;"~'" perimental steps which can be taken to reduce hadx -• 90'' ( · - ), a 20 i.:urvc measured at r,=q,o, x=90" ground intensity : (I) Capillaries should ideally he cho( - - ) ; and a 20 curvc calculated from the ¥J1oX,o and 'Pn, Xw curves. representing ,, 2() curve which would be obtained if sen lo be smaller in diameter than the radius or the rp~ i(cp11 + 91.) (). The light l'.urve ( • --·· l rcpr<:scnts illl incid,·nt X-ruy beam. This renders 'I' dependen,·c inideal <iaussian llisll ihution . of error. As in Fig. 9, the area signilic:anl, diminishes x dependence or the background un1.lcr each curve rcp11,."SCnh 1956 data points distrihulcd and reduces the high-1111gle background due 10 capillary ,ivcr the rangL' 4 ·6 ". 20 -: 4() '. I I 42. 748 DATA COLLECTION IN PROTEIN CRYSTALLOGRAPHY scatter. (2) Helium-filled tubes are often used to reduce absorption by air in the input or output beam ,Jirections. However, as the main problem is generally one of peak-to-background ratio, these measures only slighlly affect this ratio, and in any case do not reduce background intensity significantly. II would seem highly advantageous In fill the 'visible volume' with helium, so affecting the hackgrnund intensity directly. This is obviously diflicull lo do as it implies either a helium-filled enclosure over tlw entire diffractometer, or a helium-lilied chamber mounted around the ,p axis which surrounds the capillary and ery,tal completely. There arc ohvious mechanical difficulties in huilding such a device. l'irst, it must be almost X-ray transparent over the angular ranges used. Second, ii must be moderately well scaled to minimize leakage of helium if it encloses the ,p drive shaft hearing. /\ mylar cylinder with solid supports mounted 0111<> the lop of the go niomcll"r head would seem to he ..i good t:nm~ promise· allowing for rigid support of the top and hottom in the X-ray shadow. (J) Any means of rc,1ricting the visible volume will reduce the hackgrnund in almost uircct proportion to the volume change. This can be achieved by placing the final restricting aperture and the scalier cap on the input collimator as clo,c lo the crystal as possible. Similarly, there should he a defining apcrlure as close I(> the crysta l as possible in the crystal· counter pathway, and a SCl:ond one dos(' to the counter. We thank Drs /\. KossiakofT and R . Swanson for generously supplying background data, Mr R . Alma"y for helpful assistance, and Dr R. E. Marsh and Mr J. Greif for valuable discussions. We are grateful to Dr Sten Samson for allowing us to collect data on diffractometer F, a machine designed in the most par! by himself, and for his supervision in the redesign of diffractometer B. We also recognize his constant attention in the rcuesign and improvements made to ditfractomeler C. References AURAHAMS, S. C. & K•v•, E.T. (1971). A,·ta Cryst. A27, 157-165. DICKERSON, R. E., KOl'KA, M. L., VARNUM, J. & WEINZIERl., J.E. (1967). A,·ta Cry.'1. 23, 511 --522. • HEN0ERSON, R. & Mo>eAn, J. K. (1971). Acta Cryst. 827, 1414--1420. lilLl., E. J. & BANAS7.AK, L. J. (1973). Acta Cry.<t. 829, 372. JENSEN, L. (1972). Personal communication. KRmo••• M., KAY, L. M. & Snoun, R . M. (1974). J . Mui. Biol. 113, 209 230. MArnews, F. s., Lev1N•, M. & ARoos, r. (1972). J. /\,fol. Biol. 64, 449--464. SAt.EMME, F. R .. l'R1 •·• • S. T., XuoNO, Ng. H., Al.llEN, R. & KRAUI, J. (1973). J. Biol. Ch,m. 248, 391~3921. WATSON, H. C., SHmTON, D. M., Cox, J. M. & Mun,HEAU, H. (1970), Nature, L,md. 225, 806-811. WvcKOff, II . w., DOSCHER, M., TSERN<Xll.</U, D., INA· 0AMI, T., J0HNS<>N, L. N., HARDMAN, K. D., ALLEWEl.l., N. M. , K,i.1.v, 0 . M. & RICHARDS, F. M. (1967). J. Mo/. Biol. 27, 563 578. WVCKOIT, II . w., TSERNOOLOII, D., HANSON, A. W., KNOX, J. R., L>.IC, II. & RICHARl>S, F. M. (1970). J. Biol. Clt,m. 2-IS, ]05 .. J28. 43 . Appendix 3 Structure-Function Relationships in the Serine Proteases 44 . Structure-Function Relationships in the Serine Proteases Robert M. Stroud, Monty Krieger, Roger E. Koeppe II, Anthony A. Kossiakoff and John L. Chambers Norman W. Church Laboratory of Chemical Biology California Institute of Technology, Pasadena, California 91109 Of the many ways available to control the biological activity of proteins, e.g., induction or repression of their synthesis at the translational (Jacob and Monad 1961 ) or transcriptional levels (Tomkins et al. I 969) , specific modification or destruction are the most direct. Many biological systems are controlled by methods such as these, and the serine protease family of enzymes plays a major role in many of these systems (Stroud 1974). The pancreatic serine proteases are digestive enzymes which show optimal activity around the neutral pH region. Their function in hydrolyzing peptide bonds and the systems of physiological control over their activity have close homology in many other biological processes, e.g. , blood clotting (Owren and Stormorken 1973; Magnusson 1971), bacterial sporulation (Leighton et al. 1973), fertilization ( Stambaugh, Brackett and Mastroianni 1969), etc. Many of the enzymes of biological control have been recognized as serine proteases, which in nearly all documented cases have amino acid sequence homology to the pancreatic serine proteases. It is therefore to be expected that these enzymes will have closely homologous tertiary structures and will share the same catalytic mechanism of action. The mechanisms by which such enzymes are activated or inhibited will also share many common features with the digestive serine proteases. In many cases the degrees to which these principles can be extended may be predicted by recognition of the chemical and structural features of the pancreatic serine proteases which appear to define their properties. In this article we will discuss recent advances in the understanding of aspects of the structures and functions of the mammalian serine proteases. THE PANCREATIC DIGESTIVE ENZYMES: TRYPSIN, CHYMOTRYPSIN, ELASTASE Intrinsic to the process of digestion in mammals is the breakdown of dietary protein by the pancreatic serine proteases. These pancreatic digestive enzymes 45. are among the most thoroughly studied of all enzymes, principally because they are extracellular enzymes that are easily separated and purified in large quantities (Kunitz and Northrop 1935). They originate in the pancreas as inactive precursors, or proenzymes, which are secreted into the duodenum. There they are activated (Kunitz and Northrop 1936; Northrop, Kunitz and Herriot 1948; Maroux, Baratti and Desnuelle 1971) by the cleavage of one critical peptide bond near the amino-terminal end of the polypeptide chain (Davie and Neurath 1955). This cleavage in turn permits a conformational change (Neurath and Dixon 1957; Sigler et al. 1968) which converts the proenzymes to active enzymes. Once activated, these enzymes catalyze the breakdown of proteins, first into fragments and ultimately into individual amino acids. Kinetic studies on a variety of amide and ester substrates have shown that the mechanism of serine protease catalysis (Eq. 1) involves a number of intermediates (Zerner, Bond and Bender 1964; Oppenheimer, Labouesse and Hess 1966; Cap low 1969; Hess et al. 1970; Fersht and Requena 1971 a; Fastrez and Fersht l 973a,b; Fersht and Renard 1974): E + S .= ES .= EST .= EP~ .= EP~' .= EP 2 .= E + P 2 • (1) + pl Here, E represents free enzyme; S, the substrate; ES and EP 2 , enzyme substrate and product complexes; EST and EP;, tetrahedral intermediates; EP:, an acyl enzyme; Pi, the amino or alcohol portion of the product; and P 2 , the carboxylic acid portion of the product. For amides, the rate-determining step is generally acylation, E + S - EP~, whereas deacylation, EP; - E + P 2 , is usually rate-determining for esters (Zerner and Bender 1964). The characteristic differences between each of the digestive serine proteases-trypsin, chymotrypsin and elastase-lie in their specificity for hydrolyzing the peptide bonds between different amino acids in the protein substrate. Trypsin, the most sharply specific of the digestive enzymes, hydrolyzes those peptide bonds that immediately follow either of the two basic amino acids, lysine or arginine. Chymotrypsin hydrolyzes peptide bonds that follow several of the amino acids with larger hydrophobic side chains, and elastase binds the small side chains of glycine, alanine or serine at the equivalent binding site (Naughton and Sanger 1961; Brown, Kauffman and Hartley 1967; Sampath Narayanan and Anwar 1969). The complete amino acid sequences and three-dimensional molecular structures have now been worked out for chymotrypsin (Sigler et al. 1968) and its proenzyme (Freer et al. 1970), elastase (Shotton and Watson 1970), and DIP (diisopropylphosphoryl)-trypsin (Stroud, Kay and Dickerson 1971, 1974) and the proenzyme (Kossiakoff, Kay and Stroud, unpubl.). These structures, along with that of the bacterial serine protease subtilisin (Wright, Alden and Kraut 1969; Alden, Wright and Kraut 1970), have been uniquely valuable in developing an understanding of how these enzymes bind a substrate and how they catalyze the subsequent chemical reaction. 46. Activating the Proenzyme The first key to activation of the pancreatic proenzymes is enterokinase, an enzyme secreted in small amounts by the mucous membrane of the stomach. Its prime function is to convert some trypsinogen to trypsin, which then activates all of the proenzymes (including more trypsinogen) (Kunitz and Northrop 1936; Northrop, Kunitz and Herriot 1948; Maroux, Baratti and Desnuelle 1971). In each case, activation involves the cleavage of a few amino acid residues from the amino-terminal end of the proenzyme (Davie andNeurath 1955;NeurathandDixon 1957). With the formation of the new amino terminus at Ile-16 1 (Oppenheimer, Labouesse and Hess 1966), the protein undergoes conformational changes (Neurath, Rupley and Dreyer 1956) leading to a catalytically active configuration. A comparison of the high-resolution structure of chymotrypsinogen with that ofchymotrypsin (Freer et al. 1970; Wright 1973) and of the highresolution structure of DIP-trypsin with the recently determined high-resolution structure of trypsinogen (Kossiakoff, Kay and Stroud, unpubl.) helps us to understand the exact nature of these conformational changes. (A detailed description of the trypsinogin structure will be published later.) In both cases, Figure 1 A comparison of the 5-A models of trypsinogen (left) and DIP-trypsin (right) shows close structural homology in most areas of the molecule. Striking structural differences are observed in only two areas of the molecule. The first is in the. binding pocket region, which is formed by residues 214-220 and I 89-192, and the second is along a loop of chain containing residues 140-151 located on the righthand side of the molecule. The small difference in size of the models is due to a difference in scale and orientation. 1 The numbering system referred lo is that of chymolrypsin, which will be adopted here as a standard for comparison of sc4ucnccs. 47. clipping the proenzyme tail permits the new, positively charged a-amino terminus at residue 16 to fold into the interior of the globular structure and form an iorl pair with the negative carboxyl group of Asp-194 (Matthews et al. 1967; Sigler et al. 1968). While this change is accompanied by movements in the region of the specificity binding pocket, there appears to be little change in the interaction between Asp-I 02, His-57 and Ser-195 at the catalytic site (Freer et al. 1970; see also Fig. I) . Thus the arrangement of these catalytic residues is preformed in the proenzyme. One major factor which contributes to the relative inactivity of the zymogen is that the binding of the normal substrates is impaired ( Kassell and Kay 1973; Gertler, Walsh and Neurath 1974) .. Enzyme Specificity and Substrate Binding The serine proteases differ in their specificities because of differences in their substrate binciing sites. Trypsin, chymotrypsin and elastase all have specific side~chain binding pockets on the surface of the protein close to the catalytic site (see Fig. 2). This pocket is lined by residues 214-220 and 189-192 and Figure 2 Stereoscopic photograph of a space-filling model showing the active site and specificity pocket in trypsin ( see Stroud, Kay and Dickerson 1971, 197 4). The imidazole side chain of His-57 is visible ; however, the carboxylic acid side chain of Asp-102 is hidden from view by several amino acid residues. The viewing direction is approximately the same as that of Figures 4 and 7. The side-chain binding pocket is located beneath and to the right of the catalytic site. Stars are placed near the active site as markers. They are located (reading from left to right) at the following positions: the Asp-102/ His-57 hydrogen bond, the ,B-carbon protons of His-57 , one of the ring protons on His-57, and the Ser-195 y-hydroxyl. 48. defines the primary specificity toward substrate side chains immediately prior to the peptide bond which is to be cleaved. Cysteine residues 220 and 191 are linked by a disulfide bond. In trypsin, residue 189 is an aspartic acid, and its negatively charged carboxyl group ( pK 11 = 4.6) (East and Trowbridge 1968) lies at the bottom of the pocket (Stroud, Kay and Dickerson 1974) . Trypsin has primary specificity for basic amino acids because their positively charged side chains bind tightly in this pocket ( Mares-Guia and Shaw 1965; Ruhlmann et al. 1973; Blow, Janin and Sweet 1974; Sweet et al. 1974; Krieger, Kay and Stroud 1974). In an attempt to determine the manner in which amino acid side chains bind, we determined the structure of benzamidine trypsin. Benzamidine is a competitive, specific and reversible inhibitor of trypsin. Figure 3 shows how benzamidine, an amino acid side chain analog, binds in the specificity binding pocket ( Krieger, Kay and Stroud 1974). In the case of trypsin, there is evidence that when side chains are bound in this pocket, they induce small conformational changes in the enzyme-substrate complex which help to accelerate catalysis (lnagami and Murachi 1964; Inagami and York 1968). In chymotrypsin, residue 189 is a serine (Hartley 1964). The pocket is now relatively hydrophobic and uncharged at neutral pH's, thus explaining chymotrypsin's specificity. In both trypsin and chymotrypsin, residue 216, a glycine, lies at the entrance to the binding pocket. In elastase, valine replaces glycine at position 216 ( Shotton and Hartley 1970). The larger hydrophobic side chain blocks the entrance to the pocket and only allows the binding of amino acids with small side chains at the primary binding site (Shotton and Watson 1970). Other parts of the enzyme are involved in binding other parts of the substrate molecule as well as the side chain, so that the susceptible substrate bond is aligned appropriately on the surface. Secondary specificity toward other Figure 3 The figure shows the structure of the trypsin binding pocket in benzamidinetrypsin. (The phenyl amidinium is depicted by heavy shading.) This view is approximately the same as that of Figure 2. (Reprinted, with permission, from Krieger, Kay and Stroud 1974. ) 49. side chains in the physiological substrate molecules can also be correlated with enzyme structure (Fersht, Blow and Fastrez 1973), although .the role of secondary specificity in the digestive enzymes is clearly much less significant than it is in highly specific enzymes of biological control. Ideally, one would like to study the three-dimensional structure of an enzyme-substrate complex by X-ray crystallography and in so doing, gain new insights into the mechanism of serine protease catalysis. Unfortunately, this has not yet been possible because the catalyzed reaction takes place almost immediately after the substrate is bound and the system becomes an enzymeplus-product complex. Data for a three-dimensional structure analysis cannot generally be collected in so short a time. Fersht and Renard (1974) have pointed out, however, that it may be possible to use equilibrium methods to trap intermediates in the reaction pathway and study their structures. Until the structures of such intermediates have been determined, crystallographers will be limited to studying the binding of substrate analogs and inhibitors, which in some limited respects resemble true substrates. Nevertheless, from such studies inferences can be drawn about the structural transformations which occur during the catalyzed reaction. Among the best analogs to true trypsin substrates are the naturally occurring trypsin inhibitors. They have evolved in parallel with the enzymes so that they bind extremely tightly to the active site. Such protein inhibitors are crucial to physiological control of the serine proteases (Tschesche 197 4). For example, if pancreatic secretory trypsin inhibitors were not synthesized along with the pancreatic serine proenzymes, one prematurely activated molecule of trypsin could start an autocatalytic chain reaction which would activate the other serine proenzymes and destroy any nearby proteins. Inhibitors are present to prevent such catastrophes and to control physiological processes mediated by proteolytic enzymes. The structure of an intracellular 6500 molecular weight trypsin inhibitor ( PTI) isolated from bovine pancreas and other organs was determined by R. Huber et al. ( 1970, 1971). Chemical modifications had already shown that Lys-15 of this inhibitor was involved in the trypsin-PT! association (Chauvet and Acher 1967; Kress and Laskowski 1967 ; Fritz et al. 1969). By combining models of PTI with the known structures of trypsin and chymotrypsin, substrate binding models (Fig. 4) were developed by us (Stroud, Kay and Dickerson 1971; Krieger, Kay and Stroud 1974) and independently by Huber et al. ( 1971 ) and Blow et al. ( 1972). High-resolution structures of the PTl-trypsin complex ( Ruhlmann et al. 1973) and of a soybean trypsin inhibitor-trypsin complex (Blow, Janin and Sweet 1974; Sweet et al. 1974) have since been determined. Several hydrogen bonds and stereochemical complementarity between enzyme and inhibitor orient the susceptible bond at the active site. There is one very important difference between the predicted substrate binding models and the structures of the trypsin-inhibitor complexes; the inhibitor and trypsin are covalently bound together via an oxygen-carbon bond between the hydroxyl of Ser-195 and the carbonyl group of Lys-15 in the trypsin-PT! complex and between the serine hydroxyl and the carbonyl carbon of Arg-63 in the trypsin-ST! (soybean trypsin inhibitor) complex. These complexes have been shown to exist as tetrahedral adducts which probably resemble normal intermediates (EST) in serine protease catalysis. 50. Figure 4 (Top) A model for the binding of a portion of the bovine pancreatic trypsin inhibitor (heavy shading) to trypsin in which the side chain of the inhibitor's Lys-15 has been replaced by an arginine side chain. The model was constructed so that the "Arg-15" side chain fitted the electron density for benazmidine in benzamidine-trypsin. This model for enzyme-substrate interaction embodies a substrate conformation that evolved to bind tightly to the enzyme and an enzyme conformation which is presumably like that induced by the binding of specific substrate side chains. ( Reprinted, with permission, from Krieger, Kay and Stroud 1974.) (Bottom) Stereoscopic view of a space-filling model of the trypsin-substrate complex described in the text. This view (approximately as above) should be compared with the same region of the enzyme alone shown in Figure 2. This figure beautifully demonstrates the intimate stereochemical compatibility between enzyme and substrate. The woolen threads indicate the end points of the portion of the oligopeptide substrate chain which is included in the model. Protons in the substrate are labeled with square dots, and atoms attached to the a-carbon atoms of the substrate fragment are labeled 4-3-2-1-1 '-2' ( along the sequence in the substrate). The peptide bond between 1 and 1' is the one to be hydrolyzed. 51. These and other crystallographic and chemical studies have produced detailed models for the association of enzyme and substrate prior to catalysis. Although based on inferences drawn from substrate analogs or inhibitor binding studies, such models do suggest ways that chemical groups on the enzyme may participate in catalysis. The Active Center: pH-Activity Correlations The X-ray crystal structures of the serine proteases have shown that their active sites are almost identical. The catalytic site of all serine proteases is characterized by a serine hydroxyl group (residue 195). Diisopropylfluorophosphate (DFP) (Jansen, Nutting and Balls 1949; Hartley 1960) and phenyl methane sulfonyl fluoride (PMSF) (Fahrney and Gold 1963; Kallos and Rizok 1964) react with this hydroxyl and irreversibly inhibit serine proteases, regardless of their substrate specificity. In the free enzyme, this hydroxyl is hydrogen bonded to the N-e~ of His-57. N-8 1 of His-57 is hydrogen bonded to the carboxyl group of Asp- I 02, which in turn is not directly accessible to solvent (Blow, Birktoft and Hartley 1969; Wright, Alden and Kraut 1969; Alden, Wright and Kraut 1970; Birktoft and Blow 1972; Stroud, Kay and Dickerson 197 4). The direct participation of these three groups in catalysis has been established, and chemical modification of any of them can greatly diminish or abolish catalysis (Jansen, Nutting and Balls 1949; Hartley 1960; Fahrney and Gold 1963; Kallos and Rizok 1964; Ong, Shaw and Schoellmann 1964; Shaw, Mares-Guia and Cohen 1965; Henderson 1971; Martinek, Savin and Berezin 1971; Chambers et al. 1974). The pH-activity profiles for hydrolysis of peptides, amides or esters by trypsin or chymotrypsin are bell shaped and reflect maximal enzymatic activity at about pH 8. The high pH limb of the curve depends on an apparent pK. of 8.8 for a-chymotrypsin (Fersht and Requena 1971b) or 10.1 for trypsin (Spomer and Wootton 1971). Fersht and Requena have demonstrated that this ionization, which controls enzyme conformation and substrate binding (Km), is directly associated with titration of the a-amino terminus of Ile-16. The internal salt bridge formed between this amino group and the side chain of Asp-194 in the active enzyme is broken at high pH, where deprotonation of the amino group was shown to favor an alternate conformation for the enzyme. The low pH limb of the profile depends on a single group of pK 0 about 6.7 in both enzymes; protonation of this group adversely affects both acylation and deacylation (Bender and Kezdy 1964) . It has often been assumed that this group corresponds to His-57. Jencks (1969) , however, has pointed out that this group need not be His-57, but might be some other group on the enzyme either controlling conformation or effecting a change in rate-determining step near this pH. The aspartic-histidine-serine system as a whole has been shown to take up a single proton as the pH is lowered below 7.0. Fersht and Renard ( 197 4) have also demonstrated that for the hydrolysis of acetyl phenylalaninep-nitrophenyl ester by 8-chymotrypsin, k,.ar or kcarl Km depends on a single ionization between pH 9.0 and pH 2.0. Thus it would seem that only one group at the active center has a pK,. in the range pH 2-9 which can be detected kinetically. Richards and his colleagues (Hunkapiller et al. 1973) have shown by 52. nuclear magnetic resonance studies that the imidazole of His-57 in a-lytic protease (a bacterial homolog of the pancreatic serine proteases) does not ionize until the pH is lowered below 4.0. Their research led them to propose that the group of pK 7 .0 was Asp- I 02. In order to study the ionization of Asp- I 02 in trypsin ( Koeppe and Stroud, unpubl. ), we monitored infrared absorbance arising from the carboxyl C=O stretch at 15 70 cm - 1 and 1710 cm -- 1 (Timasheff and Rupley 1972) as a function of pH. To diminish the number of titratable carboxyls, accessible carboxyl groups in trypsin were modified with semicarbazide (Fersht and Sperling 1973). The modified enzyme was found to contain 2.5 molar equivalents of free carboxyl groups. These were identified as Asp-102, Asp-194 ( 1.0 molar equivalent each) and the a-carboxyl terminus of Asn-245 (0.5 equivalent). The spectrum shown in Figure 5 indicates differential absorbance at 1600 cm - 1 and at 1680 cm- 1 between semicarbazide-trypsin solutions of pD 6.50 and 7 .13. Each peak is shifted toward the other by about 30 cm - 1 from the value found for other carboxyls-a result which is to be expected for hydrogen-bonded carboxyls ( Susi 1972) such as Asp- I 02. The titration .c urve of semicarbazide-trypsin compiled from a series of infrared difference spectra is shown in Figure 6. Based on the number of free carboxyl groups, we assume that the low pH titration of average pK 11 2.9 corresponds to titration of 1.5 carboxyl groups, while the titration of average pK11 6.8 corresponds to one carboxyl. The gradient of the low pH titration is approximately 1.5 times that of the upper one, which is consistent with the assumption. Both titrations, however, appear sharper than expected for single or noninteracting groups. Binding of Cu ++ ions displaces the upper limb of the titration downward from pH 6.8. Martinek et al. ( 1969, 1971) have shown that Ag+ ion:=; are powerful competitive inhibitors of trypsin, and that Cu + + and Ag+ compete with each other in inhibiting chymotrypsin. Either species competes with pro11 Figure 5 Infrared difference spectrum for semicarbazide-trypsin. The path length was 0.150 mm. The sample cell was at pD 7.13, and the reference cell at pD 6.50. pD values in all cases correspond to uncorrected pH meter readings. Concentrations were 1.5 mM enzyme, 6 mM NaNO 3 , and 12 mM benzamidine. The peak positions at 1680 cm- 1 for C= 0 in COOD and at 1600 cm - 1 for C-O in COO- are closer together than for other trypsin carboxyls due to the effect of hydrogen bonding. The detected difference in species concentration judged from peak heights corresponds to about 0.4 carboxyl equivalent. 100 tsz % .. ..:. . . . ; ll . . .. ~•D_•~::3-G~;J <X f- i 80 1:11 I.I) z :: :1 <X a:: f- ~ w 1--1 60 u 1---4 0 '.'.!!!!!~! !iiiii!i a:: w .......... ... .. Cl. 40 0 0 CX) ~ cm-• ~ 53. 0.00 -0-......:,_1 'o' I I I 0.02 I ©\ \® 1710 cm.-• \ \ I' '',, \ © corboxyl ',\, I I I I c:i o 0.04 <l <?........ 0 - 0 - - 0 ' \ t Errors --....:0 0 ,680,m' \ 0.06 pD • 2.9 t\ I 0.08 0 p0•6.8 I I 2 3 4 pO 5 6 7 8 Figure 6 Titration curve for free carboxyls in semicarbazidetrypsin obtained by plotting differential absorbance at 1710 cm - 1 and 1680 cm - 1 from a series of infrared difference spectra. Reference solutions were at pD 3.2, 3.8, 4.4 and 6.6. The gradient of the titration at low pD was 1.5-2.0 times greater than the gradient for the group of pKa 6.8 and corresponds to titration of about 1.5 carboxyls. tons for a site on the enzyme with pK 01, 1, '.:::'. 