key: cord-0050918-snxyn6kh
authors: nan
title: Symposia
date: 1987-06-03
journal: Protein Eng
DOI: 10.1093/protein/1.3.225
sha: 4664a2453c3ea64ac7138f273da16a5086ce0bca
doc_id: 50918
cord_uid: snxyn6kh

nan

marupulated by established relationships to yield the entropic and enthalpic components. The alternative approach is to consider all the atoms (or at least many of them) explicitly as polarizable dipoles This removes the need to use a dielectric constant and the energy so calculated is a conformational potential, which may be man~pulated by standard statistical mechanical procedures to yield the enthalpy, entropy and free energy associated with a given change. In practice these approaches are often mixed, and compromise models are used with the result that what is calculated 1s often difficult to assign as a thermodynamic entlty. By use of the first model it w i l l be shown that the thermodynamics of ionpair formation in proteins is far from straightforward and that the driving force (enthalp~c or entropic) depends upon the position within the globule relative to the solvent.

Chris Sander Biocornputing, EMBL, 0-6900 Heldelberg, FRG During a recent 3-week workshop held at the EMBL, 24 researchers and students joined in an attempt at & now protein des~gn using available computer tools. The six groups at the workshop produced these p r e m designs: two different TIM barrels, followmg the topological motif of triose phosphate isomerase; two alfl protelns, i.e. proteins with an alternating sequence of the structural elements a-helidfl-strand, one of them a redesign of flavcxioxin, with the original nucleotide bmding site replaced by a binding site for a fluorescent dye; two a-helical bundles, w~th four helices each, one of them designed to bind calcium, the other designed to bmd copper ions. The design strategy used at the workshop represents a departure from classical protein-folding theory. Classically, the main objective is to calculate the correct threedimensional structure for a given amino acid sequence. Here, in contrast, the problem was restated: given a protein structure, in the form of its basic topology of secondary structure elements, design an amino acid sequence which wlll give rise to that structure as the protein folds. The results are in the form of a list of the new amino ac~d sequence arad coordinates of the corresponding protein structure. The design workshop has identified the deficiencies in our present tools and marks the beguuung of an experimentalltheoretical cycle of design and test of newly created proteins. Tobacco mosaic vlrus (TMV) is the best studied example of a self-aggregating system. Sequence data are available for the a t proteins of six tobamoviruses, with homologies rangng from 33 to 82 %; atormc coord~nates are known for tobacco mosaic vlrus wild-type. The constraints on the overall size and shape of the protein subunit and on the character of those regions of the s u b unit surface involved in quaternary structures should be reflected in the nature and pattern of acceptable amino acid substitutions. A significant spatial relationship has been found between groups of residues with identical amino acid substitution patterns. This strongly suggests that after mutation they have not become stabilized independently of each other, and that their locatlon is linked to a particular function, at least in viruses identical to the disulphide for these residues. The most conserved feature of TMV is the RNA binding regon. Core residues are conserved in all viruses or show mutations complementary in volume. The specificity of inter-subunit contacts 1s ach~eved in different ways In the three more distantly related viruses. The strategy used here for detectlng coordinated substitut~ons has worked well within the tobarnov~rus family where the protem has extensive quaternary stntdure. If this approach can be applied equally successfully to other families of protems, it could contribute to the understanding of protein folding and interactions.

J.C.Swaffield, 1.J.Purvls and A.J.P.Brown

Many highly expressed genes from Saccharomyccs cere~siae show a strong bias in their choice of codons for the 20 amino acids, this c d o n bias correlating strongly with the relative abundance of the iso-accepting tRNAs. (The set of preferred codons varies between orgarusrns.) When 'rarely used' cdons are clustered withn a gene, a pause in the rate of translation is predicted that would result in the accumulation of nascent polypeptide chains of a discrete length. As pmtems are synthesized from the N terminus and initial folding reactions probably occur before translation is completed, it is possible that translational pauses influence the folding of some proteins by allowing regions of the growing polypeptide chain to fold correctly before C-terrninal regions are synthesized. A major problem in relating potential translational pauses to protein foldmg is the lack of knowledge about the tertiarylquaternary structures of many proteins for which the gene sequence is available. However, there is a tight correlation between the presence of potential translational pauses and the interdomain regions of the arom multifunctional enzyme in S. cerevisiae. Potential translational pauses have been observed in the genes for other multifunctional enzymes in yeast (e.g. TRF9, TRPS), but m these cases the pauses do not seem to lie in ~nterdomain reg~ons.

