The Structure Determination of M. denitrificans Cytochrome C₅₅₀ Citation Timkovich, Russell (1975) The Structure Determination of M. denitrificans Cytochrome C₅₅₀. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/xmcg-d794. https://resolver.caltech.edu/CaltechTHESIS:10202022-180817376 Abstract The protein sequence and the three-dimensional folding structure for cytochrome c 550 from Micrococcus denitrificans are reported. The tertiary structure resulted from an x-ray crystallographic determination. Standard protein chemistry techniques were applied to sequence all the fragment peptides from a tryptic digest of c 550 . This information was combined with a 2.45 Å electron density map of the protein where identifying the tryptic fragments in the map led to the correct ordering of the peptides and the final complete sequence. The crystallization of c 550 is described. Three chemical derivatives of the protein were found, PtCl 4 2- ; UO 2 2+ , and Pt(CN) 4 2- , and these aided in the isomorphous replacement solution to the crystallographic phase problem. The interpretation and refinement of the derivatives to produce an accurate protein map is discussed. This bacterial c 550 possesses a structure similar to the known cytochromes from eukaryotic sources and from a photosynthetic bacterium. The three are contrasted to conclude that for the cytochrome family, there exists a core structure of polypeptide folding. Current proposed mechanisms for the action of c-type cytochromes are rejected as being incompatible with the c 550 structure. The view is advocated that the exposed heme edge in cytochrome may be the sole active site for the molecule. 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): Dickerson, Richard E. Thesis Committee: Unknown, Unknown Defense Date: 6 September 1974 Record Number: CaltechTHESIS:10202022-180817376 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:10202022-180817376 DOI: 10.7907/xmcg-d794 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 15044 Collection: CaltechTHESIS Deposited By: Benjamin Perez Deposited On: 20 Oct 2022 19:04 Last Modified: 07 Aug 2024 18:55 Thesis Files PDF - Final Version See Usage Policy. 59MB Repository Staff Only: item control page

Copyright © by RUSSELL TIMKOVICH 1974 THE STRUCTURE DETERMINATION OF M. DENITRIFICANS CYTOCHROME ~ 50 Thesis by Russell Timkovich In Partial Fulfillment of the Requirements For the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California 1975 (Submitted September 6, 1974) - i ACKNOWLEDGMENTS I would like to thank my wife for her help and understanding, her sacrifices and patience, during a very busy four years. Profess or A. Tulinsky "of Michigan State University began my scientific training, and it was indeed an excellent beginning, by his advice and support. Professor R. E. Dickerson, my thesis advisor, has contributed invaluable guidance and inspiration to this work, to my education, and, perhaps most importantly, to my outlook on science. It is a joy and a pleasure to acknowledge a debt of gratitude to Profess or E. Margoliash for the opportunity to work in his laboratory. The research described in these covers would not have been possible without the assistance of George McLain and Dr. Lucile Smith of Dartmouth Medical School. During my pack-rat days, I relied heavily on equipment and facilities in the laboratories of Dr. R. Stroud and Professor L. Hood, for which I wish to express thanks. The author 1 s graduate program was made possible through financial support from the National Science Foundation and the California Institute of Technology. - 11 - ABSTRACT The protein sequence and the three-dimensional folding structure for cytochrome .£ from Microco.ccus denitrificans are 550 reported. The tertiary structure resulted from an x-ray crystallo- graphic determination. Standard protein chemistry techniques were applied to sequence all the fragment peptides from a tryptic digest of ~ 550 . 0 This information was combined with a 2. 45 A electron density map of the protein where identifying the tryptic fragments in the map led to the correct ordering of the peptides and the final complete sequence. The crystallization of .£ 550 is described. derivatives of the protein were found, PtCll-; Three chemical uoz2+, and Pt(CN)l-; and these aided in the isomorphous - replacement solution to the crystallographic phase problem. The interpretation and refinement of the derivatives to produce an accurate protein map is discussed. This bacterial .£ possesses a structure similar to the known 550 cytochromes from eukaryotic sources and from a photosynthetic bacterium.. The three are contrasted to conclude that for the cyto- chrome family, there exists a core structure of polypeptide folding. Current proposed mechanisms for the action of ~-type cytochromes are rejected as being incompatible with the _£ 550 structure . The view is advocated that the exposed heme edge in cytochrome may be the sole active site for the molecule. - iii TABLE OF CONTENTS CHAPTER I. II. PAGE ACKNOWLEDGMENTS i ABSTRACT ii LIST OF TABLES V LIST OF FIGURES vi INTRODUCTION I. 1. General 1 I. 2. Historical Background 2 1. 3. The Cytochrome Family 3 1. 4. Micrococcus denitrificans £ 1. 5. References 550 5 10 CRYSTAL STRUCTURE DETERMINATION 2. 1. Source of £ 12 2. 2. 550 Crystallization 2. 3. Heavy Atom Derivatives 17 2. 4. Data Collection 21 2. 5. Data Reduction to Structure Factor 13 Amplitudes 27 2. 6. Solution of the Phase Problem 31 2. 7. Refinement of the Heavy Atom Derivatives 49 2. 8. Refinement Weighting Functions 69 2. 9. Crystalline £ 75 2. 10. 2. 11. as a Function of pH 550 Ammonium Selenate Binding to £ 550 References 77 80 - Ill. IV. lV - PRIMARY SEQUENCE OF c 550 3. 1. Introduction 83 3. 2. Materials and Methods 85 3. 3. Results 89 3. 4. Discussion 110 3. 5. References 113 TERTIARY STRUCTURE OF c 550 4. l. Ordering the Tryptic Peptides 115 4. 2. Molecular Structure - General 120 4. 3. Molecular Structure - Details 127 4. 4. Molecular Structure - Some Net Effects 134 4. 5. Evolutionary Relationships 138 4. 6. Current Proposed Mechanisms for the Reduction of Cytochrome £ 141 4. 7. Alternate Mechanisms 142 4. 8. References 152 PROPOSITIONS 155 - V - TABLE OF TABLES TABLE PAGE CONTENT 1. 1. Some Characteristics of Cytochromes 2. l. Molecular Weight Determination of £ 2. 2. 4 550 on Sephadex Gel 13 Crysta] Forms of £ 16 2. 3. 550 Heavy Atom Derivatives 20 2. 4. Collected Data 30 2. 5. R-Factor Comparisons 32 2. 6. Comparison of Centric and General Space 51 Refinement 0 2. 7. 4A Parameters and Statistics 53 2. 8. Refinement of Minor Sites 56 2. 9. Final Parameters, Averaged Data 61 2. 10. Final Parameters, Eight-Derivative Data 62 2. 11. Final Refinement Statistics 63 2. 12. Heavy Atom Binding Sites 70 2. 13. Ammonium Selenate Binding Sites 79 3. l. Amino Acid Compositions 93 3. 2. Amino Acid Composition of £ 3. 3. Sequence of Tryptic Peptides 101 4. l. Assessment of Residues 117 4. 2. C 118 4. 3. Sequence 5 50 Cytochrome Sequence Comparisons 4. 4. Hydrogen Bond Comparisons 130 4. 5. Bacteriophage Coat Proteins 133 4. 6. Cytochrome Aromatic Ring Orientations 136 4 . 7. Sequence Similarity-Difference Matrix 140 4. 8. Chemical Modifications of Cytochrome 147 550 100 128 - vi TABLE OF FIGURES FIGURE PAGE CONTENT 1. 1. Comparative Reaction Kinetics 9 2. 1. Crystal Habit of £ 18 2. 2. 550 Radial Distribution of Native Data 2. 3. Centric PtC1 22 2. 4. 2- Patterson Map 4 Patterson Harker Sections for PtC1 2. 5. Patterson Harker Sections for uo 35 2 4 2 + 2. 6. 2 Patterson Harker Sections for Pt(CN) 2. 7. Pseudo-Symmetry of PtC1 36 - 39 2 4 - 42 47 2. 8. 2 - Sites 4 Thermal Ellipsoids of the Heavy Atom Sites 2. 9. Radial Distribution of the Figure of Merit 65 2. 1 o• . Heme Density Section, Three Derivative Phasing 66 2. 11. Heme Density Section, Eight Derivative Phasing 68 2. 12. Weighting Scheme Comparis6n 73 3. 1. Purification of Tryptic Peptides 90 3. 2. C 92 57 4. 6. Fingerprint of Peptides 550 a-Carbon Diagram of £ , Front View 550 C ,Side View 550 c 550 , Top View a-Carbon Comparison of £ and Tuna c 550 Schematic £ 5 50 Crystal Packing Diagram 4. 7 . Lysine Residues About the Heme Crevice 144 4 . 8. Heme Edge Binding of Transport Enzymes 150 4. 1. 4. 2. 4. 3. 4. 4 . 4. 5. 121 122 123 124 125 132 - 1 - CHAPTER 1. 1. 1. INTRODUCTION General Imagine plotting a graph of biochemical information content versus the nurr1ber of protein crystal structures which have been solved. If the information content was evaluated in terms of knowledge gained on the general principles of molecular architecture, the resultant curve might show a dramatic initial rise followed by a · gradual turn toward a smaller slope. This might follow from the bursts of knowledge which the first few structures provided on the elements of protein folding-hydrophobic and entropic factors, salt bridges, hydrogen bonds, oil drop packing and so on. Later work has provided fresh examples without necessarily revealing astounding new principles. That statement is not quite accur?,te,for recent advances in the structural basis for allostery may give the hypothetical curve delta-function jumps. On the other hand, if the information content were gauged in terms of the specific knowledge of how a protein or enzyme functions or what are the relations among a family of proteins, the curve might still appear steeply rising . Crystallographic results continue to be carefully compared with proposals for mechanisms of protein function,as the results continue to present the most stringent constraints against which models are judged. Structural studies on related proteins, such as the serine protease family, have revealed the common architectural features required to build a - 2 - catalytic active site. In a sense, the relevance of protein crystal- lography to biochemistry has shifted toward the solution of specific biological problems. The motivating force for the research to be described in these pages is the idea that the second type curve, that of increasing specific knowledge, will continue. cytochrome~ proteins. At issue here is the family of What is there to be learned about structural variations within the family? What are the limitations imposed by structure on the workings of the molecule? 1. 2. Historical Background The early stages of the work to be described here were set against a background of previous structural work on cytochrome~The 2. 8 A crystal structure of horse cytochro~e ~ had been submitted for publication in 1970 (Dicker son, et al. , 1971 ), although the main results had been available since 1969. A structure determination was in active progress on the reduced form of the tuna protein with the species change brought about by the availability of x-ray quality crystals. The data on hand were provoking interest in the biochemical community on the mode of action of cytochrome and it was possible to detect an increase in the number of research papers appearing on the properties and action of cytochrome. Already at this early date, there existed a definite promise for comparative structural biochemistry in the cytochrome family. In Dr. J. Kraut' s laboratory at the University of California, San Diego, F. R. Salemme and co-workers were investigating the crystallographic - 3 - structure of a bacterial .£-type cytochrome from Rhodospirillum rubrum. This molecule, called cytochrome .£ , showed a typical 2 .£-type spectrum and molecular weight, but a redox potential some 60 mv higher and an isoelectric point some 4 pH units lower than the mitochondrial protein. The physiological role of .£ was strikingly 2 different as it participated in a photosynthetic electron transport chain,in contrast to the eukaryotic cytochrome in a respiration chain. Some sequence homology with mitochondrial cytochrome led to predictions concerning possible structural congruity (Dickerson, 1971), but it was not at all clear how similar the two molecular versions might be. The structural analysis on .£ was completed during the early 2 stages of work on .£ (Salemme, et al., 1973) and the results, which 550 will be discussed in detail later, increased interest in .£ purposes of comparison. 550 for It began to look as if there would be a characteristic folding pattern within the cytochrome family. To recognize the pattern, a high resolution structure for three representative family members would be necessary. 1. 3. The Cytochrome Family The ubiquity of .£-type cytochromes is the ubiquity of electron transport chains. As such it extends to all known organisms with the exception of certain primitive bacteria which survive on a pure fermentative metabolism (e.g., Clostridia). The most widely known cytochrome role is in oxygen respiration as it occurs in the mitochondria of eukaryotes. As sequence studies show extensive - 4 - homologies among the eukaryotic cytochromes, and physiochemical characterizations reveal common properties, the class is taken as nearly identical in structure and function. On the prokaryotic side of life, the uniformity which has been seen in the eukaryotes is obscured by a wealth of diversity. As a class, the prokaryotic cytochromes exhibit remarkable extremes in molecular weight, redox potential, spectra, · isoelectric point, and, most significantly, biological function (see Table 1.1). It remains to be seen what structural variation can produce such a wide range of properties. TABLE 1. 1 SOME CHARACTERISTICS OF CYTOCHROMES Molecular weight P. aeruginosa £ 551 82 residues Reduction potential Desufovibrio desulfuricans £ 135 residues 3 -205mv Isoelectric point horse c R. spheriodes cc' a- band @ 548 nm Function 2 +320mv 550 4.5 R. palustris £z a-band@ 555 nm R. rubrurn £z horse c in o Rhodospirillum rubrum £z M. denitrificans ~ 10.5 Spectra M. denitrificans £ 550 respiration in photosynthesis As far as cytochrome lore goes, the situation is like a man from some previous century attempting to understand transistors. He may study some one transistor in some one radio in great detail - 5 - and still lack the knowledge of what is ''transistor" and what is "transistor being in radio". By studying components of televisions or computers or what not, this level of understanding could be reached. Hence one looks at a variety of cytochromes ~ to see how structural variations can lead to the known diversity of function and physiochemical properties. 1. 4. Micrococcus denitrificans ~ 550 Granted that there is validity in looking at structure in a bacterial cytochrome, the next point is which one? denitrificans ~ 550 Micrococcus was an attractive candidate and a compelling reason was that its characterization, isolation, and purification to a high degree had been worked out by Scholes, Mcl..a.in, and Smith (1971). Besides the practical factors, the possible contributions to the previously mentioned information content curve were considered. An early realization of the cytochrome evolutionary implications was due to the thoughts and influence of Professor R. E. Dickerson. By comparing the known eukaryotic cytochrome structure with that of the prokaryotic ~ 550 , one could obtain some idea of the evolution which had taken place since the prokaryotic-eukaryotic divergence, an event dated some 2-3 billion years in the past (Barghoorn, 1971). With the forthcoming structure of photosynthetic ~ , it would be 2 possible to ask questions concerning the history of respiratory and photosynthetic transport proteins. It might be within reason to test ideas such as those of Gaffron (1962) which hold that photosynthesis - 6 predates respiration and that components of the latter may be evolution products of the former. At issue was the thought that by comparing the two primitive bacterial cytochromes (photosynthetic £ and respiratory £ 550 2 ) with the modern mitochondrial counterpart, some inkling of the structural folding of the ancestor protein might be gained. To explain the differences in between £ 550 physiochemical characteri_stics and other cytochromes was another goal. molecular weight of circa 15, 000,£ 550 At a differed markedly from£ (12, 500) and bacterial cytochromes such as Pseudomonas aeruginosa £ 551 (8,000). With an acid isoelectric point, 4. 5, it contrasted with the strongly basic (10 . 5) mitochondrial protein. This was held to be a key point,for it has been well-documented that positive charges are involved in cytochrome recognition by oxidase and reductase (Smith and Minnaert, 1965; Davies, Smith and Wasserman, 1964). So even at this early date the possibility arose for identifying these charges as an island of positive lysines on a molecular surface of negative charge. Subtle spectral and redox potential differences are seen in comparing £ 550 , £ 2 , and eukaryotic £· For example, the respective redox potentials range from +260 mv for £ and £ 550 to ~320 mv for ~ 2 . For the first few points , the body of this thesis can provide at least partial answers. For explaining the delicate adjustments in spectra and reduction potential, protein crystallography may be inadequate . Consider the crystallographically similar structures for - 7 - ferrodoxin and the HIPIP iron-sulfur protein (Carter, et al., 1972) which differ in potential by 750 mv. M. denitrificans is not a photosynthetic organism so ~ differ in function from ~ . 2 550 must The protein has been implicated in the classical form of oxygen respiration and so it resembles eukaryotic ~ in that respect (Vernon, 1956). physiological function of~ and ~ to ~ 550 Are there any differences in the 550 ? Some role has been ascribed in-the nitrate reduction step of anaerobic respiration in M. denitrificans (Sato, 1956), but this function remains to be conclusively proved, in light of growing evidence for a b-type cytochrome mediating nitrate reduction in bacteria (Knook, Van t'Riet, and Planta, 1973). It has been more firmly established that ~ 550 participates in the nitrite reduction step (Newton, 1969). The most intriguing aspect of the comparative biochemistry lies in the realm of relative kinetics of electron transport. As purified and isolated components, the bacterial and eukaryotic cytochromes may be interchanged with their respective oxidases and reductases. As a general rule among the eukaryotic systems, if you take any eukaryotic cytochrome~ (say from horse), it will react with any other eukaryotic oxidase (e.g., cow oxidase) or reductase at the same rate as it would with its own natural oxidase or reductase. When one starts to mix bacterial prokaryotic components with eukaryotic components, the general trend is that of very slow or no reaction (e.g., see Yamanaka and Okunuki, 1964). - 8 - Even reduced cytochrome ~ 550 , reacting in a eukaryotic oxidase system, undergoes essentially no oxidation (2. 6% specific activity taking eukaryotic cytochrome~ as 100%). The same sort of experiment with eukaryotic reductase and oxidized ~ 28% to 50% activity. 550 gives about The true surprise comes when the converse experiment is performed, reacting eukaryotic cytochrome~ with M. denitrificans oxidase and reductase. With oxidase the specific rate is 410% and with reductase 220%! (For references see Smith, Newton, and Scholes, 1966; the data actually quoted here were the result of recent experiments to update the first reference given and will ·be the subject of a future L. Smith publication. ) The general picture is that the eukaryotic oxidase and reductase are quite specific in their acceptance of transport partners, suggesting a binding or recognition site for cytochrome ~- By inference it can be assmned that bacterial oxidase and reductase have a recognition site tailored along the same lines. binding to bacterial ~ 550 But here, in the process of evolving , the site has somehow become compatible with an unnatural redox component - eukaryotic cytochrome~- A schematic way of thinking about this is shown in Figure LI for oxidase kinetics which show the most striking differences. If the structure of ~ 550 were known, it could be contrasted with the already known eukaryotic structure taken as a substrate analog of ~ 550 . By looking for similarities and differences some idea of the oxidase/reductase binding site may be gained. In point of fact, the bacterial system electron transport protein interface would Figure 1. 1. CYT C □ C550 -::, I Eukoryotic oxidose > FAST N.R. Schematic of the comparative reaction kinetics. M. d. oxidose + Q+ ➔ HYPOTHETICAL REACTIONS I '° I - 10 - be sought at the level of cytochrome ~ by looking at kinetically acceptable structural variations between the substrates~ and ~ 550 . It should be a simple step to say that the principles which govern the interface for the bacterial system also apply to the eukaryotic system, but with some changes in specificity. then, the £s 5 Broadly speaking o structure may say something about how electrons are passed in transport chains. The next chapter deals with the x-ray crystallographic details for the solution of the crystal structure. Chapter 3 recounts the protein chemistry applied to sequence the cytochrome. Chapter 4 returns to biochemistry and discusses the results and implications of the complete structure. 1. 5. References Barghoorn, E. S., (1971). Scientific American, 225, no. 5, 30. Carter, C.W., Jr., Kraut, J., Freer, S.T., Alden, R.A., Sieker, L. C., Adman, E., and Jensen, L. H., (1972). Proc. Nat. Acad. Sci., ~, 3526. Davies, H. C., Smith, L., and Wasserman, A. R., (1964). Biochem. Biophys. Acta, ~, 238. Dickerson, R. E., (1970). J. Mol. Biol., 22_, 1. Dickerson, R. E., Takano, T., Eisenberg, D., Kallai, 0. B., Samson, L., Cooper, A., and Margoliash, E., (1971 ). J. Biol. Chem. , 246, 1511. Gaffron, H., (1962). In Horizons in Biochemistry, p. 59-89, Academic Press, New York. Knaak, D. L., Van t'Riet, J., and Planta, R.J., (1973). Acta, 292, 237. Newton, N., (1969). Biochim. Biophys. Biochim. Biophys. Acta, 185, 316. - 11 - Salemme, F. R., Freer, S. T., Xuong, Ng. H., Alden, R. A. and Kraut, J., (1973). J. Biol. Chem., 248, 3910. Sato, R., (1956). In Inorganic Nitrogen Metabolism, eds. W. D. McElroy and B. Glass, Johns Hopkins University Press, Baltimore. Scholes, P. B., McLain, G., and Smith, L., (1971). _!Q, 2072. Smith, L., and Minnaert, K., (1965). Biochemistry, Biochim. Biophys. Acta, 105, 1. Smith, L., Newton, N., and Scholes, P., (1966). In Hemes and Remo-proteins, B. Chance, R. W. Estabrook, and T. Yonetani, eds., Academic Press, New York. Vernon, L. P., (1956). J. Biol. Chem., 222, 1035. Yamanaka, T., and Okunuki, K., (1964). J. Biol. Chem., 239, 1813. - 12 CHAPTER 2. 2. 1. Source of _£ CRYSTAL STRUCTURE DETERMINA_TION 550 A small scale preparation of the purified cytochrome was attempted at Caltech according to the procedure of Scholes (Scholes, McLain, and Smith, 1971). The yield, less than 1 mg per 20 £ of M. denitrificans culture, was so low that large fermentation batches (300.Q) were subsequently necessary. The cultivation of the organism and subsequent extraction and purification of the cytochrome were performed in the laboratory of Dr. School. Lucile Smith, Dartmouth Medical The author is indebted to Dr. Smith and Geor.g e McLain for gifts of the protein, which were used without further purification. The small scale Caltech preparation confirmed qualitatively the spectral properties reported previously (Scholes, McLain, and Smith, 1971). An anomaly of the protein was noted in agreement with unpublished data of McLain and Smith. In molecular weight determina- tion by gel filtration on Sephadex G- 75 (Andrews, 1964), the protein shows a spuriously high molecular weight, indicative of a dimer (Smith data) or a tetramer (Table 2.1). Smith and McLain have analyzed their preparation by sedimentation techniques and have found the monomeric weight exclusively (unpublished results). Thus the aberration appears to be due to cytochrome-gel interaction and not due to the actual presence of polymeric forms, as has been observed for mammalian cytochromes (Margoliash and Schejter, 1966). - 13 TABLE 2. I GEL FILTRATION DATA Protein Molecular Weight V( elution)- V(void) Hemoglobin (marker) 67,000 7. 