7.0. Since we have shown that Ag+ binds specifically between Asp-102 and His-57 in the orthorhombic crystal form of DIP-trypsin ( Chambers et al. 197 4 and Fig. 7), in trypsinogen and in trigonal DIP-trypsin ( Kossiakoff, Kay and Stroud, unpubl.), the pK. of 6.8 has tentatively been assigned to the carboxyl of Asp-102. (A detailed description of the experimental procedures and results will be published elsewhere.) Copper ion binding shows that Asp-194 and Asn-245 cannot be responsible for this pK". These data suggest that the average apparent pKa of Asp-194 and Asn-245 is 2.9. Control experiments eliminate the possibility that imidazole-stretching frequencies could account for the infrared bands at 1680 cm- 1 and 1600 cm- 1 in Figure 5. However, imidazole titration may perturb a neighboring carbonyl and thereby conceivably be responsible for the infrared difference peaks observed around pD 6.8. This possibility is the subject of continuing investigations in our laboratory. There are two arguments against this possibility and in support of the assignment of a pK,, of about 7 .0 to Asp-102. First, using an average extinction coefficient for the other seven carboxyl groups in ,8-trypsin, derived from infrared difference spectra of the unmodified enzyme, the low pH limb of the titration shown in Figure 6 corresponds to 1.5 carboxyl equivalents. This implies that the remaining carboxyl group must titrate outside of the range pH 2-pH 5. Second, in the presence of Cu+ + ions, there is no differential 54. Figure 7 A stereo ORTEP drawing of the catalytic site of silver DIP-trypsin, shown in the same orientation as Figure 1. The DIP group has been omitted for clarity. The y-oxygen of Ser-195 is in the position found for DIP-trypsin, close to that found in tetrahedral intermediates or in acyl enzymes (see text). (Reprinted, with permission, from Chambers et al. 1974.) absorbance in the region of pH 7 .0, which suggests that either of the two imidazoles at His-40 or His-91 which may titrate around pH 7.0 do not induce changes in their neighboring carbonyl groups which would be detected by the technique. The Mechanism of Hydrolysis by Serine Proteases Following their determination of the active-site structure of chymotrypsin, Blow, Birktoft and Hartley ( 1969) first proposed that proton transfers between His-57 and Asp-102 were important in catalysis. The studies of the microscopic pK.,'s of His-57 and Asp-102 referred to in the previous section are consistent with this proposal. From their studies of the histidine ionization in a-lytic protease, Hunkapiller et al. ( 1973) explained the sequence of proton transfers between Asp-102 and His-57, discussed here and included in Figure 9 (below), in terms of pK;s. In this discussion we assume ( see previous section) that the pKa of Asp-102 is 6.8, and that the imidazole of His-57 is essentially neutral above pH 4.0. This leads to the ionization of the active center around pH 7.0 (Fig. 8). The mechanistic importance of these assignments is that the aspartate ion of residue 102 can act as a chemical base which can readily accept a proton from the histidine side chain during catalysis (Hunkapiller et al. 1973). Together, Asp-102 and His-57 shuttle protons back and forth from enzyme to substrate, and so the mechanism can best be described as nucleophilic attack with general base catalysis by His-57 (Bender and Kezdy 1964; Inward and Jencks 1965) and Asp-102 . .The important differences between this reaction and a nonenzymatic hydrolysis are the binding to the enzyme and the efficient proton shuttle. Interfering with this shuttle inhibits catalysis. For example, by methylating the N-E~ of His-57 in chymotrypsin, the shuttle can no longer operate normally 55. Active form• above pH 6.7 Figure 8 Inactive below pH 6.7 Ionization of the active center in the range pH 4.0-8.0 as discussed in the text. and the rate of catalysis drops by a factor of 5000 to 200,000 for specific substrates (Henderson 1971). Silver ions bind specifically to trypsin between Asp-102 and His-57 (Chambers et al. 1974). By adding silver, the shuttle is blocked and catalysis is inhibited (Martinek, Savin and Berezin 1971). The mechanistic scheme shown in Figure 9 is consistent with most experimental data relating to hydrolysis of peptides, esters or amides by trypsin or chymotrypsin. In the first step (I), substrate and enzyme form a Michaelis complex. Nucleophilic attack by the hydroxyl group of Ser-195 follows. As the reaction proceeds, the hydroxyl twists around the C,.-C/l bond and forms a covalent chemical bond to the substrate carbon at step 1-11 (Steitz, Henderson and Blow 1969) . Concerted with this, a proton is transferred from the serine hydroxyl group to the N-e 2 of His-57. From there, it is eventually delivered to the nitrogen of the peptide bond in the substrate. As a result of this proton transfer, the proton previously bound to the N-6 1 of His-57 is transferred to the carboxyl group of Asp-102, which acts as a base in this reaction (Hunkapiller et al. 1973). Although these proton transfers are rapid, deuterium isotope effects show that proton transfer is involved in the rate-determining step of the catalysis (Bender et al. 1964; Pollock, Hogg and Schowen 1973). Whether the Asp-His-Ser proton shuttle is concerted or stepwise remains in question. If the mechanism is concerted, the negative charge of Asp-102 would be neutralized while negative charge develops on the carbonyl oxygen of the substrate. The imidazole would remain neutral throughout the reaction; thus unstable intermediates due to charge separation would be avoided (Jencks 1971; Hunkapiller et al. 1973). In fact, charge development in the transition state in chymotrypsin catalysis does appear to be small (Jencks 1971). The shuttle may be stepwise if the energy requirements of charge separation (negative charges on the substrate and Asp-102 and a positive charge on His-57) are offset by a more favorable entropy of activation in a two-step process (Jencks 1972). One might favor the concerted mechanism because it might be expected that the precise alignment of the shuttle, which has been observed in all serine protease structures, evolved so that the entropic advantage of the • 57 h{ ser 195 ~ m. Acyl enzyme H .JJ• ljl I -_..::., ~ - Jr. Tetrahedral ~ R, t7X intermediate H I n-..R1 intermediate H o /8_.wN II •• ;_N.11.. ~~-~-- - ~ J.o:: .. H-N -.:.,__O-H·· ~ · ~c, · n. Tetrahedral 0 ---..._o-w·· ~-R-N-- , // _.,N.,.. .... ~-!~7~;--°c<e ✓ 0·-..wNH-N 0 -_..::., _..::., N H e,P ~ / o=-"C 'R, / N·- · - · H - 0 ~ ~ :m. Enzyme Product complex o r D~ m. Acyl enzyme Ro 'H 0 8----H- N~ ~ leaving group / -<°8····H-~ 0 ~N A mechanism for serine protease hydrolysis of peptides or amides. In this representation, the proton shuttle is concerted. Figure 9 0 \\ I. ES complex H A~ -<e:'---H~~- 0 ~ ~ ~..9 N----H,,P _fl0&,---H-~~ R -N-C 0 0 I ~R osp 102 ) °' u, 57. two-step process over the concerted process was minimized. Thus the enzyme could exploit for increased reaction rate the energy saved in eliminating charge separation. If this were not the case, it would seem unnecessary to use both an Asp and an His for the general base catalysis. The Asp could be eliminated and the His could act as the base. After the attack by the serine on the substrate, a short-lived tetrahedral intermediate is formed (II). This intermediate is stabilized by the covalent bond to the enzyme and by a number of hydrogen bonds. The following structural features of the tetrahedral intermediate are primarily based on the crystallographic determination of many different protease-inhibitor structures. The negatively charged substrate oxygen in the tetrahedral intermediate is stabilized by hydrogen bonds from the amide nitrogens of residues 195 and 193. The importance of the amides of Ser-195 and Gly-193 was first noted by Steitz, Henderson and Blow ( 1969), and their role in transition-state stabilization was postulated by Henderson ( 1970) and by Robertus et al. (1972) . Another hydrogen bond forms between the carbonyl group of Ser-214 and the a-N of the substrate (Steitz, Henderson and Blow 1969 ; Segal et al. 1971). Comparison of the kinetics of hydrolysis of specific trypsin and chymotrypsin substrates with and without the hydrogen bonding capacity of the a-N suggests that the Ser-214-a-N bond may not form in the Michaelis complex (Ingles and Knowles 1968; Caplow and Harper 1972; Kobayashi and lshi 1974). These results show, however, that this hydrogen bond does play a role in the transition states between intermediates and possibly in the tetrahedral and acyl enzyme intermediates. One explanation for the exceptional catalytic powers of enzymes is that enzymes have evolved so that they can optimally bind the transition-state structures in the reactions they catalyze rather than the substrates themselves (Pauling 1946; Wolfenden 1972). The hydrogen-bonded structure in the serine protease-substrate transition state may be an example of transition-state stabilization, for the oriented hydrogen bonds can help to speed up the reaction by smoothing down the highest energy barriers between intermediate states. The stability of the tetrahedral adduct in the trypsin-trypsin inhibitor complexes is consistent with the transition-state stabilization hypothesis. At step 11-111, the now unstable carbon-nitrogen bond is broken, and the first product of hydrolysis, an amine, is free to diffuse away, taking with it a proton from the enzyme. At the same time, the bound part of the substrate rearranges to a covalently modified acyl enzyme intermediate (III). At pH 8, N14 /N 1 " kinetic isotope effects (O'Leary and Kluetz 1972) show that the C-N bond rupture is partially rate-determining for the hydrolysis of acetyl tryptophanamide by chymotrypsin. The rate-determining step for amide hydrolysis, however, may vary from the formation of the tetrahedral intermediate to its breakdown, depending upon the pH and the structure of the substrates (Fastrez and Fersht 1973a) . The breakdown of the acyl intermediate (lV-VI) is the microscopic reverse of steps 1-111; this time water is the attacking group. At step V-VI, the second product is formed. It is an acid which loses a proton to the solution and becomes negatively charged. For the first time (if the proton shuttle is concerted), there are two charges in the system. These two negative charges repel each other and so help to dissociate the second product from the enzyme (Johnson and Knowles 1966), regenerating free enzyme. 58. The presence of a carboxyl group of high pKn and a neutral side chain of His-57 with a low pK,. would suggest two compelling evolutionary reasons why the Asp-His-Ser arrangement should be universal to serine proteases (Hunkapiller et al. 1973). First, by neutralizing a negative charge on Asp102, rather than generating a positive charge on His-57, during formation of the tetrahedral intermediate, there would be no unfavorable charge separation. This would contribute to reducing transition-state internal energies, and thus to rate enhancement (Jencks 1971 ) . Second, if the charged Asp-102 is to be a proton acceptor at physiological pH values, its pK,. must be raised and it must have access to the proton donor. The imidazole of His-57 is ideally suited both to insulate Asp-102 from solvent (so raising the pK. of the buried carboxyl group) and to serve as a proton conductor, transferring charge from the carboxyl group to the substrate. It is also important to note that both the reverse separation of the pK" values of Asp-102 and His-57 and the structure of trypsin at pH 7 and pH 8 (Stroud, Kay and Dickerson 1974; Krieger, Kay and Stroud 1974; Huber et al. 1974; Sweet et al. 1974), which shows a symmetric interaction between the charge on Asp-102 and His-57 ( see Fig. 3), are unlike the situation expected in aqueous solution and reflect a unique microenvironment for these groups. As far as we know, all the serine proteases use the same three chemical groups to hydrolyze peptide bonds. They, like trypsin, are active catalysts only when the aspartic acid is negatively charged. Against the active site, the reaction goes on in an unique way as the enzyme smoothes the transition from one intermediate state to another. This emphasizes the importance of exact stereochemical fit and correct orientation ( Koshland 19 58) of the substrate as the reaction takes place, rather than simply the generation of an especially reactive site. After all, a very reactive site could react in many less specific ways. It is better to have a moderately efficient catalytic site coupled with a very selective binding requirement (Fersht and Sperling 1973). With this in mind, the subtle differences between serine proteases involved in biological control can be understood in terms of differences in their specific substrate binding properties. Acknowledgments This is contribution number 4990 from the Norman W. Church Laboratory of Chemical Biology. We would like to acknowledge our indebtedness to the National Institutes of Health for their support by means of grants GM-19984 and GM-70469. R.M.S. is the recipient of an NIH Career Development Award, M.K. is the recipient of a Danforth Foundation Fellowship, R.K. and J.C. are recipients of NIH Predoctoral Traineeships, and A.K. is the recipient of an NIH Postdoctoral Fellowship. REFERENCES Alden, R. A., C. S. Wright and J. Kraut. 1970. A hydrogen-bond network at the active site of subtilisin BPN'. Phil. Trans. Roy. Soc. London B 257:119. Bender, M. L. and F. J. Kezdy. 1964. The current state of the a-chymotrypsin mechanism. J. Amer. Chem. Soc. 86:3704. 59 . Bender, M. L., G. E. Clement, F . J . Kezdy and H . D'A. Heck. 1964. The correlation of the pH (pD) dependence of the stepwise mechanism of a-chymotrypsin-catalyzed reactions. J. Amer. Ch em. Soc. 86:3680. Birktoft, J . J. and D. M. Blow. 1972. Structure of crystalline a-chymotrypsin. V. The atomic structure of tosyl-a-chymotrypsin at 2 A resolution. J. Mol. Biol. 68:187. Blow, D . M. , J. J. Birktoft and B. S. Hartley. 1969. Role of a buried acid group in the mechanism of action of chymotrypsin. Nature 221:337. Blow, D. M., J. Janin and R. M. Sweet. 1974. Mode of action of soybean trypsin inhibitor ( Kunitz) as a model for specific protein-protein interactions. Nature 249:54. Blow, D. M., C. S. Wright, D . Kukla, A. Ruhlmann, W. Steigemann and R. Huber. 1972. A model for the association of bovine pancreatic trypsin inhibitor with chymotrypsin and trypsin. J. Mo/. Biol. 69:137. Brown, J. R., D. L. Kauffman and B. S. Hartley. 1967. The primary structure of porcine pancreatic elastase: The N-terminus and disulphide bridges. Biochem. J. 103:497. Caplow, M. 1969. Chymotrypsin catal ysis. Evidence for a new intermediate. J. Amer. Chem. Soc . 91:3639. Caplow, M. and C. Harper. 1972. Discrete effects of the acylamino proton in a chymotrypsin substrate on different processes in catalysis. J. Amer. Chem . Soc. 94:6508. Chambers, J. L., G . G. Christoph, M. Krieger, L. Kay and R. M. Stroud. 1974. Silver ion inhibition of serine proteases : Crystallographic study of silver trypsin. Biochem . Biophys. Res. Comm. 59:70. Chauvet, J. and R. Acher. 1967. The reactive site of the basic trypsin inhibitor of pancreas. J. Biol. Chem . 242:4274. Davie, E. W. and H. Neurath. 1955. Identification of a peptide released during autocatalytic activation of trypsinogen . J. Biol. Chem . 212:515. East, E. J . and C. G. Trowbridge. 1968. Binding of benzamidine and protons to trypsin as measured by difference spectra. Arch. Biochem. Biophys. 125:334. Fahrney, D. E. and A. M. Gold. 1963. Sulfonyl fluorides as inhibitors of esterases. J. Amer. Chem. Soc. 85:997. Fastrez, J. and A. R. Fersht. 1973a. Mechanism of chymotrypsin. Structure, reactivity, and nonproductive binding relationships. Biochemistry 12:1067. - - - . 1973b. Demonstration of the acyl-enzyme mechanism for the hydrolysis of peptides and anilides by chymotrypsin. Biochemistry 12:2025. Fersht, A. R. and M. Renard. 1974. pH dependence of chymotrypsin catalysis. Biochemistry 13: 1416. Fersht, A. R. and Y. Requena. 197 la. Mechanism of the a-chymotrypsin hydrolysis of amides. pH dependence of k 0 and Km. Kinetic detection of an intermediate. J. Amer. Chem. Soc. 93:7079. - - - . 1971 b. Equilibrium and rate constants for the interconversion of two conformations of a-chymotrypsin. J. Mo/. Biol. 60:279. Fersht, A. R. and J. Sperling. 1973. The charge relay system in chymotrypsin and chymotrypsinogen. J. Mo/. Biol. 74: 137. Fersht, A. R., D. M. Blow and J. Fastrez. 1973. Leaving group specificity in the chymotrypsin catalyzed hydrolysis of peptides. A stereochemical interpretation. Biochemistry 12:2035. Freer, S. T., J. Kraut, J. D. Robertus, H. T. Wright and Ng.- H. Xuong. 1970. Chymotrypsinogen: 2.5 A crystal structure, comparison with a-chymotrypsin, and implications for zymogen activation. Biochemistry 9:1997. Fritz, H., H. Schult, R. Meister and E. Werlo. 1969. Herstellung und eigenschaften 60. von derivaten des Trypsin-Kallikrein-Inhibitors aus Rinderorganen. z. Physiol. Chem. 350:1531. Gertler, A., K. A. Walsh and H. Neurath. 1974. Catalysis by chymotrypsinogen. Demonstration of an acyl-zymogen intermediate. Biochemistry 13:1302. Hartley, B. S. 1960. Proteolytic enzymes. Annu. Rev. Biochem. 29:45. - - - . 1964. Amino acid sequence of bovine chymotrypsinogen A. Nature 201:1284. Henderson, R. 1970. Structure of crystalline a-chymotrypsin. IV. Structure of indoleacryloyl-a-chymotrypsin and its relevance to the hydrolytic mechanism of the enzyme. J. Mo/. Biol. 54:341. - - - . 1971. Catalytic activity of a-chymotrypsin in which histidine-57 has been methylated. Biochem. J. 124: 13. Hess, G. P., J. McConn, E. Ku and G. McConkey. 1970. Studies of the activity of chymotrypsin. Phil. Trans. Roy. Soc. London B 257:89. Huber, R., D. Kukla, A. Ruhlmann and W. Steigemann. 1971. Pancreatic trypsin inhibitor : Structure and function. Cold Spring Harbor Symp. Quant. Biol. 36:141. Huber, R., D. Kukla, A. Ruhlmann, 0. Epp and H. Formanek. 1970. The basic trypsin inhibitor of bovine pancreas. I. Structure analysis and conformation of the polypeptide chain. Naturwissenscha/ten 57:389. Huber, R., D. Kukla, W. Steigemann, J. Deisenhofer and A. Jones. 1974. Structure of the complex formed by bovine trypsin and bovine pancreatic trypsin inhibitor. In Bayer Symposium V: Proteinase Inhibitors (ed. H. Fritz et al.), p. 497. Springer Verlag, Heidelberg. Hunkapiller, M. W., S. H. Smallcombe, D. R. Whitaker and J. H. Richards. 1973. Carbon nuclear magnetic resonance studies of the histidine residue in a-lytic protease. Biochemistry 12:4732. Inagami, T . and T. Murachi. 1964. The mechanism of the specificity of trypsin catalysis. J. Biol. Chem. 239: 1395. lnagami, T. and S. S. York. 1968. The effect of alkylguanidines and alkylamines on trypsin catalysis. Biochemistry 7:4045. Ingles, D. W. and J. R. Knowles. 1968. The stereospecificity of a-chymotrypsin. Biochem . J. 108:561. Inward, P. W. and W. P. Jencks. 1965. The reactivity of nucleophilic reagents with furoyl-chymotrypsin . J. Biol. Chem . 240:1986. Jacob, F. and J. Monod. 1961. Genetic regulatory mechanisms in the synthesis of proteins. J. Mo/. Biol. 3:318. Jansen, E. F., M. D. F. Nutting and A. K. Balls. 1949. Mode of inhibition of chymotrypsin by diisopropylfluorophosphate. I. Introduction of phosphorous. J. Biol. Chem. 179:201. Jencks, W. P. 1969. Catalysis in Chemistry and Enzymology, p. 218. McGrawHill, New York. - - - . 1971. Reactivity correlations and general acid-base catalysis in enzymic transacylation reactions. Cold Spring Harbor Symp. Quant. Biol. 36:1. - - - . 1972. General acid-base catalysis of complex reactions in water. Chem. Rev. 72:705. Johnson, C. H. and J. R. Knowles. 1966. The binding of inhibitors to a-chymotrypsin. Biochem. J. 101:56. Kallas, J. and D. Rizok. 1964. Heavy atom labeling of the serine of the active center of chymotrypsin: Pipsyl-chymotrypsin. /. Mol. Biol. 9:255. Kassell, B. and J. Kay. 1973. Zymogens of proteolytic enzymes. Science 180:1022. Kobayashi, R. and S. Ishi. 1974. The trypsin-catalyzed hydrolysis of some L-aamino-lacking substrates. J. Biochem. 75:825. 61. Koshland, D. E., Jr. 1958. Application of a theory of enzyme specificity to protein synthesis. Proc. Nat. Acad. Sci. 44:98. Kress, L. F. and M. Laskowski. 1967. The basic trypsin inhibitor of bovine pancreas. J. Biol. Chem. 242:4925. Krieger, M., L. M. Kay and R. M . Stroud. 1974. Structure and specific binding of trypsin: Comparison of inhibited derivatives and a model for substrate binding. J. Mo/. Biol. 83:209. Kunitz, M. and J. H. Northrop. 1935. Crystalline chymo-trypsin and chymotrypsinogen. I. Isolation, crystallization, and general properties of a new proteolytic enzyme and its precursor. J. Gen. Physiol. 18:433. - - -. 1936. Isolation .from beef pancreas of cystalline trypsinogen, trypsin, a trypsin inhibitor, and an inhibitor-trypsin compound. J. Gen. Physiol. 19:991 . Leighton, T . J ., R. H. Doi, R. A. J. Warren and R. A. Kelln. 1973. The relationship of serine protease activity to RNA polymerase modifications and sporulation in Bacillus subtilis. J. Mo/. Biol. 16: I 03. Magnusson, S. 1971. Thrombin and prothrombin. In The Enzymes ( ed. P. D. Boyer), vol. 3, p. 277. Academic Press, New York. Mares-Guia, M. and E. Shaw. 1965. Studies on the active center of trypsin : The binding of amidines and guanidines as models of the substrate chain. J. Biol. Chem. 240:1579. Maroux, S., J. Baratti and P. Desnuelle. 1971. Purification and specificity of porcine enterokinase. J. Biol. Chem. 246:5031. Martinek, K. , Y. V. Savin and I. V. Berezin. I 971. Kinetic manifestations of trypsin's active center during inhibition of its enzymatic activity by Ag+ ions. Biokhimiya 36:806. Martinek, K., Kh. Viii', Z. A. Strel'tsova and I. V. Berezin. 1969. Kinetic manifestations of the active-center structure of a-chymotrypsin during inhibition of enzyme activity by Ag + ions. Mo/. Biol. (U.S.S.R.) 3:554. Matthews, B. W., P. B. Sigler, R. Henderson and D. M. Blow. 1967. Threedimensional structure of tosyl-a-chymotrypsin. Nature 214:652. Naughton, M. A. and F. Sanger. 1961. Purification and specificity of pancreatic elastase. Biochem. J. 78:156. Neurath, H. and G . H. Dixon. 1957. Structure and activation of trypsinogen and chymotrypsinogen. Fed. Proc. 16:791. Neurath, H., J. A. Rupley and W. J. Dreyer. 1956. Structural changes in the activation of chymotrypsinogen and trypsinogen. Effects of urea on chymotrypsinogen and delta-chymotrypsinogen. Arch. Biochem. Biophys. 65:243. Northrop, J. H., M. Kunitz and R. M. Herriot. 1948. Crystalline Enzymes, p. 96. Columbia University Press, New York. O'Leary, M. H. and M. D. Kluetz. 1972. Nitrogen isotope effects on the chymotrypsin catalyzed hydrolysis of N-acetyl-L-tryptophanamide. J. Amer. Chem . Soc. 94:3585. Ong, E. B., E. Shaw and G. Schoellmann. 1964. An active center histidine peptide of a-chymotrypsin. J. Amer. Chem. Soc. 86:1271. Oppenheimer, H. L. , B. Labouesse and G. P. Hess. 1966. Implication of an ionizing group in the control of conformation and activity of chymotrypsin. J. Biol. Ch em. 241:2720. Owren, P. A. and H. Stormorken. 1973. The mechanism of blood coagulation. Ergeb. Physiol. Biol. Chem . Exp. Pharmacol. 68:1. Pauling, L. 1946. Molecular architecture and biological reactions. Chem. Eng. News 24:1375 . Pollack, E. , J. L. Hogg and R. L. Schowen. 1973. One-proton catalysis in the deacetylation of acetyl-a-chymotrypsin. J. Amer. Chem. Soc. 95:968. 62. Robertus, J. D., J. Kraut, R. A. Alden and J. J. Birktoft. 1972. Subtilisin; a stereochemical mechanism involving transition-state stabilization. Biochemistry 11:4293. Ruhlmann, A., D. Kukla, P. Schwager, K. Bartels and R. Huber. 1973. Structure of the complex formed by bovine trypsin and bovine pancreatic trypsin inhibitor. J. Mol. Biol. 77:417. Sampath Narayanan, A. and R. A. Anwar. 1969. The specificity of purified porcine pancreatic elastase. Biochem. J. 114:11. Segal, D. M., J. C. Powers, G. H. Cohen, D. R. Davies and P. E. Wilcox, 1971. Substrate binding site in bovine chymotrypsin Ay. A crystallographic study using peptide chloromethyl ketones as site-specific inhibitors. Biochemistry 10:3728. Shaw, E., M. Mares-Guia and W. Cohen. 1965. Evidence for an active-center histidine in trypsin through use of a specific reagent, l-chloro-3-tosylamido-7amino-2-heptanone, the chloromethyl ketone derived from N 0 -tosyl-L-lysine. Biochemistry 4:2219. Shotton, D . M . and B. S. Hartley. 1970. Amino-acid sequence of porcine pancreatic elastase and its homologies with other serine proteases. Nature 225:802. Shotton, D . M. and H. C. Watson. 1970. The three-dimensional structure of crystalline porcine pancreatic elastase. Phil. Trans. Roy. Soc. London B 257:111. Sigler, P. W., D. M. Blow, B. W. Matthews and R. Henderson. 1968. Structure of crystalline a-chymotrypsin. II. A preliminary report including a hypothesis for the activation mechanism. J. Mo/. Biol. 35: 143. Spomer, W. E. and J. F. Wootton. 1971. The hydrolysis of a a-N-benzoyl-L-argininamide catalyzed by trypsin and acetyl trypsin. Dependence on pH. Biochim. Biophys. Acta 235:164. Stambaugh, R., B. Brackett and L. Mastroianni. 1969. Inhibition of in vitro fertilization of rabbit ova by trypsin inhibitors. Biol. Reprod. 1:223. Steitz, T. A., R. Henderson and D. M. Blow. 1969. Structure of crystalline a-chymotrypsin. III. Crystallographic studies of substrates and inhibitors. J. Mo/. Biol. 46:337. Stroud, R. M. 1974. A family of protein-cutting proteins. Sci. Amer. 231:74. Stroud, R. M ., L. M. Kay and R. E. Dickerson. 1971. The crystal and molecular structure of DIP-inhibited bovine trypsin at 2.7 A resolution. Cold Spring Harbor Symp. Quant. Biol. 36: 125. - - - . 1974. The structure of bovine trypsin: Electron density maps of the inhibited enzyme at 5 A and 2.7 A resolution. J. Mo/. Biol. 83:185. Susi, H. 1972. The strength of hydrogen bonding: Infrared spectroscopy. In Methods in Enzymology (ed. S. P. Colowick and N. 0. Kaplan), vol. 26, p. 381. Academic Press, New York. Sweet, R. M., H. T. Wright, J. Janin, C. H . Chothia and D. M. Blow. 1974. Crystal structure of the complex of porcine trypsin with soybean trypsin inhibitor ( Kunitz) at 2.6 A resolution. Biochemistry 13:4212. Timasheff, S. N. and J. A. Rupley. 1972. Infrared titration of lysozyme carboxyls. Arch. Biochem. Biophys. 150:318. Tomkins, G . M., T. D. Gelehrter, D. Granner, D. Martin, Jr., H. H. Samuels and E. B. Thompson. 1969. Control of specific gene expression in higher organisms. Science 166: 1474. Tschesche, H. 1974. Biochemistry of natural proteinase inhibitors. Angew. Chem. Int. Ed. Eng. 13:10. Wolfenden, R. 1972. Analog approaches to the structure of the transition state in enzyme reactions. Acc't. Chem. Res. 5: I 0. 63. Wright, C. S., R. A. Alden and J. Kraut. 1969. Structure of subtilisin BPN' at 2.5 A resolution. Nature 221:235. Wright, H. T . 1973. Comparison of the crystal structures of chymotrypsinogen-A and a-chymotrypsin. J. Mo/. Biol. 79:1. Zerner, B. and M. L. Bender. 1964. The kinetic consequences of the acyl-enzyme mechanism for the reactions of specific substrates with chymotrypsin. J. Amer. Chem. Soc. 86:3669. Zerner, 8., R. P. M. Bond and M. L. Bender. 1964. Kinetic evidence for the formation of acyl-enzyme intermediates in the a-chymotrypsin-catalyzed hydrolysis of specific substrates. J. Amer. Chem. Soc. 86:3674. 64. Appendix 4 0 Structure of Bovine Trypsinogen at 1. 9 A Resolution 65. 1'<1•1u·1111t•d fro111 Hiodl1'1111~lf'\', t W7/J 16, lifl-l.j ( '1111vr1µh1 I i,1·11,y I ht• ,\11wri1·111t l 'lw1ni1·11I ~111·i1•1~· ,111,I n·1,ri11tcd 11.v 111•r111i~i1111 ,,f tht• 1·c,1,.v ri.:ht 11w1u•r. Structure of Bovine Trypsinogcn at 1.9 A Rcsolutiont Anthony A. Kos.,iakoff. 1John L. Chambers.I Lois M. Kay, and Robert M. Stroud*·, T: The three-dimensional crystal structure or bovine trypsinogen at ~pH 7.5 was initially solved at 2.6 A resolution usin!