Proteins are well designed for their functions. They may be rigid or flexlble to various degrees as required for optimal performance. Flexibility at the level of amino acid sidechains occurs universally and may be mprtant for some fumons. Large-scale flexibdity where large parts of a proteln rearrange or move coherently are particularly interesting and will be discussed here. We may differentiate behveen different categories of flexibility (Bennett and Huber, 1984) . Order-disorder transitions of domains and domain motions. The domains may be flexibly linked to allow rather unrestricted motion or the motion may be constrained to certain mcdes by hinges. The connecting segments and the hinges show characteristic structural features. The followmg examples w~ll dlustrate various aspects. Small proteinase inhibltors are essentially rigid molecules to provide tight complementary binding to their cognate protease (Huber and Bode, 1978) . The large plasma inhibitors, however, exhibit large conformational changes upon interaction with proteases, probably for regulatory purposes (Lobermam et al., 1984) . The pancreatic wine proteases exhibit a disorder -order transition of their activation d o m between proenzyme and enzyme forms as a means to regulate enzymic activity (Huber and Bode, 198) . Immunoglobulms show rather unrestricted and also hinged domain motlons In d~fferent parts of the molecule, probably to allow cross-linlung of antigens (Huber, 1984; Huber et al., 1976) .

Citrate synthase adopts open and closed forms by a hinged domain motion to bind substrates and release products and to perform the catalytic condensation reaction respectively (Rernington et al., 1982) . In the multi-enzyme complex riboflawn synthase in which two consecutive enzymic reactions are catalysed by two distinct enzymes restricted motions by engaging one enzyme in a capsid formed by the other, and directed substrate delivery seems to play an important role (Ladenstein et al., 1986) . In contrast to the previous examples, motion would be deleterious to function in the light-harvesting complexes (Schirmer et al., 1985) and the reaction centres involved in the photosynthetic light reactions (Deisenhofer et al., 1984, 1985) . These are huge protein complexes whlch serve as matrixes to hold the pigments active in light absorption and electron conduction. Motion would deactivate the exclted functional states of the pigments and destroy the proper geometric arrangement.

Taiji Imoto Faculty of Pharmaeutical Sciences, Kyushu Univerity 62, Maidashi, Fukuoka 812, Japan Lysozyme was variously modified and the stabilities of the derivatives were determined with thermal denaturation experiments. Contributions of salt bridge, hydrophobic interaction and cross-linkage were evaluated. The stabilities against proteolysis were also considered. For the latter stability it might be important to depress the rate of unfolding, i.e. to stabilize native conformation. As a rule, salt bridges and hydrophobic interactions stabilize native conformation and cross-linkages destabilize denatured conformation. However, cross-linkages are apt to introduce strain in native conformation and only suitable lengths of cross-linkages can stabilize protein. The stabilization was shown to be generally effective at improving functionality of proteins. Regeneration of disulphide bridges was examined by employing several chemically cross-linked lysozymes. The derivative cross-linked between Lys-13 and Leu-129 showed 2.3 times accelerated regeneration rate compared with that of unmodified lysozyme. Mutation of Ala-31 to Val completely hindered the regeneration of lysozyme into active conformation. Such a minor mutation can lead to critical damage in the folding of proteins which is an indispensable process in genetic engineering of proteins. Disulphide linkages, electrostatic interactions and hydrophobic contacts that increase the stability of subtilisin BPN' were first separately and then collectively introduced into the protein structure through the use of site-directed mutagenesis. The disulphide linkages were designed with the aid of a computer program which scored various sites according to the amount of distortion that a disulphide bond would introduce into a 1.3 A resolution Xray model of wild-type subtilisin BPN'. The electrostatic changes were designed to increase the Coulombic interaction between the protein and divalent metal cations at an already existing calcium binding site. The hydrophobic changes were designed to introduce aliphatic residues at positions buried within the tertiary structure. The thermodynamics of unfolding for all these new variants of subtilisin were studied by differential scanning calorimetry. Moreover, detailed structural analysis through X-ray crystallography was possible for many of these more stable proteins. Finally, in putting several of these stabilizing changes together simultaneously in one subtilisin molecule, it was possible to show that the AG of unfolding due to each individual change is, to a first approximation, additive, so that the total change is equal to the sum of the parts. We are currently studying an hierarchical approach to protein folding (see abstract, this meeting, Somorjai and Narang). Towards this goal, we have chosen T4 lysozyme as a model enzyme. T4 lysozyme DNA (495 bp) was redesigned to contain several endonuclease restriction sites and the most efficient codon se-quence for production in Escherichia colt. It was chemically synthesized by a multiple fragment hgation procedure with a linearized plasmid followed by transformation. High-level expression of T4 lysozyme DNA was obtained under /ac-promoter. T4 lysozyme was purified and crystallized under conditions very similar to those published by Remington et al. (1978) The unit cell appears to be very similar to that of the wild-type crystals. We have measured X-ray diffraction data to a resolution of 2.9 A and we are using the T4 lysozyme coordinates from the Brookhaven Protein Data Bank to generate a structure of the synthetically generated molecule. Progress on this aspect of the project, as well as any structural differences identified between the two molecules, will be reported.