0 ml Candida krusei cyt ~ (marker) 12, 500 49. 4 ml "'"'60, 000 7. 4 ml ~ 2. 2. 550 preparation Crystallization As a preliminary to the x-ray work, it was necessary to grow large protein crystals with sufficient internal order to produce x-ray scattering. conform Special crystallization experiments were designed to to the constraints imposed by the amount of material, 20 mgs, available initially. small Variations on a technique first introduced by Zeppezauer were used exclusively (Zeppezauer, 1968). Ekland, and Zeppezauer, capillary tube (length 30 mm, A small bore glass 0. D. 6. 5 mm, I. D. 1 mm) was fitted at one end with a small square of dialysis tubing,held over the opening by an 0- ring cut from Tygon tubing with an inner diameter just smaller than the outer diameter of the tube. This cell was filled from a syringe with a small volume, 0. 01 ml, of an appropriate protein solution,and sealed with parafihn covering the other end. The entire cell was then placed into solutions designed to slowly precipitate, i. e., crystallize, the protein. Best results were obtained when the cell was placed membrane side up in an upright - 14 - test tube of precipitant. tion. A simple rationale may exist for the observa- Since all precipitant solutions were of higher density than the initial protein solution, the dense precipitant diffused across the membrane, and gravity helped to mix it with the dilute protein solution, thereby gradually increasing the concentration of precipitant throughout the cell. With the membrane down, a layering of light protein solution on heavy precipitant would lead to a high local concentration and rapid precipitation rather than slow crystallization. Each cell was treated as a single crystallization experiment and, since each experiment required so little protein, extensive sets of conditions were explored. The nature of the precipitant varied from organic solvents to ammonimn sulfate solutions. Among the latter trials, variable parameters consisted of protein concentration, salt concentration, pH, and temperature. Other experiments also explored the effect of buffers and salts added, in addition to the ammonimn sulfate. A general strategy was adopted for these trials. A small nmnber of cells were employed with all conditions constant but one, which was varied over a large range. For example, at constant pH, temperature, and protein concentration, the concentration of ammonimn sulfate was varied from 10% to 100% in intervals of 10% . The results of these trials were examined by microscope for any indication of crystallinity. Around the values of the varied parameter giving the most promising results, further trials were attempted on a finer scale. By permuting all parameters in this manner, conditions suitable for the growth of x-ray quality crystals were found. - 15 Optimum conditions for crystallization are given in the following procedure. The stock protein, as a precipitate in ammonium sulfate, was collected by gentle centrifugation and redissolved in 20% saturated ammonium sulfate at pH 7. 5 to give a 2% (w/v) protein solution, oxidized form. Microdiffusion cells containing this were placed in 95% saturated ammonium sulfate adjusted to pH 7. 5 with concentrated aqueous ammonia, and containing 1. 0 ~ NaCl. Crystal growth was complete in one week at room temperature, care being taken to a void disturbing the cells during this time. The requirement for NaCl was reminiscent of the crystallization of horse heart cytochrome_£ (Margoliash and Walasek, 1967). However, when the above procedure was duplicated with CsCl replacing the NaCl, crystals giving a diffraction pattern identical to the NaCl grown crystals were obtained. When Nal replaced NaCl only an oily, amorphous precipitate was formed. Thus it is the Cl- which is necessary for crystallization. A variant crystal form, type II, was grown if the above procedure was followed, but with the addition of maleate at a concentration of 0. 02 ~ and pH 6. 0. This result was attributable solely to the malea te, for ammonium sulfate solution adjusted to pH 6. 0 still produced type I crystals. Precession photographs were taken with crystals sealed in capillary tubes, with a small amount of mother liquor to preserve humidity. Crystallographic parameters observed are listed in Table 2,2. Measurement of crystal density was by the flotation method (Stout and Jensen, 1968). For _£ 550 , advantage was taken of the fact that for - 16 both crystal forms, the density was only slightly less than that of the precipitant solution. Saturated CsCl was added to a mixture of crystals in precipitant until the crystals attained neutral buoyancy, and then the solution was measured pycnometrically. Using this method, the reproducibility for Type I crystals was ±0. 003 g/ml. The density quoted for Type II crystals reflects difficulty in measurement due to the small number and size of these crystals. The resolutions quoted were visual estimations from photographic films of short exposure. TABLE 2. 2 CRYSTAL FORMS OF MICROCOCCUS DENITRIFICANS CYTOCHROME C Type II Type I Class Orthorhombic · Tetragonal Space group 0 0 Cell dimensions 42. 7 X 82. 1 X 31. 6 A Cell volume llO, 799A3 135, 722 A. 3 Molecules/asymmetric unit 1 1 Crystal density 1. 275 g/ml. 1. 3 g/ml. Mass (%) of protein in the unit cell 70% 56% Maximum resolution of observable reflections 2. 7A 2. 7A 0 37.5X37.5X96.7A 0 Type I crystals, oxidized cytochrome, were used for all subsequent work. This choice was based on the presence of three centrosymmetric projections in space group P2 2 2 , which are of 1 1 1 - 17 - value in solving the phase problem, and the extreme stability of this crystal form to x-ray irradiation. One crystal was exposed for more than 250 hours to nickel-filtered copper radiation with no photographically observable decrease in intensities. Figure 2. 1 shows an idealized crystal morphology, Type I, in relation to the crystallographic axes. This crystal habit was rarely attained, since growth against the inner capillary walls gave highly distorted crystals. Usable crystals possessed dimensions in the range (0. 02 ➔ 0. 10 mm) X (0. 05 ➔ 0. 25 mm) X (0. 2 ➔ 0. 5 mm). 2. 3. Heavy Atom Derivatives For M. denitrificans ~ 550 , a de novo solution of the structure was desired and hence an isomorphous replacement solution to the crystallographic structure was sought (Phillips, 1966). It was realized that an alternative approach existed, that" of the rotation function (for such an application, see Greer, 1971), but since the final ~ 550 structure was to be compared with eukaryotic S it seemed best to follow as independent a crystal analysis as was pas sible. The survey for derivatives followed a general experimental scheme. An appropriate derivatizing reagent at high concentration was added to solutions of protein precipitant, 95% saturated ammonium sulfate, pH 7. 5, 1. 0 ~ in NaCl. Several crystals (5-10) were added to 0. 5 ml of this solution and allowed to stand. At various time intervals, from one day to one month, crystals were removed and centric zone precession photographs taken,to compare the resultant diffraction patterns with the native pattern. Many compounds \ Figure 2. 1. a* 550 crystal. - HABIT An idealized ~ CRYSTAL .. . . .. . _____ : C* ◄ I :) .. ._ ................. -·.... ·:•......... ................... . . ... .. .. .. b* ..... 00 - 19 produced no changes, even in saturated solutions, or at the other extreme destroyed the internal order of the crystal. More hopeful candidates were examined as a function of concentration and soaking time, in an attempt (1) to produce maximum intensity changes consistent with isomorphous cell parameters, (2) to preserve the native "mosaic spread" (James, 1965), and (3) to preserve high resolution scattering data. Conditions for derivative production and the properties of the modified crystals are given in Table 2. 3. It will be noted that there were only three chemically distinct derivatives. Variations of concentration led to derivatives differing mainly in the number of sites of binding per protein molecule and the occupancies of the sites. Reference to the refinement section will illustrate that even under identical conditions, the binding produced different effective atomic numbers at the sites. This may have been due to a certain inherent lack of reproducibility in the liquid-solid reactions as has been frequently noted {Tulinsky, et al., 1973). Alternatively, the differences in calculated occupancies may reflect lack of precision in the measurements or a certain degree of insensitivity and indeterminancy in the refinement procedure. - 20 TABLE 2. 3 HEAVY ATOM DERIVATIVES Concentration Soaking Time K PtC1 2 4 l.3mM 4 days 42.81 82.45 31. 64 (NH ) Pt(CN) 4 2 4 5. 6 mM 7 days 42. 90 82. 36 31. 11 (NH ) Pt(CN) 4 2 4 11. 2 mM 7 days 42.80 82. 30 31. 17 UO (NO ) 2 3 2 saturated 3 days 42.76 82.76 31. 62 UO (NO ) 2 3 2 saturated (see text) 8 days 42.67 82.03 31. 65 42. 70 82. 17 31. 56 Derivative ~ ~ Native 0 Cell Parameters (A) a b C For all compounds reported, binding was complete by the indicated time, and no further reaction was observed up to one month later. The two uranyl derivatives produced from saturated uranyl nitrate arose from a side reaction peculiar to the uranyl ion. Initially solµble in mother liquor, the uranyl ion precipitated from concentrated ammonium sulfate solutions, if the pH were above approximately 5. Although the product was not characterized, it may have arisen from the replacement of an atjuo-coordination sphere with sulfate The difference in preparation of the two forms of uranyl protein was based on the assmnption that even at low concentration of uranyl nitrate the precipitation reaction proceeded and the product would not bind to the protein. A solution of 5 mg of UO (NO ) • 6H O in 2 2 3 2 1 ml of precipitant buffer (95% ammoniunl sulfate, pH 7. 5, 1 M NaCl) was allowed to stand for 24 hours. The resultant precipitate was - 21 removed by centrifugation and 0. 5 ml of the supernatant used for derivatizing as described previously. For the low occupancy form, simple standing for the necessary time affected reaction. For the high occupancy form, every two days for a period of eight days, fresh supernatant was prepared and it replaced the old soaking media. If protein crystals were placed in the uranyl solution before the precipitate had formed, the crystals accelerated the formation of the uranyl solid and in turn became coated by a fine layer of the yellow precipitate. While an alchemist's delight, this produced powder diffraction lines which obscured the protein diffraction measurements. 2. 4. Data Collection The bulk of the diffraction data was collected in two stages. The first included several redundant measurements of native crystals 0 and all the derivatives to 4. 0 A resolution. This approach, with the resolution cut-off being dictated by consideration of a convenient number of reflections to process, was adopted so that the known chemical derivatives could be analyzed as to interpretability before exhaustive data collection was initiated. Once a derivative had been successfully interpreted, data were recollected, including remeasure0 ment of the 4. 0 A shell, to provide a complete set from a single crystal. 0 The resolution limit of 2. 45 A (or a Bragg 2 8 angle of 36. 00°) was imposed by the inherent scattering from the protein crystals,as shown in Figure 2. 2. This graph,illustrating the behavior of the mean structure factor amplitude as a function of data resolution, reveals - 22 - C ... ... 0 ·-::> 00 .c 0 ·en ' -- °'v9i7·i 0 °'v08.G 0 +- a, ·-0 > ·-0 z ·-0 ... -0 0 0:: .o . r<) 0 0 't- . 0 0 0 0 0 0 0 N .--.. • CD °'vOO·t, 0 °'v9G·v 0 N C N (/) ~ Q) ---- ' µ; II · Cf) vv·g 0 0 -. "100"6 0 ---------'----------'L-------~0 0 0 0 LO LO I!) r<) tj- LL N :::$ b.O •r-1 - 23 0 a rapid diminution beginning just beyond 2. 8 A. This was somewhat unexpected, since crystals of comparable size, containing protein of comparable molecular weight and even less mass percent protein per unit cell,have been reported as scattering to a much higher resolution (Wyckoff, et al., 1970; Blake, et al., 1965). Anticipating some comments from the structure results, the discrepancy may be explained by noting that certain parts of the molecule appear highly disordered between two or more possible configurations. Thus the num.ber of scattering centers ordered to within the limits of the data resolution, relative to the total mass per cell, falls off as the resolution increases. Unless otherwise noted the following comments are equally 0 0 applicable to the 4A and the 2. 45 A data sets. Precautions taken prior to intensity measu·rements improved accuracy. Crystals intended for diffractometry were carefully chosen to be single, without cracks or spurs, and of the largest available size. Heavy atom derivatives were prepared as noted previously, and at the end of the optimal soaking time, all were mounted in capillary tubes at the same time, thus insuring equal exposure time to the deri vatizing agent. The capillary tubes employed, of standard lithium. borate glass (1 mm diameter), were pretreated by cleansing in fum.ing nitric acid and by coating with a water-repellant film. ("Desicote", Beckman Instruments, applied per manufacturer's directions.) This process enabled excess solvent to be removed from the crystal and capillary, - 24 - thereby lessening background x-ray scatter and improving signal-tonoise features. Crystals were sealed in the capillaries with a drop of mother liquor to preserve humidity. They were invariably mounted with the c~~ axis parallel to the spindle direction (azimuthal axis) of the diffractometer, since in this orientation the shape of the crystals most closely approximated a cylinder leading to a minimization of absorption effects. Reflection intensities were measured on a General Electric quarter chi-circle XRD-490 diffractometer equipped with a graphite monochromator and controlled by a PDP-8 computer. The controlling programs were modifications by Dr. J. N. Brown of the standard GE control system, and are discussed in context below. Preliminaries to a production mode data collection included a survey of the diffracted peak intensities as a function of the machine w-angle (equals the 8 Bragg angle). This revealed that incorporation of heavy atom derivatives led to the appearance of asymmetry and a broadening of the peak profiles, suggesting some disorder. For the native crystals and the PtC1 2 -derivatives, peak width at half-peak 4 height was circa 0. 13° in w, relatively independent of 2 8 over the 0 angular range of 2. 45A resolution data, while for the Pt(CN) U0 2 2 -t- derivatives, this number could be as high as 0. 45°. 2 4 -and This necessitated an independent choice of peak scan widths for each crystal observed. Monochromator adjustments (monochromator diffracting plane perpendicular to 2 8: 8 plane of the diffractometer) and incident beam - 25 collimation (1. 0 mm) were chosen to bathe the crystal in a beam of uniform cross-sectional intensity. The detector collimator, located 20 cm from the crystal, consisted of a system of 1. 0 mm and 2. 0 mm pinholes, chosen to minimize reception of background scatter while containing the entire diffracted ray. Unit cell parameters were determined from a least squares fit of calculated and observed Bragg angles according to the ± 2 0 method of Arndt and Willis (1966). The theoretical least squares 0 standard deviations (Lipson and Cochran, 1966) were: a(a) = 0. 02A, 0 0 cr(b) =0. 03A, a(c) =0. 02A; but a more realistic estimate is given by the statistical unbiased standard deviations of parameters from 0 independent native crystals: 0 O cr(a) =0. 04A, cr(b) = 0. 05A, cr(c) =0. 02A. The intensities of (0, 0, £) reflections were measured at X = 90° as a function of the azimuthal angle for future use in absorption correction. While the dependence of intensity upon azimuthal angle may in general change as a function of the index£, for ~ 550 crystals all (0, 0, £) reflections showed the same absorption profile and hence the most intense reflection, (0, 0, 4), was employed to construct the empirical absorption behavior. Integrated intensities were obtained by an omega step-scan procedure,always scanning through the peak from low to high omega to avoid gear backlash. The scan width was chosen as the average full peak width plus 0. 30° to provide a margin for wider than average peaks and peak center mis-setting. Steps were generally 0. 02° to 0. 04 °, with the larger steps used with the broadest peaks only. average time spent scanning a reflection was 96 seconds. The Background - 26 was measured by stationary counting both at the beginning and end of the scan,each for one-half of the total scan time, with the total background taken as a sum of these counts. To understand the order of data collection (the path taken by the instrument through reciprocal space) two cases must be considered. 0 All the 4A data sets were collected in a straightforward manner, with little attention paid to optimizing the task. In general, the reflections (h, k, £) were obtained by holding h and k constant and progressing through reflections of increasing £ up to the two-theta limit imposed 0 by 4A resolution. Then k and finally the £ index were permuted, leading to a scheme of collection consisting of nested loops of reflections with£ changing most rapidly, h least. For the high resolution data, the decision was made not to collect anomalous pairs, in spite of their potential usefulness (Matthews, 1966a, b; North, 1965 ), so as to complete data collection within the decay life-time of a single crystal. It was, however, necessary to collect sufficient re-measurements of some reflections during the crystal life, so as to ascertain the course and extent of decay. The data were therefore collected in segments according to the following scheme: 0 1. A "before" data zone of (h, k, 0) reflections from oo ➔ 2. 45 A with h, k obeying the relationship h+k=3n, n=integer. 0 2. A high resolution shell consisting of reflections between 2. 45 A 0 and 2. 80 A resolution. 0 0 3. An intermediate shell, 2. 80 A to 4. 0 A. 4. A completion shell, 4. 0 A ➔ oo. 0 - 27 5. An "after" zone consisting of the same reflections as in step 1. 6. Throughout steps 1-5 the same three,high two-theta reflections were monitored every 75 data points. A comparison of data observed during steps 1 and 5 established the over-all extent of crystal x-ray damage and any systematic pattern in the destruction process. The 3n condition limited the amount of reflections to a convenient number while preserving, for this space group, a random sampling of reciprocal space. The most likely victims of irradiation damage are high order data, describing molecular fine structure. Therefore, these were collected in the beginning before x-ray exposure of any duration in the indicated shell order. 2. 5. Data Reduction to Structure Factor Amplitudes To the net intensities, standard Lorentz and polarization corrections were applied ( Levy and Ellison, 1960), as were approximate absorption corrections (North, Philips, and Matthews, 1968) obtained from the empirical absorption curve previously described. The transmission coefficient for x-rays in the least absorbing direction divided by that in the most absorbing gave an average ratio of 1. 3. The stated method of monitoring x-ray decay revealed that under the given experimental choice of scanning times, irradiation damage was minimal. The classical R factor, defined as between 11 before 11 and 11 after 11 data zones averaged 0 . 03 over all sets and was never higher than O. 036. Perusal of these zones failed to - 28 - disclose any systematic effect, and the check reflections confirmed the stability. In consideration of this behavior, no time decay corrections were applied to the data. Reflections processed were assigned a standard deviation in percentage units defined by 3( cr = 100% • 2 • lobs+ BKG+ (0. 02 • lobs) 2)i /Inet 2 where I b 0 S = total scan counts BKG = total background counts I ne t=\Ib o s -BKG\ The square root dependency upon I b 0 S + BKG follows from counting statistics and represents a lower estimate to the actual standard deviation, which may be higher due to (a) machine error, (b) data collection error (in crystal orientation, cell parameters, crystal slippage within the capillary), and (c) data reduction error (in absorption and decay). The term (0. 02 • I b / 0 S attempts to alleviate somewhat any underestimation and dampen the steep descent of cr with increasing I b . 0 S The value of 0. 02 was used in a similar scheme in the solution of the trypsin structure (R. Stroud, personal communication), and was employed without any test of validity. The factor of 3/2 arises from a consideration of the statistical unbiased standard deviation for the monitor reflections, which were measured 50 to 100 times per data collection. This 3/2 factor was required to equate the calculated cr with that statistically determined. - 29 A somewhat critical use must be made of these standard deviations, for there is obviously a point below which net intensity counts are meaningless and a reflection is accordingly unobserved, within experimental limits. To arrive at some measure of the lower limit of detectable intensity, the space group systematic absences were measured redundantly and processed as normal observations. Most gave, of course, I ne t,..__, 0 to within ± 50 counts while the highest, , the limits of random fluctuation, were 100 to 150 net counts. These spurious reflections, at the level of total scan intensity and background measured, convert to percent standard deviations in the range of 33% to 50%. Hence 33% was taken as a realistic indication of an observable data point, with such observables for all data sets. criteria leading to 75% to 85% In the parlance of crystallography this scheme amounts to retaining all reflections 30 above noise. Table 2. 4 lists the data sets which were instrUinental in the solution of the crystal structure, a subset of the data actually taken. The ammoniUITI selenate, the high pH, and the low pH data pertain to some ancillary experiments on the protein structure and will be discussed later. measurements. The data sets listed amount to 32,400 reflection - 30 TABLE 2. 4 SOME DATA SETS COLLECTED I. Native Data EH Resolution 14N 7.5 2. 8A 17N 7. 5 2. 8A ((h, 0, £)only) 20N 7. 5 2. 45A Ac5. 1 5. 1 2 . 8A ((h, 0, £) only) NH OH 4 8. 9 2. 8A ((h,k, 0) and (0,k,£)) Reagent Resolution (NH ) Seo 4 2 4 2. 8A ((h, k, 0) and (0, k, £)) 0 0 0 0 0 IL Derivative Data 0 ASe2a 0 2- 4. 0A 2- 2 . 45A Cl9b PtCl 2 4 2. 45A CN7c Pt(CN) 2 Pt(CN) 2 Cl 7a PtCl Cl9a PtCl CN9a 4 4 0 0 0 - 4. OA - 2. 45A 4 (high concentration) 0 4 (low concentration) 0 UNlla (UO )(NO ) 2 3 2 (low concentration) 4. OA UN12IIa (UO )(NO ) 2 3 2 (high concentration) 2. 45A 0 - 31 - Table 2. 5 reports some representative R-factors, thus assessing the precision of the measurements. From Table 2. 5 it can be seen that data set pairs (14N, 20N), (CNlOb, CN9a), and (Cl7a, Cl9b) agreed well within the pair. Thus they were averaged together to give a master native set and a set of "average" data for 2 the PtC1 4 - and Pt(CN) 2 4 -derivatives. As will be seen, the various uranyl sets have extensively differing substitution and hence in this case averaging was not possible. 