( the multiple isomorphous replacement method. Preljminary refinement cycles or the atomic coordinate., for trypsinogcn have been carried out first to a resolution or 2.1 A. and later to I .'I A. using constrained difference Fourier refinement. During the process, structure factors F, and >Pe were calculated from the trypsinogen structure and final interpretation was based on an electron-density map computed with terms (211",,I - IF,i) and phases >Pe al a resolution or 1.9 A. Crystal• or trypsinogen grown from eihanol-watcr mixtures arc trigonal. with space group P3121. and cell dimensions a • SS.11 Aand r • I09.25 A. The structure is compared with the bovine diisopropylphosphoryltrypsin structure at ---pH 7.2, originally determined from orthohombic crystals by Stroud ct al. (Stroud, R. M.. Kay, L. M.,and Dickerson, R. E. (1971), Cold Spring Harbor Symp. Quant Biol. Jti, I 25- 140; Stroud, R. M.. Kay, L. M.,and Dickerson, R. E. (1974),J. Mot. Biol. 83, 185-208). and later refined at 1.5 A reaolution by Chambers and Stroud (Chambcrs,J. L.,andStroud, R. M. (1976), A~ta Cry.,tallogr. (in press)). At lower pH, 4.0-5.5, diisopropylphosphoryltrypsin forms crystals which arc isomorphous with tryr.lnogcn. with cell dimensions a • SS.0S A and r • 109.45 . Thia finding was used in the solution of the aix lryplinogcn heavy-atom derivatives prior to isomorphous pltasc analysis. and as a further basis of comparison between trypsmo,cn and the low pH trypain structure. There arc amall differences between the two diisopropylphosphoryltrypsin structures. Bovine trypsinogcn has a large and accessible cavity at the site where the native enzyme binds specific side chains or a substrate. The conformation and stability or the binding site differ from that found in trypsin at ---pH 7.5, and from that in the low pH form or diisopropylphosphoryltrypsin. The catalytic site containing Asp-102, His,57, and Ser-195 is similar to that found in trypsin and contains a similar hydrogen-bonded network. The carboxyl group of Asp-I 94, which is salt bridged to the amino terminal of lle-16 in native trypsin or other serine proteases, is apparently hydrogen bonded to internal solvent molecules in a loosely organized part of the ,ymogen structure. The unusually charged N-tcrminal hexapcptidc of trypainogcn, whose removal leads to activation of the zymogen, lies on the outside surface of the molecule. There arc significant structural changes which accompany activation in neighboring regions, which include residues I 42- I 52, 215-220, l.8 8A-195. The NH group of Gly-193, normally involved in stabilization of reaction intermediates ( Steitz, T. A., Hcndcnon, R., and Blow, D. M. ( 1969), J. Mal. Biol. 46, 337 - 348; Henderson, R. (1970), J. Mol. Biol. 54, 341-3S4; Robcrtua, J. D., Kraut, J., Alden, R. A., and Birkoft, J. J. ( 1972), /Jloc~emistry I I, 4293-4303) in the enzyme, is moved 1.9 A away from its position in trypsin. Alterations in the binding site for substrate side chains arc prime candidates for cxplai~ing tbc relative inactivity of trypsinogcn. Lyptic activation of the family of pancreatic procnzymcs or zymogclll!. which include trypiinoacn, chymotrypsinogcn, procarboxypcptidusc. and proelaataae, takca place in the duodenum by limited proteolysis (Davie and Ncur_a th, 1955; Ncurath, 1975). The primary activating clcawgc for the serine prot~ is hydrolyais of the. bond between a basic residue 15, 1 and ilolcucinc (vu line in clastasc and plaamin). This cleavage ia accompaniod by a conformation change which generates, or at least greatly cnham,cs, enzymatic activity (Ncurath and Dixon, I957; Sigler ct al., 1968). The convcnion of trypiinogcn to trypiin is the fint atcp of the digestive cascaded-activation sequence, and it relics on the known high specificity of the activator (cntcrokinasc) for the unusual N-tcrminal sequence Val-Asp-Asp-Asp-Asp-Lys unique to trypsinogcns (Marou• ct al., 1971 ), The activation of further trypsinogcn by trypsin at the second-cascade step requires that the most favored tryptic dcavagc of that zymogen be the a~tivating split between residues I 5 and 16. Yct, at the same time, it has been shown that the prcscnce of four aspartic acids in the activation peptide (residues 10--15) hasa markedly negative effect on this tryptic cleavage (Abita ct al., 1969). In all species, the activation peptide n:iust generate a good substrate quality for the activator, and a poor, but nonetheless most favorable, tryptic cleavage site on thc entire trypiinoacn molecule. In this light, it is not too 111rprising to find that the activation peptide sequence of trypsinogcn, with its dual role, is so hlgbly conserved throughout all species (De Haen ct al., 1975). It seems likely that the fundamental basis for inactivity of the procnzymes will be shared by all of the trypsin-like en• zymll!I, The comparison between the structures of chymotrypsinogen (dctmnined by Freer ct al., 1970) and lryplinogcn should aervc to isolate characteristics common to both procnzymca, which will limit the primary reasons for inactivity, to the lo-t COIJ\IIIOn factor. In thi1 rapect, there arc real and signifilllnt difTorcncca botwcon the two proenzymca. Taking for granletl that the enzymes thcmachel are catalytk:ally active AIISTRA< ' From the Norman W. Church Laboratory ol Chemical Biolosy, C.Hfornia l1111itute orTi!cbnulagy, hadcna, California 91115. R«tlo«I Jw,w 16, 1916. Contriboation No. 5342. Supponod by Nalional lnatltutea or Health Grant Cl M, I9984and National Science Foundation Grant No. BMS75-01405. I Recipient or a NIH Ptlltdoctorlal Fcllowahip. Curren! addreu: Departme.nt or BiolCJly. llrookba..., National Laboratories, Uptoo, Lona laland, N.Y. I 1973. t Rcclpiont or a NtH Prcdoctorial Trai,-hip. , Rccipienl or. NIH Career Development Award •nd I Sloon Foundatiun Fellow. , 1 The n•mbcrln1 rorc,,.i to 11 that ol i:hymotrypainopn sequence (Hartley, 1970). The oeqoence fot t'Y)lllftClietl II that ol Watah and N111rath (1964). and Mika et at. (1967). 654 BIOCIIF.MtllTltY, VOi .. 16, NO. 4, 1977 66. ST R l IC T lJ K I; O I-' Ho V I N F T It Y PS I N O <i F N A 'I I . 9 - A R t• SO I ll T I o N P2,2,2 1 i ,◄ e PS 111 I ca :~ r r" ' I ( I 1Q;:)1V ! ,,,, -I , i=·. I Fl<itJMI '. I : 'l"hc p.icking dial(ram for lryp,iinugc:n (right) and trypsin (i.:cntirl rc¥cal1 a common intcrmolc~ular armngcmcnt between molecules around a particular 2 1 ~~is. (something which would not be obvious if the structures alone were viewed in the absence of such knowledge), we will compare the two procn,ymes emphasizing the question as to why they should be relatively inactive. The answer to this question implies acceptance of the many well-established details or substrate binding and activity summarized, for example, by Rlow ( 1974n). Ex per imen ta I l'roccd u res Crysta/li:ation and Deri11ative Preparation. Bovine trypsinogen (Worthington) was chromatographically purified using the procedure of Schroeder and Shaw ( 1968). Crystals were obtained by vapor diffusion of a 3% solution or protein against 30% ethanol- water. The crystals grew at 5 °C, as modified rhombohcdra which generally averaged 0.25 mm acros.<. A trypsin inhibitor, ben1.amidine, was added to the solution in a concentration of 2.5 mg/ml to prevent possible trypsin-mediated autolysis of the zymogen during the several weeks required for crystallization. It was essential to eliminate divalenl cations from the crystallizing solutions, since they mimic the effects ofCaH and lead to autoactivation and autolysis of the trypsinogcn in solution (Northrop ct al., I 948). Crystals used for structure determination were grown from solutions adjusted to about pH 7.5, although isomorphous crystals were obtained throughout the pH range 5 -8. Crystals grown this wuy arc trigonal, space group P3 121 with cell parameters a = 55.11 A and c = I09.25 A. The umt cell volume is approximately 288 000 A.-1 and contains one molecule per asymmetric unit. In view of the possible autolysis problem, the protein content of the crystals was checked by chromatography and by endgroup amino acid analysis. Crystals were dissolved and found to contain pure trypsinogen by chromatography (Schroeder and Shaw, 1968). End-group analysis, using a modified Edman procedure, showed t~at the protein contained an N-terminal valinc with no detect.1blc alternative amino acid. This analysis rules out the possibility that any trypsinogen had been converted to trypsin during crystalli1.ation, since trypsin has an N-tcrminal isolcucinc. At low pit, DI P-trypsin 2 crystals were obtained which were isomorphous with the trypsinogen crystals. These crystals could be grown at pH 4.0- 5.5 from either ethanol-water mixtures or from MgSO4 solutions. At higher pH (5.7-8.5), trypsin 1 Abbrcvia1ions us~-ct urc: OIP, diisopropylphaiphoryl; DIPT. L.pT, Tg, ( 'T, ,md ('g refer 10 the cry~tallogrnphically determined 11ructure.~of di- iM>propylphtMiphoryl tryp~m 11l ~pH 7.5 from onhorhombic crystals (C'hnmbcn, and S1roud, 1«>76), low pli (~5.0) f>IP-tryp.,in from trigonal crystal:i1, lt)'fl!Cinn~cn (~rll 7.,) from 1ri1on11I cryHhtls iAomorphic with l.pT cryiu11l:i1, 1osyl-u-chymo1ry~in {Rirktofl ond Blow, 1972). and chymolryp!linot1,cn (Freer ct nl., 1970; Rirk.10(1 ct ul., 1976), respectively; NP<iK. p-nitrophcnyl-p'-)lu.11nidinubcnwestc; Rl'fl. the bu~ic J'Mlncrcatic lf)'P'in inhibitor. crystalli1.cs exclusively in the orthorhombic space group P2,2 12 1 (Stroud et al., 1971, I 974). At pH's between 4.0 and 5.5, the trigonal form of DIP-trypsin crystallized first. After several weeks, orthorhombic crystals grew from the same solution. The trigonal crystals were found to be unstable above pH 5.5, which ,uggcsts that there is some type of structural reorgani,.ation which occurs around pH 5.5. In rorc cases, O11'-trypsin crystals grew as trigonal-orthorhombic twins at about pH 5.5. These twins, in which the orthorhombic a axis was parallel to the trigonal a axis, suggested that the two crystal forms had some common symmetry clement. The packing diagrams obtained from the structure solutions show that molecules arc similarly disposed with respect to 2 1 screw axe.< in the two crystal forms. The 2 1 axes arc parallel to the a axe., in both the trigonal and orthorhombic crystals, as indicated in Figure I. The unit-cell dimensions parallel to the 2, axes in the two crystal forms are 54.84 A. in orthorhombic DIP-trypsin and 55.17 A. in trigonal trypsino- gcn. A vector parallel to the c axis of the orthorhombic cell in the tryp.,in molecule is tilted by a 13.49° rotation around the crystallographic 2, axis in the trigonal structure (eq I, below). Derivative Preparation. lsomorphous derivatives were prepared by placing the crystals into ethanol - water sol~tions containing heavy-atom salts. Optimum-diffusion conditions and soaking times were e.,tablishcd by varying the pH from 6.0 to 8.0 and determining photographically the time necessary for a stable and reproducible diffraction pattern change to occur. Several of the derivatives showed a variation of pattern over an extended period of time. The initial study of the Hgl 4 2-, Pt{NO 2 ), 2 -, and mersalyl derivatives showed that pattern stability had been reached after I to 2 weeks of crystal soaking. However, these same crysta l soaks showed a further distinct change in diffraction pattern arter 2 years. When the derivatives were solved, it was found that these changes were caused by appearance of several new minor sites, and nuctuation in the occupancies of several of the original sites. The overall differences between the short- and long-soak derivatives were large enough to allow them to be treated as independent derivatives in the phase refinement. Data Collection. 1.8 A. data for native trypsinogen and 2.6 A. data for six derivatives were collected on a Syntex PT automated diffractometer using graphite-monochromatizcd Cu Ka radiation. The detector was located 40.6 cm from the crystal, and a helium-filled tube was placed between the detector and the crystal to limit air absorption losses of the diffracted beam. Intensity data were collected using the Wyckoff step-scanning method (Wyckoff cl al., 1970). An optimal scan width was chosen for each crystal separately, based on the criterion that the first steps outside of the usable peak should not be less than 95% of the peak step counts. Typically, the measurement of each rencction consisted of live steps centered on the diffracR I lll' H ~MISTRY. VO t. I 6, NO . . 4 , I977 655 67. KOSSl,\~Ol·T tion maximu, lakcn over an omega range of0. 10°. The highcsl sum of three, or of four. adjacenl slcps was lhcn lreated as lhe inlegruled inlcnsily. Scan rales ran11,ed from 0.15° 10 0.25° /min. In m,,sl cases, lhe nalive and derivalivc cryslals were sla blc enough lo collect a full 2.1> A dala sci from a single cryslal before any of lhe slandard rcncclion intensities had chan11,cd by more than 10%. The method used for data reduction and absurptiun correction wus similar lo that described by Stroud cl al . (IY74), but wa~ altered fur four-circle geometry. Individual background measurements were approximated by interpolation from empirical background curve.,, which were calculated as a function of 28, )(, and ,p by lhe mclhod of Krieger ct al. (1974a) . Pha.,e Analysis Difference Pa11erso11 Method.,. I nilial phases for lhe trypsinogen f 'i hlil were determined using the multiple-isomorphuus-rcplaccmcnt method. Three-dimensional difference Palterson syntheses (Blow, 1958) were cakulaled for each one of the derivatives. However, of lhe len derivatives examined, nol one of the difference l'uttcrson maps could be easily solved . The reasons for this diniculty became apparent only later when it wus reali,ed thnl lhere were nu fewer lhan 12 sites per ltsymmelric unit in cnch one of lhe derivulivc.,. A difference PatterS<m mup for such u derivalivc ~-onlains in execs., of 2500 vcclors. Typically, lhe lrypsinugen maps conluincd KOO - 1400 peaks of sufficient height lo warranl investigation. Even with 1his large number of peaks, the ratio of 1he highc.,1 peuks on the map lo lhe origin peak encouraged nltempls to find al leas\ 1hc major sites by dtcunvolulion of the Patterson map. Therefore, vector superposition methods were applied in altempls to solve the difference syntheses. In no case was an unambiguous solution lo any derivative found. The primary source of difficulty was that a large number of cross vectors between nonC<Juivulcnl "A--8" atom sites mimic the symmetry of "A . A" type vectors between symmetry-equivalent atoms around the threefold axes al u ., I - v • 1/.1, ¾. Asa result, they have double weight and so frustrate allcmpts al solution . Solution of tht Derivu1i11,.,. At this stage, heavy-atom derivatives of the isomorphou• low pH crystals of DI P-lrypsin were investigated . A silver ion derivative was prepared, since a silver derivative of the neutral pH orthorhombic DIP-trypsin had previously shown a simple substitution pattern (Chambers ct al., l'174) . The trigonal trypsin derivative was unambiguously solved by the superposition programs and contained only two major sites and five minor sites. Since this form of trypsin is highly isomorphous with trypsinogen, an initial set of single i•omorphuus phases was obtained and Ibey were used directly to determine the heavy atom positions in the trypsinogen de- rivatives. J>ljfnnice El,ctrn11 Dnuity Map., and Rt/intmtnt. Three-dimensional single isomorphous replacement phases derived from the major Ag+ site in the isumorphous DIPtrypsin derivative were used first lo locale the major sites in the HgJ,' - and mersalyl derivatives of lrypsinogen. The maps were computed using '1f' = IF.,111 - IF el for each derivative, and phases from the Ag• (trypsin) site structure factors. Three H11,1 4 2- sites and one mcrsulyl site were found using this procedure. Heavy-atom parameters were refined using a minicomputer (Data General Corp., NOVA 800), by the method of Dickerson ct al . ( I961, 1968) in alternating cycles of phase calculation and refinement of parameters. The positional parameters and occupancic., of the two trypsinogen derivatives were refined along with th= of the Ag+ trypsin derivative. 656 klOt'lll·. MISTMY , VOi. . 16, NO 4, 1977 ~Thi.. At a later stage of t he refinement, the Ag+ trypsin derivative was discarded to prevent bias in the phases towards a tryp- sin~like structure. Phases from these derivative., were used to find the prima ry sites in the other 1rypsinogcn derivatives. Secondary sites were located using a double-difference or error synthesis, which uses as structure factors the luck-of-closure error, computed during phase refinement, with the calcululed phases of the derivative data (if'1•1d- IF,,+f11i)exp(i<PJ•11) (Bloke ct al., 1963; Siroud ct al.. 1974), wheref11 is the current value for the calculated heavy atom structure factor. In most derivatives, several minor sites were not included in the phasing calculations as their contributions were below the errors in measurement. Finally, anisotropic temperature factors for the sites of higher occupancy were included in the refinement. The correctness of each derivative solution was checked by permutation of pairs of derivatives which were corefined together, lo generate phase., which were lhen used to calculalc difference map., for the other derivative.,. Summary s1a1is1ics for the six derivatives used arc listed in Table I. 1-:1,,·tron-Den.,ity Ma11s. An electron density map was 11,cncratcd usin11, multiple isumorphous replaccmcnl phases (MIR map) obtained from the derivatives listed in Table I. The fir~t map was calculated using (hkl) data lo a minimum interplanur spacing of 5 A (figure of merit, FM = 0.88), and revealed the location and conformation of each molecule in the unit cell. The second MIR map was computed using data to 2.6 A, and a model was built from Kendrew - Watson components to match the density map using an optical comparator (Richards, 1968). The mean figure of merit for all 5700 rene<:tions with Fu,> 3 "from"' to 2.6 A was FM= 0.71 (0.91 for centric rencc1ions) when calculated according to Dickerson ct al. (1968). A histogram shown in Figure 2 indicates the distribution of rcnections according to their figures of merit. The map was easily interpreted in most areas, although difficulties were encountered in the regions between residues I 0--15, 143- I 51, and I 86 --188, all of which lie on the surface of the molecule. Coordinate.~ of ull lhc reliably determined amino acid residue., in trypsinogen were measured using an automated coordinate-measuring device which operates on the principle used by Sa lemme and Fehr ( 1972). Amplitudes Fco1cd and phases ""'""d were calculaled from the measured coordinates, and a difference map was computed using terms: t:,,F= IF.. 1- IFcl,<Pc The well-determined parls of the molecule (90% of the structure) were adjusted using an automated procedure for constrained difference Fourier refinement and model building, which was developed to operate entirely on a minicomputer (Chambers and Stroud, 1976). Individual temperalure factors were assigned to each heavy atom or ion (C, N, 0, S, Ca>+) based on electron density al atomic centers. No solvent molecules were included. Structure factors were calculated a second lime and the procedure was repeated once more. At this stage, hkl data to 2.1 A resolution were incorporated and an electron density map was calculaled using terms l2F0 - Fd, "I:· Areas of structure which were ambiguous in the 2.6 A MIR map were rebuilt using this map. Data were included to 1.9 Aresolution, and the molecule was subjected to several cycles of difference Fourier refinement. Isotropic temperature factors were refined so as to account for elcc1ron .densi1y at atomic centers in the 68. STRtH · ruRF OF 110\' IN I'. IKVPSINO(il '. N i\T I . 9 - A k FSOl.lJTION TABLl; I: Sfatislical Analy:iis of Dcrivttlives. No. of Derivative Sites Occupancy• llgl,'·· 25 62 61 49 43 108 18 61 Pt(NO,)/- 14 Mersalyl I 14 Mersalyl 2 9 Merulyl 3 13 Ag• 15 No. of {IAFl)d •• R,• Rc.olution (A) Rcncctions 99 68 (93 hi) 0.67 (0.37-0.84) 2.6 5450 (242-81) 91 (211 - 1>4) 64 (87 -~3) 0.64 (O.JS-0.91) 2.6 4213 103 (195- R4) 57 (87- 55) 0.57 (0.41 -0.65) 3.6 4908 82 (130-70) 44 (57 42) 0.57 (0.43-0.61) 3.6 2184 110 (217 -98) 18 (74- 79) 0.70 (033- 0.72) 2.6 5043 98 (180-74) 61 (75 -53) 0.60 (0.42-0.72) 2.6 5102 54 74 53 52 31 65 42 32 23 59 39 38 36 16 49 45 22 "Oocupancy of the four moot major sites. Scale is arbitrary (approximately 1.4 X aboolute sc-ale). • , a ( 11'•111 - IFp +/ 1, I) for all reOcctions. )/(l:•i•AI-·•., ). ' Numbers in parcntlieoe., give the average values of parameters at the lowest and highest values of the scattering ,· R, - n.: ..,,.., a ogle 28 used . 1000 ,oo I N "' - 1t<illkl· 2: lfi."1usr11m 11howin1t distribution of fip;urc11 of merit for rencctiom,. lli~tuirnm for the ccntrusymnuHric; rcnections alone is shad - ,-.!. current jF., - ,.-,.j, <A synthe,iis. Therefore, large temperature factors for individual atoms give a measure or lhc possible error, or therm"I motion or each group. The RI' value at this stage wns 31% for all renections from m to 1.9 A, where R I= }JF 0 - F,! l:jF.. I While most or the molecule is well defined in the electron density (coordinates arc probably defined to within about ±0.4 ,\). there arc small regions between residues I0-· 14, 143- 150, and 186 · I 94 where there is clearly a large degree orlibration . These arc external and solvent-aecessiblc chains in the molecule. The electron density in the 2F,, - F, synthesis is diffu,icd and low in comparison with other regions, both when these residues were lcrt out or the structure factor calculation and when included in the calculalion. In addition. a density map for isomorphous trypsin, computed uRing lrypsinogcn phases, with amplitude., for trigonal trypsin wu• clearly interpretable for every residue in the region or residue., 143 --150 ( 10- 14 is l"l<ilJKF J: The s1r uc1urc around As~I02 . His-57 . and Scr-195 in 1ryp:1inogcn is shown superposed onlo the electron-density map at 1.9 A rCJ10lution. no longer present). and at lle-16- Val-17. Thus, there is clearly nnd unequivocally variability in the orientations or the lrypsinogcn residues 10-- 14. 143-150, and 186- 194. A region or the 1.9 A electron density map around the active center is shown in Figure 3. The structure in this region is superposed on the density map. Most the carbonyl groups are clearly identified in the map and six-membered rings orten have dimpled density in their centers, as, for example, docs the side chain of Trp-215. The sulrur atoms in disulfide bridges arc clearly resolved from each other. Some I0-20 ordered solvent molecules arc contained within the molecular boundary, and, in cases where they exist as isolated molecules, these KIOCHF.MtSTRY . VOi. 16, NO . 4, 1977 657 69. KOSSIAKOH" ET A L . .I ,1 ~; ~·! ~ i \·'.~ .. ! :J L~ ) r• I rl ;~ 1i~ l~ I tc;t :HI· 4· (a) Jlis1osr:in1 showi118 the difference (Ad) between n-carbon poiiitions in t>IPT :ind CT. (b) llistogram showing the diffcn:ncc ('1.J) between rM,'arbon po!.ilions of the Tg and l>IPT strui:lurc~ after suilablc rouuion . l.arttc diffcrcncc.<i in structure arc M>Cn Hn>und residua lt:i - 19, 142 1~2. IK6 - l94, and 21 l 220. (c) Hist"l!r•m showing 111< mcun difrcrcncc (.1d) in position for all uloms in each rci1idue bc1wccn Ts and Dll'T. can usually be resolved. Solvent in the cavity which becomes occupied by llc-16 upon activation was generally not resolved into discrete peaks. Results and Discussion The coordinates for trypsinogcn arc regarded as only preliminarily refined (R 1 ~ 31%). Typically, wire-model coordinates of a protein give ri•c to Rr values of ~40- 45%, and can often contain occasional large errors of several angstroms. As· refinement continues, the Rr value decreases and a value of <25% for a constrained struc'lurc often seems to represent convergence onto a rcs,.,..1ablc set of coordinates and i!Kllropic tcmpcrnturc factors which account for thermal motion. Approximately, ten well-ordered solvent molecules were included in the later phase calcula11ons. The model is clearly accurate 658 ttl0CIIH,41STRY, VOi . 16, NO . 4, 1977 for nearly all of the structure to within about 0.4 A. Comparison of Structures. The lrypsinogcn structure (Tg) was compared both with the orthorhombic DIP-trypsin structure (DIPT) and with the trigonal low-pH DIP-trypsin structure (LpT) . The trypsinngen structure will be described in reference 10 the DIP-trypsin structure. The Tg and DIPT structures were determined and refined completely independently. The differences in structure between Tg and DIPT will, at the same time. be compared with the differences between the structures of chymotrypsinogen (Cg; Freer et al., 1970) and tosyl-n-chymotrypsin (CT; Birktofl and Blow, 1972). Current Status of Trypsin Studies. The DIPT structure determined initially by Stroud et al. (1971, 1974) has since been refined using an automated procedure for constrained difference Fourier refinement (Chambers and Stroud, 1976) . The agreement index was R = 23 .5% for all re0cctions observable lo a resolution of 1.5 A at the lime this comparison was made. The crystals were grown from 5.8% MgS04 solutions at about pH 7.5. X-ray data are observable to I .I A resolution and, so far, recorded 10 1.5 A (22 117 observable renections out of 35 566 independent rencclions). There are well-defined nnd small conformational changes between this structure and that of ben,.amidine-trypsin (Krieger cl al., 1974b). These changes have been shown by recomputing the maps with more recent phases not to be an artificial feature of the difference map, as suggested by Bode and Schwager ( 197 5a ), based on a comparison of their native trypsin structure with the 1.8 A refined structure ofbenzamidine-inhibitcd trypsin . The orthorhombic DIP-trypsin structure (Chambers and Stroud, 1976) is now among lhe most highly refined of any enzyme structure (R = 23.5% to 1.5 A resolution), as is the more recently determined structure for benzamidine-trypsin of Bode and Schwager ( 1975a) (R = 22.9% to 1.8 A resolution). No comparison between these two independenlly refined tryP5in structures has yet been made. The comparison of Huber cl al. (1974) was made between their refined structure and our initial unrefined wire-model coordinates for DIP-trypsin derived from the first 2.7 A MIR density map. Wire-model coordinates such as these arc inadequate for detailed comparison, even when the map is of the !).ighesl quality (Chambers and Stroud, 1976). • A comparison between our trypsin structure and a-chymotrypsin (Birktoft and Blow, 1972), similar lo that made by Huber ct al. (1974) between their trypsin structure and achymotrypsin, was carried out using a rigid-body least-squares computer program to superpose lhe two structures. Atoms compared in this manner included all a -carbon atoms, except in regions where amino acid deletions or insertions occur. The residue numbers included were thus the same as those used by Huber et al. (1974). The average positional difference for a-carbon atoms of these residues was 0.80 A, a value similar to that of 0.75 A obtained by Huber et al. (1974) in their comparison of all main chain atoms for these residues (see Figure 4a). The similarity between this result and that of Huber ct al. (l 974) suggests that the DIP-trypsin refinement has resolved the discrepancies reported by lho&e authors between the 2.7 A unrefined DIPtrypsin structure and their 1.9 A refined lrypsin-PTI structure. The LpT structure crystallizes preferentially below pH 5.5 and the crystals arc isomorphous with those of Tg. This ~truclurc was compared directly with Tg by computing difference maps with amplitudes /lF • IFT,1 - IFi.pTI, and phases for Tg to a resolution of 2.1 A. This difference map is 70. STRll<'TlJRl; ot-· IIOVINF TRYPSINOOEN I\T 1.9-A a more sensitive lcsl of small conformulional changes than comparison of coordinates, and is correspondingly les., prone to interpretative error. Such n map can clearly show changes which arc smaller than the estimated error in the currcnl set of atomic positions. Thus, "identical position" in the comp.irisnn of atomic positions between Tg and l.pT probably implies identity within :t:0.1 A (which for II single carbon atom would result in II p.1ir of peaks of 5<1 above lhc noise level in the mup), whereas lhe same phrase applied lo the Tg/DIPT comparison implies equivalence lo within about :t:0.3 A. It musl be noted lhat there is n strong possibility lhal the low-pH form of DIP-trypsin has a different structure from the nculral-pll form, as do low- and neutral-pH a-chymolrypsin (Tulinsky et al., 1973; Mavridis ct al., 1974). Further, there is the pos.,ibilily that the isomorphic low-pH trypsin structure is more like trypsinogen in some ways, and thus the difference map procedure would fail to identify any difference between 1hc two proteins at such sites. In fact, the difference map is very clean in many areas where the neutral pH structures appear h> be slightly different. However, the activity of the enzyme depends on only a single ionization between pH 2 and 8, and this has been identified as the ionization of Asp- I 02 at the active site (Koeppe and Stroud, 1976). Therefore, any pHdependent structure change which docs occur between pH 7 and 5 in trypsin is not a change which is rcnccted in the inherent activity of lhe en,.ymc, unless it can be associated dircclly with ioni,.ation of Asp-I 02. Thus, the changes between the isomorphnus structures as seen in the difference Fourier arc more likely to pick out the important differences for the change in catalytic activity in a direct way. This adds emphasis to this comparison as an assay for the important catalytic change. Similarities in Structure. The atomic coordinates for DIPT were rota led to match those of Tg by least-square., minimi1.ation of the distances between equivalent atoms in the two molecules. Regions of major change in structure, i.e., residues It, -19, 14.l 152, 18'1 195, and 215 · 220, were left out of the rotation calculation, as were residues 76 --80 where the differences arc most probably due lo errors in lhe model (sec below). The relined equation, which described rotution of the DIPT coordinates x. )', z (listed in angstroms with respect lo the a, h. c axes) 10 match 1hc trypsinogen sci (x', y', z'h, (listed in angstroms with respect to an orthogonal axial set, such lhal x' is perpendicular to the crystallographic b, c axes; y' and z' are parallel to the hand,- axes, respectively), is: x') ( y' z' E 0.9724 (-0.0017 1.0000 0.0052 0.0028 -0.2333 -0.2333)(x) 0.0046 y -0.9724 z + ( 1.82) -II.Ob -11.47 (I) Thi• rotation matrix cl=ly approximates the form of a simple ro1atiun of I .l .5° abuut 21 axes in the two structure.