Reference Remington et al (1978) J Mol Biol , 118, 8-98. 13. The in vitro and in vivo stability of the nucleoprotein complexes formed between a thermosensitive VP1 and SV40 chromatin Veronica Blasquez, Maria Behm, Henry Lowman and Minou Bina Chemistry Department, Purdue University, West Lafayette, IN 47907, USA We have studied the SV40 assembly mutant tsB265 as a model system to gain insight into how the stability of structural proteins could influence protein-protein and protein-DNA interactions, and the assembly of nucleoprotein complexes formed between chromatin and proteins. The tsB mutations map to the gene coding for the major virion protein (VP1). Sequencing of the DNA fragment which spans the tsB265 mutation revealed that a single point mutation results in the substitution of Thr for Ala at amino acid residue 71 in VP1. We find that the virions assembled in the tsB265-infected cells at the permissive temperature (33 °Q remain stable when incubated in vitro at the non-permissive temperature (40°C). In contrast, we can readily detect the effects of temperature perturbations on VP1 stability in vivo. The protein undergoes reversible changes in conformation when the infected cells are incubated at temperatures ranging from 33 to 40°C. However, the results of pulse-chase kinetic studies and temperature-shift experiments indicate that, in vivo, thermally induced conformational changes in the protein cannot be accommodated after VP1 has polymerized around the minichromosome. Thus, nucleoprotein complexes assembled at a given temperature must first dissociate and go through a 'chromatin pathway' before being converted into another type of nucleoprotein complex.

14. Atomic structure of thymidylate synthetase: target for rational drug design R.M.Stroud Department of Biochemistry and Biophysics, University of San Francisco, CA 94143, USA The atomic structure of thymidylate synthase from Lactobacillus casei was determined at 3 A resolution. The native enzyme is a dimer of identical subunits. The dimer interface is formed by an unusual association between five-stranded /3-sheets present in each monomer. Comparison of known sequences with the L casei structure suggests that they all have a common core structure around which loops are inserted or deleted in different sequences. Residues from both subunits contribute to each active site. Two arginine side-chains can contribute to binding phosphate on the substrate. The side-chains of several conserved amino acids can account for other determinants of substrate binding.

15. Effect of deleting functional domains, cysteine residues or salt bridges necessary for temperature stability on the folding and transport of haemagglutinin Mark S.Segal, Jackie Bye, Jianhua Zhang, Joe Sambrook and Mary-Jane Gething Department of Biochemistry, University of Texas Health Science Center, Dallas, TX 75235, USA Attainment of correct trimeric structure is required for export of influenza haemaglutinin (HA) from the endoplasmic reticulum (Gething et al., 1986) . The close linkage between folding, oligomerization and transport to the Golgi apparatus raises the questions of how mammalian cells discriminate between folded and unfolded proteins and how and in what compartments are unfolded proteins retained The cell might use a general feature of unfolded proteins, such as exposed hydrophobic surfaces, or a specific signal, such as unpaired, reduced cysteine residues, to recognize a malfolded protein. To test these hypotheses three sets of HA mutants have been made. One that lacks specific cysteine residues, another with whole functional domains deleted. The third was made using the crystallographic structure of HA to choose amino acids whose alteration would create mutants which are temperature sensitive. Following expression in mammalian cells, we plan to unfold the mutant proteins in the successive compartments of the secretory pathway for transferring the cells to a higher temperature at different times after their synthesis. The analysis of the folding, assembly and intracellular transfer of the variant HA molecules will be presented.