2. 6. Solution of the Phase Problem Data sets collected from various crystals were placed on a common scale by the method of Dickerson, et al. (1967a). An additional multiplicative scale factor was calculated and applied to the heavy atom derivative sets according t~ the method of Kraut, et al., (1962) to account for the theoretical increase of scattering matter per unit cell. The solution to the phase problem was essentially completed 0 with the 4A data. As will be seen) the higher resolution data were processed in a similar manner, but this mainly confirmed the original solution, improved the phases somewhat, and revealed features of the heavy atom derivatives lacking from the earlier work. So, until 0 noted, the ensuing discussion refers solely to the 4A data. 0 Two 4A sets of data were collected from crystals modified with Pt(CN) 2 4 - at two different concentrations. Raw intensity differences indicated somewhat different substitution between the two, and so they were treated as independent derivatives. Later results will indicate - 32 TABLE 2. 5 SOMER-FACTOR COMPARISONS Type Redundant zones, same crystal R-factor ~:< Sets 14N vs. 14N 0 . 02 20N 20N 0.03 vs. UN l 2IIa vs. Equivalent data, different crystals UN l 2IIa 0. 04 UNl la vs. Unl la 0. 04 ASe2a vs. ASe2a 0. 03 14N vs. 17N 0. 03 14N vs. Ac5. 1 0 . 04 14N vs. NH 0H 4 0. 05 20N 0. 04 14N vs. CNl0b vs. Cl 7a vs. CN9a Cl9b 0. 04 0. 05 LIF 1 - F 2 \ R = - - - - - , the summation over all comparable terms in the set. LI F1 I - 33 - that although the sites of attachment were the same, the occupancies differed. For each of the derivatives, centric and full three-dimensional Patterson syntheses were calculated using difference coefficients (Blow, 19 58), where F PH' F p refer to the derivative and native structure factors. Considerable experimentaion was conducted in an effort to improve the quality, i.e., feature-to-noise ratio, of these syntheses. For instance, various low order terms were omitted from the calculation under the assumption that they might arise from general salt or 2 solvent differences. Artificial temperature factors of the form e -Bs were applied to eliminate series termination effects or lack of isomorphism in the high orders. While manipulations led to changes in peak shapes and heights, it was not clear that any improvement in signal-tonois e ratio was obtained and certainly the maps were not easier to interpret. Such editing was abandoned and subsequent discussions refer to standard syntheses. Comparisons between centric (6F) 2 maps and those for the same derivative in three-dimensions revealed a common theme three-dimensional maps were easier to interpret and showed a higher ratio of peak height to general noise. Figure 2. 3 shows an (h, k, 0) centric difference Patterson synthesis calculated with centric reflections from the complete data set used to calculate the map shown in Figures 2. 4. 1 through 2. 4. 3. Figure 2. 3 has been contoured so that - 34 the projected Harker peaks, designated by "H", were on the same scale as the Harker peaks in the three-dimensional map. The centric map contained general Patterson cross-vectors and noise peaks projected onto its limited area, hence adding background and obscuring interpretation. It also represented a Fourier summation over 150 terms, compared to 1100 in three-dimensions, making it more susceptible to random error in a few terms. From Patterson syntheses a single site was found for the PtC1 2 - and uo 2 + derivatives and three sites for each Pt(CN) 2 4 2 4 derivative. Figures 2. 4, 2. 5 and 2. 6 present Harker sections from these maps. Only one Pt(CN) virtually identical. 2 4 -map is shown, since the other was In each case the three mutually perpendicular sections are contoured at equal, arbitrary intervals with zero and negative contours omitted. P2 For a single site of substitution in z z , peaks are expected at (1) ½, ½-Zy, 2z; (2) 2x, ½, ½-Zz; 1 1 1 (3) ½-Zx, 2y, ½- The identification of these vectors is indicated by the letter H for the single sites in PtC1 2 - and uoz2+and by the set H , H , 4 2 1 H 3 (decreasing occupancy) for the multiple site Pt(CN) Many strong features in the Pt(CN) the solution indicated. 2 4 2 4 - derivative. -map were unexplained by They may be definitively ruled out as random experimental error, for their heights and positions reappear in all the (6F/ maps calculated from independently measured data sets. They may arise in part from the relatively large cell changes of Pt(CN) 2 4 -crystals versus native crystals. Further, the large amount of substitution could have caused local protein deformations and lack of isomorphism. 0.5' UI 0.0 1' ' 0 ~ )11111 ( I I I\\\\ Figure 2. 3. I 11111£ (@) a ~ '' ' '' 1 II \ (6F) 2 , JJ I I Ill ))Ill ~ I II ( I f L I I -,., > I I I IIIIIIII) © ........ , 1111 Patterson synthesis of (h, k, 0) PtC14 2 - data. rg; \ t It 11 V 0. 5 illillllllli I I u, l,..) 0.5~ WI 0.0 \ 0 Figure 2. 4. 1. ...... I ( '-, @) 7 '(((H:? 0~ Harker section for (.6F/ synthesis based on three-dimensional PtC1 2 - data. 4 U= 0.5 / '® @ C2) 0 //////~"\."\.'\'\\ ® V \~~~ 0.5 0.5 0 0 ~ As in 2. 4. 1. V =0.5 ~ Figure 2. 4. 2. G w; / 0.0~ u I 0.5 I -.J l,v 0.5 I U, o.ol ~ I I /I/ff/Iii ,t\\\\\\ ~(Q) W= 0.5 Figure 2. 4. 3. L>\ © oQOO V As in 2. 4. 1. \\\ 0 JJJ C\ ~ V ~ \\ I J 11 I I 00 l,-) 0.5 I WI 0.0 Figure 2. 5. 1. ~ U= 0.5 ....--.... 0 / "' j) C) 2 Harker section for (.6F) Patterson synthesis based on 2 three-dimensional uo + data. 2 ///11-H~,\ \ 0 ~ 0 V 0 0 I~ 0.5 0.5 w 0.0, C> Figure 2. 5. 2. u Jllill C J 0 As in 2. 5. 1. V= 0.5 ® 11 0j5 ~ 0 0.5 u 0 ~ () As in 2. 5. 1. W = 0.5 Figure 2. 5. 3. 0 V 0 0 0.5 I I ...... ~ 0.5 w 0 0 Figure 2. 6. 1. • 0 J ') ((5))) -~ ~ a ~ ~(9/~, ~ v~ 2 Harker section for (.6F) Patterson synthesis based on three-dimensional Pt(CN) 2 - data. 4 V I I N ~ I 0.5 0.5 w Figure 2. 6. 2. As in 2. 6. 1 V= 0.5 ~Cb 0 I I w ,.p.. 0.5 uI 0.0 ~ 0 Figure 2. 6. 3. ~ ~ As in 2. 6. 1. C:=7 ~\ ~ ~ V 0 "-./ o\\ fa I I ~ 0.5 - 45 Up to this time, Wilson-type statistics (Wilson, 1942) had not been used to place the protein data on an absolute scale, because the method is unsuitable for protein crystals where the atomic scattering contribution of bound and interstitial solvent cannot be known a priori, and therefore cannot be included into a summation of squared atomic scattering factors over the crystal cell. Instead, the single site PtC1 2 - derivative was used to scale approximately native and platinum 4 data by determining the constant required to match observed I6FI terms with terms from a centric difference structure factor calculation based on a site of 78 electrons. The atomic number of platinum was used rather than the sum of electrons in PtC1 2 -, because there 4 was evidence that the chloride ions had been replaced by interstitial solvent. This will be further substantiated in the discussion of anisotropic temperature effects. Multiple solutions for atomic co-ordinates are possible from the PtC1 2 4 - Harker vectors, and selection of one such set established a crystallographic origin for future work and fixed an enantiomorph. This arbitrarily selected "handedness'' led to a calculated structure containing L-amino acids and right-handed helices and thus was fortuitously correct. The remaining task was to determine atomic co-ordinates for the other derivatives relative to the same origin. First the single PtC1 2 - site was refined over centric data for 4 several cycles according to the method of Dickerson, Weinzierl, and Palmer (1968). These authors advocate using atomic scattering 2 factors of the form Ze-Bs , where Z is an effective occupancy in - 46 electrons and B is a pseudo-temperature factor to include actual temperature-disorder effects and the scattering incoherence with increasing s = sin 8. Compared to Thomas-Fermi scattering centers, such a form,for typical values of B, was found to agree to better than 1. 6% over the resolution range of the entire protein data. Initial values were obtained (Dickerson, et al., 1967a) and refinement adjusted atomic co-ordinates and occupancy. Attempts to refine B 0 at this 4A stage led to unrealistic and wildly fluctuating values, so 0 for this and all other 4A refinement, the initial value was held fixed. 2 It is suggested that the scatter of values (.6F)/(Ze-Bs ) (averaged in 2 zones of s ) about a line of zero slope is too great for the narrow resolution range to define B adequately. For centric reflections with phase angles constrained to O or 1T (also ,r/2 or 3,r /2 in P2 2 2 ), the ready knowledge of the magnitudes, 1 1 1 lFpl and IFPHI, and just the phase of the heavy atom contribution allow unambiguous determination of the native protein phase. Therefore, the refined single PtC1 2 - site permitted calculation of a 4 set of native centric phases, which were in turn combined with appropriate terms .6F = (IF PH l - IF p I) from the Pt(CN) 2 4 - and uo 2 2 + data to produce a series of Fourier density syntheses of these heavy atom sites, of necessity referred to the PtC1 2 4 - origin. The outstanding characteristic of the resultant calculations was the presence of a non-crystallographic pseudo-symmetry element at . 1 1 o 1 ce 11 co-or d 1nates an d 43 , 43 ,2, 4, 4, the source; Reference to Figure 2. 7 reveals the dark circles mark the PtC1 2 4 - sites, four symmetry t X L- .£_ ♦ ♦ ~ + 4 lL- 4D + 4 I l. 4 l Z up ~ 0 0 ~ ♦ -2 I t -- Figure 2. 7. SJ - - SJ 82 I t -- -Y~ PSEUDO-SYMMETRY ~ 02 I I 02 ~ DIAGRAM I 2++ 4 ~~ 4 l~l I _.::.I I -.J >J:>. - 48 - equivalent sites per unit cell, which are seen to possess an x co-ordinate near zero (+0. 01). This produced an accidental centro- symmetric distribution of the PtC1 2 - sites about inversion centers at 4 11 1. 0 an d 3 4,4, 4 ,3,,.,2, structure factors. The effect is also seen in the calculated heavy atom If one takes the structure factor expressions collapsed over the four symmetry equivalent positions for P2 2 2 , 1 1 1 h-k) cos 21T ( ky - ~ k-£ ) cos 21T ( £z - ~ £ -h ) A = 4 cos 21T ( hx - ~ h - k) sin 21T ( ky - ~ k- £ ) sin 271" ( £z - ~ £- h ) . B = -4 sin 271" ( hx - ~ It can readily be seen that x = 0 leads to the conditions: (0, k, £) no conditions (h, 0, £) h + £ odd, A = B = 0 (h, k, 0) k odd, A = B = 0 . Examination of heavy atom structure factors for the refined PtCI 2 4 site revealed such pseudo-extinctions were approximately obeyed for both calculated -and observed .6F terms. Resultant Pt(CN) 2 4 - and uo 2 2 + .6F maps were constrained to show this pseudo-symmetry, one of the pseudo-symmetric pair being the correct site and the other spuriously introduced by the phasing abnormalities. For the Pt(CN) from false images. 2 4 - maps, true peaks could not be differentiated For the uo 2 2 +maps, the slight departure from perfect symmetry (x not identically zero) allowed the proper uo 2 2 + single site to appear at a peak height four times the spurious image. Tha t this choice was correct was confirmed by first refining the uo 2 + 2 data and site as a single derivative as had been done with the PtC1 2 4 - 49 data, then including the two into phase refinement concurrently and observing that the two derivatives refined congruently, i.e., predicting the same native protein centric phase. The protein centric phases, now calculated from two -derivatives, one not posses sing pseudo- symmetry, were again combined with Pt(CN} 2 4 - differences to generate maps in which the three expected sites could be found with no evidence of the former spurious images. The higher concentration Pt(CN} 2 4 - data led to slightly higher Fourier peaks suggesting that the two derivatives differed solely in occupancies. 2. 7. Refinement of Heavy Atom Derivatives With an origin-consistent set of atomic coordinates, refinement could be extended to the three-dimensional data using all derivatives. Weighting functions, a crucial aspect of the process, will be discussed in a separate sec~ion. Under refinement, the heavy atom parameters were varied in several combinations and permutations to determine the extent of correlation among the interdependent parameters, such as over-all scale and the isotropic temperature factor. As before, the thermal parameters could not be refined beyond their initial values. Refinement did not significantly alter the heavy atom data scale factors, yet caused fluctuations in occupancies, and so these scale factors were held fixed from the early stages onward. Refinement of the heavy atom sites in one derivative, holding the other derivative sites constant, invariably led to improvement in reliability indices (see Table 2. 7} for the adjusted derivative, at the expense of the same statistical index for the others. As the average figure of merit (Blow - 50 - and Crick, 1959) then suffered, it seemed obvious that all sites must be refined concurrently. The computational experiments in refinement were too numerous to be listed here. After the experiments, it was found that the same set of refined parameters were obtainable by five cycles of phase analysis-least squares adjusting concurrently co-ordinates and occupancies for all sites, starting with Patterson map initial co-ordinates and initial Z and B terms as discussed for PtC1 2 - . 4 It is instructive to compare the results of three-dimensional refinement with refinement over centric data only. For purposes of comparison the PtC1 2 -and uo 2 + sites were refined in general 4 2 reciprocal space and the results compared with the equivalent refinement, mentioned previously in connection with the relative problem, over centric data only. origins Table 2. 6 lists the corresponding parameters and reveals the marked similarity between the two methods. Were general space refinement for some protein project to be prohibitively expensive in computation cost or time, it appears that restriction to the limited centric data would be justifiable. At this stage, native protein phases were calculated from the refined parameters and a series of three-dimensional difference Fourier syntheses was produced using these calculated native phases and the observed ~F's for each derivative in turn. Although inclusion of all derivatives leads to the possibility of some feedback in this Fourier approach, the number of derivatives used for phasing lessens this tendency (Dickerson, et al., 1967 b). - 51 TABLE 2. 6 COMPARISON OF CENTRIC AND GENERAL SPACE REFINEMENT Refined Co-ordinates X y_ z z PtCl4 2 - , centric method 0. 0123 0.0519 0.2130 75. 84 , general method 0. 0135 0. 0526 0.2135 74. 66 uo 2 2 +, centric method 0. 0966 0. 3011 0.7415 37 . 29 , general method 0.0975 0. 3027 0. 7378 33. 16 Refinement Statistics ,:, Centric Refinement Method General Refinement Method Mean figure of merit, centric · 0.822 0.826 Mean figure of merit, general 0.619 0.626 PtCl 4 2- uo 2 2 + PtCl 2 - uo 2 + 4 2 Rk, centric 0.090 0. 058 0.091 0. 056 Rk, general 0. 087 0. 055 0.087 0. 051 R 0.481 0. 656 0.485 0. 629 58.9 40.4 59. 3 37. 9 C ERMS _,_ -··see Table 2.11 for d e finitions. These statistics follow from the final parameters and are independent of the refinement method. The statistics quoted as "centric'' or "general" refer to their compution over respecti v ely centric or g eneral reflections . This presenta tion esta blish e s that the refinement method does bi a s the s ta tis tics in favor of the data subs et refined. - 52 The maps were examined for minor sites not found from the Patterson, but now possibly revealed by the reasonably accurate protein phases coupled to the minor site contribution to 6F. From the frequency and distribution of features in each map, it appeared that the general noise level was approximately 8% of the peak height for the major site in that derivative. Therefore most attention was directed toward any features higher than this cut-off. and Pt(CN) 2 4 - For the u_o 2 2 + data there were none, and it was concluded that no minor sites were present in these derivatives . For the Ptc1 2 - data, two minor sites w ,e re possible, each at 4 13% of the major site peak height. At their indicated occupancy, their Harker vectors would be obscured by noise in the (6F/ map, but, their cross vectors should appear, and indeed peaks were found, at about the Patterson noise level, at the calculated vector positions. The two minor sites were therefore included in phase refinement where they immediately (one cycle) refined to zero occupancy (1 and 4 electrons versus 69 for the main site), and hence were dropped from consideration. 0 Table 2. 7 lists the final parameters at this 4A stage. 0 A 4A electron density map was computed using figure of merit adjusted coefficients (Blow and Crick, 1959). A discussion of results will be postponed except for a comment on series termination errors. Since the data for the map are limited to a small subset, with significant terms omitted, there is the possibility that series termination has introduced false features into the synthesis. To test this possibility, 2 artificial temperature factors of the form e ( - Ds ) were applied to the - 53 TABLE 2. 7 HEAVY-ATOM PARAMETERS AND . 0 REFINEMENT STATISTICS, 4A REFINEMENT R i k R i Heavy-atom sites C X y z z2 B2 PtCl 2 4 0.091 0.512 0.0087 0.0528 0. 2147 68. 6 4. 0 uo 2 2 + 0.053 0.619 0. 0992 0. 3018 0.7384 31. 7 2. 3 0. 108 0.458 0. 1008 0.4307 0.7954 76. 9 1. 6 0.4622 0.4664 0. 8657 46. O 1. 6 0. 3078 0. 2810 0.7530 33. 0 1. 6 0. 1017 0.4312 0.7952 84.4 1. 8 0.4635 0.4672 0.8654 53.8 1. 8 0. 3056 0.2812 0.7573 43.4 1. 8 Pt(CN) 2- 4 (5. 6mM) Pt(CN) 2 - 4 (11. 2mM) 0. 123 0.499 1 See definitions accompanying Table 2. 11. 2 2r . atomic . sea tt er1ng . f actors = Z e -Bs sotrop1c - 54 coefficients to generate a series with decreasing terms at the resolution limit. Let (F) be the average native structure factor amplitude over 0 0 all the 4A data before application of the artificial factor and (F) be the average computed over a narrow range of s limit (the highest resolution 13% of the data); chosen such that (F) 4 2 4 .A near the 4.A. then the D factor was .A was 75%, 50%, or 25% of (F) 0 . Several sections of an electron density map were calculated in each case and compared, yielding the generalization that no change in features was observed, only a gradual broadening of peaks and troughs. It was held that series termination was not a serious error and the artificial factors were omitted. 0 0 For the high resolution refinement at 2. 45 A, the 4A process was essentially duplicated. In this instance, initial refinement used the averaged data sets for native, PtC1 2 -, and Pt(CN) 2 - and the high 4 4 concentration uo 2 2 +set {Table 2. 4, set UN12Ila}. Several cycles were performed beginning with the 4.A parameters and it was now possible to refine temperature factors realistically for the individual heavy atoms. 0 The resulting protein phases were used to compute 2. 45 A .6F maps again matching the various derivatives in turn with the same master phase set. As in the previous case, these same-phase-cross- difference maps were searched for minor sites now with a minimum Fourier peak height of 6% of the main heavy atom site in the case of the PtC1 2 -and uo 2 +derivatives and 8% of the main site in Pt(CN) 2 -. 4 2 4 Two possible minor sites for PtC1 2 -, four for uo 2 ~ and six for 2 4 - 55 Pt(CN) 2 4 - satisfied calculated Patters on cross -vector predictions and were entered into least squares. Table 2. 8 gives the result of occu- pancy shifts and indicates that only one minor site in uo and one in Pt(CN) 2 4 - 2 2 + (site (site h) behaved consistent with the data. b) These were therefore accepted as valid. The now complete list of heavy atom sites was converted to an anisotropic temperature format and [3 .. parameters adjusted with lJ the phase program. strongest Pt(CN) The light uo 2 2 2 4 + and - . The main PtC1 2 - and uo 2 + sites and the two 2 4 sites showed reasonable, i. e. small, shifts. Pt(CN) 2 4 - sites yielded non-positive definite tensors,werereset to their isotropic equivalent, and subsequently were held there. This failure raised the question of whether the main sites were meaningfully represented by the anisotropic parameters or had they been shifted as extra degrees of freedom to accommodate random error in the data? To satisfy this dilemma an OR TEP plot was prepared graphing the heavy atom sites as directed thermal, vibrational ellips aids specified by the [3 •. 1 s and given in Figure 2. 8. lJ Below and to the left of each ellipsoid is presented an equivalently orientated section of the heavy atom as it appeared in the cross -difference Fourier synthesis. Recalling that these syntheses were computed from protein phases in turn based upon isotropic heavy atoms, the general correspondence in shapes lends credence to the anisotropic refinement. The over-all billiard ball or egg-shape to the heavy atoms is 0 at variance with their ligand arrangements. At 2. 45 A the ligands - 56 TABLE 2. 8 OCCUPANCY REFINEMENT OF POTENTIAL MINOR SITES1 Derivative Site Initial Occupancy 2 (e ) Final Occupancy (e ) a 65. 13 64. 9 b 5 C 5 - 0. 3 2.8 a 82. 93 86. o · b 6.9 8.4 C 9. 0 3. 6 d 6.5 1. 3 e 5. 3 1.9 a 81. 73 82.4 b 50. 03 52.0 C 36. 43 36. 2 d 10.4 1. 7 e 7.5 2.2 £ 7. 0 - g 7. 7 h 13. 8 i 7. 9 - 4. 8 16. 9 1. 3 PtCl 2 4 uo 2 2 + Pt(CN) 1 2 3 2 4 - 1. 3 Test sites were entered with the same isotropic temperature £actor as the main site, and this was also refined. Computed from the ratio of Fourier peak heights. Main, established sites. - 57 - ~ ·Y-RXIS ::---:: - TCNl .~ PTCNLJ @ Figure 2. 8. Thermal ellipsoids of the heavy a tom sites. - 58 would be expected as at least protrubances on the spherical central atom. The failure to detect the ligands has been attributed to the following: ( 1) there may be disorder at the heavy atom binding site on the protein surface where, for example, rotation about the proteinheavy atom bond could obscure substituent atoms; (2) the oxygens and cyanide ions are comparable in electron density to the interstitial sulfate solution and in a difference map could be subtracted out along with the solvent; and (3) ligand substitution is quite possible for PtCI 2 ~ (Langford and Gray, 1965), so this derivative may also have 4 substituents at the same electron density as the surrounding solvent. Use of the averaged data sets for refinement at this stage was clearly a compromise. Wbile averaging the various derivative sets to give one master native set and a master set for each chemical modification gi v es statistically more accurate F PH's, it may be that there are slight but physically real differences in at least occupancies among the non-averaged derivative sets. 0 and low 4 A Pt(CN) 2 4 In the extreme there are the high O - 0 sets and the 4A v ersus the 2. 45 A uo 2 + set 2 where occupancies can change by a factor of two. Thus averaging may have obscured some useful differences between almost redundant data sets, differences that could be exploited to improve the native phases. If two derivati v es have the same sites of chemical substitution and differ only by a common ratio between all the occupancies of the identical sites, no new nati v e phase information is provided by using both derivati v es over that provided by either one. This is best seen by adopting a Blow-Crick (1959) formulation for the single isomorphous - 59 replacement (SIR) calculated native protein phase. Then the native phase satisfies the condition where P(cp) is the Blow-Crick phase probability distribution, cp the formal native phase, cpb the best calculatable phase, and e the lack of closure error. Letting F PH be the observed derivative scattering factor magnitude, F p the same for the native structure, f the magnitude contributed by the heavy atoms, et the phase of f, and D the calculated derivative scattering factor magnitude, the above condition implies IFPH - D(cp) IFpf sin(cpb -0!) = 0 D(cp) which for non-trivial cases says Thus a single derivative predicts a protein phase independent of f. Derivatives differing only by a common ratio of occupancies will have different values of f, but the same O!, and hence give the same cpb. Nevertheless, a joint Blow-Crick phase probability distribution function for the two combined is much sharper than for either single derivative and this sharpening influences the final "best" protein phase (see Proposition I). Setting aside the final parameters from the work with averaged data sets, the process was duplicated treating all data sets as independent derivatives - one 4 A and two 2. 