~ 0 ~ I ( 0 0 -0.23 -~23) -I It therefore explains the formation of the orthorhombic-tri- l{f:Sol.UTION gonal twinned crystals and the congruence of the two a axial dimensions which are within 0. 7% of each other (see Figure I, and Experimental Section). The histogram shown in Figure 4b indicate., the distance between all a carbons for residues common 10 both structures after rotation. and clearly identific~ the regions of greatest structurnl change. The mean deviation between equivalent "-carbon aloms in the two structures was calculated by leaving oul the 32 residues wilh obviously different structure, and was ti.d,-,, • 0.38 A. Thus, the difference in conformation of the main chain for all these residues is less than for the a carbons of the best region (the C-tcrmina! a helix) of the Cg/CT comparison (Freer el al., 1970). The mean deviation of position between all atoms in the main chain and side groups for the same residues was ti.d = 0.45 A ( Figure 4c). Since the two structures are so similar in these regions, these dislances presumably represent the sum of errors in the two coordinate sets, coupled with real differences in structure. The errors in the trypsinogen coordinates arc clearly much greater than those of the more relined DIP'f set. Real differences in structure could be due to (I) crystal packing forces, (2) different solvent or counterion effects, and (3) lo genuine consequence., of the difference in covalent structure ofTg and DIPT. Most of the differences in side-chain orientation for structurally homologous regions do occur on the surface of the protein. It is notable that DIPT and Tg are much more homologous in structure than arc CT and Cg. In a rccenl analysis. Birktofl cl al. ( 1976) identified several regions of large difference between CT and Cg, associated with a relative movement of the two slruclural cylinders which arc formed by lhe first and second halves of the chymotrypsin sequence. There is no corresponding difference between DIPT and Tg. However, it must be noted that the CT structure was determined al low pH, ~4.5, where the enzyme has different structure from the neutral-pH form (Tulinsky el al., 1973), and crystallized as a dimer, where lhc'active center lay in the dimer interface. The Acti1>e Cmter. The structure around the active center of Tg is shown in Figure 5a. Crystallographic studies of chymotrypsinogen (Freer cl :ii., 1970) showed thal the orientation and immediate environment of the active-site residues, Asp102, His-57, and Scr-195. were very similar lo that found in chymotrypsin. However, a small change in the orientation of llc-99 pcrmilled limited access of Asp-102 and His-57 to solvent in CT, and the hydroxyl of Scr-214 was not hydrogen bonded to Asp-I 02 in Cg, raising a question as 10 the importance of subtle changes in this region for zymogen activation. Similar features are not seen in Tg, where both Leu-99 (t,,d•• = 0.5 A) and the 0-y ofSer-214 (M = 0.2 A) are in identical positions in the zymogen and enzyme; therefore, they cannot be common features of the mechanism of activation. The a carbon of Ser-195 differs by only 0.4 A between Tg and DIPT, and the 0-y of Scr-195 in Tg is hydrogen bonded to the ,N of llis-57 (Figure 5a) as it is in the benz.amidinelrypsin structure (Kreiger cl al., 1974b). The Tg-LpT difference map also indicates a small shift in the orientation of the imida,olc of flis-57 between the two structures. The presence of a DIP group al Scr-195 may be, in part, responsible for the observed shift of 0.3 A in the ring. This movement of lhc imidazole ring is similar 10 the shift observed previously in a comparison of ben,.amidine-trypsin and DIP-trypsin (Krieger cl al., 1974b). It is also clear from the latter comparison, and from more recently computed difference maps using improved phases, that there are several other small differences between these two structures. The fact that Bode and BtOCH~MISTRY. VOi. 16, NO. 4, 1977 659 71. KOSSIAkOFF ET Al. . a b FlfaJHli ~: Strut.:turc uf the adivc liitc in trypsinottcn (a) ii compared with strut.:lurc for lhc sumc region of trypsin (b) . Schwager ( 1975a) did nol lind any such differences between benz.imidinc-lrypsin and native trypsin suggests that the presence of the covalently bonded l>IP group may be respon· siblc for small differences in the DIP structure. Nevcrthclcs.s, no signifi,•ant movement of the o carbon of Ser-195 was ob•crvcd in l>I l'-lrypsin when compared with bcnzamidinc• trypsin. llirklofl cl al. ( 197(,) found that the hydrogen bond between the N, 1 of His-57 and the 0-y of Ser-195 was •trained in CT, while it was relatively normal in Cg. Our present results with the trypsin system differ in that the 0-y of Ser-195 (even though allachcd to the DIP group) in DIP-trypsin is still only 1. 15 A from the ideal position (described by Birktoft cl al., 1976) and only 0.1>7 A from the ideal position in trypsinogen. This is 10 be compared with the corrc.,ponding values of about 2.5 A for o-chymolrypsin and 0.7 Afor chymotrypsinogen. Since the hydrogen-bond distances between Hi•-57 and Scr-195 in lrypsinogcn arc normal (2.5 A) and can be made normal in trypsin without the DIP group (2.8 A in DIP-trypsin), it doc., not seem that significant distortion al this site can be con,idercd as u common feature of the activation mecha• nism of serine proteases. Th,• Sid,•·< 'huin Hindin11 Pock~t. Freer cl al . ( 1970) ~ucstioned whether the absence of a •pccific binding pocket could c•pluin the inactivity of t'g. Trypsinogen has a large and accc....sible binding pocket, although the structure und stability of the cavity is different from that found in DIPT (or in ben• 66() HIOt'lll ' MISTMY. VOi . 16, NO . 4, 1977 1.amidinc-1rypsin ( Krieger ct al., I 974b)) . The region of chain between 186 and I 94 is loosely orga• 11i1.ed and adopts several different conformations in Tg. The present structure for this chain was determined carefully. Evidence for this structure as the most predominant conformation depended on (I) the isomorphous-replacemenl density map; (2) 2F., - f ", or f'., - f ',. electron-density syntheses using calculated phases for a model from which this region was omilted; (3) the LpT-Tg difference map, which gives the most direct measure of the loose organiution of the 186- 194 chain. This difference map shows unequivocally that the chain structure is variable and that the low density is not due 10 errors in the phases. Two hydrogen bonds present in DIPT, between Yal-17 NH and the main-chain carbonyl of Asp-I 89, and between the Val- 17 C=O and Asp-189 NH, arc absent in Tg. These may contribute relative stability to the chain in DIPT. Ben1.amidinc is a competitive inhibitor of trypsin (K; = 1.8 X 10-< M) (Mares-Guia and Shaw, 1965) which binds inside the side-chain specific binding pocket of the enzyme (Krieger cl al., 1974b). Benzamidinc was present at 20 mM concentration (at least 10 1 times the K; for trypsin) in the trypsino• gen-crystallizing solution; yet, there was no bcnzamidinc bound to the procnzymc in the Tg structure. This emphasizes the fact that the structure for the binding site, as seen in the crystals, reprcscnta a functionally significant impediment to the binding capabilities of the site for benzamidinc, which is an analogue 72. ST K lJ ( T \JR (· U I-' IH> V I N 1: TI< Y PS I NO li EN AT I . I} - .i\ of the side chain of a substrate for trypsin. The binding constant for ben1.amidinc binding 10 trypsinogcn in solution was found to be about 10 1 times larger than for binding to trypsin (Ger1ler cl .ii., 1974). The binding sile seen in lhe Tg structure is large enough to acccpl a ben,amidinc molecule, and the facl 1h111 none was bound implies that the differences bctwcrn 1.yrnogcn and cn1.y111c struclurcs (us shown in Figure 5) arc quite udc~uatc to :u.:i.:ount for a large chnngc in spt:cifkity. A rclal ivcly large 111<>vc111enl of the backbone chain from rc!-!iducs IM 7 to 194 in the Cg/CT t.:umparison w.is correlated with lhc formal ion of the spccir.c binding site for substrate side chains in CT. a cavity which was only partially formed in C'~ ( hccr ct al.. I'/70) There arc much smaller but significant change, between Tg and l>IPT (or 1.pT) in the chain from rc,iducs 188,\ to l'IS (Figure)). The migration of Mcl-1 92 from a cmnplc1cly buried position (in the site occupied by the inn pair between lk- 16 and Asp-I 94) in Cg to the outside of the CT molc..:ulc also h:.ts no compara ble counterpart in the Tg/1.pTcomparisnn. The" carbon of(iln-192 in Tg moves by ~I A on activation . The side chain uf Asp-189 inside the specific binding pocket uf trypsin and responsible for the substrate specificity of trypsin lies within ~I A of the position found in Tg. ild, .. ,.,, = 0.'I A. Thus, it would seem that modificutiun, rather than generation or the spe\:ific cavity, musl be considered as a po~sihlc reason for 1.ymogcn inactivity in general. Studies of lhe residual activity of lrypsinugcn have gcncrnlly been carried out on simple eslers where hydrogen bonds 10 lhe main chain of 214 219 arc nol made. For NPGB hydrolysis by Tg, lhc binding constant was found tu be al least 10 1 times worse lhan for trypsin (Robinson ct HI., 1973). Changes in lhc binding-pocket slructurc could well be responsible for some or all of thi• poorer binding. He.tidue.v 2/4 210. In Cg, changes occurred in lhc main chain belwccn Scr-214 and Cys-220, a region known to be involved in hydrogen bonding toa peptide subslratc (Segal cl al., 1971; Swccl ct al., 1974; Huber ct al .. 1974; Blow, 1974b) . In Tg. the chain structure diffeni slightly from trypsin as shown in Figure 5. The C=O of Ser-214 (ild ~ 0.3 A). and 1hc NH (M ~ 0.5 A) and C'•·~O (dd = 1.4 A) of Gly-216 still point in the same direction. The peptide bond be1wcen Gly-219 and Cys-220 is apparent ly rotated by 180° wilh respect 10 trypsin al neutral pll (DIPT) . There is a difference in residues 214 -2 I 7 of Cg in thal the NH and C=O of Gly-216 would be unable 10 par1icipa1c in substrate binding (Freer ct al., 1970). This and the olhcr difference., in binding-site slructurc between C'g and Tg arc not inconsistent with the fact that the basic pancreatic lrypsin inhibitor (BPTI) will form a stoichiometric I: I complex with trypsinogcn and wilh trypsin or ehymolrypsin, hut will not so complex with chymotrypsinogen (Dlouha und Keil . 1%9). There was an error in the original placement of lhc side chain ofTrp-21 ~ in DIPT (Figure 8 of Krieger ct al., 1974b). Refinement of thal struclure revealed that the indolc ring is reversed, such ihal , N poinls in toward the center of the molecule ( Chambers and Stroud, 1976 ). The same configuration is found in Tg . This oricn1:11ion is now the same as thal found in C'T (Rirktoft and Blow, 1972). Modification ofTrp-215 in trypsinogcn blocks the binding of RPTI by lrypsinogcn (Kei l-Dlouha and Kei l, 1972). ASf'•/ 94. The side chain of Asp-194 forms a salt bridge with lie-lo in trypsin . The N terminal of lie-lo is not available in Tg, since ii lies in 1hc 1.ymogcn-activa1ion peptide. Asp-194 is in a similar oricntalion inside the trypsinogcn molecule, although the" carbon is moved by ild • 0.9 A. The side chain R 1: .SO I l IT I() N is surrounded by internal solvent molecules. There appears to be only one in1ramolccular hydrogen bond to Asp- 194 in Tg between 1hc06-I of Asp- 194and iheOyofSer-190. The side chain of Asn-143 (in Tg) poinls in toward 1hc solvent cavity around Asp- I 94. Freer cl al. ( I '/70) poslulated lhRI inlcraclion bc1wecn Arg- 145 and Asp-194 could contribute to charge stabili,.111ion of lhc ca rhoxyl group of Asp-194 in Cg possibly through waler molecules. In a more detailed analysis, Wright ( 197.1a.b) found that the guanidi nc group of Arg- 145 was too far from Asp-194 to participate directly in this role, but could nculrulizc lhe charged ca rbonyl via waler molecules hydrogen bonded lo the 6 nitrogen of His-40. The, nitrogen of His-40 was found lo be 3.5 A from the carboxyl group of Asp-I 94, and so was indircc1ly implicated as providing a possible pathway for proton transfer from the surface of the molecule to the buried carboxyl group. In Tg, neither l.ys- 145 nor His-40 lies close to the side chain llf Asp-194 (sec Figure 6), and, therefore, cannot be involved in stabili1.ing Asp-194. The side chain of l.ys-145 is on the outside of the molecule and the (N of Lys-145 is about IO A from Asp- I 94 . Indeed . !here arc no basic side chains close enough 10 contribute directly to charge ncutrali,.ation of Asp-I 94. His-40 is found in the same oricntalion. in Tg as in l.pT. and almost cxac1ly the same as in DIPT (t:.dc., = 0.4 A). The Nn- 1 of I lis-40 is hydrogen bonded to the 0-y of Ser-32 in both Tg and DIPT (and C'T, but not Cg). N,-2 of His-40 is hydrogen bonded 10 the C=O of Gly-193 in DI PT and CT, but is hydrogen bonded to an cxlcrnal solvent molecule in Tg. This change corresponds tu lhc movement of the Ca of 194 (ild = 0.9 A) and 193 (M = 2.0 A) toward 1he solvent-filled cavity around Asp-I 94 . Wright ( I973b) emphasized the apparent conservation of a basic group al rc.,iduc 145 of the trypsin-like enzymes, although 1hcrc are now exceptions in the sequences of dogfish trypsin (where the residue is mclhioninc), and in all the bacterial enzymes ( Bradshaw ct al., 1970; De Haen ct a l., 1975) . His-40 is also nol a conserved residue among trypsinli ke enzymes, and is Leu in bovine thrombin. Gly in bloodclot1ing factor X, and Met, Arg, or Leu in bacterial trypsin sequences (De Haen el al. , I 975). The "Auto/y.,;., Loop" (R,sidues 141- 15/ ). The largest conformational difference between Tg and DIPT or LpT occurs between residues 142 and 151. This sequence was referred 10 as 1hc "autolysis loop" by Blow ct al. ( 1969), since it contains a region where two amino acids are excised in c,-chymotrypsin. The difference map between Tg and LpT (Tg- LpT) showed clear negative density, which picked out the conformation of the Rulolysis loop in LpT. There was much less (~50%) positive density in 1hc map for this part of the chain, establishing that Ihe chain is ncxible in Tg and adopts a quite different con. formation there . (The LpTstructurc was identical to that of DI PT in this region .) A further change in the structure of this loop or of Lys-15-lle-16 would be required if an activating en1.ymc was to approach the Lys- I 5-llc-16 bond in the Tg conformation. Trp-141 is in csscnlially the same orientation in Tg as found in DIPT (ildc-.. = 0.6 A). The side chain of Asn-143, which was solvent accessible in DI PT, points in toward the cavity around Asp-I 94 in Tg (ildc-,. = 1.4 A) . The side chain of Thr- 144 (Tg) is moved by severa l angstroms toward the disulfide Cys-191 - Cys-220 (ildc,. "' 4.0 A), and residues 145 - 151 arc in 4ui1e different conformation in the two structures. They arc all solvent accessible in both Tg and DIPT. There may be a functiona l correlation belY•ccn the loose structure of this chain in the zymogcn, and the necessity for RIOCHEMISTNY . YU i. 16, NO . 4 , 1977 661 73. KOSSIAKOFf ~T AL . FIGURE 6: (a) Structure uound A•~ 194 and His-40 in Tg. (b) Structure around Asp-194 and His-40 in OIPT. or autolysis (Northrop et al., I 948). Figure 7 shows a schematic of the primary Ca2+ ion site in trypsinogcn and in trypsin (Chambers and Stroud, 1976). Coordination of the ion in both models is approximately octahedral and the ligands arc identical in the two structures, although there arc differences of up to 0.7 A in the absolute positions of the ligands involved (6dc,1♦ • 0.1 A). It seems likely that these differences may be due to error rather than to any real structure chongc, since there were no peaks in the Tg-LpT difference map around the ion. The peak of electron density assigned to the Ca 2+ ion in trypsinogen was the highest peak on the entire Fourier map, even during the first stages of Fourier refinement at 2.1 A resolution and site occupancy refined to I8 electrons. The primary Ca 2+ binding site in trypsin was first identified and described in detail by Rode and Schwager ( 1975b) based on analysis of an independently refined trypsin structure seen at 1.8 A and the Cu,. content of their crystals. No comparison of their coordinates with ours has been made, although the structures for the Ca2+ site in all three determinations seem to be essent ially the same. In our case, no Ca2+ was added to any of the crystallizing solutions and no attempt has been made to determine the Ca2+ content in the crystals. The ion found at the primary site in DIPT is probably a magnesium ion present in the crystallizing liquor (5.8% MgSO.). The electron density at this site refined to that expected for a water molecule ( IO electrons) or a magnesium ion (Chambers a nd Stroud, I 976). Since Ca2+ accelerates autoactivation of trypsinogen to trypsin, reasonable care was taken to exclude it from the trypsinogen-crystallizing solution. ca>+ was present in the buffer (0.02 Min CaCl2) used during the chromatographic purification, although the solution was s n,a,RI'. 7: /\ schematic of the coordination of ions at lhc primary Ca.'• ion bindm1e site in tryp,.in (lopl and tryr,'linoacn (bullom). Di:o;tanccs between the ion and chelating ligandi; arc indicalcd in ang~lroms. access of the activating enzyme to the Lys-15- lle- 16 bond. The sequences for residues 146 -- 154 arc extremely variable in related enzymes (De Haen ct al., 1975), and no specific roles can be assigned to any of the side chains of these residues in trypsinogcn, as judged from the Tg structure. The Primary Culdum Ion Binding Sitt. There arc two calcium ion binding sites in trypsinogcn. The primary site, with higher affinity for calcium ion (Kn~ 10-u M, Abita ct al.. t 969), is common to both trypsinogcn and trypsin. Occupancy of this site stabilizes the protein toward thermal dcnaturation 662 8 IO l' It EM IS TRY. VOL I 6, NO . 4 , I 97 7 74. STRIJ('TIJRE 01 .- HOVINI : TRYPSIN(HiFN AT 1. 9 later dialy,.cd using lurge volume., of Io - ' M HCI. Thus, it was surprising to find that the bound ion at the primary ion site in Tg refined to an occupancy of IXcleclruns. We presume this 10 be a single site, fully occupied by u single, tightl y-bound, Ca" ion. which remained as nn integral component of the ,ymogen thrnughoul. The difference map calculated between the isomorphous crystals of Tg a nd 1.pT showed no electron density in the CaH region. Therefore, the primary ca lcium ion binding sites in trypsinogcn, low-pH trypsin, or DIP-trypsin at about pH 7.5. a rc similar to one another and Lo the st ructu re described by Bode and Schwager (1975b) in bcnzamidinc-trypsin. The Second Cal+ Binding Site and the N -terminal Sequena. A secondary Ca ' • ion binding site of K" ~ I0- 1-• M exists only on the ,.ymogen . Binding of Cal+ at this site is essential for complete and efficient activa tion of trypsinogcn ( Northrop cl al., 1948). The effect of binding is to improve the substrate character of the activation peptide, so, favoring lryptic hydrolysis of the 1.ys-15 --llc-16 pe_plidc bond. (K,.. decreases by n factor of th ree, while k,,. 1 remains unchanged (Abila cl al .. 191\9).) Tryptic hydrolyses of the nonapcptidc Val-(Asp)4-l. ys-l le-Va l-Gl y and of the hcxapcptidc Val(Asp),-1.ys-lle-Val have been shown to be Ca" dependent in n very similar way, while hydrolyS<.-s nf peptide., containing only one uspurtntc do not depend on cu>+ (Delaage ct al., 1967; Abita ct al., 191,'J). Thus, Cu·'• inn binding to A•p-lJ and Asp-14 together is suflicient to explain the 1.ymogcn Ca,. ion effect (Abita ct al., 1969) . There is no corroborative crystallographic evidence as yet, since Cul+ was excluded from the crystallizing solution. The overall weak electron density for much of the N-tcrminal hexapeptidc of trypsinogen must be attribu ted to loca l disorder or loosely organi,.ed structure. rather than to phasing error. This conclusion is implied by the (Tg-LpT) difference map computed using phases for trypsinogen. In regions where there is difference in structure, this map should show a positive-density image for trypsinogcn, and a somewhat weaker negative image for trypsin. Yet, the positive density for trypsinogen residues I0-15 was consistently much lower in amplitude than the negative images for trypsin structure. The loose arrangement of this hcxopcptidc on the surface of the procn,.ymc is consistent with the finding that the synthetic nonapeptidc Val-(Asp).-Lys-llc-Val-Gly is hydrolyzed by trypsin at almost the same rate (k,,,, ~ 10-J s- 1) as the Lys-lle bond on the same sequence in Tg for which k,,,, ~ 2.5 X Io- ·1 s- 1 (Abita ct al.. I969) . Thus. it is clear that the slow hydrolysis of the I .ys - lie bond by trypsin docs not depend on any unique tertiary structure of these re.siducs. The four aspartatcs arc all ava ilable to modification by carbodiimides, and such modification docs not destroy activatibility of trypsinogen (Radhakrishnan ct a l.. 1967). This is in agreement with our finding that the Val- I0-Asp-14 sequence is not a determinant of tertiary structure. Thus, the binding of Cal+ to Asp- 13 and A•p-14 is probably independent of the protein structure as an ion-specific conformer, and is adequately explained by an induced structural change in the peptide ( I0-15) a lone, or by ncutrali1.ation of charge at Asp-13 and Asp-14. There is a clear necessity for trypsinogen among all pancreatic zymogcns to develop immunity to autolysis or activation by its own residual activity, since trypsin activates other zymogcns. The loosely structured hcxapeptidc is presumably protected against trypsin more by its negative charge than by tertiary structure. The Ly~ I 5-llc-16 is not inacceasible, although aomc reorientation of this chain or of the autQly1is loop must accompany enzyme binding to this bond. The fact that A Rl:SOl.lJTION this bond is still the most readil y hydroly1.ed is consistent wi th the fa\..'l thnl comp:11..·t native slructures are generally mor~ resistant to degradation thun arc denatured ones. The 'I' and ,J, angles on either side of l.ys-15 ure not dissimilar Lo those found for the susceptible I .ys-15 bond on Lhc trypsin inhibitor (a good substrate model). Since it is clcarly proven that both this chain and the neighboring autolysis loop have little to restrain their configuration, the reorientation nCC<.-ssary for access by the activating enzyme should be energetically easy to accomplish. Because of disorder. no real density for the side chains of Asp- I I or Asp- I 2 exists. The Reasons for Zymogen Inactivity. Freer ct al. ( 1970) and later Wright ( 1973b) identified four possible factors which might be responsible for inactivity of chymotrypsinogen . We now sec that two of them do not seem to be shared by trypsinogen, since (I) the potential source of a hydrogen bond from Gly-216 NH is not excluded, although this group is moved by 0.5 A from its "substrate-binding" position in the free enzyme, and (2) the side chain of Gln-192 occupies a fairly innocuous position outside the trypsinogcn molecule, whereas the corresponding Met- 192 of chymotrypsinogcn was shown to undergo a complete rcorgani,.ation upon activation. The third factor. an incompletely formed binding site in Cg. doc, have a counterpart in Tg, a lthough the binding pocket is open und accessible in Tg, whereas it was less so in Cg. Nevertheless, alterations in the binding site arc su flicienl to alter or remove specificity. The remaining factor was the lack of an important hydrogen-bonding group in the NH of Gly-193. Steitz ct al. ( 1969). Henderson ( 1970), and later Robertos el al. ( 1972) postulated that the NH groups of Gly-193 a ndof Scr-195 could play an important catalytic role in stabilization of reaction intermediates. In chymotrypsinogen, the NH of Gly-193 points away from the position for the carbonyl oxygen of the substrate (Freer ct al., 1970). In trypsinogen, the NH ofGly-193 and of Scr-195 both point toward a site not dissimilar to that found in trypsin and chymotrypsin. Nevertheless, this site, termed "the oxyanion hole" by Robertus et al. ( 1972), is removed by 1.9 A away from the position of the site in trypsin. Thus, oxyanion stabilization is possible in trypsinogcn, but not at the ideal site. Peaks in the (Tg-LpT) difference map for this region arc further testimony to the differences between these two molecules and also indicate a shift of about 2.0 A.. Loss of this component of the catalytic apparatus may well account for the slower reaction rate of the zymogen. There is a plausible connection between the movement of this site away from the cavity around Asp-194 and the formation of the salt bridge to lle-16 upon activation. There is also plausible connection between the small changes in the binding pocket and the relocation of lie! 6. Similar changes in the structure of the binding site could account for the poorer binding of substrate by the zymogens in general. Conclusion The structure of trypsinogcn is generally much closer to that of trypsin than in chymotrypsinogen to chymotrypsin. The structure of trypsinogen docs not exclude the possibility of substrate binding in a mode similar to that found for trypsin, although changes in the structure of this region contribute to an impaired or altered substrate binding mode-certainly for bcnzamidine and, most probably, for a substrate side chain. If the procnzyme is cons idem! to be rigid, then the general base cataly1t (the ,N of His-57) and the oxyanion-binding site formed by the NH groups ofGly- 193 and Ser-195 are too far BIOCHl!MISTRY. VOL. 16 , NO. 4, 1977 663 75. apart to cooperate in substrnte hydrolysis. F.vcn if the procnzymc structure was lo chungc on substrate binding, as seems likely, nonproductive binding may provide another important component for 1ymogen inactivity . Such binding might sti ll leave too ~reut a distunce, or an unfavoruble interaction between the substralc and clements of the cntn lyt ic center. The altered position of the chai n between 1.ys- l 8KA and Scr-19~ and the main chnin between Trp-21~ und Ser-217 could be responsible fur competitive, yet nonproductive, , ubstrute bindin~. as could lhc NH group of Oly-193, since they ure normally involved in orienting the substrate. Thc;c possibi lities remain as prime candidates in a universal scheme for inactivity of the trypsinogcn-like 1.ymogens. Acknowledgments We thank M. 0. Jones and R. Price for assista nce with calculations for the comparisons of structures. References Abita, J. P .. Dclaage, M .. Lazdunski, M., and Savrda, J. (1969), f:ur. J . Biochem. 8, 314-324. , Birktoft , J. J .. and Blow. D. M. (1972). J . Mo/. Biol. 68, I 87-240. Birktoft, J. J., Kraut, J.. and Freer, S. T. ( 1976), Biochemistry, /6, 4481. Blake, C. C. F .. Fenn, R.H., North, A. C. T .. Phillips, D. C., and Poljak, R. J. (1963), Acta Crysta/log,. 16. A77 . Blow, D. M. ( 195K), Proc. R. Sor. London. Sn. A 247, 302· 336. Blow, D. M. (l974u), /sr. J. Chem. 11, 483 - 494. Blow, D. M. ( 1974b). Bayer-Symp. 5, 473-483. Blow, D. M., Birktoft. J. J., and Hartley, B. S. ( 1969), Nature (London) 121. 337 - 340. Bode, W .. and Schwager, P. (1975a), J. Mo /. Biol. 98, 693717. Bode, W .. and Schwager, P. (1975b), FEBS Lm. 56, 139 143. Bradshaw, R. A., Ncurath, II., Tye, R. W .. Walsh , K. A., and Winter, W. P. (1 970), Nature (J..ondon) 116, 237-239. Chambers, J . L., Christoph, G. G., Krieger, M., Kay, L., and Stroud, R. M. (1974), Biochem. Blophy.,. Res. Commun. 59. 10- 14. Chambers, J. I... and Stroud, R. M. (1976), Acta Crysta/log,. (in press). Davie, E.W., and Ncurath, H. (1955), J. Biol. Chem. 112. 515-529. De Haen. C., Ncuruth. H .. a nd Teller, D. C. ( I 975), J . Mo/. Biol. 91. 225 -259. Delaage, M .. Dcsnuclle. P., Lazdunski, M .. Bricas, E .. and Savrda, J. ( 1967), Biochem. Biophy.,. Re.,. Commun. 29. 235--240. Dickerson , R. E .. Kendrew, J. C., and Strandberg, B. E. ( 1961 ), Acta Cry.,tal/ogr. I 4, 11 88 -1195. Dickerson, R. E.. Weinzierl, J.E., and Palmer, R. A. ( 1968), Acta Cry.<tal/ogr .. Sect. BU. 997 - 1003. Dlouha, V.,and Keil. B. (1969),FEBS Lett. 3. 137-140. Freer. S. T .. Kraut, J., Robertus, J. l>., Wright. H. T .. and Xuong. Ng. H. ( I970), Biochemistry 9. I997 - 2009. Gertlcr, A., Walsh, K. A., and Neurath, H. (1974), Biochemistry IJ. 1302 .. 1310. Hartley, B. S. ( 1970), Phi/os. Trans. R . Soc. London, Ser. B 664 RIOCIIEMISTRY. VOL 16, NO 4, 1977 257, 77 - 87. Henderson, R. (1970),J. Mo/. Bio/. 54. 