Ligand binding to the active site of haemoglobin is an ideal modelsystem to study the formation of enzyme-substrate complexes. Using site-directed mutagenesis, we have replaced the two highly conserved residues, Val-Ell and His-E7, in the ligand binding pocket of the /3-subunits with several other amino acids. We have studied the properties of the mutants by X-ray crystallography, equilibrium and kinetic ligand binding and resonance Raman spectroscopy. The rate of carbon monoxide binding to deoxyHb was found to increase with decreasing size of residues at E-7 nposia substitution of Lys-99 (the site of covalent adenylation in the first step of the enzyme mechanism) by Asn or His leads to inactivity in adenylation as well as in the joining reaction, whereas an Arg-99 mutant retains activity. A total synthesis approach has been used to construct the gene for bovine caltrin (seminalplasmin), a 47 amino acid protein with anti-fertility and anti-bacterial activity. The gene has been expressed in Escherichia coli as several different fusion proteins with a view to efficient production of caltrin for structural and mutagenesis studies. 21. Protein engineering by chemical means?

R.E.Offord Dipartement de Biochimie medicate, Centre me 'dical universitaire, 9 avenue de Champel, 1211 Genive 4, Switzerland Of two recent reviews, one gave a definition of 'protein engineering' that explicitly included chemical methods, whilst the other explicitly excluded them. Questions of methodology and terminology seem of limited importance beside the realization that, together, chemical and recombinant techniques have put within our reach areas of exploration that seemed unthinkably remote a very few years ago. Concrete examples will be given from the author's laboratory of the preparation by chemical and recombinant methods of proteins of modified structure, and of their use to address problems of academic and clinical relevance.

Biochemisches Institut der Universitdt Zurich, CH-8057 THrich, Switzerland Based on secondary structure prediction rules and model building, a 24-residue DDT-binding polypeptide (DBP) was designed and synthesized. Preliminary evidence for the proposed folding [a four-stranded antiparallel /3-pleated sheet, Moser et al. (1983) ] comes from CD studies and from the formation of high mol. wt aggregates in dilute acid. In 0.05 M NHtHCOyethanol (5:6, v/v) the dissociation constant of the peptide-DDT complex was -10~6 M, > 1000-fold lower than that of a BSA-DDT complex. The location of the DDT-binding site of the artificial polypeptide in the model was largely confirmed by the synthesis of DBP analogues. In the presence of DBP, the rate of DDT degradation by a hemin-cysteine cytochrome P-450 model system was increased by a factor of three. In collaboration with E.-L. Winnacker and R.Mertz (Genzentrum, Martinsried, FRG) the synthetic gene of the designed DBP was cloned in Escherichia coli using plasmid pUR291 and was expressed as a /3-galactosidase-DBP fusion protein. The biological properties of the DBP + strain are under investigation. Moser et al (1983) FEBS Lett , 157, 247-251 23. Determination of three-dimensional structures of proteins in solution G.M.Clore and A.M.Gronenborn