45 A PtC1 4 2 - sets, two - 60 0 0 4A and one 2. 45 A Pt(CN) set. 2 4 - set, 0 0 and one 4 A and one 2. 45 A uo 2 2 + The course of refinement followed ground rules established 0 earlier, such as only constant isotropic B-values at 4A. Tables 2. 9 through 2. 11 record pertinent results from which it can immediately be seen that an improvement resulted in the figures of merit although co-ordinates and occupancies were consistent across the sets. A theory too complex to be recounted here (but see Proposition I) predicts increases in figure of merit in the eight derivative refinement over the three derivative refinement due to purely statistical effects and not an improvement in phases. As noted, the mean figure of merit has risen,and in an attempt to calibrate this rise and assess the pure statistical contribution, a pseudo-eight derivative phasing model was prepared. This consisted of the averaged data from the three derivative model copied a number of times to give the same redundancy for a given chemical derivative as in the true expanded eight derivative approach. The eight derivative model contained three PtC1 2 -, three Pt(CN) 2 -, and two uo 2 + sets, 4 2 4 so the pseudo-model contained (F)PtCI42 _ three times, (F)Pt(CN) three times and (F)UO 2 + twice. 24 This copying, equivalent to 2 exponential multiplicative factors in the Blow-Crick phase probability distribution of 3, 3, and 2 for the three derivative model, was done 0 for all reflections, but only the results to 4 A should be compared, as 0 the actual non-averaged derivative model contained mostly 4A remeasurements of the same chemical derivative. 0. 3367 0. 0179 84.4 8.7 0. 7388 0. 7034 39.6 20.4 89.4 55.2 67.4 z Effective 0. 92 1. 13 4. 20 3. 42 0.68 2. 52 0. 75 0.93 4.54 3. 35 1. 02 1. 08 ~2a 5. 20 2. 97 ~l l Anisotropic 1 6. 24 6. 68 7. 67 4. 60 2. 04 4.64 2.84 ~33 0.00 -0. 56 0. 00 0.00 0.85 -0.43 -0. 10 ~la 0. 00 0. 23 0. 00 0. 00 -0.95 -1. 20 0.69 ~23 0.00 0. 30 0.00 0.00 -2.45 2.26 1. 24 ~l 3 thermal parameters X 103 1 2 2 2 Atomic scattering factor= Ze -(~11h +~22k +~33£ +~12hk+~2skH~ 13 hl) 0. 3013 0.0979 0. 3951 0. 0997 0. 7651 0. 2833 0. 3072 uo 2 2 + 0.8628 0.4692 0.4684 0.7758 0.7928 0.4302 0.0979 0. 2095 z Pt(CN) 4 2- 0.0520 y 0. 0107 X PtCl4 2- -- Derivative Co - ordinates FINAL PARAMETERS THREE AVERAGED DERIVATIVES TABLE 2. 9 I ,_. 0-- . 0996 . 4680 . 3050 . 3862 . 0991 . 4681 . 3049 . 3879 . 0995 . 0169 . 0994 Pt(CN) 4 2 14A) Pt(CN) 4 2 -(4A) uo 2 2 + (2. SA) U0 2 2 + (4A) 0 0 . 3008 . 3012 . 3368 . 4309 . 4681 . 2835 . 0985 . 4304 . 4681 . 2829 . 0980 . 4307 . 4690 . 2834 . 0976 . 0522 . 0522 . 0524 1 See definition in Table 9. 0 . 0979 . 4690 . 3074 . 3999 Pt(CN) 4 2 -(2. SA) 0 . 0104 PtCl4 2 -(4. OA) 0 . 0107 PtC14 2 -(2. SA) 0 . 0110 PtC14 2 -(2. SA) 0 Derivative . 7405 . 7396 . 7014 . 7943 . 8637 . 7623 . 7908 . 7941 . 8635 . 7598 . 7828 . 7934 . 8628 . 7649 . 7727 . 2122 . 2104 . 2097 Co- ordinates X z y 31. 9 80. 3 7. 1 85.9 61. 5 45.4 25. 8 81. 4 52. 6 · 36. 0 16 . 7 86. 8 53. 9 39. 3 17.4 71. 4 71. 5 84. 2 9. O - - 7. 0 7.0 7. 0 12.8 6. 3 6. 3 6. 3 12. 8 - 3. 22 3.42 5, 87 5. 35 3. 71 4. 20 2. 76 16. 0 3. 02 1311 0.69 0. 92 0.97 1.01 0. 94 1. 13 1. 09 1. 05 1322 Anisotropic - Effective Isotropic z B FINAL HEAVY ATOM SITES EIGHT DERIVATIVE TABLE 2. 10 1 7.93 6. 24 5. 09 1. 82 1. 46 7. 67 2.52 2. 55 -0.80 0. 00 -0.04 0. 10 0. 66 0. 00 0. 04 0.24 0. 28 0. 00 -1. 35 -1. 17 0. 99 0. 00 1. 39 1. 34 5.50 0. 00 4. 65 -2, 67 I 1. 34 a0.00 N 2. 32 2.54 thermal parameters X 103 l33 3 131 a 131 3 132 3 - 63 - TABLE 2. 11 FINAL REFINEMENT STATISTICS Rl Derivative C R!(centric) RZ k E3 (m)4 (m)4 C EIGHT DERIVATIVE METHOD 0.92 0. 67 0. 87 0. 61 0 2 PtC14 l2. SA) 0 PtC14 2 -(z. SA) 0. 643 0. 153 0. 118 89. 6 0.595 0. 115 0.092 69. 8 0.497 0. 080 0.071 58. 9 0.587 0. 149 0. 127 93.8 0A) 0.421 o. 106 0.094 80.5 Pt(CN)4 2 -(4. OA) 0.420 0. 113 0. 100 86. 7 0. 542 0. 110 0. 094 65.4 0. 574 0. 049 0. 045 34. 7 2 PtCl4 -(4. OA) 0 Pt(CN) 4 2 -(2. SA) Pt(CN) 4 2 -(4. 0 0 U0 2 2 + (2. SA) 0 uo 2 + (4. oA) 2 THREE DERIVATIVE METHOD PtCl4 2 - 0. 606 0 . 114 0.096 69. 6 Pt(CN) 4 2 - 0.550 0. 140 0. 125 90. 2 uo 2 2 + 0.541 0. 110 0.094 64.8 1 R I:(lack of closure error) over centric data (Cullis , et. al., 1961). C - LI ,6F \ ZR_ _ I: (lack of closure error) _K I: FPH I 3 4 \ (Kraut, et. al., 19 6 2). Root mean square lack of closure (Blow and Crick, 195 9). Respectively, centric and all data figure of merit (B l ow and Crick, 1959) . - 64 A number of criteria are available to examine these models. b b b Let cp , cp , cpp, stand for the Blow-Crick "best" protein phase for 8 3 the three, eight, and pseudo-eight derivative models, then <jcp~ -cp~l > = 16. 9° <l cp3b - cppbl > = 12. 0° <lcp~ -cp~l > = 12. 9° and further, those reflections for which cpb differed by large amounts (~b> 20) correlated with very low figures of merit, and so would be eliminated from a Blow-Crick "best' 1 Fourier calculation. Increases in average figure of merit are revealed as purely statistical by Figure 2. 9. The most striking comparison of the models is given by Figures 2. 10 and 2. 11 which show the same protein electron density section calculated with the different phasing. coefficients The Blow-Crick mF p have in each case been divided by the over-all average figure of merit for that particular phasing model to compensate for the decrease in term magnitude as (m) drops, thus placing all sections on the same density scale. The virtual identity of the sections establishes the net insensitivity of the final map to manipulations in the choice of derivatives. Phase refinement thus beaten to death, the final protein density map was calculated using Blow-Crick coefficients mF p with phases from the a v eraged derivative refinement. Two separate n'laps were calculated with sectioning along the x and y crystallographic directions for use in a Richards optical comparator which gives two - 65 - - Radial Distribution of Figure of Merit 0.90 0.85 4- '- a, ~ 0.80 a, ~ 0.75 O" LL ·- 0.70 • 0.65 8 derivatives X pseudo 8 derivatives 0.04 0.06 0.08 0.10 0.12 2 sin 8 Figure 2. 9. 0.14 0.16 0.18 · - 66 - cy -a------=-- e~ ~ ~. 3 DERIVATIVE Figure 2. 1 O. 1. , ~ - ~ ,l i l1 . JI <?tLJ ~! t0 COEFFICIENTS : m F/.607, cp (best) Density section through the heme plane. - 67 - ~ · @> ~~ ~ 3 DERIVATIVE Figure 2. 10. 2. .· ~ COEFFICIENTS : F, cf, (max) As in 2. 10. 1 but with Blow-Crick ~(max) phasing. - 68 - @> ~ ~~Q ~ ~ - 11 8 11 DERIVATIVE ' ~ COE FF IC I ENTS : m FJ.675, cp (best) I Figure 2. 11. Comparison with Figures 2. 10. 1 and 2. 10. 2 shows there is no improvement in the map when the expanded data set is us ed. - 69 different views of the model and density simultaneously. Contouring was performed in an arbitrary manner, but with the goal of maximizing internal continuity without creating excessive cross-boundary connections. The results of the structure determination are discussed in depth in the fourth chapter. Table 2. 12 lists the identification of the heavy atom binding sites with respect to chemical functional groups on the protein. The binding of PtC1 2 - to the methionine has 4 been documented (Dickerson, et al., 1969). The binding of the uranyl ions suggests a chelation by oxygen atoms involved in unsaturated bonds. This same type of binding has been noted in chymotrypsin (Vandlen and Tulinsky, 1973), where a uranyl ion is poised between two acidic side chains of a glutamic and aspartic acid. chemistry of Pt(CN) 2 4 - The binding is obscure, with carboxylic oxygen, carbonyl oxygen, and amine groups involved. 2. 8. Refinement Weighting Functions The inclusion in this thesis of some few words on the weighting functions of least squares is justified by a colloquialism - you find what you seek. The experimenter can control to a remarkably fine degree the final set of refined parameters, the least squares solution, by choice of a weighting scheme applied to the observations, such a choice stemming from an astute knowledge of the reliability of the data or from a desire to obs cure poor measurements or a lack of the proper solution. The basic difficulty and the distinction from the normal crystallographic least squares problem lies in that the observation to - 70 TABLE 2. 12 HEAVY ATOM BINDING SITES Derivative Co-ordinates Binding Site PtCl 2 4 0. 011 0.052 0. 210 Adjacent to sulfur of Met 90 uo 2 + 0. 010 o. 301 0. 739 Between carbonyl oxygen of As n 1 side chain and carbonyl oxygen of main chain peptide bond 2 0. 018 0. 337 0. 703 Between carbonyl oxygens of main chain peptide bonds 2 and 3 0. 098 0.430 0. 792 Near carbonyl oxygen and hydroxyl OH of Thr 98 0.468 0.469 0.863 Near € -amine of Lys 89 0. 307 0.283 0. 765 Near side chain carboxyl oxygens of Glu 124 0. 395 0. 098 0.776 Near e-amine of Lys 10 2 Pt(CN) 2 4 - - 71 be fitted is actually a difference, I F(derivative) - F(native) I, the magnitude of which is uncorrelated with the magnitude of the parental terms. The lack of an obviously rational weighting scheme is reflected by the proliferation of schemes employed in the field. Dickers on, Weinzierl and Palmer (1968) have suggested assigning weights proportional to 1/(e:!kJ')r- where e:hkJ' is the lack of closure of reflection (h, k, £) for a particular derivative (Dickerson, Kendrew and Strandberg, 1961), and the root mean square average is computed over all j derivatives. Lips comb, et al. , ( 1966) have used weight = ·n. ,_JjJ [ J~ + 9(E./ l FPH.J l /] J € h.tU'.. where ~k.Q is the number of derivatives on which (h, k, £) was measured; E. is mean lack of closure; and FPH. is the observed J J structure factor of heavy atom derivative j. Dreuth, Jansonius, and Walthers (1967) proposed the use of a weight equal to the square ofthefigureofmerit. Adams, etal., (1969)used weigh~kl! . = q./(E~ e + E~) ,J J J, where q. is the inverse of the scale factor required to place derivative J is the root mean square j on the same scale as the native data; E~ J, e lack of closure for derivative j computed over a narrow range of sin 9/},. appropriate to (h, k, £); and E~ is a resolution independent root mean square lack of closure averaged over all derivatives. Cohen, et al. (1969) and Ludwig, et al. (1971), among others, have avoided the is sue by using unit weights. - 72 In the early stages of the ~ 5 50 phase refinement, three possible weighting schemes were employed: 1) the lack of closure scheme of Dickerson, et al. mentioned above; 2) a counting statistics based scheme where and cr(F p), cr(F PH) are the standard deviations as defined earlier of the native and the derivative being refined; and 3) unit weights. A test was devised wherein each of these schemes was applied for four cycles of refinement of the major site of the PtC1 2 - derivative. 4 For computational ease, only the centric .6F(PtC1 2 -) data was refined. 4 Figure 2. 12 demonstrates the course of refinement for the co-ordinates and occupancy from which some generalizations may be made. For most of the parameters (x, y, z) aV schemes are tending toward a common solution, while for some parameters (Z) the results may differ considerably. The lack of closure scheme is drastically slower to converge, which is understandable, as after every cycle the the PtC1 2 4 - e: for data changes, and hence to some extent so does the weight. This latter point is a serious problem in the early stag es of refinement when many tests must be performed. Little distinction in centric refinement could be seen between results from unit weights and counting statistics weights, yet, since the counting statistics do reflect the accuracy of scheme was taken as most valid. I.6F I , the latter In three-dimensional refinement, however, this scheme led to poorer refinement than unit weights, as judged by the various reliability indices . This may be due to the - 73 - ~ 0.012 ~-~-~ / X o.oos 0.004 CJ) a, .,_ 0.000 • Weights based on R. M.S. Epsilon 0.0532 X Weights based .on counting statistics 0 C ·"'O '"- A Unit weights 0.0528 0 I 0 l) Y 0.0524 u E 0 ,._ 0.0520 <r 0.0516 0.211 z 0.209 0.207 X I X X -A /A_Aac-- •· • Initial I 2 A • • 3 4 Cycles of Refinement Figure 2. 12. Weighting scheme comparison. - 74 - 78 x--x--x a • Weights based on R.M.S. Epsilon ~ u X Weights based on C 0 counting statistics §- 74 u u 0 A Unit weights Q) .....> ~ 70 '4'4- w A- 66 Initial I A A 3 4 / 2 Cycles of Refinement Figure 2. 12. (continued) - 75 fact that counting statistics results in a bias toward large .6F terms. For centric reflections, large .6F terms must be matched with large atomic f (calculated) values which are correspondingly sensitive to parameters, while in the general case f may be large, yet .6F = 0 (f perpendicular in phase to F p). Hence a few large .6F terms dominate refinement and the majority of small .6F terms are not well matched. Consideration of these points led to the following use of weighting schemes. Centric refinement was performed exclusively with counting statistics weights, and three-dimensional refinement with exclusively unit weights until a final convergence was reached. At that time, additional cycles the lack of closure scheme. of refinement were performed under In this manner the rapid convergence characteristics of unit weights could be merged with the inherent data evaluation of the e-scheme. Interestingly, when the jump between functions was so performed, the e - refinement converged very quickly (two cycles) with minimal readjustment of parameters. coordinates 2. 9. shifted by less than 10 Crystalline _£ 550 -5 Fractional and occupancies by about 3%. as a Function of pH Crystal structure experiments were undertaken to study the effect of pH on the conformation of the protein. question was asked - does _£ the native form at pH 7. 5? 550 In particular the differ in structure at pH 5 and 9 from The motivation for the experiments sterns from recurrent observations on eukaryotic cytochrome suggesting some change at both high and low pH, (Margoliash and Schejter, 1966). - 76 - The high pH form of cytochrome £_ involves displacement of methionine 80 and is extensively invoked to explain a host of kinetic and structural data. Even though this high pH form is non-physiological, it has established that the integrity of the pH 7. 5 form, which probably means a stable methionine-iron bond, is absolutely necessary for the in vivo redox mechanism. For example, it has been proposed that di-carboxymethylated methionine 80, methionine 65 cytochrome£_, inactive toward cytochrome oxidase, mimics the high pH form in losing sulfur co-ordination (Schejter and Aviram, 1970). For the present experiments, £_ 550 crystals at pH 7. 5 still in the crystallization capillaries were transferred to soaking solutions at pH 5. I (mother liquor adjusted with acetic acid) and pH 8. 9 (mother liquor adjusted with concentrated aqueous ammonia),and as slow diffusion took place, the crystals equilibrated to the new pH. Centric zone data sets were collected and compared with the native data (see Tables 2. 4 and 2. 5). The identity of the pH 5. I and 7. 5 data sets (R-factor = 0. 4) immediately established that there is no structural change. Perhaps the result is not too surprising in that in the protein there is only one functional ·group with a pK a in this range, the fifth ligand histidine, and its protonatable nitrogen is already involved in a stable hydrogen bond. The R-factor of 0. 05 between native and pH 8. 9 suggested a potential slight change. Difference Fourier syntheses, using observed .6.F's and native protein phases failed to reveal any specific feature of structural change, certainly nothing drastic in the Met 80 region. However, the pH of 8. 9 is a sharp experimental limit, for attempts - 77 - to observe data at higher values saw complete disintegration of the crystals. · The strongest statement that can be made by combining these two points is the claim that a sharp pH transition exists belowwhich the native form predominates, above which some modified form exists, the modification being extensive enough to force the cytochrome molecules apart from their packing array in the crystal. of a lysine side chain, with a pK a The protonation above 9, is certainly consistent with this view. With hindsight, it can be seen that the above experiments really offer little conclusive evidence on the nature of the supposed structural change and border on the verge of worthlessness. The main problem is that the crystal lattice may be locking the cytochrome into its pH 7. 5 conformation by crystal packing forces and thus preventing a change which would spontaneously occur in solution. Alternatively, the lattice might be is elating or sealing off from the external pH some group or groups which mediate the conformation change. 2. 10. Ammonium Selenate Binding to _£ 550 Tulinsky and Wright (1973) have recently published results which may provoke profound repercussions in the field of protein crystallography. These authors made use of the remarkable similarity in chemical properties (size and pK ) between sulfate and selenate a anions to study salt binding to chymotrypsin. By substituting selenate for sulfate in the mother liquor, measuring intensity changes, and calculating D.F maps, they were able to identify ordered binding sites for sulfate/selenate on the surface of the protein. The usefulness of - 78 - such information lies in interpreting the native protein map and being able to recognize density features as bound solvent rather than forcing something like a lysine side chain into the region. An additional potentiality lies in using the selenate-substituted crystals as derivatives for the phasing. This is a fascinating possibility in light of the fact that a large number of protein crystals are grown from ammonium sulfate solutions. The substitution reaction was attempted on native ~ already grown from sulfate solution. 550 crystals Commercial ammonium selenate is an extremely dirty preparation and must be repeatedly recrystallized from hot water-ethanol before use. By soaking experiments similar to those in the pH work, various solutions of differing selenatesulfate concentrations were explored; the tightly packed protein crystals were highly inert to exchange and, eventually, solutions were used containing only selenate (95% saturated) and no sulfate. Interestingly, no gross crystal damage occured, although the 0 mosaic spreads broadened considerably. A three-dimensional, 2. 45 A data set was collected (see Tables 2. 4 and 2. 5) and a full 6F Fourier map calculated. The resultant synthesis was very clean showing only a few peaks above the general noise level. This map was studied with the knowledge of the location of the molecular surface so that potential sites of selenate binding could be realistically deciphered from the general noise. Six sites were located and correlated with the molecular surface (see Table 2. 13). - 79 - TABLE 2. 13 AMMONIUM SELENATE BINDING SITES Co-ordinates Molecular Lo cation 0. 302 0. 363 0.890 Near carbonyl oxygen of Glu 135 0.254 0. 072 0.072 Interstitial, greater than 4 A from any protein group 0.048 0.041 0. 345 Near the e - amine group of Lys 97 0. 309 0. 093 0. 633 Interstitial, greater than 4 A from any protein group 0. 195 0. 184 0.829 Near the hydroxyl OH of Thr 27 0. 273 0. 089 0. 371 Near the acidic side chain oxygens of Glu 64 0 0 The disappointingly weak appearance of the substitution sites may be attributed to several factors. ( 1) Close protein packing may lead to only a partial occupancy of the binding sites. (2) The binding sites may be disordered or possess unusually large temperature factors. (3) The change, sulfate to selenate, of 18 electrons may be inherently too small to be seen by this method. The small number of sites found was not surprising in light of point (1) above and a realization that most solvent sites must be so labile as to be essentially unordered and therefore non-existent in the x-ray analysis. It was anticipated that replacing the non-ordered interstitial sulfate with selenates of higher electron density would lead to greater 6F changes in the low order terms, those most sensitive to solvent effects. - 80 - However, a plot of R-factor calculated over ranges of sin 8 /).._ versus sin 8 /\ demonstrated a line of zero slope. Possibly, lack of isomor- phism at high resolution balances the loose solvent effect at low resolution to make .6F changes so uniform. The two strongest sites were refined over the centric differences as a single derivative to respectively 18 and 6 electrons. However, the Cullis R-factor (0. 75) and the figure of merit (0. 36) clearly indicated that predicted protein phases were inaccurate. The final nail for the coffin was provided by a cross phase .6F map using uo .6F' s and phases predicted by the selenate data. the uo 2 2 - 2 2 This failed to show site, but did give a number of other nonsense sites. In summary, the ~ 550 selenate derivative, while indicating potential bound solvent, was not suitable for phasing. This latter failure may be attributable to the very low occupancies. This author feels that the potential for phasing via selenate still exists for other proteins in other circumstances. For instance, in some other case there may be a larger number of high occupancy sites, or, if one other strong, simple derivative is available, the selenate derivative may allow solution of the phase ambiguity by indicating which of the alternatives offered by the strong derivative is correct. 2. 11. References Arndt, U. W., and Willis, B. T. M., Single Crystal Diffractometry, University Press, Cambridge, 1966. Adams, M. J., Haas, D. J., Jeffrey, B. A., McPherson, A., Mermall, H. L., Rossman, M. G., Schevitz, R. W., and Wonacrott, A. J., (1969). J. Mol. Biol., !!_, 159. + - 81 Andrews, P. (1964). Biochem. J., ..21., 222. Blake, C. C. F., Mair, G. A., North, A. T. C., Phillips, D. C., and Sarma, V. R., (1965). Proc. Roy. Soc., B., 167, 365. Blow, D. M., (1958). Proc. Roy. Soc., 247, 302. Blow, D. M. and Crick, F. H. C., (1959). Acta Cryst., ..!._; 794. Cohen, G. H., Silverton, E.W., Matthews, B. W., Braxton, H., and Davies, D.R., (1969). J. Mol. Biol., 44, 129. Cullis, A. F., Muirhead, H., Perutz, M. G., Rossman, M. G., and North, A. C. T., (1961). Proc. Roy. Soc., London, A265, 15. Dickerson, R. E., Kendrew, J.C., and Strandberg, B. E., (1961). Acta Cryst., .!