341 - 354. Huber, R., Kukla, D.. Bode, W .. Schwager, P., Bartels, K., Dciscnhofer, .I., and S teigcmann, W. ( I 974), J. Mo /. Biol. R9. 73 · 101. Kcil-l>lou ha, V.. and Keil, B. ( I972). J. Mo/. Biol. 67. 495 505. Koeppe, R. E., II , and Stroud, R. M . ( 1976), Biochemistry 15. .14~0 3458. Krieger, M., Chambers, J. I.., Christoph, G. G., Stroud, R. M., und Trus, B. L. (1974a). Acta Crysta/log,. Sect. A 30, 740-748 . Krieger. M ., Kay, L. M., and Stroud, R. M. (1974b), J. Mo/. Biol. 83, 209-230. Marcs-Guia, M., and Shaw, E. (1965), J. Biol. Chem. 240, 1579-1585. Maroux, S ., Baratti, J., and Dcsnuelle, P. ( I 971 ), J. Biol. Chem. 246, 5031 - 5039. Mavridis, A., Tulinsky, A., Liebman, M. N. (1974), Biochemistry 13. 3661-3666. Mikes, 0., Tomasek, V., Holeysovsky, V., and Sorm, F. (1967), Collect. Czech. Chem. Commun. 32, 655-677. Neurath, H. ( 1975), in Cold Spring Harbor Symposium on Proteases and Biological Control, 1975, Reich, E., Rifkin, D. B., and Shaw, E., Ed., Cold Spring Harbor, N.Y., Cold Spring Harbor Laboratory, pp 51-64. Neurath, H., and Dixon, G . H. (1957), Fed. Proc .. Fed. Am. Soc. /:.'xp. Biol. 16. 791-801. Northrop, J. H., Kunitz, M., and Herriot, R. M. (1948), Crystalline Enzymes, New York, N.Y., Columbia University Press, pp 133- 136. Radhakrishnan, T. M., Walsh, K. A., and Neurath, H. (1967), J. Am. Chem. Soc. 89, 3059-3061. Richards, F. M. (1968), J. Mo/. Biol. 37, 225-230. Robertus, J. D., Kraut, J., Alden, R. A., and Birktoft, J. J. ( 1972), Biochemistry I I . 4293-4303. Robinson, N. C., Neurath, H., and Walsh, K. A. (1973), Biochemistry 12. 420-425. Salemme, F. R., and Fehr, D. G. (1972), J. Mo/. Biol. 70. 697 -700. Schroeder, D. D., and Shaw, E. (1968), J. Biol. Chem. 243, 2943-2949. Segal, D. M., Powers, J.C., Cohen, G. H., Davies, D.R., and Wi lcox, P. E. ( 1971 ), Biochemistry 10, 3728-3738. Sigler, P. 8., Blow , D. M., Matthews, B. W ., and Henderson, R. (1968),J. Mo/. Biol. 35, 143-164. Steitz, T. A., Henderson, R., and Blow, D. M. ( 1969), J. Mo/. Biol. -16, 337-348. Stroud, R. M., Kay, L. M., and Dickerson, R. E. (1971), Cold Spring Harbor Symp. Quant. Biol. 36, 125-140. Stroud, R. M., Kay, L. M., and Dickerson, R. E. (1974), J. Mo/. Biol. 83, 185- 208. Sweet, R. M., Wright, H. T., Janin, J., Chothia, C.H., and Blow, D. M. (1974), Biochemistry 13. 4212-4228. 'fulinsky, A., Vandlen, R. L., Morimoto, C. N., Mani, N. V., and Wright, L. H. (1973), Biochemistry 12. 4185-4192. Wa lsh, K. A.,and Neurath, H. (1964) , Proc. Natl. Acad. Sci. U.S.A. 51. 884 - 889. Wright, H. T. (1973a), J. Mo/. Biol. 79, 1-11. Wright, H. T. (1973b),J. Mo/. Biol. 79, 13-23. Wyckoff. H. W., Tscrnoglou, D., Hanson, A. W., Knox, J. R., Lee, B., and Richards, F. M. (1970), J. Bio/. Chem. 245. 305-328. 76. Appendix 5 Mechanisms of Zyrnogen Activation 77. MErnANISMS OF ZYMX;EN ACTIVATION Robert M. Stroud, Anthony A. Kossiakoff1 , and John L. Chambers Nonnan W. Church Laboratory of Chemical Biology California Institute of Technology Pasadena, California 91125, U.S.A. (Ann. Rev. Bioph. Bioeng.--in press) 1Present address: Brookhaven National Laboratory Upton, Long Island, New York 11973, U.S.A. 78 . Introduction It is now well recognized that proteolytic enzymes play key roles in the regulation of or control over the action of other pro teins (1,2,3) . Such enzymes can be found in all species from bacteria to man , and in control of diverse systems which include honnone production, bacteriophage assembly (4), development, fertilization (S), digestion, defense against invading organisms (6), and tissue repair (7,8) . In most cases the proteolytic enzymes are known to be synthesized as inactive precursor proenzymes , or zymogens (9) . They are activated by proteolytic cleavage of a single peptide bond in the proenzyme, and so become catalytically active. Further control over the degree of specificity for a target molecule or molecules is determined by the degree of specificity inherent to the enzyme. Further control over the time and location of action is often carried out by protein inhibitors of the requisite specificity . There are several different classes of proteolytic enzymes which were first classified according to their susceptibility toward different inhibitors (10), and later according to their homologous amino acid sequences or structures within each class (11). There are many excellent papers and reviews which deal with all aspects of the collected information of the proenzymes of proteases (for example 1,2 , 3, 9,11). Therefore in this review we shall consider the single question of why the zymogens of the serine proteases--the enzyme class about which most is known--are relatively inactive . We consider this in the 79. light of most infonnation about the known structures of t wo proenzymes, chymotrypsinogen (12) and trypsinogen (13) , and their activated counterparts , a-chymotrypsin (14) and trypsin (15). Inactivity of t he Proenzyme Several reasons have been proposed for the relati ve inactivity of these proenzymes . The question became particularly intriguing when i t was found that the active site in chymotrypsinogen appeared to be preformed and accessible , much as it is in chymotrypsin (12; and Figures 1 and 2). The finding of similar structures at the active center would suggest that the ubiquitous basis for the unusual reactivity of the catalytic groups in the enzyme should be present in the proenzyme ; that the proenzyme should be catalytically active. Indeed, some re- activity does exist in the proenzyme (9 , 16), although the catalytic rates for hydrolysis of simple esters are at least 10 4 - 10 7 times lower than for the enzyme. Attention was focused (12) on the sub- strate binding site , which appeared to be incompletely formed in chymotrypsinogen , although in trypsinogen the binding site for substrate side chains is open and quite large enough to accept the side chain of a normal substrate of trypsin (13; and Figures 3 and 4) . It is not at all obvious why a trypsin substrate should not bind to trypsinogen in a mode similar to that found for trypsin, although it is clear that substrate binding is impaired. 80. Birktoft, Kraut &Freer (17), as the result of a closer look at the comparison between active center structure in the chymotrypsinogen/chymotrypsin pair, proposed that the development of a high degree of strain in the hydrogen bond between His 57 and Ser 195 1 in the enzyme (Figure 2) but not in the proenzyme (Figure 1) may be a further contributor to the reactivity of the enzyme . Yet in the trypsin/ trypsinogen comparison such differences are certainly much less apparent if they are present at all (13; and Figures 3 and 4). There are two hydrogen bond donors in the enzyme which may stabilize the high energy transition states during substrate hydrolysis. the N-H groups of Gly 193 and Ser 195 (18). These are The N-H group of Gly 193 is not available for this role in chymotrypsinogen, while in trypsinogen the N---H of Gly 193 still points in the same direction as 0 in trypsin but is moved by 1.9 A from its site in the enzyme. Together, it seems that (i) the contribution of correct substrate binding, and (ii) the ability to stabilize the intermediates in catalysis are implicated most strongly as prerequisites for the enzyme to work efficiently. This leaves open the question of how much, if any, small changes in or around the groups that participate most directly in covalent catalysis do contribute to the difference in reactivity of proenzyme and enzyme. Since there is no clear answer to the question of the basis for such a significant change in the reactivity of proenzyme upon activation (a change which is crucial to the function of the many related serine proteases in a general sense), the status of the structures 81. and their accuracy and the conditions under which the structures for comparison were determined become particularly important . of these factors is included in Table I. A sunnnary At this stage the trypsin and trypsinogen structures have been subjected to more complete refinement and at a higher resolution than have the corresponding chymotrypsin pair. Indeed, the trypsin structure is probably the best determined of any enzyme structure (15,19). A portion of the most re- cent electron density map around the active center of DIP-trypsin, obtained by Chambers &Stroud (15), is shown in Figure 5. The pH-optimum of the enzymes lies between pH 7 and pH 9. The trypsin pair were determined at pH 'v 7, while a-chymotrypsin was initially determined at pH 'v 4.5 where the molecule has low activity and exists mainly as a dimer (20). However, the a-chymotrypsin structure has since been studied up to pH 'v 8.6 by Mavridis, Tulinsky &Liebman (21) who reported a number of significant pH-dependent changes which occur between pH 'v 4. 5 and pH 7.5. The fact that the a-chymotrypsin structure crystallized as a dimer is a further factor which must be taken into account in a detailed structure comparison since Tulinsky et al. (22) also found the two molecules which form the dimer to be of somewhat different structure within the dimer interface . This is a critical region for this analysis, since it includes the reactive center of a-chymotrypsin. We therefore consider the trypsin/trypsinogen pair of structures most suitable as the primary basis of comparison between enzyme and proenzyme for the purpose of this review. Further, the comparison 82 . between chymotrypsinogen and trypsinogen should serve to isolate characteristics common to both proenzymes , which will limit the primary reasons for inactivity to the lowest common factor. In this respect, there are real and significant differences between the structures of the two proenzymes . Taking for granted that the en- zymes themselves are catalytically active (something which would not be obvious if the structures alone were viewed in the absence of such knowledge), we will compare the two proenzymes emphasizing the question as to why they should be relatively inactive. The answer to this question implies acceptance of the many well-established details of substrate binding and activity sunnnarized, for example, by Blow (23). This question cannot yet be answered simply. Some of the small- est alterations in structure may be crucial for the answer, while others may be merely consequences of crystal packing forces, etc . Therefore, we first describe the conditions under which the trypsinogen and trypsin structures were determined . It must be noted that both trypsin and a-chymotrypsin were crystallized from high salt solutions, while trypsinogen and chymotrypsinogen were crystallized from ethanol solutions. Trypsin can also be crystallized from ethanol solution . Trypsino..,,e,cn structure--Crystals of bovine trypsinogen were obtaine<l by vapor diffusion of a 3% solution of protein against 30 90 ethanol/water (13) . Benzamidine was added to the solution in a 83. concentration of 2.5 mg/ml to prevent possible trypsin-mediated autolysis of the zymogen during the several weeks required for crystallization. It was essential to eliminate divalent cations from the crystallizing solutions, since they mimic the effects of Ca++ and lead to autoactivation and autolysis of the trypsinogen in solution (24). Crystals used for structure detennination were grown from solutions adjusted to about pH 7.5, although isomorphous crystals were obtained throughout the pH range 5 - 8. Crystals grown this way were trigonal, space group P3121 with cell parameters~= 55.17 A and~= 109.25 A . The unit cell contains one molecule per asyrrnnetric unit (13). Trypsin structure--At low pH, DIP-trypsin 2 crystals which were isomorphous with the trypsinogen crystals were obtained. These crys- tals could be grown at pH 4 - 5.5 from either ethanol/water mixtures or from MgSO4 solutions . At higher pH (5.7 - 8.5) trypsin crystal- lizes exclusively in the orthorhombic space group P2 12121 (15,25,26). The trigonal crystals were found to be unstable above pH 5.5, which suggests that there is some type of structural reorganization which occurs around pH S.S. Thus there are three trypsin-related structures used as the basis for comparison between proenzyme and enzyme. The neutral pH, or orthorhombic fonn of DIP-trypsin--the most highly refined structure-is compared with the neutral pH, trigonal fonn of trypsinogen. The fact that DIP-trypsin crystallizes at low pH, in a fonn isomorphous with trypsinogen, is useful in that it provides an accurate means of 84. assay for structure change between the two molecules by difference Fourier map techniques. These methods can pinpoint differences in structure which are on the same scale or smaller than the average expected errors in atomic coordinates of the refined structures from different crystal forms. Thus, there is an independent means of assay for structure change which is almost independent of the absolute accuracy of the orientation of, for example, a single side chain which may not be accurately placed by the constrained difference Fourier refinement procedures in either structure alone. There is a further complexity to this comparison, which may be used to advantage. There is a strong possibility that the low pH fonn of DIP-trypsin has a different structure from the neutral pH fonn, as does chyrnotrypsin (21,22). Further, there is the possibility that the low pH trypsin structure is more like trypsinogen in some ways, and thus the difference map procedure would fail to identify any difference between the two proteins at such sites. In fact, the difference map is very clean in many areas where the neutral pH structures appear to be slightly different. However, the activity of the enzyme depends on only a single ionization between pH 2 and pH 8, and this has been identified as the ionization of Asp 102 at the active site (27). Therefore, any pH-dependent structure change which does occur between pH 7 and pH Sin trypsin is not a change which is reflected in the inherent activity of the enzyme, unless it can be associated directly with Asp 102. Thus, the changes between the isomorphous structures are more likely to pick out the important changes for the change in 85. catalytic activity, the subject of this review, in a direct way . Similarities in Structure--The atomic coordinates for DIPT were rotated to match those of Tg by least-squares minimization of the distances between equivalent atoms in the two molecules . Regions of major change in structure, i.e . , residues 16-19, 143-152, 186-195, and 215-220, were left out of the rotation calculation, as were residues 76-80 where the differences may be due to errors in the model (13). The histogram shown in Figure 6a points up the distance between all alpha carbons for residues corranon to both structures after rotation, and clearly identifies the regions of greatest structural change . The mean deviation between equivalent alpha-carbon atoms in the two structures was calculated by leaving out the 32 residues with obviously different structure, and was ~dca= 0.4 A. The difference in conformation of the main chain for all these residues is about the same as for the alpha carbons of the best region (the C-tenninal alpha helix) of the Cg/CT compar ison (12) . The mean deviation of position between all atoms in the main chain and side groups for the same resi0 dues of Tg and DIPT was ~d = 0. 45 A. Since the two structures are so similar in these regions, the distances preslUilably represent the SlUJl of errors in the two coordinate sets, coupled with real differences in structure. Most of the differ- ences in side chain orientation for structurally homologous regions occur on the surface of the protein, and are probably not of primary importance for our purpose here . 86. The Active Center--The structure around the active center of Tg is shown in Figure 3. Crystallographic studies of chymotrypsinogen (12) showed that the orientation and :immediate environment of the active site residues, Asp 102, His 57, and Ser 195, were very similar to those found in chymotrypsin. However, a small change in the ori- entation of Ile 99 permitted limited access of Asp 102 and His 57 to solvent, and the hydroxyl of Ser 214 was not hydrogen bonded to Asp 102 in Cg, raising a question as to the importance of subtle changes in this region for zymogen activation. Similar features are 0 not seen in Tg, where both Leu 99 and Ser 214 (~dca = 0.27 A, ~d 0 = 0.3 A, respectively) are in identical positions in the zymogen ca and enzyme, therefore they cannot be connnon features of the mechanism of activation. The alpha carbon of Ser 195 differs by only 0.4 A between Tg and DIPT, although the Oy of Ser 195 is still hydrogen bonded to the EN of His 57 (Figure 3) as it is in the benzamidine trypsin structure (28). The Tg/LpT difference map also indicates a small shift in the orientation of the imidazole of His 57 between the two structures. The presence of a DIP group at Ser 195 may be in part responsible for the 0 observed shift of 0. 3 A in the ring. This movement of the imidazole ring is similar to the shift observed previously in a comparison of benzamidine trypsin and DIP-trypsin (28). The fact that Bode and Schwager (19) did not find any such differences between benzamidine trypsin and native trypsin suggests that the presence of the covalently bonded DIP group may be responsible for small differences in 87. the DIP structure. Nevertheless, no significant movement of the alpha carbon of Ser 195 was observed in DIP-trypsin when compared with benzamidine trypsin. The change in position of O:x Ser 195 seen in Tg may be of some importance for the relative inactivity of trypsinogen. It lies at the end of the sequence of residues 186-195, which all occupy different positions in Tg and DIPT as they do in the comparison of Cg with CT. Birktoft, Kraut &Freer (17) observed a striking difference between the structures of CT and Cg in the active center. They defined an "ideal" site for the Oy of Ser 195 relative to the side chain of 0 His 57. This was chosen to lie 2.8 A away from the sN of His 57 and in the direction of the sp 2 orbital of the sN. While they found that 0 the Oy of Ser 195 was within 0.7 A of this site in Cg, it could not be 0 . brought closer than about 2.5 A to this site in CT by any manipulation of these side chain orientations (17). They raised the question of whether a strained hydrogen bond at this site in the enzyme could contribute to enhanced catalytic efficiency. The Oy of Ser 195 in Tg and in DIPT (where the Oy is covalently bonded to the DIP group) are not so dramatically different. Although the actual position of the Oy 0 differs by almost 1 A between these structures, the Oy lies at 0.7 and 0 1.1 A, respectively, away from the "ideal" position, and it is not clear that this difference is enough to account for any significant contribution of distortion at this site to the catalytic mechanism. Birktoft, Kraut f, Freer (17) have also identified other quite large differences between CT and Cg in this region of interface between the 88. two "structural cylinders" (14) of CT . ference in the Tg-DIPT comparison. We see no corresponding dif- Perhaps the fact that the er structure existed as a dimer is in part responsible for alteration in the structure of a -chymotrypsin, relative to the neutral pH form of the enzyme? This does not rule out the possibility that even a small change in the active center structure could contribute to the generation of enzyme activity, but at present there is no connnon factor which can be identified . The Side Chain Binding Pocket--Freer et al . (12) questioned whether the absence of a specific binding pocket could explain the inactivity of Cg . Trypsinogen has a large and accessible binding pocket (see Figure 3), although the structure of the cavity is different from that found in DIPT or in benzamidine trypsin (see Figure 4) . Benzamidine is a competitive inhibitor of trypsin (K. = 1.8 x 10-S M) (29), which binds inside the side-chain specific 1 binding pocket of the enzyme (19 , 28) . Benzamidine was present at 20 rrM concentration (at least 10 3 times the K. for trypsin) in. the 1 trypsinogen crystallizing solution, yet there was no benzamidine bound to the proenzyme in the Tg structure . This emphasizes the fact that the structure for the binding site, as seen in the crystals, repre sents a functionally significant impediment to the binding capabilities of the site for benzamidine, which is an analog of the side chain of a substrate for trypsin. The binding constant for benzamidine binding 89 . to trypsinogen in solution was found to be about 10 5 times larger than for binding to trypsin (30). The binding site seen in the Tg struc- ture is large enough to accept a benzamidine molecule , and the fact that none was found implies that the differences between zymogen and enzyme structures (as shown in Figures 3 and 4) are quite adequate to account for a large change in specificity. A relatively large movement of the backbone chain from residues 187 to 194 in the Cg/CT comparison was correlated with the formation of the specific binding site for substrate side chains in CT, a cavity which was only partially formed in Cg (12). There are TIR1ch smaller but significant changes between Tg and DIPT (or LpT) in the chain from residues 188A - 195. The migration of Met 192 from a completely buried position (in the site occupied by the ion pair between Ile 16 and Asp 194) in Cg to the outside of the CT molecule also has no comparable counterpart in the Tg/LpT comparison. 0 Gln 192 in Tg moves by 0.9 A on activation. 0 moves by about 4 A. The alpha carbon of The side chain of Gln 192 The side chain of Asp 189 inside the specific binding pocket of trypsin and responsible for the substrate specifico ity of trypsin lies within 0.9 A of the position found in Tg, 6dcal 89 = 0.9 A. Thus, it would seem that modification rather than generation of the specific cavity must be considered as a possible reason for zymogen inactivity in general. Studies of the residual activity of trypsinogen have generally been carried out on simple esters where hydrogen bonds to the main chain of 214 - 219 are not made. For NPGB hydrolysis by Tg the 90. binding constant was found to be at least 10 3 times worse than for trypsin (31). Changes in the binding pocket structure could well be responsible for some or all of this poorer binding. Residues 214-220--In Cg changes occurred in the main chain between Ser 214 and Cys 220, a region known to be involved in hydrogen bonding to a peptide substrate (32-35). In Tg the chain structure differs slightly from trypsin as shown in Figure 3. The C=O of 0 9 of Gly 216 still point in the same direction. The peptide bond be- 0 Ser 214 (td = 0.3 A) and the N-H (td = 0.5 A) and C=O (td = 1.4 A) tween Gly 219 and Cys 220 is apparently rotated by 180° with respect to trypsin at neutral pH (DIPT). There is a difference in residues 214 - 217 of Cg in that the N-H and C=O of Gly 216 would be unable to participate in substrate binding (12). This and the other differ- ences in binding site structure between Cg and Tg are not inconsistent with the fact that the basic pancreatic trypsin inhibitor (BPTI) will form a stoichiometric, 1:1 complex with trypsinogen and with trypsin or chymotrypsin, but will not so complex with chymotrypsinogen (36) . There was an error in the original placement of the side chain of Trp 215 in DIPT (Figure 8 of reference 28). Refinement of that struc- ture revealed that the indole ring is reversed, such that £N points in toward the center of the molecule (15). found in Tg. The same configuration is This orientation is the same as that found in CT (14) . Modification of Trp 215 in trypsinogen blocks the binding of BPTI by trypsinogen (37). 91. Asp 194--The side chain of Asp 194 forms a salt bridge with Ile 16 in trypsin. The N-tenninal of Ile 16 is not available in Tg, since it lies in the zymogen activation peptide. Asp 194 is in a similar orientation inside the trypsinogen molecule, although the 0 alpha carbon is moved by ~d = 0.9 A. internal solvent molecules. The side chain is surrounded by There appears to be only one intramolecu- lar hydrogen bond to Asp 194 in Tg between the 06-1 of Asp 194 and the Oy of Ser 190. The side chain of Asn 143 (in Tg) points in toward the solvent cavity around Asp 194 (see Figures 7 and 8). Freer et al. (12) postulated that interaction between Arg 145 and Asp 194 could contribute to charge stabilization of the carboxyl group of Asp 194 in Cg--possibly through water molecules. In a more de- tailed analysis Wright (38,39) found that the guanidine group of Arg 145 was too far from Asp 194 to participate directly in this role, but could neutralize the charged carboxyl via water molecules hydrogen bonded to the oN of His 40. The sN of His 40 was found to be 3.5 A from the carboxyl group of Asp 194, and so was indirectly implicated as providing a possible pathway for proton transfer from the surface of the molecule to the buried carboxyl group (see Figure 9). In Tg neither Lys 145 nor His 40 lies close to the side chain of Asp 194, and therefore cannot be involved in stabilizing Asp 194. The side chain of Lys 145 is on the outside of the molecule and the sN of 0 Lys 145 is about 10 A from Asp 194. Indeed, there is no basic side chain close enough to contribute directly to charge neutralization of Asp 194 . His 40 is found in the same orientation in Tg as in LpT, and 92. 0 almost exactly the same as in DIPT (6dca = 0.4 A) . The No-1 of His 40 is hydrogen bonded to the Oy of Ser 32 in both Tg and DIPT (and CT , but not in Cg). NE-2 of His 40 is hydrogen bonded to the C=O of Gly 193 in DIPT and CT, but is hydrogen bonded to an external solvent molecule in Tg. This change corresponds to the movement of the Ca of 194 (6d = 0.9 A) and 193 (6d = 2. 0 A) toward the solvent-filled cavity around Asp 194. Wright (39) emphasized the apparent conserva- tion of a basic group at residue 145 of the trypsin-like enzymes, although there are now exceptions in the sequences of dogfish trypsin (where the residue is methionine), and in all the bacterial enzymes (40,41). His 40 is also not a conserved residue among trypsin-like enzymes, and is Leu in bovine thrombin, Gly in blood-clotting factor X, and Met, Arg or Leu in bacterial trypsin sequences (41). The "Autolysis Loop" (Residues 142-151)--The largest confonnational difference between Tg and DIPT or LpT occurs between residues 142 - 151. This sequence was referred to as the "autolysis loop" (42) since it contains a region where two amino acids are excised in achymotrypsin. The difference map between Tg and LpT (Tg-LpT) showed clear negative density which picked out the conformation of the autolysis loop in LpT. There was much less c~ 50%) positive density in the map for this part of the chain, establishing that the chain is flexible in Tg and adopts a quite different confonnation there. structure was identical to that of DIPT in this region.) (The LpT A further change in the structure of this loop or of Lys 15 - Ile 16 would be 93 . required if an activating enzyme were to approach the Lys 15 - Ile 16 bond in the Tg confonnation . Trp 141 is essentially in the same orientation in Tg as found in DIPf (~dca = 0.6 A) . The side chain of Asn 143, which was solvent accessible in DIPT, points in toward the cavity around Asp 194 in Tg (~dca = 1.4 A) . The side chain of Thr 144 (Tg) is moved by several angstroms toward the disulfide Cys 191 - Cys 220 (~d ca. 0 4. 