A prerequisite for successful protein engineering is a knowledge of the three-dimensional structure of the protein under investigation. To this end use may be made of two techniques: X-ray crystallography in the solid state and n.m.r spectroscopy in solution. The determination of three-dimensional structures of proteins in solution using n.m.r. spectroscopy comprises three stages: (i) the assignment of proton resonances by two-dimensional techniques to demonstrate through-bond and through-space connectivities; (11) the determination of a large number of short (< 5 A) interproton distances using nuclear Overhauser effect (NOE) measurements; and (lii) the determination of the three-dimensional structure on the basis of these distances. Our approach for step (lii) has involved the use of restrained molecular dynamics and distance geometry calculations. The principles of the restrained dynamics approach will be illustrated for model calculations on crambin (Brunger et al., 1986; Clore et al., 1986a) and the determination of the three-dimensional structures of the following proteins in solution will be presented: purothionin (Clore et al., 1986b ), phoratoxin (Clore et al., 1987a ), hirudin (Clore et al., 1987b , the globular domain of histone H5 (Zarbock et al., 1986) , growth hormone releasing factor (Clore et al., 1986c) , secretin and potato carboxypeptidase inhibitor. The lack of detailed structural information on the nature of antibody combining sites, particularly in the complex environment with antigen, has resulted in the widespread application of modelling techniques to the study of these systems. An understanding of the nature of the specific interactions between antibody and antigen is fundamentally important if one is to hope to design antibodies with a given specificity and affinity for a particular antigen. The latter is perhaps the ultimate goal of such studies, and of immense importance in clinical application. The modelling of a monoclonal antibody, Gloop2 (class IgG), on the basis of DN A sequence information and a structural database of known immunoglobulin structures, has yielded a preliminary model for the structure of the 'anti-loop' binding site (de la Paz et al., 1986) .

The docking of this model with the loop region of HEL (determined crystallographically at 1.6 A resolution), and subsequent energy minimization of the resulting complex with GROMOS (van Gunsteren, Groningen), has given a model for the antibodyantigen complex which may be used, in parallel with site-directed mutagenesis studies on the antibody itself, to probe the nature of the interactions between the two molecules. Prediction of the effects of single and double mutations engineered within the antibody combining site have been compared directly with experiment, and the results point to important considerations for the design of any future mutagenesis experiments. Concurrent with the study of the 'anti-loop' monoclonals, modelling of a series of antibodies directed against the influenza virus (A/PR/8/34) has been undertaken. In the first instance, a pair of antibodies which show markedly different affinities for the antigenic variant CaT3 (H28-A2: high affinity; H37-4O: low affinity) have been modelled. Docking of these antibodies with the Ca epitope of the protein suggests one possible origin of the observed difference. In addition, a number of possible amino acid changes within the combining site region of one of the antibodies, H37-4O, which the model would predict to result in improved binding to the antigen, are discussed.

A set of amber suppressor genes has been constructed for use in amino acid substitution studies as well as protein engineering. These genes were assembled from four to six oligonucleotides, which were annealed and ligated in vitro into a plasmid; they are expressed constitutively from a synthetic promoter. We have constructed high-efficiency suppressors that specifically insert alanine, arginine, cysteine, histidine and phenylalanine in response to an amber codon. Among the additional amber suppressors that we have synthesized, some that were inefficient or inserted an unwanted amino acid have been improved in various ways, while others are still being tested for efficiency and specificity of insertion. We are examining the effect of a large series of amino acid substitutions in specific proteins by creating amber mutations at known positions and expressing these mutated genes in the Escherichia coli strains that produce the suppressor tRNAs. The lac repressor is one object of our study and we have monitored the effects of > 1000 amino acid substitutions in the repressor by employing all of the existing suppressors, including those constructed in this study. The 1.5 A X-ray structure of Staphylococcal nuclease suggests that glutamate-43 is appropriately positioned to act as a general base in facilitating the attack of a water molecule on an internucleotidic phosphodiester bond of a substrate. Site-directed mutagenesis has been used to mutate residue 43 to aspartate, glutamine, asparagine, serine and alanine such that the importance of glutamate-43 as a general basic catalyst might be assessed.