_±, 1188. Dickerson, R. E., Kopka, M. L., Bordner, C. L., Varnum, J.C., Weinzierl, J.E., and Margoliash, E., (1967a), J. Mol. Biol., ~, 77. Dickerson, R. E., Kopka, M. L., Varnw:n, J.C., and Weinzierl, J.E., (1967b). Acta Cryst., 23, 511. Dickerson, R. E., Weinzierl, J.E., and Palmer, R. A., (1968). Acta Cryst., B24, 997. · Dickerson, R. E., Eisenberg, D., Varnw:n, J., and Kopka, M. L., ( 1969). J. Mol. Biol., 45, 77. Dreuth, J., Jansonius, J. N., and Walthers, B. G., (1967). J. Mol. Biol., 24, 449. Greer, J. (1971). J. Mol. Biol., 59, 107. James, R. W. (1965). The Optical Principles of the Diffraction of x-rays, Cornell University Press, Ithaca, New York, 1965, 43-46. Kraut, J., Sieker, L.C., High, D.F., Freer, S.T., (1962). Nat. Acad. Sci., 48, 1417. Proc. Langford, C.H., and Gray, H.B., (1965). LigandSubstitution Processes, W. A. Benjamin, Inc., New York. Levy, H. A., and Ellison, R. D., (1960). Acta Cryst., .!1_, 270. Lipscomb, W. N., Coppola, J.C., Hartsuck, J. A., Ludwig, M. L., Muirhead, H., Searl, J., and Steitz, T.A., (1966). J. Mol. Biol., ~, 42 3. - 82 Lipson, H. and Cochran, W., The Determination of Crystal Structures, Cornell University Press, Ithaca, New York, 1966. Ludwig, M. L., Andersen, R. D., Apgar, P. M., Burnett, R. M., LeQuesne, M. E., and Mayhew, S. G., ( 1971 ). In Cold Spring Harbor Symposium on Quantitative Biology, XXXVI, Cold Spring Harbor Laboratory, 1972. Margoliash, E., and Schejter, A., (1966). In Advances in Protein Chemistry, 1_!, (C. B. Anfinsen, M. L. Anson, J. T. Edsall, and F. M. Richards, Eds.), Academic Press, New York, 114. Margoliash, E., and Walas ek, 0. F., ( 1967). In Methods in Enzymology, X, Es ta brook, R. W. , and Pullman, M. E. ; (eds. ), Academic Press, New York, 339-348. Matthews, B. W., (1966a). Acta Cryst., 20, 230. Matthews, B. W., (1966b). Acta Cryst., 20, 82. North, A. T. C., ( 1965 ). Acta Cryst., ~, 212. North, A. T. C., Phillips, D. C., and Matthews, F. S., (1968). Acta C rys t. , A24, 351. Phillips, D. C., (1966). Methods, _£, 75. Advances in Structural Research; Schejter, A, and Aviram, I., (1970). Diffraction J. Biol. Chem., 245, 1552. Scholes, P.B., McLain, G., andSmith, L., (1971). _!_Q_, 2072. Biochemistry, Stout, J. H., and Jensen, L. H., (1968). X-ray Structure Determination, A Practical Guide, Macmillan Co., New York, 79-82. Tulinsky, A., Mani, N. V., Morimoto, C. N., and Vandlen, R. L., (1973). Acta Cryst., B29, 1309. Tulinsky, A., and Wright, L.H., (1973). J. Mol. Biol., g, 47. Vandlen, R. L., and Tulinsky, A., (1973). Biochemistry, .!..£, 4193. Wilson, A. J . C., (1942). Nature, 150, 151. Wyckoff, H. W., Tsernoglou, D., Hanson, A. W., Knox, J. R., Lee, B., and Richards, F. M., (1970). J. Biol. Chem., 245, 305. Zeppezauer, M., Eklund, H., and Zeppezauer, E. S., (19 68). Arch. Bio chem. Biophys. , 12 6, 5 64. - 83 CHAPTER 3. 3. 1. PRIMARY SEQUENCE OF s 550 Introduction Even a cursory glance at such compilations of protein sequence data as the Dayhoff Atlas (Dayhoff, 1972) is sufficient to reveal that £-type cytochromes, especially from the higher organisms, are the proteins most extensively investigated in primary structur~. It must be admitted that this is due in large part to their ubiquity and the relative ease, as biochemical substances go, with which they are isolated. At a somewhat higher plane, there are the contributions made by such exhaustive data toward understanding and reconstructing evolution at the precise molecular level, through sequence comparisons (Margoliash, Fitch and Dickerson, 1968). Among the closely related eukaryotes, the various cytochromes may be taken as_ a set of natural protein derivatives, showing the allowable modifications of side chain residues which still produce fully functional molecules. On the prokaryotic side of life, the available sequences reveal the astounding diversity of cytochromes, (for example, compare the four heme/protein cytochrome .£ 3 of Desulfovibrio with horse heart cytochrome ~) and also the extent to which homology may be strained (for example, the comparison of the 82 amino acid Pseudomonas £ 551 with horse cyto- chrome~ (Needleman and Blair, 1969; Dickerson, 1971)). Since, of the cytochromes sequenced, the vast majority have been from eukaryotes, the primary order of a prokaryotic cytochrome such as M. denitrificans would be of obvious value in extending this comparative biochemistry. Since the denitrificans respiratory chain shows - 84 - cross-reactivity with that from eukaryotes at the level of cytochrome _s (Smith, Newton and Scholes, 1966) the information becomes of interest. special Potentially, a faltering step may be made toward unraveling a molecular phylogeny for microbes. The sequence determination was accordingly undertaken for the _£ 550 protein. Of immediate concern was that this was necessary 0 to compliment in a reliable manner the already existent 2. 45 A electron density map for the molecule available through an x-ray crystallographic determination. From this map, many of the side chain residues could be identified,although it was impossible to identify all with total confidence. A similar situation in other protein structure determina- tions has been encountered, in which a crystallographic electron density map has been used to deduce sequence information (Kendrewet al.,1961; Herriott et al., 1973). For ': h e _£ 550 map,as in these other pure crys- tallographic approaches, a complete, reliable sequence could not be deduced, which generally would necessitate an independently determined sequence to compliment the x-ray data. In this chapter, the preparation, isolation, and sequence characterization of the tryptic peptides from _£ Subsequent 550 M. denitrificans is described chapters will deal with the combination of these peptide fragments into a full, ordered primary structure, this ordering being deduced from the identification of the unique peptides in the electron density map. This approach is novel in that it combines both chemical and crystal data. The actual sequence of residues determined as described in this paper is merged with the density map serving to align - 85 - the peptides, a function normally performed by overlap of peptides from a second and possibly a third limited hydrolysis of the protein or from sequence homology. 3. 2. Materials and Methods Purified ~ 550 was the generous gift of Drs. G. McLain and L. Smith, prepared according to their procedure (Scholes, McLain and Smith, 1971). Since the isolation of the cytochrome requires -large culture volumes not readily accessible on the laboratory scale, experimentation was designed to complete the tryptic peptide work on the available stock of 10 µM. Three times crystallized preparations of chymotrypsin and £-1-tosylamide-2-phenylethyl chloromethyl ketone {TPCK) - treated trypsin were obtained from Worthington Biochemical Corporation, as was a frozen solution of diisopropylphosphorofluoridate - treated carboxypeptidase B. Carboxypeptidase A (crystal suspension), leucine aminopeptidase (ammonium sulfate suspension), pepsin ( crystalline), and thermolysin (lyophylyzed powder) were the highest grades available from the Sigma Chemical Company. Carboxypeptidase C (powder) from orange leaves was purchased from the Henley Company of New York. Reagents for the Edman and dansyl chloride reactions were purchased from Pierce Chemical as 11 Sequenal 11 grade. Dowex resin was obtained from Bio-Rad Laboratories, gel filtration media from Pharmacia (Sephadex) and Bio-Rad Laboratories (Bio-gel), and cellulose and silica gel TLC plates from Eastman. Pyridine was - 86 re-distilled before use and anhydrous hydrazine prepared by azeotropic vacuum co-distillation with toluene. All other chemicals were of the highest reagent grade. Tryptic digestion of 10 µM of desalted, oxidized native protein, 0. 67 mM in 0. 05 M ammonium bicarbonate, was affected by I. 2 mole % ...,..._ ""-J trypsin for 6. 5 hr. at 25 °C. The course of the digestion was monitored to assumed completion by cellulose TLC chromatography of 1 A aliquots of the mixture - the release of a high Rf red heme peptide and the disappearance of a zero Rf red protein indicating digestion. Chymo- tryptic digests were performed in 0. 1 ~ ammonium bicarbonate 0. 5 mMin peptide at mole percentages which will be discussed where ...,.._,..,_. pertinent. Thermolytic digestion followed the procedure of Matsubara and Sasaki (1968) at a mole% of 0. 2 for 2 hours. All the above reactions were quenched by rapid freezing and lyophilization. Kinetic studies with leucine aminopeptidase (Delange and Smith, 1971 ), carboxypeptidases A and B (Ambler, 1967) and carboxypeptidase C (Tschesche and Kupfer, 1972) were patterned after established procedures. Partial acid hydrolysis in 0. 25 °M- acetic acid was performed at 105 ° .m_ ~ in sealed ampuoles at a peptide concentration of 2. 6 ~ for 2 hours. The conditions for the initial separation of tryptic peptides on Dowex 50 • XZ have been documented elsewhere (Matsubara and Smith, 1963). Subsequent use of Dowex 50 resin employed the "Macroporous AG-MP50" form in a short column (0. 9 X 50 cm) at 40 ° C developed with (1) 60 ml of pyridine acetate (0. 05 Min pyridine, pH 3. 1), (2) 100 ml of pyridine acetate (0. 2 Min pyridine, pH 3. 1), (3) a linear - 87 gradient formed by 150 ml of 0. 2 M pyridine acetate plus 150 ml of the same buffer (2 1)1 in pyridine, pH 5. 1), (4) 100 ml of the latter buffer, 100 ml of pyridine acetate buffer (4 ~ in pyridine, pH 5. 6). and (5) Fractions of 1. 5 ml were collected under a pumped flow rate of 15 ml/hr. Gel filtration columns (0. 9 - 2. 0 cm X 100 cm) were eluted with 0. I M ammonium bicarbonate allowing the detection of peptide containing fractions from the absorbance at 220 nm. Preparative paper chromato- graphy and electrophoresis on Whatman 3MM paper have been described (Chan and Margoliash, 1966 ). Amino acid compositions of peptides were determined from samples run on a Durrum Model D500 or Beckman Model 121B analyzer after in vacuo 24 hour hydrolysis with 6 N HCL Peptide fingerprinting was performed on cellulose TLC plates with the electrophoresis (first dimension) and chromatography (second dimension) using the same buffers as for preparative work. Appropriate staining reagents, in addition to ninhydrin, as given by Easley ( 1965) were applied to visualize peptides containing tyrosine, histidine, cysteine, methionine, typtophan or a blocked N-terminus. Amino-terminal residues were identified by the dansyl method of Gray and Hartley (Hartley, 1970). The combination Edman- dansylation sequencing procedure (Gray, 1972) was reserved for sequential degradation of peptides available in small amounts, as will be noted. The majority of sequencing operations were performed by a manual Edman reaction with direct product identification (Bitar, et. al. , 1972). One peptide, derived from the carboxyl terminus of the protein (29 residues long) was sequenced automatically on a Beckman Sequenator Model 890B employing a peptide program. Direct phenylthiohydantoin - 88 amino acids present in sufficient yield (> 50 nM) were identified via published chromatographic methods (Jepps son and Sjoquist, 1967), and those in low yield on the Durrum analyzer after HI hydrolysis (Smithies et al. , 1971 ). For the more hydrophobic peptides the terminal-COOR derivatization of Foster et al. ( 1973) or the lysine deri vatization of Braunitzer et al. ( 1970) were invaluable in attaching a hydrophilic group onto the peptide thus anchoring it into the aqueous phase during organic solvent extraction. No side reactions were evident and the free acid side chains of aspartate and glutamate were still detectable. For example, the hydrophobic peptide T6 (see Results) could be sequenced only four steps before total loss of the remaining material whereas the same starting amount after treatment according to Foster et al. gave high yields on nine of the twelve residues. As will be noted later, certain atypical standards were required for identification of products encountered. The amino acid 3, 5- dichlorotyrosine was synthesized according to Bouchilloux (1955) and characterized via chromatographic properties. The synthesis of 1-acetyl, 2-dansyl hydrazide followed Schmer and Kreil (1969) and the product was confirmed by elemental analysis (Calcd.: 54. 50% C, 5. 40% H, 13. 83% N. Found: 54. 30% C, 5. 44% H, 13. 27% N.) For complete characterization of the heme containing peptide, the heme was removed by the method of Ambler ( 1963) and the cysteine • residues carboxymethylated as described by Crestfield, Moore, and Stein (1 9 6 3). - 89 - 3. 3. Results Figure 3. 1 outlines the purification of the tryptic peptides and other peptides derived from these by the indicated means, emphasizing, of course, that these were the procedures that ultimately worked, a subset of those attempted. 3. 3. 1. Nomenclature For the clarity of reporting results, a consistent nomenclature is adopted for the peptide fractions. Major peptide-containing fractions from the Dowex column on the whole tryptic digest are numbered Tl through T 14 in the order of elution. Peptides further purified from these with no further digestion are labeled according to the manner of purification, uch" designating preparative paper chromatography, •el" paper electrophoresis, and "S"Sephadex gel filtration. Peptides derived from tryptic peptides are indicated by the method of subsequent digestion; "C" for chymotrypsin, ''.Ac" for acetic acid partial hydrolysis, and "Th11 for thermolysin. The final symbol refers to a fraction number of the preceding purification column (eg. C3 - third main fraction off the column used to separate the chymotryptic peptides) or the relative Rf in chromatography (eg. ch2 has second highest Rf of the mixture), or the net charge in electrophoresis (+, -, or N= neutral). Thus Tllch1C2 is a peptide from a chymotryptic digest of another peptide purified by paper chromatography after Dowex fractionation. Numbering is not necessarily complete as there were some fractions isolated in insufficient yield for further work or as intermediates between ill-resolvedmajor components. C3 ch2 I I I I r1. Poper Chromatography r1. chi ch2 G-25 I ch3 G-50 r.C2 Sephodex I Chymotryptic Digest I chi C3 I P-2 Bio-gel ch2 I Poper Chromatography or Sephodex I L, chi ch2 ch2 ch3 e-l+ I I Poper Chromatography Poper Chromatography Poper Electrophoresis T chi I I I Thi Th2 I I SI G-50 Sephodex G-25 r-1--, 50 I Dowex Thermolytic Digest Sephodex I Corboxymethylotion Heme Removal I (heme) TS T9 TIO TII Tl2 Tl3 Tl4 I Poper Chromatography I I T7 Purification of tryptic peptides. I I T6 el- e~+ Ac! AclI ~ 50 Dowex I I Acetic Acid Hydrolysis I Poper Electrophoresis I I I I T2 T3 T4 T5 Figure 3. 1. Cl rL, G-25 Sephodex I Chymotryptic Digest I Tl I 50 Dowex I Tryptic Digest I C550 0 '° - 91 3. 3. 2. Analytical Results Figure 3. 2 presents a simulated peptide fingerprint map of the tryptic products. Table 3.1 gives the amino acid composition of relevant peptides. The purity of peptide products was judged by the stoichiome- try of the residues in the compositions, by peptide fingerprints, and by dansyl end-group analyses. The sum composition of the residues in the unique peptides agrees well with the reported amino acid composition (see Table 3.2)and establishes 135 residues per mole of __s 550 . The sequence determinantion of residues within the tryptic peptides is recounted below and the tryptic peptides are listed in Table 3. 3. 3. 3. 3. Peptide Tl As this 29 residue peptide is the only major fragment not containing a trypsin specific Lys or Arg, it may be as signed to the carboxyl-terminal portion of the protein. Automatic Edman degrada- tion establishes the sequence Asn-Glu-Ala-Asp-Val-Val-Ala-PheLeu-Ala-Gln-Asx-Asx-Pro-Asx-Ala-Gly-Glx-Gly, leaving ten unsequenced residues. Chymotryptic digestion at 0. 6 mole% for 22 hours produced only two fragments through specific cleavage of the Phe-Leu bond. Sequence analysis of T1C3 gave Asn-Glu-Ala-Asp- Val-Val-(Ala, Phe). Further sequence analysis of the fragment TlCl would not yield usa.ble results beyond the tenth residue although a variety of Edman procedures were attempted. Very severe over-lap was encountered after each Asn or Asp residue with the dansyl-Edman technique. Gray (1972) has noted this as a diagnostic indicating an a-f3 peptide shift at Asx residues (see also Bornstein and Balian, 1970). - 92 - Simulated Fingerprint Csso Tryptic Digest T7,8 el+ Solvent Front Tl I Ch1 \ Tl4 T7Ch2--E) 0 T4 C) T2"'0 T3--0 j) T1 ◄ rJ C) ~ T11Ch2 u 0-Tl 3 T1C3-£) 2nm Sample 400 Volts 40 Minutes Tl2 TIICh3 0_ TIO Ch 2 ~T9Ch 1 ;_0Ch3 0 (3 cf5el+ 0 T8Ch 2 ~T9Ch 2 jchrom~ography EI ectrophoresis (+) (-) Figure 3. 2. · X Origin Simulated fingerprint. Spot shapes are only approximate. The true line of electroneutrality is displaced 1. 7 cm toward the cathode by endosmosis and runs through T8ch2, Tl0ch2. - 93 TABLE 3. 1 Amino Acid Composition of Isolated Peptides 2 Tl TlCl TlC3 T2 6. 13 (6) 4. 04 (4) 1. 96 (2) 4. 06 (4) Ser 1. 83 (2) 1. 76 (2) Glu 5. 22 (5) 4. 24 (4) Pro 1. 00 (1) ND (1/ Gly 4. 16 (4) 4. 11 (4) Ala 7. 22 (7) 5. 07 (5) Asp Thr 1. 68 (2) 1. 11 (1) 3. 94 (4) ND 4. 00 (4) 2. 05 (2) 4. 99 (5) Cys Val 1. 57 (2) 1. 84 (2) Met Ile Leu 1. 02 (1) 0.99(1) 0.69(1) Tyr Phe 1. 04 (1) 1. 04 (1) His Lys Arg Trp Totals 29 21 8 21 Net charge (-) (- ) (-) (-) 1 2 ND, not determined. Subsequent Edman degradation revealed one Pro. rntegral moles per mole peptide in parenthesis. Net charge was determined from electrophor etic mobility as discussed in the Methods section. - 94 - TABLE 3. 1 (continued) Asp 3 T3 T3AcI 1. 91 (2) 1. 18 (1) T3AcII T4 0. 94 (1) 2. 13 (2) 1. 01 (1) Thr Ser Glu 0. 98 (1) 1. 00 (1) Gly 0. 96 (1) 0.54 0. 91 (1) Ala 2. 22 (2) 0.23 2. 15 (2) Pro Cys Val Met 0. 56 (1) Ile Leu Tyr Phe His Lys 0.92 (1) 0.90(1) 0.86(1) 7 5 5 (0) (-) Arg Trp Total Net charge 3 {-) Known mixture of peptides. mole per sample ana ly zed. (-) Actual moles of Glu defined as 1. 00 per See text. - 95 - TABLE 3. 1 (continued) T5el T6 T7, 8chlel Asp 1. 95 (2) 1. 06 (1) Thr 1. 06 (1) 1. 04 ( 1) + Ser T7ch2 0.84 (1) 2. 13 (2) 4. 02 (4) Glu Pro Gly 2. 00 (2) Ala 1. 04 (1) 0. 97 (1) 4 2. 95 (3) 1. 04 (I) 0. 96 (1) Cys 0.97(1) Val Met 1. 65 5 (2) 0. 67 (1) 1.02(1) Ile 0.98(1) Leu 1.06(1) 0. 97 (1) Tyr 1. 02 (1) 1.00(1) 1. 08 (1) Phe His Lys 0. 99 (1) 0. 93 (1) 0.92(1) 1. 24 (1) Arg Trp Total Net charge 4 5 5 12 11 8 (-) (-) (+) (-) From a 72 hour hydrolysis. Shorter times erroneously indicate 4 gly /peptide. Other values agree between the 24 and 72 hour hydrozylates, except for Tyr which is degraded to 0. 4 tyr /peptide. Edman degradation revealed a Val- Val bond which is resistant to acid hydrolysis. - 96 TABLE 3. 1 (continued) T8ch2 T9chl T9ch2 T10ch2 Asp 0. 92 (1) Thr 0. 12 Ser Glu 1. 00 (1) 0. 94 ( 1) Pro Gly 1. 01 (1) Ala 1.03 (1) 2. 08 (2) 0. 96 (1) Cys Val Met Ile Leu Tyr 1. 01 (1) Phe His Lys 0. 99 (1) 1. 01 (1) 0.92(1) I. 13 (1) 3 3 3 4 (O) (+) (+) (0) Arg Trp Total Net charge - 97 TABLE 3. 1 ( continued) Tl0ch3 Asp Thr 1. 06 (1) 7 Tllchl TllchlC3 Tllch2 3. 88 (4) 0.86(1) 0. 21 3. 04 ( 3) 0. 76 (1) 1. 13 (1) Ser Glu 2. 23 (2) Pro 3. 00 (3) 0. 37 . 1. 57 (2) 0. 82 (1) Gly Ala 1. 09 (1) Cys Val 1. 98 (2) 2. 10 (2) Met 0.86(1) Ile 1. 01 (1) Leu 2. 91 (3) Tyr 1. 03 (1) 1.17(1) Phe His Lys 0. 94 (1) 2. 04 (2) 2. 08 (2) 1. 19 (1) Arg 1. 36 Trp (1) 2 23 9 4 (+) (-) (+) (+) Total Net charge 6 Determined by base hydrolysis according to Noltmann et al. (1962) and placed on the scale of the acid hydrozylates by equating the leucine detected in each method. 7 Isolated in low yield. Probable met-sulfone form of peptide T12. - 98 TABLE 3. 1 ( continued) Tllch3 T12 Tl3 Tl4 (heme) Asp o. 11 0. 14 Thr 1.02(1) 0. 10 1. 92 (2) 0. 95 9 (1) Ser 0. 72 (1) o. 18 Glu 2. 16 (2) Pro Gly 0.97(1) 1.05 (1) 1. 03 (1) Ala 0.09 1.17(1) 0. 99 (1) 0.08 0. 98 (1) 1. 71 (2) Cys 0. 43 (2/ Val Met 8 0 0.21 0. 89 (1) 0. 73 9 (1) 0.81 ( 1) Ile 0. 07 0. 93 (1) Leu 0.08 0. 14 1. 11 (1) 0. 88 (1) 2. 07 (2) Tyr Phe 0. 73 9 (1) His Lys 1. 07 (1) 1.06(1) 1. 88 (2) 3 4 9 16 (+) (+) (0) N. D. 11 1.45(2) Arg Trp Total Net charge 8 96 hour hydrolysis. Subsequent work indicated 2 moles Lys /mole peptide, but only between 1. 4 and 1. 6 moles Lys could be liberated from the intact heme peptide by any given acid hydrolysis. 9 Extrapolated to zero time. lONot liberated quantitatively due to covalent attachment to the heme. 11 Heme dominates electrophoretic behavior and precludes a clear charge determination. - 99 TABLE 3. 1 (continued) Tl4S1 Tl4Thl T14Th2 Asp 2. 13 (2) 1. 98 (2) 1. 90 (2) Thr 1. 05 (1) 0.95 (1) 0.94(1) Glu 1. 13 (1) 0.97 (1) 1. 03 (1) Pro ND (1) 1. 11 (1) 1. 03 (1) Gly 1. 07 (1) 1. 02 (1) 0.98(1) Ala 1. 98 (2) 12 1. 72 (2) 0. 93 (1) 1. 00 (1) 1.05 (1) 2. 12 (2) Ser Cys Val Met 0. 56 (1) Ile 2. 00 (2) Leu Tyr Phe His 0.86(1) Lys 1. 78 (2) 0. 68 (1) Arg Trp Total Net charge 12 13 16 ( _) 13 8 10 (-) (-) Detected as carboxymethylcyteine after heme removal and treatment with iodoacetate. carboxymethyl groups add two negative charges. - 100 - TABLE 3. 2 Amino Acid Composition of ~ 1 2 3 550 From Tryptic Peptides From Intact Protein Asp 18 18 Thr 8 8 Ser 3 3 Glu 17 16 Pro 6 8 Gly 17 17 Ala 15 15 Cys 2 2 Val 7 7 Met 4 4 Ile 5 6 Leu 6 6 Tyr 3 3 Phe 4 4 His 1 1 Lys 17 16 Arg 1 1 Trp 1 ND Totals 135 3 Scholes, McLain, and Smith, 1971. Assuming 1 Trp. Assuming 2 moles free Lys per tryptic digest. 136 2 1 - 101 - TABLE 3. 3 Sequence of Tryptic Peptides Tl: Asn-Glu-Ala-Asp-Val- Val-Ala-Phe-Leu-Ala-Gln-Asx T3: -Asx-Pro-Asx-Ala-Gly-Glx-Gly-(Asp , S er ,Gix , 2 2 1 Gly , Ala ) 2 3 1 Ac -Asn-Glu-Gly-Asp-Ala-Ala-Lys T4: Met-Thr-Asp-Asp-Lys TS: Lys T6: Tyr-Gly-Glu-Gly-Ile-Leu-Glu-Val-Ala-Glu-Glu-Lys T7, 8chlel +: Thr-Gly-Pro-Asn-Leu-Tyr-Gly-Val-Val-Gly-Arg T7ch2: Ile -Ala -Ser - Glu-Glu- Gly-Phe- Lys T8ch2: Gly-Glu-Lys T9chl: Gly-Ala-Lys T9ch2: Gly-Gly-Lys Tl0ch2: Glu-Phe-Asn-Lys T 1 Och3: Thr-Lys Tl lchl: Asn-Pro-Asp- Leu-Thr-Trp-Thr-Glu-Ala-Asn-Leu-Ile 2 -Glu-Tyr-Val-Thr-Asp-Pro-Lys -Pro- Leu-Val-Lys Tl lch3: Met-Gly-Lys Tl 2: Met- Thr -Phe - Lys Tl 3: Lys -Ile-Ala-Ser-Glu-Glu-Gly-Phe-Lys Tl4: Cys - Lys -Ala-Cys -His -Met-Ile -Gln-Ala-Pro-Asp-Gly -Thr -Asp-Ile - Ly s 1 2 Ac is the acetyl group . Two moles - 102 Severe decreases in yield were also noted at Gln or Glx residues, attributable to the formation of pyroglutamic acid. Finally, cleavage was noticeably slower at Gly residues in keeping with previous observations (Konigsberg and Hill, 1962). As these three problem amino acids account for over 5 0% of the peptide, lack of complete Edman degradation is not surprising. Further attempts to generate usable shorter fragments were unsuccessful. The peptide TlCl was resistant to digestion with pepsin. Papain at a variety of concentrations, times, and pH produced at least 14 major sub-fragments of similar chromatographic and electrophoretic mobilities. Partial acid hydrolysis for periods as short as 2 hours still produced 16 major sub-fragments, which, by a combination of Dowex 50 and Dowex 1 chromatography, could not be purified in sufficient yield for further characterization. The successful completion of the sequence of this peptide must await the isolation of sufficient quantities of material for extensive purification of the nonspecific sub-fragments. 3. 3. 4. Peptide T2 Direct Edman degradation established the initial sequence Leu-Ala-Gln-Asn and comparison with Tl indicated this peptide was derived from the carboxyl-terminal region of the larger fragment by cleavage between Phe-Leu. The yield of purified peptide was 10% and the amino-terminal fragment was not recovered. tryptic-like cleavages were observed. No other chymo- The supposed anomalous hydrolysis by TPCK-treated trypsin of a Try- Leu bond in ~-type - 103 - cytochromes has been observed in several previous instances, (Nolan and Margoliash, 1966; Chan and Margoliash, 1966; and Matsubara and Smith, 1963) suggesting that an aromatic to hydrophobic linkage or some structural feature of the intact protein makes this bond susceptible to tryptic action. From the known three-dimensional structures for the eukaryotic cytochromes (Takano et al., 1973) and, as will be shown, the known structure of this protein, the Tyr - Leu ( or Phe- Leu) bond occurs near the center of a long et-helix, with other tryptic cleavage points at least 6-8 residues removed. Thus during enzymatic digestion of the whole protein, the segment may retain a helical conformation and this structural arrangement may in some way prime the bond toward tryptic hydrolysis. 