0 A) , and ~ residues 145 - 151 are in quite different confonnation in the two structures. They are all solvent accessible in both Tg and DIPT. There may be a functional correlation between the loose structure of this chain in the zymogen, and the necessity for access of the acti vating enzyme to the Lys 15 - Ile 16 bond. The sequences for resi- dues 146 - 154 are extremely variable in related enzymes (41), and no specific roles can be assigned to any of the side chains of these residues in trypsinogen, as judged from the Tg structure . The Effect of Divalent Cations on Trypsinogen--There are two calchnn ion binding sites in trypsinogen . The primary site, with higher affinity for calcit.nn ion(~'\., 10- 3 • 2 M, (43)), is conunon to both trypsinogen and trypsin. Occupancy of this site stabilizes the protein toward thermal denaturation or autolysis (24). The primary Ca++ binding site in trypsin was first identified and described in detail by Bode and Schwager (44) based on analysis of an independently refined trypsin structure seen at 1.8 crystals. A, and the Ca++ content of their The difference map calculated between the isomorphous 94. crystals of Tg and LpT showed no electron density in the Ca++ region . Therefore, the primary calcium ion binding site in trypsinogen and low pH trypsin are identical. There are small differences between this Ca++ site in Tg and DIPT (13). • b.ind.mg site • of KD ~ 10-l.S M exists · A secondary Ca ++ ion only on the zymogen. Binding of Ca++ at this site is essential for complete and efficient activation of trypsinogen (24), and this site is more pertinent to the subject of this review. The effect of binding is to improve the substrate character of the activation peptide, so favoring tryptic hydrolysis of the Lys 15 - Ile 16 peptide bond. by a factor of three while kcat remains unchanged (43).) m decreases Tryptic (K hydrolysis of the nonapeptide Val-(Asp) 4-Lys-Ile-Val-Gly, and of the hexapeptide Val-(Asp) 2-Lys-Ile-Val, have been shown to be Ca++ dependent in a very similar way, while hydrolyses of peptides containing only one aspartate do not depend on Ca++ (43,45). Ca++ ion binding to Asp 13 and Asp 14 together is thus sufficient to explain the zymogen Ca++ ion effect (43). There is as yet no corroborative crystallo- graphic evidence for binding at this site since Ca++ was excluded from the trypsinogen solution. The overall weak electron density for much of the N-terminal hexapeptide of trypsinogen must be attributed to local disorder of loosely organized structure rather than to phasing error. This con- clusion is implied by the trypsinogen/DIP-trypsin difference map crnnputed using phases for trypsinogen. In regions where there is dif- ference in structure, this map should show a positive density image 95. for trypsinogen, and a somewhat weaker negative image for trypsin . Yet, the positive density for trypsinogen residues 10 - 15 was consistently ITRlch lower in amplitude than the negative images for trypsin structure. The loose arrangement of this hexapeptide on the surface of the proenzyme is consistent with the finding that the synthetic nonapeptide Val-(Asp) 4-Lys-Ile-Val-Gly is hydrolyzed by trypsin at almost the same rate (kcat ~ 10 -3 -1 sec ) as the Lys - Ile bond on the same sequence in Tg for which kcat = 2.5 x 10- 3 sec-l (43). Thus, it is clear that the slow hydrolysis of the Lys - Ile bond by trypsin , a factor which is important in preventing premature activation of all trypsinogens, does not depend on any unique tertiary structure of these residues. The four aspartates are all available to modification by carbodiimides, and such modification does not destroy activatibility of trypsinogen (46). This is in agreement with our finding that the Val 10 - Asp 14 sequence is not a determinant of tertiary structure (13). The binding of Ca++ to Asp 13 and Asp 14 is probably in- dependent of the protein structure as an ion specific conformer , and is adequately explained by an induced structural change in the peptide (10 - 15) alone, or by neutralization of charge at Asp 13 and Asp 14. There is a clear necessity for trypsinogen among all pancreatic zymogens to develop i.rram.mity to autolysis or activation by its own residual activity, since it activates other zymogens. The loosely structured hexapeptide is presumably protected against trypsin more 96. by its negative charge than by tertiary structure. The hexapeptide sequence also provides a good substrate character for the physiological activator, enterokinase (47). The hexapeptide, with its dual role as most favored substrate for enterokinase, and a poor but still most favored substrate site for activation by trypsin, is highly conserved throughout all species (41). The Lys 15 - Ile 16 bond is not inac - cessible, although some reorientation of this chain or of the autolysis loop rrrust accompany enzyme binding to this site. The fact that this bond is still the most readily hydrolyzed by trypsin in the presence of Ca++ ions is consistent with the fact that compact native structures are generally more resistant to degradation than are denatured ones. The¢ and~ angles on either side of Lys 15 are not dissimilar to those found for the susceptible Lys 15 bond on the trypsin inhibitor (a good substrate model). Since it is clearly proven that both this chain and the neighboring autolysis loop have little to restrain their configuration, the reorientation necessary for access by the activating enzyme should be energetically easy to accomplish. In many of the trypsin-like enzymes involved in biological control mechanisms, the primary activating cleavage occurs at a site analogous to the Lys 15 - Ile 16 bond with respect to the catalytic component of the protein. This split may be far from the N-terminal of the whole protein since the activation peptides, or proteins released, often have separate functions (11). In prothrombin, for example, the activating cleavage occurs between residue? Arg 323 and 97. Ile 324 in the sequence. The B-chain (residues 324 - 582) is highly homologous with trypsin, while the A-chain, which remains attached to the B-chain via a disulfide bridge, is thought to have an important role in the blood-clotting mechanism (48). The Reasons for Zymogen Inactivity--Freer et al. (12) and later Wright (39) identified four possible factors which might be responsible for inactivity of chyrnotrypsinogen . We now see that three of them do not seem to be shared by trypsinogen, since (in Wright's notation): (i) the binding site of trypsinogen is open and accessible while it is not quite so open in chyrnotrypsinogen; (ii) the potential source of a hydrogen bond from Gly 216 NH is not excluded, although this group is moved by 0.5 A from its "substrate binding" position in the free enzyme; and (iii) the side chain of Gln 192 occupies a fairly innocuous position outside the trypsinogen molecule, whereas the corresponding Met 192 of chyrnotrypsinogen was shown to tmdergo a complete reorganization upon activation. The remaining reason was the lack of an important hydrogen-bonding group in the N-H of Gly 193. It has been pointed out that the N--H groups of Gly 193 and of Ser 195 could play an important catalytic role in stabilization of reaction intermediates (18,49,50). In chyrnotrypsinogen the N--H of Gly 193 points away from the position for the carbonyl oxygen of the substrate (12). In trypsinogen the N--H of Gly 193 and of Ser 195 both point toward a site not dissimilar to that fotmd in trypsin and chyrnotrypsin. Nevertheless, this site tenned the "oxyanion hole" by Robertus et al. 98. 0 (18) is removed by 1.9 A from the position of the site in trypsin . Thus, oxyanion stabilization is possible in trypsinogen, but not at the ideal site. The loss of this component of the catalytic apparatus may well account for the slower reaction rate of the zymogen. There is a plausible connection between the movement of this site away from the cavity which is subsequently occupied by Ile 16 upon activation, when Ile 16 forms the salt bridge to Asp 194. There is also plausible connection between the small changes in the binding pocket, and the location of Ile 16. Similar change in the structure of the binding site could account for the poorer binding of substrate by the zymogens in general. Conclusion The structure of trypsinogen is generally nruch closer to that of trypsin than is chymotrypsinogen to chymotrypsin. The structure of trypsinogen does not exclude the possibility of substrate binding in a mode similar to that found for trypsin, although changes in the structure of this region contribute to an impaired or altered substrate binding mode--certainly for benzamidine, and most probably for a substrate side chain. If the proenzyme is considered to be rigid, then the general base catalyst (the EN of His 57) and the oxyanion binding site formed by the N-+I groups of Gly 193 and Ser 195 are too far apart to cooperate in substrate hydrolysis. Even if the proenzyme structure were to change on substrate binding, as seems likely, 99. nonproductive binding may provide another important component for zyrnogen inactivity. Such binding might still leave too great a distance, or an unfavorable interaction, between the substrate and elements of the catalytic center. The altered position of the chain between Lys 188A and Ser 195, and the main chain between Trp 215 and Ser 217, could be responsible for competitive yet nonproductive substrate binding, as could the N--H group of Gly 193, since they are normally involved in orienting the substrate. These possibilities re- main as prime candidates in a lllliversal scheme for inactivity of the trypsinogen-like zyrnogens. Acknowledgments We thank Melvin Jones and Rebecca Price for assistance with calculations used in the comparison of structures. We gratefully ac- lmowledge the support received from research grants from the National Institutes of Health and the National Science Foundation. One of us (RMS) is the recipient of a National Institutes of Health Career DeVelopment Award, another (AAK) was the recipient of a National Institues of Health Postdoctoral Fellowship, and another (JLC) holds a National Institutes of Health Predoctoral Traineeship. This is contribution #5445 from the Norman W. Church Laboratory of Chemical Biology, California Institute of Technology. 100. References 1. Neurath, H. 1957. Adv. Protein Chem. 12:320-86. 2. Ottesen, M. 1967. Ann. Rev. Biochem. 36:55-76. 3. Neurath, H. 1975. In Cold Spring Harbor Symposium on Proteases and Biological Control, eds . E. Reich, D.B. Rifkin, E. Shaw , pp. 51-64. 4. New York: Cold Spring Harbor Laboratory . 1021 pp. Leighton, T.J., Doi, R.E., Warren, R.A.J., Kelln, R.A. 1973 . J. Mol. Biol. 76:103-22. 5. Stambaugh, R., Brackett, B., Mastroianni, L. 1969. Biol. Reprod. 1:223-27. Ann. Rev. Biochern. 38:389-414 . 6. Muller-Eberhard, H.J. 1969. 7. Owren, P.A., Stonnorken, H. 1973. Ergeb. Physiol. Biol. Chern. Exp. Phannakol. 68:1-53. 8. Magnusson, S. 1971. In The Enzymes, ed. P.D. Boyer, 3:277-321 . New York: Academic Press. 886 pp. 9. Kay, J., Kassell, B. 1971. J. Biol. Chem. 246:6661-65. 10. Hartley, B.S. 1960 . 11. Walsh, K.A. 1975. Ann. Rev. Biochern. 29:45-72. In Cold Spring Harbor Symposium on Proteases and Biological Control, eds . E. Reich, D.B. Rifkin, E. Shaw, pp. 1-11. 12. 1021 pp . Freer, S.T., Kraut, J., Robertus, J.D., Wright, H.T., Xuong, Ng. H. 1970. 13. New York: Cold Spring Harbor Laboratory . Biochemistry 9:1997-2009. Kossiakoff, A.A., Chambers, J.L., Kay, L.M., Stroud, R.M. 1976 . Biochemistry 16:654-64. 101. 14. Birktoft, J.J., Blow, D.M. 1972. J. Mol. Biol. 68:187-240. 15. Chambers, J.L., Stroud, R.M. 1976. 16. Morgan, P.H., Robinson, N.C., Walsh, K.A., Neurath, H. 1972. Acta Crystallogr.--in press. Proc. Nat. Acad. Sci. USA 69:3312-16. 17. Birktoft, J.J., Kraut, J., Freer, S.T. 197. Biochemistry 15:4481-85. 18. Robertus, J.D., Kraut, J., Alden, R.A., Birktoft, J.J. 1972. Biochemistry 11:4293-4303. 19. Bode, W., Schwager, P. 1975. J. Mol. Biol. 98:693-717. 20. Sigler, P.B., Blow, D.M., Matthews, B.W., Henderson, R. 1968. J. Mol. Biol. 35:143-64. 21. Mavridis, A., Tulinsky, A., Liebman, M.N. 1974. Biochemistry 13:3661-66. 22. Tulinsky, A., Vandlen, R.L., Morimoto, C.N., Mani, N.V., Wright, L.H. 1973. Biochemistry 12:4185-92. 23. Blow, D.M. 1974. Israel J. Chem. 12:483-94. 24. Northrop, J.H., Ktmitz, M., Herriot, R.M. 1948. Enzymes, pp. 133-6. 25. Crystalline New York: CollUilbia University Press. 352 pp. Stroud, R.M., Kay, L.M., Dickerson, R.E. 1971. Cold Spring Harbor Symposium on Quantitative Biology 36:125-40. 26. Stroud, R.M., Kay, L.M., Dickerson, R.E. 1974. J. Mol. Biol. 83:185-208. 27. Koeppe, R.E., Stroud, R.M. 1976. 28. Krieger, M., Kay, L.M., Stroud, R.M. 1974. 30. Biochemistry 15:3450-57. J. Mol. Biol. 83:209- 102. 29. "Mares-Guia, M., Shaw, E. 1965. J. Biol. Chem. 240:1579-85. 30 . Gertler, A., Walsh, K.A., Neurath, H. 1974. Biochemistry 13:1302-10. 31. Robinson, N.C., Neurath, H., Walsh, K.A. 1973. Biochemistry 12:420-5. 32. Segal, D.M., Powers, J.C., Cohen, G.H., Davies, D.R., Wilcox, P. E. 1971. 33. Sweet, R.M., Wright, H.T., Janin, J., Chothia, C.H., Blow, D.M. 1974. 34. Biochemistry 10:3728-38. Biochemistry 13:4212-28. Huber, R., Kukla, D., Bode, W., Schwager, P., Bartels, K., Deisenhofer, J., Steigemann, W. 1974. 35. Blow, D.M. 1974. J. Mol. Biol. 89:73-101 . In Bayer-Syrnposil.Ull V "Proteinase Inhibitors", eds. H. Fritz, H. Tschesche, L.J. Greene, E. Truscheit, pp . 473483. New York: Springer-Verlag. 751 pp. 36. Dlouha, V., Keil, B. 1969 . 37 . Keil-Dlouha, V., Keil, B. 1972. 38. Wright, H.T. 1973a. J. Mol. Biol. 79:1-11. 39 . Wright, H.T. 1973b. 40. J. Mol. Biol. 67:495-505. J. Mol. Biol. 79:13-23. Bradshaw, R.A., Neurath, H., Tye, R.W., Walsh, K.A., Winter, W.P. 1970. 41. FEBS Letters 3:137-40. Nature 226 : 237-9 . DeHaen, C., Neurath, H., Teller, D.C. 1975. J. Mol. Biol. 92:225-59. 42. Blow, D.M., Birktoft, J.J., Hartley, B.S. 1969. 40. Nature 221:337- 103. 43. Abita, J.P., Delaage, M., Lazdunski, M., Savrda, J. 1969. Eur. J. Biochem. 8:314-24. 44. Bode, W., Schwager, P. 1975. 45. Delaage, M., Desnuelle, P., Lazdunski, M., Bricas, E., Savrda, J . 1967. 46. FEBS Letters 56:139-43. Biochem. Biophys. Res. Connnun. 29:235-40. Radhakrishnan, T.M., Walsh, K.A., Neurath, H. 1967. J . .Amer. Chem. Soc. 89:3059-61. 47. Maroux, S., Baratti, J., Desnuelle, P. 1971. J. Biol. Chem. 246:5031-39. 48. Magnusson, S.L. 1975. In Cold Spring Harbor Sympositnn on Pro- teases and Biological Control, eds. E. Reich, D.B. Rifkin, E. Shaw, pp. 123-49. New York: Cold Spring Harbor Laboratory. 1021 pp. 49. Steitz, T.A., Henderson, R., Blow, D.M. 1969. 46:337-48. SO. Henderson, R. 1970. J. :Mol. Biol. 54:341-54. J. Mol. Biol. 104. Footnotes 1Numbering scheme used will be that of chymotrypsinogen throughout the manuscript. 2Abbreviations used: DIP-trypsin, diisopropylphosphoryl trypsin; NPGB, p-nitrophenyl-p'-guanidinobenzoate; DIPT, DIP-trypsin; LpT, low pH DIP-trypsin; Cg, chymotrypsinogen; Tg, trypsinogen; CT, chymotrypsin; BPTI, basic pancreatic trypsin inhibitor. 105. TABLE I. Sunmary of corresponding zymogen and enzyme structures used for comparison in the text. Structures Resolution a-Chymotrypsin 2 ' 3 tosyl-, pH 4.2 Chyrnotrypsinogen4 ' 5 , pH 6.3 Trypsin DIP- 4 ' 6 ' 7 , pH 5.0 DIP- 6 ' 8 , pH 7.5 Benzamidine 2 ' 9 , pH 7.0 Trypsinogen4 , 10 , pH 7.5 (A) State of Refinement1 2.0 R 43% 2.5 R 43% 2.1 1.5 R = 23% 1.8 R = 23% 1.9 R = 31% 1state of refinement is indicated as the value of R = EIF -F I/EF, 0 C O representing the agreement between observed and calculated structure factors to the resolution indicated. 2crystallized from ammonium sulphate. \iodel building; real space and energy refinement (14). 4crystallized from ethanol. 5calculated phases; rebuilt model (12,17). 6crystallized from magnesium sulphate. 7Difference Fourier versus trypsinogen (13). 8constrained difference Fourier refinement (15). 9constrained difference Fourier refinement (19). 10 eonstrained difference Fourier refinement (13). 106. Figure Legends FIGURE 1. sinogen . The structure around the active center in chymotryp- Coordinates were those obtained from the Brookhaven Protei n Data Bank (12). The orientation of this view differs from that in succeeding figures. FIGURE 2. Structure around the active center in a-chymotrypsin FIGURE 3. The structure around the active center in trypsinogen FIGURE 4. Structure around the active center in DIP-trypsin (14). (13). (15). The DIP group has been omitted for clarity . FIGURE 5. A portion of the electron density map for DIP-trypsin in the region of the active center (15). FIGURE 6. 0 (a) Histogram showing the distance 6d (A) between equivalent alpha-carbon atoms in Tg and DIPT after the two coordinate sets have been rotated and translated to achieve optimum superposition of the two coordinate sets . (b) Histogram showing the average positional difference of all atoms within each residue of the Tg/DIPT comparison . 107. FIGURE 7. Structure of trypsinogen in the region of His 40 . The No-1 of His 40 is hydrogen bonded to the Oy of Ser 32 as it i s i n trypsin. The C=O of Gly 193 is moved from its position in trypsin . Asp 194 adopts a similar conformation to that seen in trypsin (13) . FIGURE 8. The structure of trypsin in the region of His 40 shows the hydrogen bond formed between No -1 of His 40 and the Oy of Ser 32, and the hydrogen bond between the N£-2 of His 40 and the C=O of Gly 193. Asp 194 is salt bridged to the amino terminal of Ile 16 (15). FIGURE 9. The chymotrypsinogen structure around His 40 and Asp 194 shows a closer interaction between these two groups, which are in quite different orientation than that seen in trypsin or trypsinogen . (12) . Coordinates were obtained from the Brookhaven Protein Data Bank 108. Figure 1 109. Figure 2 110. Figure 3 - -- - - -- - --·-· - - •. 111. Figure 4 112. Figure 5 113. Figure 6a 0 II) N II) N N 0 0 N II) ,- a: w O(D ~::i: :J z w :J ~□ -Cf) w a: 0 0 II) ,- oo ·s1 oo ·z1 oo ·s oo ·s CSW0~1S9N~J 3JN~lSIO ~J oo "E oo ·o 0 114. Figure 6b 0 U) "' U) "' N 0 0 N .... U) a: w O(D ~~ ::J z w ::J ~□ -CJ') w a: 0 0 - .... IJ) 0 II) U) N oo·s1 oo·z1 oo·s C~J S~Ol~ 77~ 1 oo·s oo·E 3JN~1S10 "3A~ oo·o 0 llS . Figure 7 116. Figure 8 ,0 m m • , ' ~ 0 JJ '2.~ 0 Gt'- ~ ~➔. ' •v (.") ~ ."l .,i . (, ¾'.~. B .." 117. Figure g ~ ~~~ 0, ~· ~ m ~.. . . , · ·c " - - ~) ' 118 . Appendix 6 Difference Fourier Refinement of the Structure of DIP-Trypsin at 1.5 A With a Minicomputer Technique 119 . Difference Fourier Refinement of the Structure of DIP-Trypsin at 1.5 A With a Minicomputer Technique* JOHN L. Cl-lAMBERS t AND ROBERT M. STROUD:j: Norman W. Church Laboratory of Chemical Biology California Institute of Technology Pasadena , California 91125 , U. S.A. (Acta Crystallogr.---in press) Running Title: 0 1.5 A Refinement of DIP-Trypsin. *Contribution #5352. Supported by National Institutes of Health grant GM-19984 and National Science Foundation grant BMS75-01405 . tNational Institutes of Health Predoctoral Trainee. fNational Institutes of Health Career Development Awardee, U.S. Public Health Service Grant G-1-70469. 120 . ABSTRACT The x-ray crystal structure of diisopropylfluorophosphateinhibited bovine trypsin has been refined to a resolution of 1. 5 A with the use of a constrained difference Fourier technique . The refinement was carried out almost entirely on a minicomputer with a set of programs written for this study to largely automate the refinement procedure . Use of the minicomputer has allowed the refinement to progress efficiently at an extremely low cost, from a starting R0 0 factor of 47.2% at 2.7 A resolution to 27 . 7% at 1. 5 A resolution . Two cycles of a least-squares refinement procedure using a first-order 0 gradient minimization have since reduced R to 23.5% at 1.5 A resolution. The refined model and electron density maps are substantially improved over the starting ones. The methods described may be very attractive when computing funds or access to a large computer are limited. INTRODUCTION During the past few years a nt.nnber of protein structures determined by x-ray crystallography have been refined to 2 A or better resolution using a variety of techniques (Watenpaugh, Sieker , Herriot , and Jensen, 1973; Sayre, 1974; Diamond, 1974; Huber, Kukla, Bode , Schwager, Bartels, Deisenhofer, and Steigemann, 1974; Freer, Alden, Carter, and Kraut, 1975; l\1oews and Kretsinger , 1975; Deisenhofer and 121. Steigemann, 1975; Bode and Schwager, 1975a). The refinement programs written by Diamond (1966, 1971) are particularly effective and are used extensively. However, the programs are large and require signif- icant amolillts of storage and central processing unit (CPU) time on a large computer. Several of these workers have successfully employed difference Fourier (6F) refinement techniques, both alone and in conjunction with Diamond's programs. The 6F method is economical in terms of progrannning complexity and cost, and is readily adaptable for use on a small computer. This paper describes the methods developed and used for 6F refinement of the structure of diisopropylfluorophosphate (DFP)inhibited bovine trypsin originally determined by Stroud, Kay, and Dickerson (1971, 1974). A set of refinement programs has been written and implemented almost entirely on a Data General NOVA 800 minicomputer. Use of the minicomputer has enabled the refinement to progress efficiently at a very low cost, and the methods described here may therefore be attractive when large amoIB1ts of time and a very low rate of cost on a large computer cannot be obtained. The refinement has resulted in a model which is greatly improved over the starting one, with substantial (i.e., several angstroms) changes in some areas. This study thus indicates, as do those cited earlier, that the improvement to be gained by refinement warrants the routine use of some refinement procedure in the determination of protein crystal structures. 122. The refined model is described only as much as necessary to illus trate the application of the methods. A detailed description of this structure will be presented in a future paper . The methods described here have also been used successfully in a preliminary refinement of the bovine trypsinogen structure (Kossiakoff , Chambers, Kay, and Stroud, 1977). METI-IODS DIP-trypsin crystals The conditions for crystallization of diisopropylphosphoryl (DIP)trypsin are described by Stroud et al. (1974). The crystals are ortho0 rhombic, space group P2 121 21 , with unit cell dimensions~= 54.84 A, Q_ = 58.61 A,£= 67.47 A, and one molecule per asyrrnnetric unit . Crys- tals of a size approximately 0. 6 nun x 0.5 mm x 0.5 nun were used for recording the high (1.5 A) resolution data; those used for data to 2 A were slightly smaller. Although these crystals contain about 40% a-trypsin (cleaved between Lys 145 and Ser 146) and 60% B-trypsin (uncleaved), they do not appear to be significantly less well ordered than the crystals of benzamidine-inhibited trypsin described by Fehlhannner and Bode (1975), which contain about 80% B-trypsin. The dif- fraction pattern from the DIP-trypsin crystals extends to at least 0 1.1 A resolution . 123. Data collection and reduction Diffracted intensities from DIP-trypsin crystals were measured with a Syntex Pi diffractometer equipped with a graphite monochromator and a helium-filled tube between the crystal and detector. The tube voltage and current for the x-ray source were 40 KV and 20 ma, respectively. Background corrections were made with an interpolation technique which accounts for variations in the background with the setting angles,¢, x, and 28 (Krieger, Chambers, Christoph, Stroud, and Trus, 1974a). The remaining data reduction and scaling together of data collected from different crystals were accomplished according to the procedure of Stroud et al . (1974) . In all, eight crystals were 0 required to provide a full set of intensities to 1.5 A resolution (28 = 62°). A total of 4072 intensities to 1.5 A were measured from different crystals for scaling purposes . The average value of R = EjF -F 1/EF 1 2 1 (where F is the current value of the structure factor in the master 1 data set, and F is the value for the crystal data to be merged into 2 the master set) was 4.6% for these multiply-measured reflections. Of 0 the 35566 independent reflections in the 1.5 A sphere, those 22117 reflections having structure factors greater than three times their standard deviation were included in the final data set. 88 . 5% of the 13,261 possible data were observed. 0 To 2.1 A, Between 2.1 and 1.76 A the number observed was 4784 of a possible 8979, or 53.3%, and between 1.76 and 1.5 A, 5602 of 13,326 possible reflections were observed, or 42%. 124. Starting model The starting coordinates for the refinement were measured from a wire model constructed in an optical comparator (Richards, 1968) with reference to a 2. 7 A MIR-phased map . This map was more recent than that described by Stroud et al. (1974), differing in that the 2.7 A data from the Ag+ derivative (Chambers, Christoph, Krieger, Kay, and Stroud, 1974) had been incorporated into the phase refinement. Atomic coordinates from the model were measured using an electronically operated device, similar in principle to the one described by Salemme and Fehr (1972). Minicomputer system The components of the minicomputer system used in this study are listed in Table I, along with typical execution times for certain types of operations and function evaluations. All computations except calculation and plotting of electron density maps were carried out using this system. An IBM 370/158 computer was employed for the elec- tron density syntheses. Difference Fourier method Detailed descriptions of the difference Fourier method have been given elsewhere (Cochran, 1951; Stout and Jensen, 1968) . The method is based on the principle that an electron density map computed using coefficients (F 0 - F )ei¢c C (where F is the observed structure factor 0 amplitude, and FC and¢ C are the structure factor amplitude and phase 125 . computed from the current model) contains positive peaks in positions where additional electron density should be added to the model , and negative peaks in positions where it should be subtractedj in order to produce better agreement between the model and the observed data . The position of an atom can thus gener al ly be improved by computing the density gradient in the ~F map at the atomic center and moving the atom toward higher density along this gr adient . For an atom which is correctly positioned , the density at the atomic center gives an indication of the shift in the temperat ure factor for that atom. The density at atomic centers is also strongly influenced by the overall scale and temperature factor between the observed and calculated structur e factors . The course of a typical cycle of ~F refinement performed in this study is schematized in Figure 1. Each cycle was begun with calcula- tion of structure factors based on the current model of the structure . The calculated structure factors were then scaled to the observed ones, which in turn had been placed on an absolute scale early in the refinement, and were then used to generate a new electron density map . This map was either a difference map (computed using coefficient s ~F = F - F and phases¢), which shows the differences between the 0 C C current model and an improved one, or a map computed with teTIJ1S (nF 0 [n - l]Fc)ei¢c which superposes n difference maps on the current structure, showing the improved structure in its entirety. Because peaks resulting from a difference synthesis are about half their theor etical height (Henderson and Moffat, 1971), n = 2 was used for 126. most of these maps. Empirically, n = 2 usually produced the best com- promise between freedom from noise and appearance of new information. Following generation of the density map, the indicated corrections to atomic positions were derived and applied to the structure. techniques were used to determine atomic shifts (Figure 1). Three The simplest and most accurate procedure was to estimate shifts from the difference maps automatically, using a program written for the NOVA 800 (the middle path in the flow chart of Figure 1). The automated technique was most useful where relatively small movements from the input structure were required. If the map indicated that gross struc- tural changes were needed, one of the other two branches was taken. The branch to the left in Figure 1, which was the simpler of the two, involved manual adjustment of atomic positions as indicated by a map made from twelve-inch square Plexiglas ® sections upon which were drawn both density contours and the atomic positions . It was very difficult to preserve accurate bond lengths and angles when making shifts in these small scale maps, and incorporation of shifts derived in this manner was relatively tedious. Using this technique, atomic shifts could be made at a rate of about 1800 atoms in two days . Once during the refinement the wire model was rebuilt and the coordinates remeasured (the branch to the right in Figure 1). Although this procedure was useful for interpreting interactions involving long regions of secondary structure and hydrogen bonding possibilities, making changes in this manner was very tedious and the resulting coordinates were not typically of very good quality. 