Although the values for V^IKm for the mutant enzymes are reduced from 1400-fold (aspartate) to 5000-fold (glutamine, asparagine, serine and alanine) relative to that measured for the wild-type enzyme, these rate reductions cannot be used to quantitate the importance of general basic catalysis. 'H n.m.r. spectroscopy reveals that the conformations of the mutant enzymes differ from that of the wild-type enzyme: the chemical shifts of and the intensities of the nuclear Overhauser correlations between the aromatic and upfield shifted methyl resonances are altered by the presence of the point mutation. Isotopic enrichment of the aromatic and aliphatic amino acid residues with 2 H, I3 C and I5 N will allow the locations and magnitudes of the conformational changes to be determined. These 'H n.m.r. studies complement the high-resolution X-ray studies of these mutants that are in progress in Professor Eaton Lattman's laboratory (Johns Hopkins University School of Medicine, Baltimore, MD). A kinetic scheme has been obtained for the Escherichia coli dihydrofolate reductase that predicts steady-state kinetic parameters and under a variety of substrate concentrations and pH. This scheme was derived by measuring ligand association/dissociation and the pre-steady-state transient using stopped-flow fluorescence and absorbance spectroscopy. A key feature is that during turnover, product dissociation follows a preferred pathway in which tetrahydrofolate (H4F) dissociates after NADPH replaces NADP + in the ternary complex. This step, H4F dissociation from the E-NADPH H4F ternary complex, is rate-limiting for turnover at low pH where k^ = k^. The rate constant for hydride transfer from NADPH to dihydrofolate (H 2 F) measured from the pre-steady-state transient has a deuterium isotope effect of 3; is rapid (fc^ = 950 s~'); essentially irreversible (K^ = 17 000); and pH dependent, pK a = 6.5 reflecting lonization of an Asp-27 in the active site. Three mutants in the folate binding site, Phe-31 -Tyr, Phe-31 -Val and Leu-54 -Gly, have been analysed according to this scheme. The k^ for both Try-31 and Val-31 enzymes have been increased owing to an increase in the R»F dissociation rate constants. This increase is partially compensated for by a reduction in k hyd , but this step is not sufficiently decreased so as to influence fc^,. The Gly-54 mutation decreases the rate of hydride transfer sufficiently to cause this step now to become limiting. 29. Importance of the catalytic His-64 in amide bond hydrolysis by subtilisin P.Carter and J.A.Wells Department of Biocatalysis, Genentech, Inc., 460 Port San Bruno Boulevard, San Francisco, CA 94080, USA Both bacterial and mammalian senne proteases contain an invariant catalytic triad consisting of Ser-His-Asp. The catalytic triad has stimuated much debate over the role of these residues in facilitating the nucleophilic attack by the serine OG on the scissile peptide bond. In all these mechanisms the catalytic histidine is proposed to function first as a general base in accepting the proton from the OG of the catalytic serine, and then as a general acid by donating it to the amine leaving group. We have estimated the contribution of this function to catalysis by mutating His-64 -Ala in the cloned subtilisin gene from Bacillus amyloliquefaciens. Depending upon the substrate assayed there is a drop of 3 x 10 6 and in k^JK^ compared with wild-type subtilisin. Most of the drop in /C^JKM appears in the k^ term, although effects appear in K M as well. Under comparable conditions, when the concentration of substrate is much below K M and at pH 8.6, wild-type subtilisin imparts a 10 M -to 10 l2 -fold increase in hydrolytic rate over the non-enzymatic reaction. Thus the His-64 -Ala mutation causes a massive drop in the catalytic efficiency to a level that is only 10 6 -to 10 7 -fold above the nonenzymatic reaction. Ribonuclease Tl (RNase Tl) specifically hydrolyses the phosphodiester linkages of guanosine 3'-phosphate of single-stranded RNA and the cleavage occurs in a two-step mechanism. It is considered that in the first step Glu-58 abstracts a proton from 2'-OH, and His-40 or His-92 adds a proton to O-5' of ribose. We succeeded in expressing the chemically synthesized genes for RNase Tl and its several mutants at base-recognition sites in Escherichia coli and reported the structure-function relationship of this enzyme (Dcehara et at., 1986) . In this paper, we changed Glu-58, His-40 and His-92 to alanine and analysed the activity of these mutant enzymes in order to clarify the role of catalytic residues. The Ala-58 mutant still retained slight activity but both His^40 and His-92 are indispensable for RNase Tl activity. We propose a new reaction mechanism, in which His-40 abstracts a proton from 2'-OH and His-92 protonates O-5' of ribose while Glu-58 enhances the basicity of His-40.