3. 3. 5. Peptide T3 This pure peptide failed to give a ninhydrin reaction although it could be visualized with the peptide-bond stain of Cl /starch/KL 2 Edman degradation, dansylation, and incubation with leucine aminopeptidase failed to liberate a free amino-terminal residue. As the intact protein concurred in these negative results, it was concluded that this peptide represented the blocked amino-terminal segment of the protein. The blocking agent was directly identified as an acetyl group through the hydrazinolysis method of Schmer and Kreil ( 1969). The 1-acetyl, 2 -dansyl hydrazide produced in the reaction was identified through chromatography on aluminum oxide TLC sheets (neutral, 200 µ; Merck Co. ) in the Schmer and Kreil solvent systems versus an - 104 authentic standard. Control experiments were performed on acetylated tuna cytochrome ,.Sand quantitation was assured through relative fluorescent intensities of product and standard. The blocked peptide at 1 ~ was degraded in 0. 25 N acetic acid for 2 hours at 105 ° C in vacuo, conditions reputed to minimize loss of amide groups (Schroeder et al., 1963). The resultant mixture was purified on Dowex 50 to give two main fractions, AcI and II plus other minor mixtures. Ac II was a pure peptide and dansyl-Edman degradation established the first two residues as Gly-Asx. From the composition and known trypsin specificity, this sequences the peptide as Gly-Asx-Ala-Ala-Lys. The net charge on the peptide under electro- phoresis at pH 6. 5 was neutral implying an acid at the second position. Fraction Ac I demonstrated three ninhydrin-negative peptides under fingerprinting, two of which were identified as undigested T3 and acetyl-asparagine through comparison with standards. Amino acid composition studies indicated mainly Asx and Glx in roughly stoichiometric amounts. This, together with the composition difference between T3 and T3AcII implies acetyl-A sn-Glx as the main component of AcI. The parent peptide T3 carried a net negative charge, and in electrophoresis was more mobile than the -1 charged subfragments of fraction A cl. As its mobility was clearly greater than that of reference peptides of similar molecular weight and charge -1, it must carry the maximum charge which is -2 consistent with the directly determined acetyl-asparagine group. Then the complete sequence must be acetyl-Asn-Glu-Gly-Asp-Ala-Ala-Lys. Incubation of AcI with carboxypeptidase C did not liberate any amino acid, but this lack of reaction may be due to the short length of the peptide. Minor fragments of uncertain purity indicated free glycine, free alanine, (Ala , Lys) and 2 (A sx, Gly, A la) in some support of the above assignments. - I 05 3. 3. 6. Peptide T4 Four direct Edman steps establish the sequence shown in Table 3. 3. Sequence and compositions established this peptide as identical to T 5el - . 3. 3. 7. Peptide T5el Fingerprint comparisons and amino acid analysis show this to be free lysine. Relative yields from the Dowex column are not helpful for establishing the yield of free lysine as fractions were selected for purity, not quantitation. A known molar amount of £ 550 was digested with trypsin and the hydrozylate applied directly to the Beckman amino acid analyzer. Free lysine was detected at the level of 0. 7 moles Lys /mole digest. Subsequent results from the x-ray work will demonstrate the presence of both a Lys-Lys and an Arg-Lys bond in £ 550 , which after maximum digestion would give 2 moles Lys /mole digest in disagreement with the above experiment. One grouping only was anticipated from the free lysine analyzed and the results of sequencing peptides T7ch2 and T 13. The discrepancy may arise from the presence of extensive minor peptides Lys -x... x-Lys or x- ... -Lys -Lys which were not recovered or an inaccuracy in the 0. 7 mole determination. This latter point is quite possible as the peptide mixture applied to the analyzer gave a high background ·to the analysis and may have distorted the free lysine elution pattern. 3. 3. 8. Peptide T6 After the modification of Foster et al. (1973) direct Edman degradation gave Tyr-Gly-Glu-Gly-Ile-Leu-Glu-Val-Ala-(Glxz, Lys). - 106 The net negative charge of the peptide is not sufficient to establish the side chain functions on the remaining two Glx residues as the already determined two Glu residues would account for the charge. Complete enzymatic hydrolysis by first papain then leucine aminopeptida.se (Margoliash et al., 1962) released 4 moles Glu/mole peptide and no Gln. The placement of Lys follows from trypsin specificity and completes the sequence shown in Table 3. 3. 3. 3. 9. Peptide T7ch2 PTH-amino acids were observed for seven of the eight residues, all in. agreement with the peptide net charge. 3. 3. 10. Peptide T8ch2 All three residues were observed as PTH amino acids. 3.3.11. Peptide T7, 8chlel+ As outlined in Figure 3.1 a major peptide had been isolated as an intermediate, split between two other consecutive major Dowex fractions. These were partially purified, recombined, then further purified to a homogeneous state. After Foster modification, ten of the eleven residues were identifiable as PTH-amino acids with Arg firmly fixed by specificity. 3. 3. 12. Peptides T9chl, T9ch2 and T10ch2 In each case, complete sequencing was affected by Edman degradation. - 107 3. 3. 13. Peptide T10ch3 Dansylation and composition complete this simple di-peptide. 3. 3. 14. Peptide Tll chl By direct degradation, the following was established: Asn-Pro-Asp- Leu-Thr-Trp-Thr-Glu-Ala-Asn-Leu-Ile-Glx-Tyr-ValThr-Asx. The triad Thr-Trp-Thr was detected in very low yields. Uphans, Grossweiner and Katz (1959) have noted that a combination of o 2 and trifluoroacetic acid leads to an X-Trp linkage rearranging to 3-carboxy-f>-carboline and this may explain the low yield. When this peptide was purified by paper chromatography, the tyrosine present was found to be converted to 3, 5 dichlorotyrosine which was detected as such by comparison synthetic standard. with an authentic, The formation of this amino acid in a side reaction involving chloride and traces of oxidizing agent has been reported (Sanger and Thompson, 1963). When the same peptide was purified via gel filtration, tyrosine was quantitatively recovered as the natural amino acid. The long peptide was hydrolyzed in a pulse digestion experiment with 0. 3 mole% chymotrypsin for 3 hours. One pure peptide, TllchlC3, was isolated and analyzed to show a sequence Val-Thr-AspPro-Lys-Pro-(Leu, Val, Lys). The sequence overlap with Tl lchl establishes this as derived from the carboxyl terminus of Tl lchl. Kinetic digestion of Tl lchl with carboxypeptidases A and B indicated -Leu-Val-Lys. - 108 Fingerprinting revealed another main fraction, Tl lch1C2, to be composed of just undigested Tl 1 chl plus a new fragment. Dansylation indkated a single amino-terminus for the fraction suggesting the new fragment to be derived from the initial portion of Tl lchl. Sequence analysis showed unique products through Asn-Pro-Asp-LeuThr-Trp. Carboxypeptidase A, which would not liberate the terminal Lys on Tl lchl, indicated the new fragment to end in Tyr, establishing it as the complement to TllchlC3. Because of low yields, the Edman procedure could not definitively establish the charge of Glx 13 (peptide numbering). There are sufficient acid groups on peptides Tllchl and TllchIC2 to account for net peptide charge. The molecular weight of Tl lchl was so large that electrophoretic migration was not able to indicate clearly a net peptide charge of -1 or -2, corresponding to Gln 13 or Glu 13. There was insufficient remaining material for an accurate enzymatic digestion. The assignment of 13 as Glu was made solely on the basis of the failure of carboxypeptidase A to liberate Gln-Tyr in kinetic digestion experiments. 3. 3. 15. Peptides Tllch3, T12 All but the terminal Lys residues were directly detected in the Edman reaction. 3. 3. 16. Peptide Tl3 This peptide differed from T7ch2 only in having an amino- terminal Lys as well as the carboxyl-terminal Lys. The previously mentioned presence of free Lys and this peptide are consistent with an Arg- Lys and/ or a Lys - Lys linkage. The lack of hydrolytic reactivity of an amino-terminal Lys has been documented (Kasper, 1970). - 109 3. 3. 17. Peptide Tl4 (heme peptide) The Edman reaction would not liberate any product until after application of the heme-removal reaction and carboxymethylation. The sequence then indicated was (Cys was detected as carboxymethyl cysteine): Cys -Lys -Ala-Cys -His-Met-Ile-Glx-Ala-Pro. The presence of the typical cytochrome heme linkage, -Cys-X-X-Cys-His-, and the applicability of the heme removal reaction are taken as indicating the normal manner of covalent heme-protein attachment, through thioether bonds on the two Cys residues. Thermolytic digestion presented four major peptides. highly insoluble column. The heme peptide was not recovered from the ion exchange An Ile-Lys fragment and peptides Tl4Thl and Th2, which differ in composition{Table 3.1) only by (Ile, Lys) were isolated as pure products. Direct or dansyl-Edman degradation of Th2 yielded Ile-Gln-Ala-Pro-Asx-Gly-Thr-Asx-{Ile, Lys). procedure established Tl4Thl as equivalent in sequence to Tl4Th2 but lacking the terminal Ile, Lys residues. step dropped markedly. A dansyl-Edman The yield after the Pro A partial explanation may be that an et ➔ (3 peptide bond shift is occuring at the succeeding A sx residue in a reaction similar to that discussed for peptide T 1. Treatment of intact peptide T14 with carboxypeptidases B and A established the terminal sequence as -Ile-Lys but failed to liberate any other amino acid in spite of high enzyme concentrations and long digestion times. Such a reactivity failure can be rationalized by the presence of Asp as the next residue in keeping with the final Edman product Asx of Tl4Th2. Both Thl and Th2 carry net negative charges - 110 and as ThZ adds one positive charge (Lys) onto Thl, the magnitude on Thl must be -2 and hence both Asx residues are Asp. In summation, the peptide T14 was sequenced as Cys-Lys-AlaCys-His -Met-Ile-Gln-Ala-Pro-Asp-Gly-Thr-Asp-Ile-Lys. 3. 4. Discussion The sequence of all tryptic peptides from ~ 550 (M. denitrificans) has been determined (Table 3. 3) with the exception of the final ten · residues at the protein carboxyl-terminus. The failure to complete these residues has been ascribed to side reactions inherent in a peptide of this composition, namely, an er ➔ ~ peptide shift at aspartyl groups and extensive formation of pyroglutamyl blocking groups. The comple- tion of the sequence is awaiting the isolation of sufficient quantities of ~ 550 protein to perform limited, non-specific hydrolysis arid succes~- ful isolation of the fragments. The sequence of other tryptic peptides has been established by conventional techniques . The placement of amide side chains has followed either direct identification as P TH products or simple arguments based upon net peptide charge. The number of amide groups so found raises an interesting point concerning the isoelectric point of the protein. Scholes, McLain, and Smith (1971) have reported a pl of approximately 4 and McLain (personal communication) has remeasured this number by gel and column isoelectric focusing as 4. 5. From the amino acid composition of the tryptic peptides, considering the number of amide groups , an approx imate isoelectric point may be calculated (Edsall and Wyman, 1958). The calculation is somewhat complicated by a - 111 , lack of knowledge concerning the amides in the unsequenced carboxyl terminal region. Nevertheless, a minimum pl can be calculated to be 4. 6 (17 Lys + 1 Arg = 18 positive charges. liganding. The only His is involved in 13 Glu + 12 Asp+ 2 propionic acids = 2 7 negative charges) by assuming that all the unknowns are acidic functions. The agreement would suggest that most of the remaining unsequenced groups are indeed acids. The isoelectric point of this molecule may ultimately be a matter of some importance in view of the known cross-reactivity between the M. denitrificans electron transport chain and that of mitochondria (Scholes, McLain, and Smith, 1971 ). Mitochond:rial cytochrome£ carries a high isoelectric point, greater than 10, and it has been shown that the positive charges are involved in binding to either cytochrome oxidase or reductase That £ 550 (Smith and Minnaert, 1965). cross-reacts with cytochrome£ clearly emphasizes that the charge effect must be due to a grouping of charge, rather than a net charge effect. In subsequent work, it will be shown that £ 550 contains a severe localization of its positive charges into a limited area on the surface of the molecule. The reconstruction of the primary structure of the intact protein follows from a knowledge of the sequence of the tryptic peptides and the availability of a high resolution electron density map; will be the subject of a later chapter . th.is result The successful reconstruction follows from the constraints obtained through the traditional sequencing of the peptides reported here, for by knowing the tryptic peptides, the interpretation of the x-ray map requires only a matching of some - 112 sixteen well characterized fragments to their unique location in the map and hence into sequence. - 113 - 3. 5. References Ambler, R.P., (1963). Biochem. I:_, 89, 349. Ambler, R. P., ( 1967). In Methods in Enzymology, XI, C.H. W. Hirs, ed., Academic Press, New York. Bitar, K., Vinogradov, S.N., Nolan, C., Weiss, L.J., and Margoliash, E., (1972). Biochem. I.:_, 129, 561. Bornstein, P., and Balian, G. , ( 1970). Bouchilloux, S. , ( 19 5 5). ~ Biol. Chem. , 245, 4854. Bull. Soc. Chim. Biol. , '}__]__, 2 5 5. Braunitzer, G., Schrank, B., and Ruhfus, A., (1970). Z. Physiol. Chem., 351, 1589. Hoppe-Seyler's Chan, S.K., andMargoliash, E., (1966). I:_ Biol. Chem., 241, 335. Crestfield, A. M., Moore, S., and Stein, W. H., (1963). I.:_ Biol. Chem., 238, 622. Dayhoff, M. 0. (1972). Atlas of Protein Sequence and Structure, National Biomedical Research Foundation, Washington, D. C. Delange, R. J., and Smith, E. L., (1971). In The Enzymes, v. III, 3rd ed., P. D. Boyer, . ed., Academic Press, New York. Dickerson, R.E., (1971). J. Mol. Biol., 57, 1. Easley, C. W., ( 1965 ). Biochim. Biophys. A eta, 107, 386. Edsall, J. T., and Wyman, J., (1958) . Biophysical Chemistry, _!_, Academic Press, New York, p. 509-510. Foster, J. A., Bruenger, E., Hu, C. L., Albertson, K., and Franzblau, C., ( 1973). Biochem. Biophys. Res. Commun., ~, 70. Gray, W.R., ( 1972). In Methods in Enzymology, XXV, 333, C.H. W. Hirs, ed., Academic Press, New York. Hartley, B. S . , (1970). Biochem. I.:_, 119, 805. Herriott, J. R., Watenpaugh, K. D., Sieker, L. C., and Jensen, L. H., (1973). I.:_ Mol. Biol., ~, 423. Jeppsson, J., and Sjoquist, J., (1967). Anal. Biochem., .!..§_, 264. Kasper, C. B., (1970) . In Protein Sequence Determination, S. B. Needleman, ed., Springer-Verlag, New York, p. 155. - 114 Kendrew, J.C., Watson, H. C., Strandberg, B. E., Dickerson, R. E., Philips, D. C., and Shore, V. C., (1961). Nature, 190, 663. Koningsberg, W., and Hill, R. J., (1962). J. Biol. Chem., 237, 2547. Margoliash, E., Kimmel, J. R. , Hill, R. L., and Schmidt, W.R., (1962). J. Biol. Chem., 237, 2148. Margoliash, E., Fitch, W. M. , and Dickers on, R. E. , (196 8). Structure, Function, and Evolution in Proteins; Brookhaven Symp. in Biol. , ~ ' 2 5 9. Matsubara, H. , and Smith, E. L. , (196 3 ). J. Biol. Chem. , 2 38, 2 732. Matsubara, H. , and Sasaki, R. M. , (196 8). J. Biol. Chem. , 243, 1 732. Needleman, S. B., and Blair, T. T., (1969). Proc. Nat. Acad. Sci., ~ ' 1227 No 1an, C. , and Margo li ash, E. , (1 96 6 ) . J. Bio 1. Ch em. , 2 4 1 , 1 04 9. Noltman, E. A., Mahowald, T. A. , and Kuby, S. A., ( 1962 ). L_ Biol. Chem. , 2 3 7, 1 146 . Sanger, F., and Thompson, E. 0. P., (1963). Biochim. Biophys. Acta, 11-, 46 8. Schmer, G., and Kreil, G., (1969). Anal. Bio chem., Scholes, P. B., McLain, G., and Smith, L., (1971). ..!.Q, 2072. ~ 186. Biochemistry, Schroeder, W. A., Shelton, J. R., Shelton, J.B., Cormick, J., and Jones, R. T. , (196 3 ). Biochemistry, ~' 992. Smith, L., and Minnaert, K., (1965). Biochim. Biophys. Acta, 105, 1. Smith, L., Newton, N., and Scholes, P., (1966). In Hemes and Hemoproteins, B. Chance, R. W. Estabrook, and T. Yonetoni, eds., Academic Press, New York. Smithies, 0. , Gibson, D. , Fanning, E. M. , Goodfliesh, R. M., Gilman, J. G., and Ballantyne, D. L., (1971) , Biochemistry, ..!.Q, 4812. Takano, T., Kallai, 0. B., Swanson, R., and Dickerson, R. E., (1973). L_ Biol. Chem., 248, 5234. Tschesche, H., and Kupfer, S., (1972). Eur. J. Biochem., 26, 33. Uphans, R. A., Grossweiner, L. I., and Katz, J. J., (1959). 129, 641. Science, - 115 CHAPTER 4. TERTIARY STRUCTURE OF c 550 4. 1. Ordering the Tryptic Peptides With the tryptic peptides of known sequence, the available x-ray map could be interpreted to align the fragments into a complete primary structure. The ordering is normally obtained through a second or possibly third enzymatic digestion of the protein, or occasionally by sequence homology to some previously determined and highly similar protein. The merging of x-ray map and peptide sequence data is quite novel, as previous sequence assignments from crystal structure data have been based solely on a serial identification of the individual amino acids from the electron density (Kendrew, et. al., 1961; Herriott, et. al., 1973). While the pure x-ray approach produces correct assignments at a rate of 50% to 75%, a reliable, complet~ sequence has never been produced in this manner. The method of aligning the peptides by quite straightforward, electron density was utilizing nothing more complex than the approach one adopts when building a jigsaw puzzle. At the resolution of the map 0 used, 2. 45 A, the main chain could be easily differentiated from the side chain groups. When first considering the array of peptides and the map, concentration was focused on patterns in the size of the side chain groups and a similar pattern was sought in the density as it was presented in a Richards box, As with any puzzle, if the choice did not reveal a clear match, another was made. Once a candidate had been found, it was coniirmed by carefully fitting a scaled Kendrew model to the box density, - 116 In this process, it was not necessary to attempt to be linear, that is, to generate a sequence from the amino to the carboxyl terminus. Rather, a strong sense of checks and balances occurred when the fitted pieces, scattered about, finally were found to be joined by the remaining pieces. Certain residues proved more helpful than others as guideposts (Table 4.1). Aromatics were especially clear and suggested many of the initial guesses. Glycine and alanine were conspicuous by an absence of density in their locale. Asparagine / asparate, serine, valine and threonine were the most difficult to match; being of moderate size and shape, they offered no distinguishing characteristics. Histidine, cysteine, and arginine are ranked low in Table 4. 1 simply because ~ contains so few of these residues. 550 In another protein, they should be quite easy to lo·cate. Table 4. 2 reports the total primary structure as signed to ~ 5 50 . From Chapter 3 it will be recalled that the final ten residues could not be sequenced chemically. The arrangement reported in Table 4. 2. represents the best guess possible from the map alone. Unfortunately, this is the poorest region of the map in which to make a pure x-ray sequence assignment. The terminal eight residues appear badly disordered in the map to the extent that several conformations appear possible. This conformation averaging smears the density and weakens it so that side chain identification is highly dubious. The remainder of the sequence is presented with total confidence. The constraints imposed by knowing from the chemical work, the size - 11 7 - TABLE 4. 1 ASSESSMENT OF RESIDUES Trp Tyr Phe Met Pro :>, .µ •r-1 Lys 1--1 rel r--l u Gly Ala bl) s:1 •r-1 U) rel Q) Ile Glx Leu 1--1 u Q) '"O ! Asx Ser Val, Thr (His, Cys, Arg) - 118 TABLE 4. 2 M. DENITRIFICANS c Ac 550 SEQUENCE 1 5 10 - Asn - Glu - Gly - Asp -Ala - Ala - Lys - Gly - Glu - Lys ~ T3 11 15 - Glu - Phe - Asn - Lys - Cys - Lys - Ala TBch:2 ~ 20 - Cys - His - Met T10ch2 21 25 - Ile - Gln - Ala - Pro -Asp - Gly - Thr - Asp - Ile 30 - Lys _ - - T14 31 35 40 - Gly - Gly - Lys - Thr - Gly - Pro - Asn - Leu - Tyr - Gly - T9ch2 ~ T7 41 45 - Val - Val - Gly - Arg - Lys - Ile -Ala Bchlet-- 50 - Ser - Glu - Glu Tl3- 51 55 - Gly - Phe - Lys - Tyr - Gly - Glu - Gly - Ile 60 - Leu - Glu TS-- 61 65 70 - Val - Ala - Glu - Glu - Lys - Asn - Pro - Asp - Leu - Thr 71 75 - Trp - Thr - Glu - Ala - Asn - Leu - Ile 80 - Glu - Tyr - Val Tllchl - - 81 85 90 - Thr - Asp - Pro - Lys - Pro - Leu - Val - Lys - Lys - Met +---+ -E-- 91 95 100 - Thr - Asp - Asp - Lys - Gly - Ala - Lys - Thr - Lys - Met T9chl T4 ~ ~T10oh3 ➔ ~ 101 105 110 - Thr - Phe - Lys - Met - Gly - Lys - Asn - Gln - Ala - Asp Tl2 ~ Tll a h3----). 111 115 120 - Val - Val - Ala - Phe - Leu - Ala - Gln - Asx - Asx - Pro Tl - 121 125 130 - Asx - Ala - Gly - Glx - Gly - Glx - Ala - Ala - Gly - Ala 131 135 - Gly - Ser - Asx - Ser - Glx Amino-terminus is acetylated. 135 residues. - 119 - and shape of even short segments of the molecule are sufficient to generate a reliable structure. Analogous to pure chemical sequencing, most information was provided by the longer peptides. The assigrunent of fragment Gly- Gly-Lys to positions 31-33 and Gly-Ala-Lys to positions 95-97 illustrates the most difficult decision which was made. The respective placements rest solely on the slight but definite methyl density seen at position 96. As long as peptides exhibit at least this amount of structural differentiation, and all others by far did, map density is sufficiently unique to determine structure unambiguously. The multiple positive charge groupings -Lys and -Arg 44 - Lys 45 88 - Lys 89 - - represent information derived more or less exclusively from the x-ray map. The presence of one such linkage was suggested (not proved) by finding free lysine in the tryptic digest and a tryptic peptide with an amino-terminal lysine. (See peptide T 13, Chapter 3. ) Ultimately, the suggestion was fulfilled by locating in the density the -Arg-Lys- linkage. The remaining group was inherently not detectable in the peptide chemistry initiated by trypsin hydrolysis. In fitting the map, the preceeding peptide -Pro- Leu- Val- Lys and the succeeding peptide -Met-Thr-Asp-Asp-Lys - were matched quite well with the density, but, a gap remained, one residue long and with a side chain density indicative of Lys. Hence, a consideration of the map and the origin of the other peptides established the correct sequence as involving a double Lys bond. - 120 - 4. 2. Molecular Structure - General Perutz and co-workers have spent more than ten years on the hemoglobin molecule and yet, parts of the fine structure are still not clear. year old ~ So this author finds no hesitancy in reporting that the one 550 structure is not complete. The answers are there; the problem has been to interpret correctly the less -than-atomic resolution map. The interpretive ambiguity and the inherent mechanical and optical inadequacies of Richard's Box techniques, referred to as model building limitations, are sufficiently severe that fine structure placement of peptide amide planes, exact hydrogen bonding, salt bridge patterns, Van der Waals contacts, etc. - must be taken critically realizing that most are correct, but some potentially wrong. What will be discussed here are only those results which are firmly established. The available information is rich in detail but is only effectively communicated in three dimensions. For the present purpose only some relevant highlights are offered, as a prelude to some conclusions which will be drawn concerning cytochrome evolution and mechanism. Figures 4.1, 4.2 and4.3 present three contrasting a-carbon diagrams of the ~ 550 structure and Figure 4.4 presents a comparison between reduced tuna~ (Takano, et al., 1973) and ~ 550 . Figure 4.5 summarizes the basic result of the entire structure determination. brief, .5:. 