127 . An automated constraints procedure was used to restore bond lengths and angles to physically reasonable values after the shifts in atomic positions had been applied . Occasionally, additional rounds of atomic shifts and constraints carried out before computing a new map resulted in further improvement , especially when nF - (n - l)FC syno The best results with difference syntheses were ob- theses were used . tained when a new map was computed after each round. If indicated in the calculated maps, new solvent molecules were added to the structure, or old ones deleted based on their occupancy and close approach distances to other atoms in the structure. A new refinement cycle was then initiated. Refinement software The refinement software and algorithms employed in the individual steps mentioned above are described below. Except where noted, all programs were written in FORTRAN IV, and execution times refer to execution on the NOVA 800 . a) Structure factor calculation Structure factors were calculated by conventional Fourier transformation of atomic coordinates, using expressions specific for space group P2 z1z . The assembly language code output by the FORTRAN 1 1 compiler for the inner loops in the program was optimized to increase execution speed. The contribution from each atom was added into every reflection before proceeding with the next atom, enabling the use of 128. recursive ITillltiple-angle relationships for rapid evaluation of the trigonometric functions. Scattering factors for the individual atoms were approximated by the five-parameter analytical expressions of Forsyth and Wells (1959). Aton~ within the coordinate list were sorted into groups which contained identical scattering factors and individual isotropic temperature factors, requiring only one evaluation of these parameters per group for each reflection. A provision was made in the program for deleting small groups of atoms from, or adding them to, the current set of computed structure factors. Such an option is essential if the entire set of F's is not to be recalculated whenever changes are made over a relatively small portion of the structure. Execution time for this program was 6.75 x 10- 4 seconds per atomreflection. For the DIP-trypsin coordinate file, containing about 1800 atoms per asymmetric unit, total execution time was approximately 0 four hours at 2.1 A (11713 reflections), seven and a half hours at 1.76 A(22309 reflections), and twelve hours at 1.5 A(35566 reflec- tions). b) Scaling and R-factor calculations A number of authors have noted that at low diffraction angles the average intensities observed from protein crystals are significantly reduced from their expected values (Watenpaugh et al. 1973; ~bews et al. 1975). T'nis effect has been attributed to the relatively large amount of disordered solvent present in protein crystals. Such 129. an effect is seen in the scaling plot for DIP-trypsin shown in Figure 2. This behavior was accollllted for by least-squares refinement of a four-parameter function to the scaling plot . The function was chosen to represent a linear fit at high angles with an exponential falloff at low angles: This procedure is similar to one described by Moews and Kretsinger (1975). The least -squares fits for both the linear and nonlinear fllllction are shown in Figure 2. The linear fit for reflections at small values of sin 28/A 2 was totally inadequate, and inclusion of these reflections scaled in this manner caused large low-frequency ripples in the resulting electron density maps. The improvement in agreement between F0 and Fe for these reflections is shown in Figure 3. For the linear fit, the residual, R (R = 61F0 - FC l/6F0 ), increased sharply at low angles, reaching over 75%; however, when properly scaled, using the nonlinear expression, the agreement for these reflections was up to three times better. It should be emphasized that, while the technique accollllts well for the difference in scale of the low-angle data, the expression used is an empirical one and is not meant to be a mathematical description of a physical phenomenon. The actual function used, as long as it conforms well to the shape of the scaling plot, does not appear to be critical to either the appearance of the density maps, or the value of R. 130 . Reflections for which > 120% were left out of the calculation of the r esidual (about 700 reflec 0 tions to 1. 5 A). These reflections were also omitted from t he calcu- lation of density maps, since the ~F term which results from the case where Fo>>Fc is likely to be poorly phased. Difference terms for those reflections where Fc>>F0 are probably correctly phased and were included in the R-factor and Fourier map calculations (Stout and Jensen, 1968). Unobserved reflections were included in the scaling with F's equal to twice the average standard deviation for unobserved F's in the appropriate zone of sin9 . They were not included in the Fourier or R-factor calculations . Absolute scaling was accomplished with the use of a standard Wilson plot technique, neglecting the reflections inside 5 A resolution. The overall temperature factor thus obtained was 14 i 2 for the 1.5 A data. All atoms included in the structure factor calculation (i . e ., pr otein and or dered solvent) were included in the absolute scaling . c) Fourier map calculatiCl!1_ Generation of electron density maps was the only step of the refinement carried out on the IBM 370/158 , with a Fourier program 131. originally written by G. N. Reeke. CPU time required to generate a map of approximately 100,000 grid points from 22117 reflections to 1.5 Aresolution was 6.5 minutes. A grid spacing of approximately 0 1.25 A along each of the crystallographic axes was used throughout most of the refinement. with 2.7 and 2.1 The coarse grid produced satisfactory results Adata, and was economical in terms of computing cost and map storage requirements. The fact that the grid size exceeded the interval corresponding to the Nyquist frequency (i.e., was greater 0 than half the resolution of the data) at 2.1 A did not appear to affect the course of the refinement adversely. 0 At 1.76 A and higher resolution, better results were obtained with a finer grid, and an 0 interval of 0.75 A was therefore used in later stages. d) Atomic shifts Because of the iterative nature of the refinement procedure, small errors introduced by making approximations in order to simplify programming or computation can be corrected in subsequent cycles. A number of such approximations were made in the atomic shifts program. Electron density at an atomic position, P, was interpolated from a three-dimensional array of either 8 or 27 grid points (depending on the coarseness of the map grid) surrounding the atom, according to the scheme Ep.w. p = . 1 1 1 Ew. i 1 132. where p.1 are densities at each of the grid points, and w.1 are weights for each grid point in the sum, which fall off exponentially (w.1 = e -ad?1) as the distance, d., between the grid point and the atom 1 increases. Values for the constant, a, were in the range 0. 25-1.0 i- 2. Gradients were computed by evaluation of the density 0.1 A on either side of the atom along each coordinate axis. In more recent cycles, the electron density was approximated using a three-dimensional polynomial, with coefficients (a 1--a 7) detennined from seven grid points around the atom center. p = This expression allows analytical evaluation of the gradients, resulting in faster execution. Both schemes produced satisfactory shifts. The correction to an atomic position indicated in a ~F map for --==- is the change noncentrosyrnrnetric structures is: --=:,,,,.. ~x = -2g/C, where~ ..... in position, g is the difference density gradient at the atomic site , and C is the curvature at that site in the corresponding F0 synthesis Cochran, 1951; Stout and Jensen, 1968). In practice the curvature has been approximated with empirically detennined values for each type of atom or ion (Watenpaugh et al. 1973; Freer et al. 1975; Moews and Kretsinger, 1975). A nlDllber of factors in addition to the atom type influence the actual values to be used, including resolution of the data, the overall scale and temperature factors, and the scale of the densities written at the grid points of the map. To account for variations in 133 . these parameters over the course of the refinement, and to give the program greater flexibility, the curvature was taken to be proportional to the atomic number, Z, and shifts were scaled by a specified constant, K, chosen to produce reasonable values for the shifts: -- ill( = ... Kg T A value of K was used which produced positional changes approximately equal in magnitude to those estimated from 2F0 -Fc maps (averaging about 0 0 0.4 A in early cycles and 0.1 A in later stages). After a shift was computed, the density at the prospective new atomic site in the map used to derive the shift was evaluated to make certain that the atom had moved into more positive density. If not, the program repeatedly cut the size of the shift by half until this criterion was met. In 0 ad.di tion, any shift exceeding a specified maximtm1 tixmax (usually O. 4 A) was reset to this value, maintaining the original direction. The above expression enabled shifts to be readily evaluated from MIR maps and nF -(n-l)F maps, as well as from difference Fourier syntheses. 0 C The electron density for an atom (p) can be approximated by a spherical Gaussian of the fonn p where r is the distance from the atomic center, Z is the number of electrons, and Bis the individual isotropic temperature factor for the atom (see Diamond, 1971). From this expression, the expected 134. change in temperature factor (tB) corresponding to a small change in electron density at the center of the atom (tp) is, 0 This expression is in practice effective only for atoms which are correctly positioned, and is also affected by scaling errors. The equa- tion was therefore modified empirically to provide more satisfactory results, and the expression used to compute corrections to the individual isotropic temperature factors was, tB = -Kz (pt - pJ t z where tB is the shift, pt is the difference electron density at the atomic site, ~ is the expected average difference electron density for all atomic sites, and K2 is a scaling constant similar to the one for positional shifts. Because atoms at the beginning of the refinement tend to reside in negative difference density due to the fact that they are not correctly positioned, there is a tendency for the temperature factors to become too large in early cycles, providing they are allowed to vary. The tenns involving pt and e-1:,x/~ enable this interaction between positions and temperature factors to be partially decoupled. The former term also compensates for errors in scaling of calculated F's to the observed ones, which can also greatly influence Lill. The latter 135. tenn reduces the shift in the temperature factor according to the size of the gradient at the atomic site; thus, the full shift is applied only when an atom is close to its correct position. The equation used produced satisfactory values for 6B, which resulted in rapid improvement in flatness of the 6F maps at atomic centers, and reasonable values for the temperature factors themselves (values of 10, 12, 14, 16, 18, 23, 28, 38, and 58 A2 were used). Larger B factors were generally associated with external side chains. Execution time for this program for a coordinate list of about 1800 atoms was approximately 30 minutes. e) Constraints Constraints were applied to the coordinates by minimization of a residual composed of sums of squares of the deviations of the atoms from "ideal" relative positions. tributed to the residual. Four types of differences con- The first was the difference between the positions of the atoms in the starting structure and those in the constrained structure. This tenn was given unit weight throughout the constraints procedure. The other three terms contained deviations from standard values of the bond lengths, bond angles, and dihedral angles to be constrained. Each of these three terms was given a rela- tive weight which was increased on each pass through the atom list, allowing greater movements from the starting structure and a closer approach to standard constraints in each subsequent pass. The pro- cedure was terminated either after a fixed number of passes, or when 136. nns deviations of bond lengths and angles had fallen below selected values . Similar procedures have been described by Hennans and McQueen (1974) and Freer et al. (1975). More recently similar techniques have been independently developed for constraining bond lengths, angles, and dihedral angles by minimization of deviations from standard interatomic distances only (Dodson, Isaacs, and Rollet, 1976; Ten Eyck, Weaver, and Matthews, 1976). Standard values for bond lengths and angles were obtained from Marsh and Donohue (1967), and the references cited therein. Standard values for the dihedral angles constrained were chosen in order to make the peptide amide groups trans- and planar. In later stages, planarity of rings and side chain amide, carboxyl, and guanidino groups was enforced by constraining dihedral angles, since bond length and bond angle constraints alone were not in practice sufficient to make these groups planar. In the case of side chain amide and carboxyl groups, dummy dihedral angles were constrained (e.g., 9 (CS, CY, 0 61 , 0 62 ) in Asp). Constraints were applied to a zone of one to five residues. A small portion of the peptide chain on either side of the zone was used as a fixed botmdary, which was necessary for convergence of the procedure at the interface between zones. A simple steepest-descent gradient minimization technique was used, and was very reliable in its ability to produce converging shifts even when atomic coordinates were far removed from ideality; it was also fast and satisfactory in its rate of convergence. (This was in contrast to the gradient-curvature 137. techniques, e.g., Newton-Raphson minimization tried in early versions of the program, which converged rapidly in the neighborhood of the minimum, but often produced unreasonable shifts for regions which were poorly constrained in the starting structure, and required substantially more computing time per cycle.) Since the molecular parameter s in real proteins, particularly the planarity of peptide amides (Rarnachandran, Lakshminarayanan, and Kolaskar, 1973; Huber et al. 1974), may deviate significantly from "ideal" values, the fact that the constraints can be given any desired degree of flexibility using this procedure is highly advantageous. It is moreover helpful to keep the constraints rather flexible when performing alternate cycles of atomic shifts and constraints in order to prevent oscillation between a constrained and an unconstrained structure . Four passes through the coordinate file were generally run after atomic shifts had been applied i..~ order to achieve the desired rigidity in the constraints. Weights for the contributions to the residual of bond lengths, angles, and dihedral angles were initially set at S, 1, and 1 for lengths in angstroms and angles in radians, and were multiplied by 3, 6, and 6, respectively, at the beginning of each new pass. 0 Resulting overall rms deviations from ideality were typically 0. 07 A in bond lengths, 3.5° in bond angles, and 4.5° in dihedral angles. Maximum deviations were usually about four times the overall rms deviations. Execution time was about 30 minutes for each pass through the coordinate file containing 223 residues. 138 . f) Peak search and close contacts The peak search and close contacts programs were used primar- ily for selection of solvent molecules from difference maps. The peak • search program locates prospective solvent straightforwardly from one asymmetric unit of the map, and runs in about five minutes for 100 , 000 grid points. Close contacts between nonbonded atoms were tabulated by first sorting the atoms based on their position in the asymmetric unit, requiring comparison between a relatively small ntnnber of atoms and their symmetry equivalents to find all distances less than a specified value. Approximately twenty minutes were required to tabulate all close contacts between atoms in an 1800-atom-coordinate file . Pro- spective solvent molecules which were closer than 2.5 A to atoms already present in the structure, or solvent which had moved within this distance of other atoms in the protein, were generally removed from the data set. In a few cases, particularly when close contacts existed between groups of the protein itself, corrections were made manually. The current structure contains 174 solvent molecules, most of them hydrogen-bonded to groups on the surface of the protein and to other solvent. About 30 of them can be considered internal. 139. RESULTS Progress of the refinement The graph in Figure 4 indicates the progress of the refinement to date. The R-factor at 2.7 A computed for the starting structure 0 (Revision 1) was 47%, rising to 50% upon incorporation of the 2. 1 A data . Manual adjustment of atomic positions with the use of small scale 2F0 -Fc maps reduced the residual to 40.2%. The automated atomic shifts program was written at this stage. The improvement in R was very rapid, and after three cycles of shifts determined from ~F maps, R was 34%. The resulting rms deviations from standard bond lengths and angles were, however, very poor--0.5 A and 22°, respectively. Moreover, 33 a-carbon atoms and 10 Ile and Thr S-carbons had adopted the wrong sense. The constraints program was subsequently implemented, and several cycles of automated shifts using a 2F -F map followed by constraints were performed. 0 C A number of the wrong-handed asymmetric carbon atoms were corrected manually using small scale maps. The resulting overall rms deviations from ideality for bond lengths and angles were 0.15 A and 6.5°; at the same time, there was only a small decrease in the average value of electron density at atomic centers. ture was 38.8% at 2.1 The R-factor for this constrained struc- A. Alternate cycles of shifts and constraints were then applied nine times, using seven difference Fourier maps (Revisions 6-12 in Figure 4). During these cycles, solvent atoms were added and deleted where 140. indicated . At Revision 12 of the coordinates (see Figure 4) , the atom file contained 220 solvent molecules. Deviations from ideal bond 0 lengths, angles, and dihedral angles were 0.1 A, 2.4°, and 3. 8°, re spectively . The sense of all asyrrmetr ic atoms had been corrected and the residual had been reduced to 26.6% at 2. 1 A. The 1.76 A data were 0 included, and two refinement cycles produced a 1.76 AR-factor of 28 . 8% (231 solvent molecules). Since it appeared that additional automated cycles would produce little further improvement in R, it was felt that a large-scale visual inspection of the structure in its current state was warranted, and the wire model was rebuilt to a 2F0 -Fc map . A ntunber of substantial changes from the starting wire model had taken place, and will be briefly described later. Some additional improvements were made at this step, including major reorientation of six amide planes and addition of seven ordered solvent molecules . Approximately 70% of the co - ordinates were replaced with coordinates measured from the rebuilt model . The 1. 76 AR-factor rose sharply to 35%. Examination of the new coordinates in a small-scale map showed that, while the atoms were not far from their intended positions, large pieces of the structure appeared randomly translated by about 0. 5 A from the center of the density, particularly in that direction perpendicular to the plane of the map sections in the Richards' box. Five difference Fourier refinement cycles rapidly corrected these 0 translations, the R-factor at 1. 76 A dropping to 26.9%. During these cycles a number of chemically unreasonable and low-occupancy solvent 141. molecules were removed, leaving a total of 174. Following these 0 cycles, 1.5 A data were added, and one subsequent refinement cycle re0 duced the 1.5 AR-factor to 27.7% for the constrained structure containing 174 ordered solvent molecules (rms deviations from ideality 0 for bond lengths, angles, and dihedral angles were 0.07 A, 3.2°, and 4.6°, respectively). Changes from the starting model A histogram indicating the movement of a-carbon atoms between the starting and current models is shown in Figure 5. The average posi- 0 0 tional change for a-carbons was 0. 9 A, and for all atoms was 1.2 A. The largest movements were in the regions of residues 60-63, 74-80 (the primary calcium binding site, first identified and described in detail by Bode and Schwager, 1975a,b), 110-121, 144-147 (the a-, Strypsin autolysis site), 186-187, and particularly in the C-tenninal a-helix. All of these residues are on the surface of the enzyme, and were not unambiguously defined in the MIR map. The problem with the terminal a-helix was largely a translational one, since it was the region farthest from the mirror of the Richards' box when building the wire model, and was therefore most prone to errors in depth perception. One type of error connnonly encountered over the course of the refinement was the 180° misorientation of an amide plane. At a resolu- 0 tion of 2.7 A the MIR map did not allow unambiguous assignment in a number of cases; however, the errors were clearly identified in the high resolution ~F maps. The regions where these misinterpretations 142. occurred were again all on the surface of the molecule: the amide planes between residues 22-23, 24-25, 34-37, 39-40, 75-76, 84-85, 110111, 117-118, 144-145, 187-188, and 222-223 all required reorientation. 0 The region of the molecule around His 40 in one of the early 2.1 A ~F maps, showing the misorientation particularly of the 'I'yr 39-His 40 amide, is pictured in Figure 6a. 0 Figure 6b is a drawing of the 2.7 A MIR map and starting model in the same region. At this resolution, an unambiguous assignment cannot be made for the Asn34-Ser 37 and 'I'yr 39His 40 peptides. Figures 6c and 6d are respectively pictures of the 0 0 current model and 1.5 A refined map, and the 1.5 A difference map in this region. The difference map is nruch flatter since the required corrections .have been made. Although large structural changes did occur in some parts of the molecule, the chemically important residues of the active site and binding pocket did not move greatly upon refinement. Positional changes for these residues resulted from limitations in resolution and clarity of the MIR map, rather than from misinterpretation. The most significant change involving the active site is the tilting of the Asp 102 carboxyl group and the His 57 imidazole into positions where the interaction between them is no longer symmetrical as previously suggested (Stroud et al. 1974; Krieger, Kay, and Stroud, 1974b). The a-nitrogen of His 57 points more toward the lower oxygen, 002, of the 0 Asp 102 carboxyl, as shown in Figure 7, fanning a 2.7 A hydrogen bond. TI1e distance between the upper oxygen atom, Oo 1 , and His 57 No is too 0 great for favorable hydrogen bond formation, 3.9 A. However, this 143. atom does fonn hydrogen bonds with the main chain nitrogen atoms of Ala 56 and His 57 of 2.7 Aand 3.2 A, respectively. As can be seen from Figure 7, the positions of these groups are well defined in the refined map. Refinement of the structure from residues 72 - 81 indicated a tightly bound positive ion and several well-ordered solvent molecules, which were previously not resolved in the 2.7 A MIR map. Bode and Schwager (1975a,b) in their independent refinement of the structure of benzamidine-inhibited bovine trypsin recognized this area as the primary calcium ion binding site. While the refined structure of DIP-trypsin arrived at from the present study is nearly identical in this area to that reported by Bode and Schwager, the density for the positive ion in the present study is about half of that expected for a fully occupied calcium ion. It is possible that this site is partially occupied by ca 2+; however, the ligands do not appear disordered as expected for such a situation. Since the crystallizing solution contained a high concentration of Mg 2+ (5.8 %MgS04), occupancy by a magnesium ion is another possibility in spite of the difference in atomic . radius between Ca 2+ and Mg 2+ . Occupancy by a water molecule is not as likely, because of the high density of negative charge surrounding this site. 144. DISaJSSION There are a ntunber of criteri a which indicate that the refinement has indeed resulted in an improved struct ure . First, the refined structure fits the observed data better than did the starting structure. The most connnonly quoted index of this agreement is the stan- dard crystallographic R-factor, which decreased from 50% at 2. 1 A to 27 . 7% at 1. 5 A, as just described. This decrease is derived , however , from a nl.Ililber of sources in addition to improvement of atomic positions, including the use of individual temperature factors, inclusion of solvent, use of the four-parameter scaling procedure, and omission of the approximately 700 reflections with F0 >>FC from the data set. The contribution of each of these factors to the decrease in R can be estimated by computing the R-factor for the refined model without its benefit . The respective contributions to the R-factor at 2.1 A are 3%, 3%, 2%, and 1. 5% . The R values for the starting and current model at this resolution are 50% and 25% , respectively, leaving about 15% of the 25% decrease attributable to improvement in atomic positions of the atoms originally in the starting model. Although the R-factor is a useful indicator of the overall progress of the refinement, it was noted in this study as well as by Huber et al. (1974) that small changes in isolated portions of the structure , such as a flipped amide plane, are not strongly reflected in the value of R. A much more sensitive device for the detection of these local errors is the ~F map. The average unsigned difference electron 145. density dropped from 0.31 electrons/A3 for the 2.1 A llF map at the start of the refinement to 0.08 electrons/A3 for the 1.5 A refined ~F map, where the integration was perfonned over an entire asynnnetric 0 The present 1.5 A difference map contains few interpretable features, with no density above 0.54 electrons/A3 in magnitude. These tmit. numbers can be compared with 2.6 electrons/A3 for a well-defined carbonyl oxygen in the refined 2F0 -F C map. The correctness of the refined structure is also indicated by the representative result shown in Figure 8, which is a picture of the 0 1.5 A 2F0 -Fc map in the same area as that of Figure 6. In this case, however, all the atoms shown (N34 C=O, S37, G38, Y39 main chain, H40, F41, C42, and G43) have been omitted from the structure factor calculation. The density for these atoms is nevertheless very clear and, although the resolution suffers somewhat, the structural infonnation is clearly contained in the observed F's. This result serves to em- phasize that the 2F 0 -FC map is not merely a reflection of the atoms included in the phase calculation. The second criterion indicating the improvement of the refined model is that it has become chemically more reasonable. Larger re- gions of the protein fit into classical patterns of secondary structure. In a number of regions fonnerly lacking good hydrogen bonding, new hydrogen bonds were fonned, many of them through solvent molecules. Forty-nine close contacts between nonbonded atoms, shorter than the extreme limits of Ramachandran and Sasisekharan (1968) have been elim0 inated. The wire model rebuilt from the 1.76 A refined map contains 146. no large empty regions within the interior (except the specific binding pocket), whereas several of these areas were present in the starting model. The MIR map also serves as a check on the correctness of the refined model, and was a very useful aid in regions where the calculated maps were noisy. The major changes made in the structure resulted primarily from difficulties in interpretation of the MIR map (except in the terminal a helix); however, these changes have not generally resulted in contradictions with this map. The refined coordinates fit the MIR electron density considerably better than did the starting ones. The mean change in phase angle, 6¢, between the "best" MIR and refined phases was 62°, and between the phases calculated from the starting model and refined phases was 58°, for all reflections to 0 2.7 A. These are large average changes, but are similar to those re- ported for other protein structures (Watenpaugh et al. 1973; Huber et al . 