Ikehara.M et at (1986) Proc Natl Acad. Sci USA, 83, 4695 

Rudi Glockshuber, Jorg Stadlmuller, Anne Skerra and Andreas Pliickthun Genzentrum der UniversiUit Munchen, Max-Planck-Institut fur Biochemie, Am Klopferspitz, FRG The genes encoding the variable domains (V H and VJ of the phosphorylchohne binding antibody McPC603, whose crystal structure with and without antigen bound is known, were obtained by DNA synthesis. In addition, we constructed genes also encoding the appropriate constant domains of each chain in order to express directly the exact F^ fragment whose crystal structure has been determined. The synthetic genes were designed for facile replacement of gene fragments (e.g. the hypervanable loops) as well as for incorporating current knowledge about efficent expression. We have investigated purifications from bacteria] expression systems and are comparing their efficiency in obtaining large amounts of protein. The essence of antibody architecture is a framework of fairly constant residues and hypervanable loops or complementarity determining regions that determine specificity, i.e. contain the antigen recognition sequences to a great number of antigens. The particularly well-studied antibody combining site of McPC603 is used by us as a model system for quantitatively investigating factors that contribute to efficient hapten binding, subunit interactions, as well as for the potential of stabilizing a transition state through the controlled modification of the protein. We are attempting to delineate protein contributions to catalysis by making an almost-catalyst (a transition state binding antibody) perform a suitable hydrolysis reaction. Histidine residues hydrogen-bonded to aspartate have been identified at the active sites of malate and lactate dehydrogenases. From X-ray studies it has been suggested that the amino acid pair serves as a proton relay during catalysis and maintains a protonated histidine, thus favouring NADH binding over NAD + To test this, we have used site-directed mutagenesis on LDH to replace the invariant Asp-168 with asparagine and then alanine. These mutations do not alter the pK of the neighbouring His-195 in the E and E-NADH forms, a result which does not support the proposal that Asp-168 is important in maintaining a protonated His-195 to favour NADH binding. However, in both modified forms of the enzyme the K m for pyruvate is 100 times greater and the k^ 10 times smaller. In the light of the X-ray structures of apo and ternary LDH these results can be explained by the formation of a strong Asp-His interaction only in the ternary (E-NADH-pyruvate) complex. Because the alanine mutant has the same properties as the asparagine, despite its inability to form a hydrogen bond with His-195, we suggest that it is the negative charge of Asp-168 rather than its orienting ability which is important in the ternary complex and in catalysis. The crystal structure of the glycolytic enzyme triose phosphate isomerase from yeast has been determined in high resolution. The structures of complexes of the enzyme with a transition state analogue inhibitor and with its actual substrate have also been determined. Based on these structures, a mechanism has been proposed for the catalytic action of these enzymes. Site-directed mutagenesis has been used to alter residues in the active site to test their functions in this mechanism. Mutagenesis has also been employed to test the function of the disordered loop that folds down over the active site when substrates bind. Finally, a series of mutants have been made to investigate the irreversible uiactivation of this enzyme at high temperatures. Based on the results of these studies it has been possible to engineer this enzyme to have greatly increased stability to irreversible inactivation at 100°C. The cyclic AMP receptor protein, CRP (also referred to as a catabolite gene activator protein, CAP) and its cofactor cyclic AMP are involved in bacterial regulation, particularly in activating transcription of catabolite operons. A model of interactions between the CRP-cAMP complex and DNA has been proposed (Ebright et al., 1984a,b) . In addition to the established contact between Glu-181 and base pairs 7 and 16 of the CRP recognition site, this model predicts a contact between Arg-180 and base pairs 5 and 18. To verify this prediction, we created a set of mutations at codon 180 of the crp gene of Escherichia coli K12, using the ohgonucleotide site-directed mutagenesis method. In order to introduce any amino acid at this position, an oligonucleotide was synthetized in which each of the three positions corresponding to codon 180 was substituted by one of the four bases. The resulting mixture served as a mutagenic primer. Mutants were identified by dot-blot hybridization and sequence determination. The yield of mutagenesis was 11%, and six classes of mutants were obtained, corresponding to changes of Arg into Cys, Ser, Pro, His, Leu or Gly. After transfer of the mutated genes on plasmids, we have undertaken to study their effect on the expression of genes under CRP control. Preliminary results confirm the predicted