550 In is a molecule composed of a core structure virtually identical with the smaller eukaryotic cytochrome but with major insertions and a few minor deletions . These changes obey a consistent rule of occurring at Figure 4. 1. An a- carbon stereo diagram of 35 60 • Heavy bonds represent the distances between consecutive ct- carbons. The light line bonds represent true ch0emical bonding in the side chains and the heme. On a scale of 0. 0559 inches/ A. I--' N I--' Figure 4. 2. As in 4. 1, rotated 90 ° about the vertical axis. N N I-' Figure 4. 3. As in 4. 1, rotated 90 ° about the horizbntal axis. w N I-' - 124 - Figure 4. 4. £ 560 (SF) and tuna c (CF) in equivalent orientation. - 125 - ( . : ... .· . ·. . . - ~ . . . . (9 . .. :--- ..· .... .. . • Figure 4. 5. • • • • . . .... . . . .. -.- ... -_:.:.·· .- . . . - -·. . . . . . . . . --• . . . . .. -:. . .. • # • A stylized £;3so molecule. - 126 - locales in the eukaryotic fold at a turn or change of direction, somewhat lengthening or abbreviating the turn, but ultimately returning to a common path. It is as if nature has been capricious with those regions of the molecule least likely to interfere with a central design for the protein. A detailed description of the £ 550 polypeptide folding and its comparison with eukaryotic cytochrome £ has been published (Timkovich and Dickerson, 1973). The structure comparison becomes more striking when R. rubrum cytochrome £ 2 (Salemme, et. al., 1973), another bacterial type cytochrome, is included, for £ the minor insertions (61 ➔ 68, 88➔ 93, and 96➔ 97; £ three of the minor deletions (12➔ 13, in £ with respect to£· 550 550 shows three of numbering) and and 106➔ 109) present To a bigh degree of accuracy, eukaryotic c stands as a core structure, £ then £ 103➔ 104, 550 2 2 adds a few insertions onto the core, makes the same additions plus some other major insertions. The striking similarities can be over emphasized leading to a false impression, in the same vein as calling myoglobin and hemoglobin identical proteins. The structural differences are sufficient to lead to the remarkable reactivity patterns outlined in Chapter 1, that £ 550 and £ can be cross -reactive in a natural electron transport chain while £ 2 is inactive. Yet recall the intriguing fact that £ mediate in structure between£ and £ 550 2 is inter- ! The most fruitful approach to this comparative cytochrome biochemistry seems to begin with the reasonable assumption that all three cytochromes employ the same mode of action. function this means that for all three For cytochrome the reductase binds, however - 127 - fleetingly, to the same vicinity; the reducing electron enters by a highly analogous path, pre-or post-reduction conformation changes are at least somewhat similar, oxidase binding and electron donation are similar, and finally the change back to the oxidized conformation follows somewhat parallel lines. structure of £ 550 With this framework in mind, the may be examined with an eye toward similarities and differences. 4. 3. Molecular Structure - Details In Table 4. 3, the sequences of the three cytochromes have been aligned to demonstrate both sequence homology (all three proteins) and structural homology (only £ 550 and£). In the latter, homology was defined rigorously as existing only when the Ci -carbons and the peptide amide planes of the respective residues agreed in orientation versus neighboring groups and the molecule as a whole. As would certainly be expected, sequence homologous residues are most often also structurally homologus, but arise (positions, in £ 550 some exceptions do numbering 3, 4, 26, 30, 65, 84, 95, and 98). Generally, these are noticed to be near insertions in £ may be the result of necessary changes forced upon £ date the new details. homologs. 550 550 , and so to accommo- Potentially, these are accidental sequence To underscore a fundamental tenet of protein folding and protein architecture, one should notice the large number of structure homologs which are not equivalent in sequence nor compatible in sequence, where by compatible it is meant that the residues would agree in general chemical properties, such as Ile-Leu, or Glu-Gln. Examples - 128 - TABLE 4. 3 Cytochrome Sequence Comparisons 1 G G a V C ~60: E ffiJ l * 20 V ~ - F I D - [QJ A p * = K Q [!] D 25 s G G * = N 40 * * * K P LL¼JV K A A E 50 PIGFIT[TITDA . yN D r--=-'N_Y.c;___, A Y S IE7 S E l._-=G__ F~I K y G LKJ G I 55 * * 60 * ~l~I~ * = p T T T D 65 F L W W W E T MCT] A N A A N I CT] 70 A 75 * 75 y * 65 * 70 L L 35 55 K N TfElM N N = 50 45 * G .....__K__T__G __P;c___N__L__. 45 Q V F E N T AIA7H V V ,~-G--R-K~! I S * K[y]G = E * * 30 I [Kl G 30 lAJ 40 = 15 = = * A D ~C K KC LAC NK CK AC - [K] - H K T G P G G T A G G - 25 H[u] . L FIG RKIT G F Y VS E F * 35 Essa: G N G D A A K G 1 --------5---=-=------==----=---i 20 C K D ---A--A---+----'AC...C....., G * ~ 1 I~l; VQ[r]C 1 5 D * * * * * I rvJ L l_yJ K E K s K M T 90 * * * * * * 95 ~1~'~1 i[IP C ~ ADVVAF .£Essa: 110 100 C £2 .£6 50 : rn~ ~ t li:_JA B Q 115 T K B * = = G s B 104 N E P 120 * * * * B A G Z G 125 z A A A 130 G s z 135 * Cytochrome c is from horse (Margoliash, 1962) and S from R. rubrum (Dus, Sletten, and Kamen, 1968). Residues in blocks indicate sequence homology b e tween _s; 60 and either ~ or ~. The small symbols beneath the _s; 5 0 sequence indic a te structural homology between Soso and residues in ~indicates that the CY-carbon and peptide are in equivalent conformation; = indicates that the side chains are also spatially equivalent relative to the main chain. * - 129 are provided by the stretches 20➔ 21, 57➔ 60, 86➔ 89, 104"4106, and 116➔ 117. The general point is that the over-all conformation is determined by some obscure energetic -entropic equation of state which has as separate inputs, the individual side groups. Some may be radically changed, yet complimentary changes elsewhere eliminate any net effect,or the particular changes are low-sensitivity parameters in directing folding. To paraphrase, the whole is surely the sum of the parts, but a few of the parts may be played with ly, willful- for there are alternatives at arriving at the same sum. Table 4. 4 completes the equating three-dimensionally with eukaryotic £· of ~550 This list of hydrogen bonds has been conservatively constructed,with the £ 550 list considerably shortened by a lack of total confidence in some nonetheless reasonable bonds. The added molecular weight of £ 550 offers the most readily discernible differences toward the eukaryotic folding. 1. The insertion at the amino-terminus (2 residues} is trivial, as it is affecting no other part of the molecule. 2. The first eleven residues form an a-helical segment of cytochrome. In £._sso the "best" model building, i.e. using density maxima, led to a distorted helix, but one not far from the ideal structures found in both~ and £2. 3. £_550 and _S2 share a common deletion relative to c of a halfturn of helix just prior to the covalent heme. 4. Residues 24➔ 3 l in £_s 50 constitute a major new feature. Reference to Figure 4. 4. shows that this is derived by a net insertion of five residues in a resultant conformation which is a clear cousin to the loop in£· 5. The folding in all three cytochromes is constant until £2 adds three residues and £_550 six at position 61 opposite the exposed - 130 - TABLE 4. 4 HYDROGEN BOND COMPARISONS c 550 AND C Bonds in £ Bonds in c 550 Cys 15 peptide O ➔ Cys 18 peptide N Cys 14 peptide O ➔ Cys 1 7 peptide N His 19 imidazole N ➔ Pro 36 peptide 0 His 18 imidazole N ➔ Pro 3 0 peptide 0 Lys 45 peptide O ➔ Ser 48 peptide N £ polypeptide folds in a different manner Glu 56 carboxylic O ➔ Front propionic acid Thr 49 hydroxyl OH ➔ front propionic. Note that Thr 49 in c is structurally equivalentto Gly 55 in Leu 59 peptide O ➔ Glu 63 peptide N Glu 60 peptide O ➔ Glu 64 peptide N Brief turn of a-helix Asp 50 peptide O ➔ Ser 54 peptide N Ala 51 peptide O ➔ Ly s 5 5 peptide N Brief turn of Of-helix Tyr 54 hydroxyl OH ➔ back propionic acid Tyr 48 hydroxyl OH ➔ back propionic Trp 71 indole N ➔ back propionic Trp 59 indole N ➔ back propionic Trp 71 peptide O ➔ Gly 43 peptide N Trp 5 9 peptide O ➔ Gly 3 7 peptide N a-helical hydrogen bonding pattern 72 ➔ 81 Q'-helical hydrogen bonding pattern 60 ➔ 68 Cys 18 peptide O ➔ Gly 35 peptide N Cys 17 peptide O ➔ Gly 2 9 peptide N Sequence analogous Thr 98 is rotated out toward solvent Thr 78 hydroxyl OH ➔ front propionic acid a -helical hydrogen bonding pattern 107 ➔ 118 CY-helical hydrogen bonding pattern 89 ➔ 101 ,S; 50 - 131 - heme edge. 6. Again £2 and _s; 50 share six residues spanning 88➔ 93 corresponding to an insertion into c between positions 75 and 76. This extra coil of protein is in intimate proximity to the exposed heme edge and constitutes a new collar of polypeptide surrounding this edge. Attention should be directed toward this feature, as in the eukaryotic cytochromes the segment 70➔ 80 is sequence cons ervedover sixty species, therefore presumably playing some mandatory role in structure-function. Slightly further along, S3 60 adds another two residues which ~ ·does not, leading to a net inserted coil slightly larger than shown by £2· 7. A minor deletion of three net residues in the region 107➔ 109 shortens the turn c 5 50 makes before beginning the carboxyl terminal a-helix. 8. A major insertion of 14 residues lengthens the tail of c 6 50 and adds an extra layer of protein onto the surface of the molecule diagonally opposite the exposed heme edge. The electron density map is extremely weak in this area suggesting high disorder. Indeed, the final five residues appear to be averaged between two possible conformations with the major form as shown in the Figures. Since these 14-residues protrude rather drastically into the solvent and do not appear to be locked into a fixed array by hydrophobic packing with the rest of the molecule or by hydrogen bonds, one would be justified in assuming that in solution they wave about as a random coil tail on a globular protein. A few additional words of some entertaining, if not enlightening, speculation on the properties of this peptide tail seen appropriate. In the crystal, the protein molecules pack in one direction in a linear chain in which the insertion on the His side (point 4 above) and the tail insertion form a complimentary jigsaw puzzle fit. This is shown schematically in Figure 4. 6. The accidental (? ) compatibility explains the extreme close packing of molecules in this crystal form; but it is doubtful that the fit has any relevancy in solution. Next, Table 4. 5 shows the amino acid composition of the tail 22 ➔ 32 Residues Figure 4. 6. c* ► Crystal Packing along c* Residues 120 ➔ 135 N w 1--' - 133 - TABLE 4. 5 AMINO ACID COMPOSITIONS 1 BACTERIOPHAGE COAT PROTEINS Residues Bacteriophage FR F2, Rl 7 Q Beta ~550 tail (115-134) Asx 15 14 15 4 Thr 9 9 12 Ser 11 13 9 2 Glx 11 11 13 4 Pro 5 6 8 1 Gly 9 9 7 4 Ala 16 14 15 5 Cys 2 2 2 Val 16 14 13 Met 2 1 Ile 6 8 4 Leu 5 8 12 Tyr 4 4 4 Phe 5 4 3 Lys 7 6 7 Arg 4 4 7 Trp 2 2 His 1 From Dayhoff, 1972. - 134 contrasted with the amino acid compositions of some bacteriophage coat proteins. As can be seen, there is some correspondence in the popularity of certain amino acids across the table. The following model, then, is proposed without further justification to reconstruct an evolutionary mutational event of some billions of years in the past. It is well known that lysogenic viral particles, by becoming passive prophag~ may infect a bacterium and insert their DNA into the bacterial chromosome without causing host demise. It is also known that extra genetic material, such as this or such as episomes or prokaryotic sex factors, can be excised from the chromosome either by taking a small piece of host DNA with it or by leaving some behind. One can, by straining, envision a phage infecting a M. denitrificans bacterium and then passively leaving behind DNA which has obliterated the synthesis termination signal for a pro-cytochrome c protein. Our ancestor organism then goes on to take over the genus as the result of some other survival trait, all the time carrying about a bit of junk peptide on its cytochrome ~4. 4. Molecular Structure ..;._ Some Net Effects In an early paper (Timkovich and Dickerson, 1973) the author stressed a preservation in the packing of aromatic and hydrophobic residues around the prosthetic group heme during the time eukaryotic ~ and bacterial ~ ancestor. 550 have evolved from some presumed common With the exception of incorrectly identifying amino acid number 86, a leucine, as a tyrosine or phenylalanine, and connecting the side chain phenyl group of Phe 114 to the wrong spot on the main - 135 0 chain, the interpretations and structural details of this 4A analysis have been successfully borne out at high resolution. In point of fact, the comparison becomes even more striking, for now the orientation of hydrocarbons and aromatic rings can also been seen to be conserved (Table 4. 6). For the eukaryotic cytochromes we have a choice of ring orientations to talk about - the oxidized or reduced structures, which show clear differences in ring alignments. ~ 550 Of the two, oxidized appears to mimic the oxidized horse cytochrome most closely, but is is somewhat intermediate between oxidized horse and reduced tuna cytochrome. Some of the discrepancy is attributable to errors and a general lack of resolution in the original horse map,and this situation is being currently remedied by a high resolution redetermination of the oxidized structure,now using high quality oxidized tuna cytochrome crystals (Swanson, Trus, and Di_ckerson, personal communication). Preliminary results are consistent with the view that the ring structures between oxidized and reduced forms are more similar than was originally supposed and the final, fine differences will be more nearly equivalent to those now seen between reduced tuna and oxidized bacterial cytochrome. The similarities here are worth belaboring,for it has been proposed (Kassner, 1972) that hydrophobic packing about the heme plays a major role in adjusting the redox potential of the iron center to its physiologically functional value. had ~ , with its 2 It would have been informative E~ 60 mv higher than~ or ~ 550 difference in heme packing compared to ~ or ~ , shown some striking 550 , but unfortunately - 136 - TABLE 4. 6 AROMATIC RING ORIENTATIONS Oxidized c £550 Phe 12, adjacent to heme and tipped perpendicular to it. 1 Phe 10, as in £5 50 • Reduced c 2 Phe 10, as in£s 50 • Tyr 39, lies in a groove His 33, although a Trp 33, another on the surface of the different ring system, different group, but the same orientation. molecule with only its lies in the identical edge exposed to orientation. solvent, like .a miniburied heme. No analog in ~ 50 • Phe 36, lies in a groove on the surface as did 33. Phe 36, same locale but rotated by 90 ° to lie flush with the surface. Phe 46 - Tyr 48. Same. Tyr 46 - Tyr 48. Same. Phe 52 - Tyr 54, adjacent and perpendicular to the heme plane With respect to each other, their planes are rotated by 60-80 °. Trp 71. Hydrogen bonding to propionic acid. TI- system 90 ° with respect to heme. Trp 59. Same as £sso. Trp 59. Similar but rotated slightly about a line perpendicular to the heme plane. Tends to place a flat face against the heme while the others place an edge. Tyr 67. TT-cloud Tyr 67. As in ~ 50 but Tyr 6 7. Same locale but rotated 90 ° to be almost parallel with mingles with heme perpendicular to the TT- could. At a bout 30 ° the heme. heme, parallel to the with respect to the Trp. heme. The analog in~ 5 0 is a leucine. Tyr 74. Parallel and , adjacent to Trp 59. . Flush with the molecular surface. Tyr 74. Rotated by 90 °. Perpendicular to the surface. - 137 {Table 4. 6 - continued) Oxidized c Phe 102. Adjacent and parallel to the heme. A methionine - ligandside equivalent to Phe 12. Uncertain. 1 Reduced c Phe 82. As in ~ 2 60 • Phe 114. Adjacent to Tyr 97. As in S5 60 but Tyr 97. As in~ 50 but Phe 12 on the side rotated to be parallel rotated to be at about away from the heme. to Ph e 10. 4 5 °. Roughly perpendicular. 1 2 Dickerson, et. al., 1971. Takano, et. al., 1973. - 138 this simplistic hope has not been realized. Things instead look uniform. The redox potential tuning must then lie in some fine structural detail; -rr - electron interactions or ligand bond strain offer two intriguing variables. A most unusual and asymmetric charge distribution can be seen on the surface of . <=_ 550 from an inspection of the distribution of acidic and basic side functions. Almost all of the lysine residues ( 15 out of 17) are grouped into a collar about the exposed edge of the heme in what shall be called the front of the molecule. This density of positive charges falls-off with the distance from the edge. There 0 are 11 lysines within 10 A of the crevice opening, another 4 within 0 17 A forming a second more or less concentric ring, and only two which seem to be outside the collar. All the lysines are grouped into the front hemisphere of surface. · The acidics show a more uniform distribution with perhaps what could be called a slight grouping on the back surface near residues 55-65, 120-135. Through default, the second hemisphere carries solely negative charges and thus . <=_ 550 exhibits properties of a 0 macro-dipole, being a rough sphere of 35 A diameter with a positive half and a negative half. The charge separation may have mechanistic implication and so will be referred to a later section. 4. 5. Evolutionary Relationships Extensive comments have been previously published (Timkovich and Dickerson, 1973) on the phylogenetic relations between..<=_, £ 550 and so, only a brief review shall be given here. -'=-z and It seems most reasonable to suppose that a common ancestral protein existed for - 139 - the three and that divergence along three lines has produced the present contemporary versions. £ 2 If the ancestor is taken as most resembling in structure, folding, and size, then additions account for £ 550 and deletions have led to S with this reconstruction being the simplest which avoids postulating parallel mutational events. By looking at in vivo function, it appears most plausible to suppose £ 550 and£ to be more closely related to each other than to ~ , using the argument that 2 respiratory proteins share more evolution history than would any respiration- photosynthesis pair. The £ 550 sequence could have provided more of a base on which to view these suppositions. Yet, a similarity/difference census between sequences (Table 4. ?)cannot clearly answer the question "Which two are most closely related?". £ 550 It is possible for the insertions in to be the product of a local gene duplication or an unequal cross - over event, but an examination of the added peptide regions shows no It may definite evidence of repetition of a neighboring sequence. simply be that time since the mutational event has allowed point mutations to scramble sequences sufficiently to obliterate local homology. The general result of sequence comparisons is neutrality things are so equidistant that the data can explain anything. It is to be expected, although disappointingly so, that time has obscured a reconstruction of phylogeny on the basis of sequence. At a very broad level, it may be stated that the respiratory mitochondrial cytochrome has diverged as far from the bacterial respiratory cytochrome as it has from a photosynthetic cytochrome. - 140 - TABLE 4. 7 SEQUENCE COMPARISON SIMILARITY - DIFFERENCE MATRIX C 43 C 57 54 46 53 Same 55 Different The Table was constructed from the sequence alignment of Table 4. 3 and thus it represents a comparison only over those positions which are structurally equivalent in the proteins. Residues aligned with deletions in the contrasted proteins are not counted in either category. The upper right triangle lists the number of residues matching, the lower right, the number different. The amides and acids Asn/Asp and Gln/Glu were counted as equivalent because in both £..z (Bartsch, et. al., 1971) and £.. (G. McLain, personal 550 communication) there exists the possibility of deamidation during isolation or in the in vivo cytochrome complement. - 141 - 4. 6. Current Proposed Mechanisms for the Reduction of Cytochrome£ Takano, et. al. ( 1973) have recently proposed a mechanism for cytochrome reduction along the following lines. In the oxidized state an aromatic residue at position 74E (subscript E refers to sequence nrnnbering according to the eukaryotic cytochromes, B will refer to bacterial £ 550 , and P the photosynthetic £z) lies parallel to the molecular surface and to the aromatic ring of Trp 59E. meanwhile lies adjacent and parallel to the heme. Tyr 67 E An electron is seen as being donated to the TI-orbital of 74E and by -rr-cloud overlap of proximal groups, it migrates to 59E' As this is going on, a stable -rr-electron from 67 Emigrates to the heme and reduces the iron center. The protein then undergoes a conformation change and Trp 59E and Tyr 67 E become parallel. 11 Through cancellation of an electron hole 11 and the antibonding electron on 59E, the process is complete. The £ 550 structure raises grave doubts concerning this proposal in so far as £ 550 and£ share a common mechanism. It can be seen that the aromatic Tyr 74E is a crucial component,for it initiates the -rr- conduction process and provides the link to the surface and the cytochrome reductase. Yet in £ 550 this residue (structurally 86B = 74E) is an aliphatic leucine with no other possible aromatics to act as replacements. By knowing the sequence and structure of £ 550 it is thus possible to eliminate consideration of the Takano, et. al. mechanism. In a similar fashion, a definitive evaluation may be made of a different scheme proposed by Salemme, Kamen, and Kraut (1973). Strictly speaking, the mechanism as discussed here is applicable only - 142 - to £ 2 with some changes in detail when it is applied to£· The slight differences in the two versions were mostly motivated by structural features on oxidized c which have since been found to have been erroneously reported (Swanson, Trus, and Dickerson, personal communication). Oxidized£ now appears to be more like £z and so the £z version will be discussed as a unified mechanism. Here, the reduction of cytochrome is seen as the result of a simultaneous protonation and electron donation. The acceptance of the electron directly by the heme iron is seen as the result of the protonation of a nearby Ser 89p(= Thr 78E = Thr 98B) which changes an elaborate hydrogen bonding network Ser 89P - Tyr 52P - Tyr 70P (or Thr 78E Asn SZE - Tyr 67 E) which in turn has been stabilizing a bent ironmethionine sulfur bond. In £ 550 , the pivotal hydrogen bond donor, 5 2P = 5 2 E = 5 8B, is an is oleucine with no possibility of forming such bonds. Further, in £ 550 approximately the same amount of bonding may be seen in the iron sulfur bond as is seen in £z, but it is bent in exactly the opposite direction. Although £ 550 is not at as high a resolution, nor has it been refined crystallographically, the author 0 feels that the small amount of bending seen(0.35 A) is more the result of noise in the analysis than in structure. 