1974; Moews and Kretsinger, 1975; Deisenhofer and Steigemann, 1975). However, the appearance of the Fourier map is ultimately de- termined by contributions from a large number of reflections and, because 6¢ contains no information concerning the interrelationships between the various reflections, it is not a very meaningful number when considering how different the MIR and refined maps really are. In order to visualize the kinds of differences involved between the MIR and refined maps, a difference synthesis was computed with 147. coefficients F0 e i<j>MIR i<P - F0 e c . While substantial difference den- sity (~p) was present over most of the map, this density was considerably small~; at atomic positions than the corresponding density (p ) 0 in an F0 e c map computed at the same resolution . This result reo fleets the strong similarity between the MIR and 2.7 A refined maps. An index of the disagreement between these maps can be computed from therms densities in t he difference and F0 syntheses: <~p> = rms Averaged over the entire asynnnetric unit, R<P was 86%. This number is smaller than the R<P of 106% estimated from the 62° value of~¢, indicating that the differences between the MIR and refined phases tend to be smaller for the larger reflections . Averaged over only those grid points surrounding the atoms in the refined structure, R<P fell to 28%. 0 The average change of 1.2 A between the starting wire model coordinates and the current set is quite substantial, and arises from two major sources of error . The first source involves errors in interpretation caused primari ly by the lack of resolution and clarity , and phase errors in the MIR map. These changes result in movements of up to several angstroms, but occur over relatively small portions of the molecule. The other major source of error is in the actual construc- tion of the wire model, particularly due to inaccurate depth perception, irregularities in the mirror, and mechanical difficulties in 148. fitting residues together, placement of supports, and bent parts. These errors are comparatively small, but are distributed over the entire molecule. Errors arising from measurement of wire model coor- dinates were very small--within 0.5 nnn, or 0.025 A. A reasonable estimate of the size of the noninterpretative errors in the wire model coordinates can be derived from the positional dif0 ferences between the coordinates of the wire model rebuilt at 1.76 A (Revision 16 in Figure 5) and those obtained from the five cycles of refinement following this rebuilding, since no gross reinterpretations were involved during these cycles. The mean change in atomic posi- tions between these two sets was 0.6 A. There were no overall system- atic translations between the data sets compared. Correction of these errors led to a substantial decrease in R--from 35% to 27% at 1.76 A, the R-factor being sensitive to small translations of large pieces of the structure. Similar types of errors, although considerably smaller in magnitude (and dependent upon the coarseness of the map grid), appeared when perfonning shifts manually in the small-scale maps. For this reason it is felt that these approaches are inadequate to provide an optimal fit to the electron density: some type of computerized scheme for atomic shifts is essential for an optimal fit with a correspondingly lower value of R. Several early problems resulted from the high content of ordered and disordered solvent. In retrospect, the presence of ordered solvent was one of the greatest factors in misinterpretation of the MIR map. At 2.7 A resolution few of the ordered solvent were separated from 149. neighboring residues, leading to inaccurate placement of these residues in the model. Although it was found that inclusion of ordered solvent and use of individual temperature factors could produce significant reduction in the R-factor in early stages, it is felt that their application in the initial stages of refinement is a mistake. Difference Fourier maps at the start were relatively noisy, making selection of peaks corresponding to real solvent difficult. Solvent added in these early stages tended to move toward the atoms of the protein, influencing the gradients and corresponding atomic shifts. In these stages it is best to add solvent only at a very high confidence level (the MIR map can be very helpful), and preferably not until the L\F maps have become easily interpretable. Because of the strong coupling between the atomic positions and the shifts indicated for individual temperature factors, individual temperature factors should also be left largely untouched at the beginning of the refinement. Unfortunately, the fact that the L\F maps were very noisy in early cycles made location of parts of the structure which had been omitted from the phasing, or which were very far from the correct position, very difficult. The technique which produced the best result in this situation was to perform several cycles of refinement in order to optimize the fit for the known parts of the structure. This improved the phases to the point where the missing features could often be recognized, and made the L\F maps considerably less noisy. One method which was not employed in this study but may be very useful in cases 150. where the early calculated maps are very noisy is the use of a weighting factor such as the one described by Sim (1959) to reduce the contribution of terms which are likely to be poorly phased. Although the R-factor and the flatness of the ~F map indicate that the refined structure is generally correct, there are a few regions which would benefit from further refinement. The first is the area around Lys 145, where the a/S-trypsin autolytic cleavage occurs. There is some disorder in this region arising from the fact that the crystals used contain about 40% a-trypsin, and that this region an external loop, with some thermal motion probably present. forms Bode and Schwager (1975a) in their refinement of the benzamidine-trypsin structure, which was analyzed as 80% S-trypsin, also suggest that there is some mobility in this region . .Another area requiring further work is the C-terminal a helix between residues 238-244. The slowness of con- vergence in this region is primarily a result of the large errors in the starting model. There are several improvements which can be made in the present refinement system. Although a refinement cycle can presently be car- ried out in a single day, it would be advantageous to shorten the time required. The rate-limiting step is presently the calculation of structure factors. It may be possible to greatly shorten this step by use of a three-dimensional fast Fourier transform (FFT) program such as that described by Ten Eyck (1973). Generation of sampled electron density from a list of atomic coordinates, suitable for transfonnation 0 to obtain 1. 5 A structure factors (:i.e. , one asymmetric unit of the 151. 0 structure on a 0.75 A grid) currently requires about twenty minutes on the minicomputer. A three-dimensional FFf program would also be use- ful for generating electron density maps, and implementation on the NOVA 800 would eliminate the need for use of a large computer for any step in the cycle . The fast Fourier algorithm would be essential for minicomputer refinement of a much larger structure, or for extending the resolution of the present structure to the limits imposed by the diffraction pattern . The treatment of low-angle reflections, currently accomplished through use of the nonlinear scaling expression, might be improved by including some physically reasonable model for the disordered solvent continulllil in the structure factor calculation. An improvement which is expected to facilitate convergence of the refinement is the incorporation of information from the electron density map into the residual minimized in the constraints program, to yield a constrained real-space refinement program. The modifications required are very simple, and the results obtained to date in tests of this program indicate that it is considerably more efficient than alternate cycles of atomic shifts and constraints. This program can also be readily adapted for energy minimization of the type described by Levitt (1974), as suggested by Hermans and McQueen (1974). Konnert (1976) has used a similar technique in reciprocal space to perform constrained structure factor least-squares refinement. One further capability which can be added to the constraints program involves the configuration at asymmetric atoms. Currently, any inversion of 152. configuration brought about by allowing atomic shifts to become too large between applications of constraints must be corrected manually. Nevertheless, it seems likely that inclusion of information about the configuration, such as the scalar triple-product of the bonds to the atom, could prevent or correct these errors. The computing cost of the DIP-trypsin refinement to date has been approximately $650, almost entirely for calculation and plotting of Fourier maps on the IBM 370/158 . About 320 hours of CPU time on the NOVA 800 have been required over the course of the refinement for programs in which computation is rate limiting, excluding program development. Considering the sevenfold decrease in CPU time that we have generally observed for cornpute-botmd programs on the IBM 370/158, this translates into about $25,000 for equivalent time on the large machine at the current overall average rate of $540/CPU hour (including use of peripherals, etc.). It nrust be kept in mind that, had the refinement been carried out on a large machine, the programs would have been written to take advantage of the greater core storage available with a corresponding significant decrease in execution time. However, since the figure above is exclusive of program development and neglects the very large ntnnber of smaller programs executed during the refinement, it is felt that this figure represents a realistic estimate of the equivalent cost on the IBM 370. The savings in computing costs with the use of the minicomputer have therefore been substantial, and considering the potential cost for refinement of a protein structure, these savings can rapidly begin to offset the initial investment for 153. the smaller machine. Since this paper was first submitted for publication, two cycles of a least -squares refinement procedure using a first-order gradient 0 minimization reduced the 1. 5 AR-factor to 23 . 5%. 0 The mean change in atomic position was small - -about 0. 07 A during each cycle. Details will be described in a later paper. ACKNOWLEDGMENTS We wish to thank Drs . Monty Krieger and Anthony Kossiakoff for valuable discussions. 154. REFERENCES Bode, W. and Schwager, P. (1975a). J. Mol. Biol. 98, 693-717. Bode, W. and Schwager, P. (1975b). FEBS Letters 56, 139-143. Chambers, J.L., Christoph, G.G., Krieger, M., Kay, L.M., and Stroud, R.M. (1974). Biochem. Biophys. Res. Connnun. 59, 70-74 . Cochran, W. (1951). Acta Crystallogr. _!, 408-411. Deisenhofer, J. and Steigemann, W. (1975). Acta Crystallogr. B31, 238-250. Diamond, R. (1966). Acta Crystallogr. 21, 253-266. Diamond, R. (1971). Acta Crystallogr. A27, 436-452. Diamond, R. (1974). J. Mol. Biol. g, 371-391. Dodson, E.J., Isaacs, N.W., and Rollet, J.S. (1976). Acta Crystallogr. A32, 311-315. Fehlhannner, H. and Bode, W. (1975). J. Mol. Biol. 98, 683-692. Forsyth, J.B. and Wells, M. (1959). Acta Crystallogr. 12, 412-415. Freer, S.T., Alden, R.A., Carter, C.W. Jr., and Kraut, J. (1975). J. Biol. Chem. 250, 46-54. Henderson, R. and Moffat, J.K. (1971). Acta Crystallogr. B27, 14141420. Hermans, J. Jr. and McQueen, J.E. Jr. (1974). Acta Crystallogr. A30, 730-739. Huber, R. , Kukla, D. , Bode, W., Schwager, P., Bartels, K., Deisenhofer, J., andSteigemann, W. (1974). J. Mol. Biol. 89, 73-101. Konnert, J .H. (1976). Acta C1ystallogr. A32, 614-617. 155. Kossiakoff, A.A., Chambers, J.L., Kay. L.M .. and Stroud. R.M. (1977). Biochemistry 16, 654-664. Krieger, M., Chambers, J.L., Christoph, G.G., Stroud, R.M., and Trus, B.L. (1974a). Acta Crystallogr. A30, 740-748. Krieger, M., Kay, L.M., and Stroud, R.M. (1974b). J. Mol. Biol. 83, 209-230. Levitt, M. (1974). J. Mol. Biol. g, 393-420. Marsh, R.E. and Donohue, J. (1967). Adv. Prot. Chern.~' 235-256. :Moews, P.C. and Kretsinger, R.H. (1975). J. Mol. Biol. 91, 201-228. Ramachandran, G.N., Lakshrninarayanan, A.V., and Kolaskar, A.S. (1973). Biochern. Biophys. Acta 303, 8-13. Ramachandran, G.N. and Sasisekharan, V. (1968). Adv. Prot. Chern. 23, 284-438. Richards, F.M. (1968). J. Mol. Biol. 37, 225-230. Salernrne, F.R. and Fehr, D.G. (1972). J. :Mol. Biol. 70, 697-700. Sayre, D. (1974). Acta Crystallogr. A30, 180-184. Sim, G.A. (1959). Acta Crystallogr. 12, 813-815. Stout, G.H. and Jensen, L.H. Practical Guide. X-ray Structure Determination: A Macmillan, New York, 1968, pp. 356-385. Stroud, R.M., Kay, L.M., and Dickerson, R.E. (1971). Cold Spring Harbor Symposium on Quantitative Biology 36, 125-140. Stroud, R.M., Kay, L.M., and Dickerson, R.E. (1974). J. Mol. Biol. 83, 185-208. Ten Eyck, L.F. (1973). Acta Crystallogr. A29, 183-191. 156. Ten Eyck, L.F., Weaver, L.H., and Matthews, B.W. (1976). Acta Crystallogr. A32, 349-350. Watenpaugh, K.D., Sieker, L.C., Herriot, J.R., and Jensen, L.H. (1973). Acta Crystallogr. B29, 943-956. 157. TABLE I Computing System Configuration NOVA 800 (Data General Corporation) 32 K (16 bit) Words Core Floating Point Hardware* 12-1/2 ips Magnetic Tape Drive 25 ips Magnetic Tape Drive 2.5 M byte Moving Head Disk Keyboard Printer Matrix Line Printer/Plotter Typical Execution Times for FORTRAN Coded Instructions OPERATION TIME (µsec) Integer Add 2 Integer Multiply 80 Floating Point Add 20 Floating Point Multiply 25 Floating Point Divide 50 SIN(X) 400 *Syntex Analytical Instruments 158. FIGURE CAPTIONS FIGURE 1. This flow chart outlines the major steps involved in a single refinement cycle, using the technique described in the text. Alternate branches are indicated in the center of the figure. The main sequence was through automated shifts in difference maps. The branch toward small-scale maps was used for major reorientation of parts of the structure. The wire model was rebuilt once during the re- finement as an additional check on its progress, as described in the text. FIGURE 2. The observed F's for DIP-trypsin fall off rapidly with respect to the calculated F's at low angles, as shown in this scaling plot for the 1.5 A data. rF0 /rFc was computed for 100 zones in sin 2S/A 2 . The solid line drawn through these computed points (O) represents the four-parameter fit . fit are symbolized by(+). Points calculated from the linear Unit weights were used in this case . For 0 the linear fit, points inside 5 A resolution were neglected. FIGURE 3. The :improvement in agreement between F0 and the scaled Fe for low-angle reflections when using the four-parameter scale is shown in this graph of overall R-factor vs . resolution. The R-factor was computed for each point using all reflections to the indicated resolution . (+)=linear fit; (O) = four-parameter f it . 159. FIGURE 4. Summary of DIP-trypsin refinement using the scheme outlined in Figure 1, indicating the improvement in agreement between F0 and Fe. The decrease in R-factor over the course of the refinement 0 0 was from 47.2% at 2.7 A to 27.7% at 1.5 A. The sharp increase in Rat 0 Revision 16, after rebuilding the wire model to a 1.76 A refined map, resulted from errors in the wire model coordinates, as described in the text. FIGURE 5. The distances between a-carbon atoms in the starting and refined structures are shown in this histogram. The largest changes all occurred on the surface of the molecule, especially in the C-tenninal a helix. FIGURE 6. The fit of the model to the electron density at dif- ferent stages of refinement in the region around His 40, viewed down the crystallographic~ axis, is pictured in these stereo drawings generated using the minicomputer. The portion of the sequence shown is 133, N34, S37, G38, Y39 (the entire ring is not shown), H40, F41, C42 (plus the Sy of C58), and G43: a) The misorientation of the Y39-H40 (and to a lesser extent the 0 N34-S37) amide is reflected in an early 2.1 A tF map. Contours are from 0.3 e-/A3 in increments of 0.15 e-/A3 . The dotted contours represent negative density. 0 b) The orientation of these amides is ambiguous in the 2.7 A - 03 - 03 MIR map, contoured from 0.375 e /A in steps of 0.25 e /A. 160. c) The reorientation of these amides allows an excellent fit to i<P 0 the 1.5 A refined (F0 e c) map, where the ambiguity no longer exists. Contours are from 0.5 e-/A 3 in steps of 0.35 e-/A3 . 0 The 1.5 A 6F map is correspondingly clean since the required corrections have been made. Contours are from 0.2 e-/A3 in 0.1 e-/A3 d) steps. FIGURE 7. A view down the b axis of the area around the cata- lytic site residues Asp 102, His 57, and Ser 195, shows the fit of the O i<P refined coordinates t o the 1.5 A refined ([2F0 -Fc]e c) density. sulfur atoms in the 42-58 disulfide bridge are well resolved. The The very large peak at the lower right arises from the phosphorus and oxygen atoms of the DIP-group. - - 03 Contours start at 0.5 e /A in 03 0.5 e /A steps. FIGURE 8. i<P (2F -Fc)e c map. 0 0 View of the same area as shown in Figure 6 of a 1. 5 A All the atoms shown in this drawing were omitted from the phasing, indicating that the structural infonnation is indeed present in the observed F's. Contours are from 0.5 e-/A3 in 0.4 e-/A3 steps. 161. Figure 1 FLOW CHART FOR DIP- TRYPSIN REFINEMENT CYCLES Monuol Sh ifts utlnQ Smoll-tcole .. . Rebu ild Wire Mod1I I Remeo1un and Edit , , Mopl Edit Coord1. + I I I I Fl nd Solvent t~ mtn.l -➔ J yes no CIOH Contoc11 110 min .) Coord1. 162. Figure 2 WILSON PLOT L INEAR VS~ PARAMETER FIT. 1.5 A REFINED MODEL 8 ♦ d di) a,P. "' ? <ct> G~ "-d ..... '"-"' 0 ...,Ill ~d z' 0 .J ~ ? §i ? "' ~ ? ~j 0. 000 0.020 Q.QijQ 0.060 0.080 0.100 (SIN[TH£TAJ/ LAMBDAl~*2 0. 120 163. Figure 3 A VS RE SOLUTION LINEA R VS ~ PA RA METER FIT. 1. 5 A REFINED MODEL 0 ~ d re d 0. OSO 4 .0 3.0 2 .5 o. 100 0.200 0. ,00 2.0 1. 7 1.5 RESOLUTION IA) 0. 150 11 0 .. 3 a: u.. I 0 \ •, I\ 1 • ." I .\ llf lilops Automoted IApplyj Shilts in Constro,nls .........~ I •""'l' """ "-. 1. l" Rebuild Wire Model '·"'·-·------~.\ . ·-·-· Chee Aufomofed Shifts in M lilops ond Constraints alter Each Cycle ""'-•--.~,_._ _ j -------·--.__. r;n,r lilop Grid Revision of Coordinotes .24 L--_._~-~:--''---:---~------:c------=----!c---::~-~-~-.,......,-';:---'--:-';---:'-:----c:':---:-'=--~::---:':---:-'=-----d-:---~...J_-=1:--J I 2 3 4 5 6 7 8 9 10 II 12 13 14 15 16 17 18 19 20 21 22 23 .30 £40 u 501/J\ 2r0 -Fc lilops Model Rebu;ld,ng ond SmolHcole ~ 2.7A -¥- - - - - -- --2.I A - - - - - - - - - - - ' - - - - - - - 1 . 7 6 A--------1.54 ➔ +'> ! -n I-' • f-' +'> °' 165. Figure 5 STARTING VS I. 5 A REFINED MOOEL <;f.:.11 .'l dl .IJl'.1 8 Sf n1 ~:. 1111 "' 8 ~ui ::,: 0 a: ...Ul8 ~,; a u 8 8 25 50 75 100 125 I SO RESIDUE NUMBER 175 200 225 250 166 . Figure 6a 167. Figure 6b 168. Figure 6c 169. Figure 6d D <'.) 170. Figure 7 171. Figure 8 172 . Appendix 7 One Picture is Wort h a Thousand Words : A Trypsin Pictorial Assay 173. This appendix consists mainly of stereo figures of the refined trypsin structure, generated with the system described in Chapter III. The electron density syntheses used for these figures ,were all computed with phases ($c) calculated from the current model. Except where noted, the observed structure factor amplitudes (F) for DIPo trypsin to 1. 5 Aresolution were used, and contour levels were approx- 03 03 imately 0.5e/A x 0.5e/A 03 03 (i.e., proceed from 0.5e/A in 0. 5e/A in- crements). The atoms of the DIP-trypsin catalytic site are shown superposed on the F0 , $c electron density in Figure 1. 0 ~ 0 The Fourier series was 0 terminated at 6 A, 4.5 A, 3 A, and 1.5 A nominal resolution in frames a, b, c, and d, respectively. There are a nunber of interesting and informative features in this series. First, these density maps are essentially "best-case" maps; the amount of detail visible at each resolution is about as much as can be expected for a protein structure of mediun size. Second, the appearance of structural information can be followed as a function of resolution. At 6 A (Figure la) *, the main chain of 0 the protein cannot be readily traced; the 310 loop containing the His 57 imidazole is seen as a pillar rather than a loop. ~reover, no side chain infonnation is visible (with the exception of a large peak for the DIP-group). 0 At 4. 5 A (Figure lb), it is possible to trace the path of the main chain, although the orientation of amide A * In view of the existence of a well-defined saddle-point in the 6 electron density above the His 57 imidazole, the high-occupancy site in Ag+-trypsin will henceforth be referred to as "high-o. silver." 174. planes cannot be easily detennined. Density is present for only the larger side chains (e.g., His 57). Ripples in the electron density, apparently resulting from series termination, can be seen; for example, there is little density for the main chain from residue 195 to 196, in back of the DIP-group. At 3 A (Figure le), there are protru- sions in the main-chain density for carbonyl groups, and all of the 0 side chains are visible, although interpretation of 3 A density is in 0 practice often ambiguous. At 1.5 A (Figure ld), individual atoms have begun to be resolved, and interpretation is straightforward. A third and very important point about this resolution series is that in the interpretation of low- and medium-resolution maps (e.g., when constructing a wire model), placement of atoms should not necessarily follow the highest density. An obvious example is that the density tends to be highest at the center of rings and disulfide bridges rather than at the atomic sites themselves, at resolutions 0 above 2.5 A. lb and le. the density. Some more subtle cases, however, are apparent in Figures At 4.5. A, the disulfide bridge is not at all centered in This arises mainly from overlap with the density of the Phe 41 side chain in the upper right and the main-chain density in 0 the upper left. In the 3 A map the CS of His 57 appears uncentered due to overlap between the main chain density and that of the imidazole ring. Overlap of the Asp 102 carboxyl-group density with that of the His 57 imidazole and of Ser 214 0y (just below the lower boundary of the picture) causes the carboxyl to look somewhat miscentered, and was a major factor in the original placement of these 175. groups such that the His 57 No 1 pointed between the two carboxyl oxygens (1). 0 Even in the 1.5 A map atoms should not necessarily be positioned in rnaxinn.nn density. For instance, in the case of carbonyl oxygens the maximum density is displaced toward the carbonyl carbon. This result serves to emphasize the value of difference syntheses in protein structure refinement performed in electron-density space . The orientation of the catalytic groups Asp 102, His 57 , and Ser 195 can be seen quite clearly in this figure . Figure 2 is a vi ew down the z axis of the DIP-serine . Its re- semblance to the postulated tetrahedral intermediate is apparent ; the oxygen atom corresponding to the carbonyl oxygen of a specific substrate is well stabilized in its position by hydrogen bonds to the main-chain amides of Gly 193 and Ser 195. Also visible in the upper right is the salt bridge between the carboxyl of Asp 194 and the amino terminus, which is formed upon activation of trypsinogen. The region arolilld His 40, approximately the same region as shown in Figures 6 and 8 of Appendix 6 but computed with more recent phases, is pictured in Figure 3. The orientation of the amide planes in thi s loop was highly ambiguous in the starting (MIR) map . The fit of some typical aromatic side chains to the refined density is illustrated in Figures 4, 5, and 6. The indole ring in Figure 4 is part of the side chain of Trp 141. The N£ 1 is hydrogen bonded to the Asp 71 carbonyl oxygen, which can be seen in the upper right. In the lower left, part of the His 40 imidazole ring and the C=O of Gly 193 are visible . Two well -ordered solvent molecules are 176. also shown. Figures 5 and 6 are two different views of Tyr 234, down x and down z, respectively. The ring and associated peptide chain ar e very well defined, to the point of resolving the individual atoms. Figures 7 and 8 are pictures of one of the regions most in need of further refinement--the primary ca 2+ site. The manner in which the polypeptide chain forms a pocket for the Ca 2+ ion can readily be seen in stereo in Figure 7. Because of the high degree of crowding , this figure was contoured at a very high level--from 0.80e/i3 in 0.5Se/A3 steps . The occupancy of the bound ion (solid black in this picture) has increased dramatically during the last few refinement cycles, and • now appears 1·k • 1s • 1n • dee d Ca 2+ . Although the 1t 1 e1y that t h"1s 10n positions of the six ligands are for the most part well defined , the orientation of the Glu 70 (to the left and below the ion) and Glu 80 (to the left and above the ion) side chains is still not completely clear. The nearly octahedral coordination of the bound ion can be observed more closely in Figure 8. Frame 8a was drawn from the cur- rent set of coordinates and phases; frame 8b is based on interpreta0 tion of the first 1. 76 A map. The view in each frame is down z. There has been a significant change in placement of the Glu 80 side chain; however, as suggested above, the interpretation for this group is still tentative. There is a high degree of negative charge sur- rounding this site, the positive ion contributing to neutralization of charge on three different Glu side chains. 177. Thus, the refinement, although in its final stages, is still continuing. Additional DIP-trypsin data have been measured to 1.35 A (28 = 69.6°, CuKa radiation), and these data will soon be incorporated. The DIP-trypsin coordinates have been deposited with the Brookhaven Protein Data Bank, Brookhaven National Laboratory, Upton, Long Island, New York 11973, from whom copies can be obtained. I am grateful to Dr. Charles Cantor for suggesting the resolution series. References 1. Krieger, M., Kay, L.M., and Stroud, R.M. (1974). J. Mol. Biol. 83, 209-230. \ 178. Figure Legends FIGURE 1. Resolution series, viewed down they axis, of the DIP-trypsin catalytic site. Each map is an F0 , ~c synthesis with terms included to the indicated resolution. 0 03 03 03 03 03 03 a) 6.0 A, contoured from O.OSe/A in O.OSe/A steps. b) c) 4.5 A, contoured from O. lOe/A in 0. lOe/A steps. 0 03 03 3. 0 A, contoured from 0.35e/A in 0. 30e/A steps . d) 1.5 A, contoured from 0.50e/A in O. SOe/A steps. 0 0 FIGURE 2. View down z of the "oxyanion hole" and N-terminal salt bridge. FIQJRE 3. View downy of the area around His 40. FIQJRE 4. Trp 141 and neighboring atoms, down z. FIQJRE 5. The region around Tyr 234, down x. 03 03 0.60e/A in steps of 0. 4e/A. FIGURE 6. Tyr 234, down z. Contours are from Part of the His 91 imidazole is visible in the lower left. FIGURE 7. . Ca 2+ b.1nd.1ng site. . The pr1JTiary This loop was almost 0 a solid ball of density in the starting 2.7 A map. There are several 179. very well ordered solvent molecules in this area, as well as the bound ion, which is totally black in this picture. FIGURE 8a. A closer view of the metal ion, which has recently refined to an occupancy nearly that expected for fully occupied ea 2+ . The six ligands are arranged in a nearly octahedral fashion . The ion is in a position to take part in charge neutralization of three different Glu carboxyl groups--those of Glu 70, Glu 77, and Glu 80. b) A drawing of the metal ion site based on interpretation of 0 the first 1. 76 A map. The largest changes have been in the occupancy of the ion and the position of the Glu 80 side chain, as described in the text. 180. Figure la 181. Figure lb 182. Figure le 183. Figure ld ''1 :.0... • • I ·••·· •-• . 184. Figure 2 185. Figure 3 186. Figure 4 / / ) ·.~ ,\ ; ----~~--, \ , )iii ~ 11 --:1 ,·,>: t JlI____ --::--, li.t@1)_ ~---, 187. Figure 5 -• -• :.-~ ~ ~ ~ , - -~ (,\ \ (\\\\, ,) \ ~:=- \1\ \ • ' x r • ~• •, • ;, - • ~ i,'fti ,. - - _,: 1sc . . -~;'r \ /;11j ~(i{,- 'f~/:[►i ;,.s~!f - ,r~t,-~;~ ~-Aj \l~~t l --" . ~,,."1; ,, .,1ii~f('.r(i ,\,~~ • / , ( - ~ ' i\ /!f l ::'~-...,,.,, 1'\ J ,_ f~ , , ,~ ,\ ~ ~-,r, ~,~ '. 'I '':~ / • =::,_.,.,. ,,, , 188 . Figure 6 ~ ,Cf;:}, ~~?}: "<...-:::.._~(. =---- 189. Figure 7 190. Figure 8 (a) (b)