4. 7. Alternative Mechanisms Before a concrete proposal for a unified mechanism of action for cytochrome £ can be made, consistent with the three known structural examples of cytochrome, the oxidized ~ 550 co-ordinates - 143 must be refined and the reduced structure investigated. However, some preliminary remarks are called for. The surface structural feature which is most strongly conserved is the collar or ring of positive lysines residues around the exposed edge of the heme. Attention was first directed toward this by Salemme, Kamen, and Kraut (1973) while Dickerson and co-workers were fostering the idea that the similarity was between a patch of lysines on the 11 right" or histidine side and a patch on the - 11 left 11 or methionine side. The severe localization of charge into a collar about the heme crevice in £ 550 favors the former view. Figure 4. 7 demonstrates in a schematic way the relevant collar grouping. The large insertions on the surface of £ 550 make certain regions on cytochrome £_ unlikely as binding sites for oxidase and reductase (bacterial or photosynthetic) in view of the partial cross -reactivity between £ 11 550 and£· The 11 right 11 side (residues 20B➔ 40B), the bottom" (residues 85B➔ 98B and 55B ➔ 70B), and the "back 11 (residues 120B➔ l35B) all contain large perturbations on the eukaryotic cytochrome surface. The "top 11 , the 1 'left 11 side, and the heme crevice remain as possible choices. The three known structures are consistent with the following mechanisms. The reductase and oxidase each utilize a different part of the lysine collar to recognize cytochrome £_ through protein-protein binding. However, each uses the same exposed heme edge as the actual pathway for electron migration, this being either through the heme -rr-orbital or directly to the iron. not the cause, of electron donation. Conformation changes are the result, Figure 4. 7. A front view of ~ 660 with all the lysine side chains. Note the grouping ·about the heme crevice. ...... 1-+:>- t-+:>- - 145 Certainly less elaborate than previous proposals, this is ahnost an "anti-mechanism" as far as most biochemists are concerned, for it says in essence that the cytochrome electron transfer is not terribly different than that of simple model system redox reactions. As such, mechanistic explanations must proceed to a higher level of detail involving orbital overlap, perturbed transition states, tunneling and/or other quantum mechanical considerations common to all redox reactions. This is not to say that the protein about cytochrome £ plays no role. Clearly, it performs certain unique, vital tasks. 1) It confers specificity upon cytochrome£ to the reductase and oxidase a vital control function for electron flow. 2) It adjusts the iron center redox potential to appropriate values. 3) It more or less isolates the iron center and this may prevent redox reactions with other physiological components which would lead to a short-circuit or by-pass of the normal chain. ( 4) The highly water -insoluble heme is made soluble by engulfing it in a hydrophilic matrix. Inevitably, but for some unknown reason, cytochrome£ is the only truly soluble protein in an extensive transport chain involving several hemoproteins. A total justification of these ideas would require at least as much space as the hundreds of papers which have appeared involving cytochrome _s. so the obvious restriction to a few salient experiments is made. Recently Margoliash and co-workers (Margoliash, et. al., 1973) have presented an investigation into the effects of the - 146 chemical modification of selected, single amino acid residues in cytochrome ~ which by their invariance among eukaryotic cytochromes has suggested a crucial function in cytochrome. In sum.mary, these workers find that purified, single-site derivatives at positions indicated in Table 4. 8 lead to a cytochrome which can function better with oxidase than with reductase, or the converse, better with reductase than with oxidas e. Their adopted explanation is that the reductas e and oxidase must utilize separate pathways through the cytochrome molecule, delineated by the surface locations of the crucial modified residues. As mentioned before, ~ 550 possesses a "derivatived'' group at the location corresponding to 74 in eukaryotic cytochrome, namely a leucine substituted for the otherwise invariant aromatic (tyrosine or phenylalanine). This suggests that it is not the aromaticity in itself at 74 which is mechanistically important but rather that the structural integrity of this region is important. The broad statement just made is best clarified by citing possibilities wherein 74 is directly recognized by reductas e as part of a binding site on cytochrome, or perhaps 74 contributes in maintaining some general conformation used in binding. The derivative experiments are consistent with the simpler view that they selectively interfere with separate oxidase or reductase binding sites without affecting the transfer event. the reported V max and Michaelis constants K m Arguments using for the derivatives cannot be used to decide between interference with binding or transfer - 147 TABLE 4. 8 SOME CHEMICAL MODIFICATIONS OF CYTOCHROME C Residue Derivative Tyr 74 Mono-iodination at ring position 3 or 5, exactly which is unknown. H I Oxidas e Reaction Reductas e Reaction normal defective normal1 auto oxidizable normal1 autooxidiz able norma11 autooxidizable Protein Trp 59 The production of 1-formyl Tryptophan oorotein H Tyr 67 /'C::,.... "'O Mono -nitration on the aromatic ring rRINOz ~ · Protein Met 80 Carboxymethylation Protein-CH -CH -S-CH -COO 2 2 2 Lys 13 Reacted with l -chloro-4-nitrobenzo 2' -oxa1 ', 3' diazole defective 40% activity - 148 (Table 8 - continued} Residue Oxidase Reaction Derivative Arg 13 Reacted with phenylglyoxal (Candida (~-C-C-H} II II Krusei 0 0 cytochrome £} Reductase Reaction defective 50% activity H I _..-::::N-C-0 Protein-NH-C ~ 0 ' CH-C-© ,, "NH-C-0/ I <b 1 only at the limit of very high substrate (cytochrome c} concentration. See Smith, Davies, and Nava (1974} for the qualifications this introduces. - 149 because it has been shown that oxidase/reductase reactions do not follow standard Michaelis -Menten kinetics (Smith, Davies, and Nava, 1974). The derivatives of ''internal" residues (67 and 59) may still produce binding abnormalities through a domino effect whereby internal alterations are transmitted to the surface conformation. A series of experiments related in principle to those of chemical modification are provided by Smith, et al. ( 1972) working with antibodies elicited against cytochrome .£· Purified, partially purified, and heterogeneous populations of Fab fragments were found which selectively inhibited either the oxidase or reductase reaction, leaving the other normal, by binding to specific (but unknown) surface areas on cytochrome.£· Here also it seems simplest to say that binding problems at or near the lysine collar produce inhibition. As with the modification work, the antibody results have been taken as indicating separate electron entry and exit pathways although the idea of just separate binding sites was mentioned. The motivation for adopting this view as opposed to a simple binding interference has been the feeling that antibody fragments are so large that for them to block binding without covering the transfer site ( e. g. the heme edge), the binding site and the transfer site must be widely separated. In rebuttal it must be pointed out that there are no data available on how much (or how little) of the Fab fragment is involved in immunological recognition and so a consideration of antibody size is an insufficient argument. A simple scheme as outlined in Figure 4. 8 would make use of the exposed heme edge for transfer and the lysine collar for binding and leave respecti v ely a hemisphere of surface for Figure 4. 8. Binding of transport enzymes to cytochrome £.· Cyt~chrome Reductase -cYs 14~Cy~ IT.·:-_._. . .... ~ Cytochrome Oxidase I-' U1 0 - 151 the anti-reductase antibody and anti-oxidase antibody to sit upon. To the extent of taking the heme edge as the electron path, this author concurs with Salemme, Kamen, and Kraut (1973) in citing the studies of Redfield and Kupta (1971) and Eaton and Hochstrasser (1967). These authors quote NMR and single crystal absorption spectra as indicating an asymmetric electron density distribution in the plane of the heme. Redfield and Kupta have further identified this as a concentration of electron density in the oxidized state such that the unpaired spin density is oriented toward the most exposed edge of the heme (toward pyrrole ring II, nearest Cys 17). The implication is that the oxidized structure placed the majority of density, to be paired with the incoming electron, at the site where donation occurs - the exposed heme edge. A variety of model reaction systems have been studied in which cytochrome is oxidized or reduced by simple, inorganic reagents (Ewall and Bennett, 1974; Sutin, 1973). Lambeth, et. al., 1973; Yandell, Fay, and Of particular interest are the experiments of Hodges, Holwenda, and Gray( 1974) on the kinetics of reduction of ferricyto2 chrome ~ by ferrous EDTA (Fe(EDTA) -), which show that this model system is clearly in line with typical reaction rates for other inorganic systems in which the transfer is 11 outer sphere 11 • By this it is meant that the reactant metal centers affect transfer without a transition state involving ligand substitution but by a simple proximity of centers with intact ligands . The point of relevancy is that the cytochrome heme edge could thus function by being in contact with an electron donor/ acceptor. - 152 Technically speaking, more elaborate proposals, such as the -rr-cloud conduction path of Dickerson, et. al., are also outer sphere. If the simple inorganic reactants make use of an electron pathway, (for example, heme edge), which is shown to be energetically, kinetically favorable, then why should physiological reactants employ a more intricate path? At stake is finding a low activation barrier reaction course ,then demanding that alternatives be shown to be clearly lower in barrier height. The situation is far from clear concerning a mechanism of cytochrome action consistent with everybody's experiments on this mo st ubiquitous of proteins. Critics can certainly off er counter - arguments to the situation as presented here. The hope for the future is that the known structures of cytochromes plus perhaps a bit more information from the forthcoming structures of reduced £ very small £s 51 550 and the of Pseudomas aeruginosa (R. Ahnassey, personal communication) as well as further sequence studies, will provide sufficient constraints to limit the pas sibilities for cytochrome mechanism and suggest conclusive dynamic experiments. system experiments are underway Model (e.g. by Gray, Bennett and others) as are new attempts at chemical modification in several laboratories (by Margoliash and Harbury). Perhaps in the near future the books can be closed on the story of a protein first observed in 1886 and purified by 1930. 4. 8. References Bartsch, R. G., Kakuno, T., Horio, T., and Kamen, M., (1971). I.:_ Biol. Chem. , 246, 448 9. - 153 - Dayhoff, M. 0., (1972). Atlas of Protein Sequence and Structure, National Biomedical Research Foundation, Washington, D. C. Dickerson, R. E., Takano, T., Eisenberg, D., Kallai, 0. B., Samson, L., Cooper, A., and Margoliash, E., (1971). _L Biol. Chem., 246, 1511. Dus, K., Sletten, K., and Kamen, M. D., ( 1968). _L Biol. Chem., 243, 5507. Eaton, W.A., andHochstrasser, R.M., (1967). J. Chem. Phys., 46, 2533. Ewall, R. X., and Bennett, L. E., (1974). J. Amer. Chem. Soc., 96, 940. Herriott, J. R., Watenpaugh, K. D., Sieker, L. C., and Jensen, L. H., (1973). _L Mol. Biol., 80, 423. Hodges, H. L., Holwenda, R. A., and Gray, H.B., ( 1974). J. Am. Chem. Soc., 96, 3132. Kassner, R. J., (1972). Proc. Nat. Acad. Sci., 69, 2263. Kendrew, J.C., Watson, H. C., Strandberg, B. E., Dickerson, R. E., Phillips, D. C., and Shore, V. C., (1961). Nature, 190, 666. Lambeth, D.0., Campbell, K.L., Zand, R., andPahner, G., (1973). _L Biol. Chem., 248, 8130. Margoliash, E., (1962) . .I_.:_ Biol. Chem., 237, 2161. Margoliash, E., Ferguson-Miller, S., Tulloss, J., Kang, C.H., Feinberg, B. A., Brautigan, D. L., and Morrison, M., (1973). Proc. Nat. Acad. Sci., 70, 3245. Redfield, A. G., and Gupta, R. K., ( 1971 ). In Cold Spring Harbor Symposia on Quantitative Biology, XXXVI, 405. Salemme, F. R., Freer, S. T., Xuong, Ng. H., Alden, R. A., and Kraut, J., (1973). _L Biol. Chem., 248, 3910. Salemme, F.R., Kraut, J., and Kamen, M.D., (1973). J. Biol. Chem., 248, 7701. Smith, L., Davies, H. C., Reichlin, M., and Margoliash, E., (1972). I.:_ Biol. Chem., 248, 237. Smith, L., Davies, H. C., and Nava, M., (1974). J. Biol. Chem., 249, 2904. - 154 - Ta~ano, T., Kallai, 0. B., Swanson, R., and Dickerson, R. E., (1973). I:_ Biol. Chem., 248, 5234. Timkovich, R., and Dickerson, R. E., (1973). L Mol. Biol., 11., 39. Yandell, J. K., Fay, D. P., and Sutin, N., (1973). J. Amer. Chem. Soc., 95, 1131. - 155 PROPOSITION 1 SOME NOTES ON THE FIGURE OF MERIT AS A COMPARATIVE STATISTIC The figure of merit is an often quoted statistic used to gauge the quality of a protein crystallographic structure determination based on isomorphous replacement. It derives its existence from the error analysis for isomorphous replacement by Blow and Crick (1959). It is empirically well known that there are correlations between the mean figure of merit and such parameters as the root mean square lack of closure and the number of isomorphs in the analysis. While the root mean square lack of closure certainly reflects the quality of the data and such critical features as the degree of isomorphism, the number of isomorphs can be a pure statistical effect quality of the phase determinations. independent of the actual In either case it is not obvious what will be the analytical result upon mean figure of merit as one changes the root mean square lack of closure or the number of heavy atom derivatives. It is the working hypothesis of these notes that, a~ currently used, the figure of merit is . a worthless comparative measure of the accuracy of a structure determination. It will be shown that statistical effects which arise from the number of is omorphs and from potential isomorphic relationships among the sites of heavy atom binding negate the usefulness of this statistic when comparing different structure determinations, although for a single structure, it continues to aid in assessing internal error. A modification in the calculation of the - 156 mean figure of merit will be proposed to normalize the statistic so as to make it independent of the number of derivatives available in any particular analysis. This treatment follows the theory of Blow and Crick (1959). Notation employed is consistent with that adopted by Dickerson, Kendrew, and Strandberg ( 1961 ). Definitions: The native structure factor magnitude The derivative magnitude The magnitude of the heavy atom at~mic scattering vector P(c.p) The probability that c.p is the correct native protein phase. pd£ The probability density function for P(c.p) c.p The native protein phase angle for Fp Maximum likelihood phase angle The Blow-Crick "best" phase angle m The figure of merit An arithmetic average or expectation value E ·Root mean square lack of closure j Running index for heavy atom derivatives Ci Heavy atom phase n The number of derivatives SIR The single isomorphous replacement solution to the phase problem Two hypothetical limiting cases will be adopted which allow a closed, analytical expression for the mean figure of merit. Real cases can be considered as some combination of these two opposite - 157 extremes. This is not an attempt to deduce expressions for (m) which would be usable for calculation, but the resultant approximations will indicate the relationsnips between m and various parameters. It is a prerequisite to show m = (cos6) where 6 = cp-cpb ( 1) Let f (x) =pd£ of any random variable x; y =H (x) be some function defined on x. Then by elementary probability theory (y) = JH(x)f(x)dx (2) X For the present case, this says (cos 6) = J (3) cos 6 P(cp)dcp cp But cos 6 = cos (cp - cpb) = cos cp cos cpb + sin cp sinc.pb (4) so ( cos 6) = cos cpb J cos cp P(cp)dcp + sincpb cp J sin cp P(c.p)ckp cp (5) =m When P(c.p) is a combined pd£ determined over several derivatives, it empirically comes to resemble closely a Gaussian curve centered about c.pm and falling quickly to zero away from the centroid. It is then called a uni-modal pd£. As SUille that P(c.p ) can be analytically written as a true Gaussian curve about c.p c.p m m given by - 180° ~CD ~ cp + 180° · m (6) - 158 Take 6 = cp- cpm. Now approximate P(cp) by ,. - 6 2 / 2 cr 2 " P(cp) = P(cp) = Ne the true P(cp) falls off quickly from cp where so both the analytical m assumption and its approximation are realistic. The latter is intro- duced solely to make subsequent manipulations easy. ,. (7) -co<cp <+co N can be readily l. evaluated to be N = 1 / (2TI) 2 cr. Then m = (cos6) "J +co 2 2 cos6e -6 /2cr dcp =N -CO co = 2Nl cos6e- 6212 cr 2 d(6) (8) 0 The integral can be evaluated in closed form to imply m=(cos6)=e -cr2 /2 (9) From this it can be seen that anything which decrease cr increases m. The empirically based approximation for the pd£ can be roughly equated to the formal Blow-Crick definition of P(cp) . ( 10) Note that N :/:. N, since - 2 . 2] -I:l e: /crm )/2Ej P(cp m ) = Ne .1 Now at values of cp :/:. cpm Assume equal 11 centric quality" derivatives. ,. = N ( 11) - 159 - E = E. J ( 13) for all j This is a good approximation, for from the M. denitrificans analysis 2 E(PtCI4 - ) = 74 } E(Pt(CN) 4 2 -) (14) = 80 E(UO 2 +) = 73 2 then ( 15) Since <l> = l;, e:~ J n __, J ( 16) j cl = ~.z E 2 / n ( <e J~) cp - <e J~) cpm ) ( 17) This allows a semi-quantitative prediction of various effects upon m and in turn upon <m). Now <eh is influenced by the quality of the data J and the degree of isomorphism and it will not change drastically as the nmnber of derivatives changes. Hence, m increases exponentially as 1/n decreases. ( 18) evaluated at some cp . The dependence of m upon n suggests that the pdf should be rewritten to remove this purely statistical effect; <m) should be normalized so that it depends only upon the phase agreement. simplest approach would be to take The - 160 - ( 19) then That this would remove the n dependence is a testable result. ~ 550 In the analysis certain heavy atom derivatives were measured independently, but as later work indicated similar substitution, equivalent (but not identical) data sets were averaged. averaged sets, 3 (m) = 0. 59. After refinement over the three A new refinement was performed whereby the redundant F PH measurements were treated as independent derivatives. Then Such an approach has been frequently used on other structures. 7 (m)= 0. 741. If the discussed dependency is correct, then (21) but 7 (m) 7 = 0. 133 and 3 3 (m) = 0. 205. In other words, the (m) did not increase enough to be explained by a 1 /n dependency. lies in one of the assumptions. The failure For chemically distinct added deriva- tives, the form of P(cp) will more closely resemble a Gaussian distribution, since the ambiguous pair to the correct protein phase will not be correlated among distinct chemical derivatives and hence will average in the product to zero. For a redundant data set, where the added set is either a strict remeasurement or is similar to some previous set, the added set can correlate with an "ambiguous pair", leading to a bi-modal distribution. - 161 - Derivative 1 Derivative 3 Derivative 2 P(cp) P(cp) P(cp} X P(cp X cp = cp cp cp This is carried to an extreme in the SIR case where a BlowCrick treatment leads to a pdf containing two equal peaks separated by the difference between cp and the incorrect alternate angle. From a consideration of cpb as the centroid angle, it is obvious that cpb is Let .6 q) = 2 (cp b - cp t the mid-angle of the two. rue ). It will be shown that in the SIR case, m is largely determined by .6 1cp. This suggests that for a pdf with a large bi-modal character, m is a function of .6cp. Let P(cp) = p for cp= cp 1 = 1 -p = q for cp=cp (22) 2 elsewhere = 0 as the limiting case of a bi-modal distribution. Then, as in Dickerson, Kendrew, and Strandberg ( 1961) m cos cpb = p cos cp 1 + q cos cp 2 m sin cpb = p sincp Solution form and taking .6cp=cp m = fp 2 +q 2 1 - cp 2 l (23) + q sincp 2 1 gives .1. - 2pq cos ( 180 ° - .6cp)} 2 (24) If one calls p a vector with magnitude p and phase cp and does the 1 same for q, then m = l p+q j. Note that if Ip l = ½ (the SIR case), then - 162 this reduces to (25) m = cos (6q)/2) Any bi-modal distribution can be treated as arising from a single SIR phase determination modulated by some not severe weighting function. so that Since, for SIR, it can be shown that 2 2 2 - I ( F PH - (F p + f ) ) q) = et± cos 2F p f (26) 2 2 2 - 1 (F PH - (F P + f )) 2 - 2 cos 2F f p (27) ~_ It follows that in this limiting bi-modal case, m will be explicitly independent of the number of derivatives. Take any actual m as approximately due to a combination of the limiting uni-modal and bi-modal cases, (28) where k , k are weighting terms describing the relative contributions. 2 1 th Attempts to normalize (rn) by raising to the n power are now seen as incorrect since only the first term should be so treated. It is proposed that the most reasonable form for the normalized (m) would be given by computing individual m's from an adjusted pd£ -~k.,)[I:(e:1 /2E.,)] P(q)) = e -' -' (29) where the modifying constants k. represent the fraction of electrons J in derivative j which are unique to the entire set of heavy atom electrons. This is best explained by means of the following example. - 163 - Example Calculation of k Derivative Salt A k 100 50 25 175 100+50+15=165 165/175 125 60 10 195 25+10+10=45 45/195 Effective electrons 1 2 3 1' 2' Salt B Unique Scattering electrons Sites Occupied 3' Salt C 4 5 75 75 150 75+15=90 90/150 Salt D 5' 6 100 60 25+60=85 85/160 Ro n=4 Under such a system, for example, a determination using six truly independent derivatives would have its pd£ adjusted by a factor of > k. = l+l+l+l+l+l = 6 to counter the implicit 1/6 in the summation .:. . . . ., J j \' 2 2 ,' ' J j J > s. /2E .. Treating two redundant measurements of the same derivative as pseudo-distinct derivatives does not have this 1/n factor and the summation ) k. = 1+0 = 1 will account for this. __J 1 j It is to be recalled that this is not necessarily a proposal to calculate m for use in figure of merit weighted Fourier analyses. is at this stage intended as a way to re- calculate m It prior to reporting - 164 - (m) so as to make the statistic comparable over different protein determinations. Additional work is clearly necessary on this proposal. It should be examined theoretically to insure that it reasonably accounts for the uni-modal and bi-modal contribution to m. It would be most interesting to calculate (m) according to this outline for a variety of protein structure determinations then compare the resulting predictions of quality with other factors which access reliability, such as R-factors, percentage E's, and electron density background-tonoise ratios. References Blow, D. M., and Crick, F. H. C., (1959). Acta Cryst., g, 794. Dickerson, R. E., Kendrew, J.C., and Strandberg, B. E., (1961). in Computing Methods and the Phase Problem in X-ray Crystal Analysis, R. Pepinsky, J.M. Robertson, and J.C. Speaknlan, eds., Pergamon Press, New York, p. 236.