role of Arg-180 in a specific interaction of CRP with DNA. Phagemid cloning vector is hybrid plasmids that contain replication origins from both a bacterial plasmid and a filamentous phage. These hybrid vectors facilitate DNA sequencing, oligonucleotide-directed mutagenesis and production of hybridization probes. The vector is usually in a double-stranded form inside the cell. Upon infection of the plasmid-bearing cells with filamentous helper phage, production of single-stranded plasmid molecules is induced. Single-stranded DNA is prepared as for M13 and can be used as template for mutagenesis or sequencing. We have constructed 'all-in-one' phagemid vectors for mutagenesis and expression in Escherichia coli of the regulatory (R) and catalytic (Q subunits of cAMP-dependent protein kinase (cAdPK) from yeast. Cloning and characterization of the genes for yeast cAdPK was done by M.Wigler and co-workers, CSHL (Toda et al., 1987) . The expression/mutagenesis vector uses the T7 phage promoter for expression (Studier and Moffatt, 1986 ) and origin DNA from phage Fl. Site-directed mutagenesis has been used to probe the interactions between the R and C subunits. Single amino acid changes have been targeted at Ser-145 of the R subunit. This residue is thought to play a role in R-C interaction since it is part of the substrate recognition sequence for the C subunit (Arg-Arg-Thr-Ser-Val). The use of phagemid vectors for mutagenesis and expression and the specific effects of SER145 mutations on R-C interactions will be reported. Studier.W. and Moffatt,B. (1986) / Mol. Bid., 189, 113 Toda.T. a al. (1987) Mol. Cell. Biol, in press. 36. Aspartate transaminase structure and function D.Ringe Massachusetts Institute of Technology, Cambridge, MA 02139, USA Pyridoxal phosphate (vitamin Bg) is one of the most important cofactors in intermediary metabolism. It is found in the active site of amino acid transaminases (ATases), decarboxylases and racemases. The L-aspartate ATase from Eschenchia coli has been well characterized in terms of its kinetic and mechanistic properties. It has now been cloned, sequenced and expressed at high levels in E. coli. Several mutants have been prepared, designed to probe the mechanistic and specificity characteristics of the enzyme. One of these mutants has now also been characterized structurally. The ammo acid residue which reacts with pyridoxal phosphate to form a Schiff base in one stage of the reaction is Lys-258. This residue has been replaced by alanine and the new protein has been characterized mechanistically and structurally. Immunoglobulins are natural precedents of protein engineering: a large spectrum of (binding) properties is realized by the variation of locally confined regions ('hypervariable loops') within an otherwise well-conserved and stable structure. Furthermore, the typical immunoglobulin fold is present in a number of seemingly unrelated proteins. The three-dimensional architecture of several immunoglobulins is known, among them dimers of light chain variable domains. Such material was originally isolated from the urine of patients suffering from Bence-Jones disease. We have chosen one such variable domain, REI-V (Epp et al., 1975) , as the starting point of a protein engineering project A gene for REI-V was chemically synthesized by a combination of known and newly developed techniques. Work currently in progress includes: (i) expression of the synthetic gene in bacterial hosts; (li) oligonucleotide-directed mutagenesis of the REI-V gene to obtain structurally predetermined variants of the protein; (iii) protein-chemical characterization of REI-V and its constructed variants.

Reference Epp et al. (1975) The determinants of antibody specificity are known to be con-236 fined to the complementarity-determining regions (CDRs) of immunoglobulins. The manner in which sequences of CDR dictate three-dimensional structure, however, is largely unknown. This relationship might be unravelled if large numbers of X-ray structures of antibodies were known but this will take many years. We have employed an approach whereby the combining site structure can be deduced from a knowledge of the CDR sequences followed by application of a model-by-homology routine. The use of this approach in the generation of a 'docked' complex between an anti-lysozyme antibody and its epitope will be described as well as the manner in which this product model has been subjected to test by site-directed mutagenesis of the antibody genes. As an example of the application of this approach we shall describe how the affinity of a naturally occurring antibody has been increased 10 times by protein engineering. Mutants in the DNA-binding helix F of the cAMP receptor protein (CRP) has been constructed by site-directed mutagenesis. Mutants in a synthetic DNA site derived from the sequence in the lac regulatory region have also been constructed. The effect of these selected amino acid and nucleotide substitutions on CRPmediated transcnptional activity and on sequence-specific DNA binding has been studied. It is shown that Arg-180 in helix F is crucial for the specific interaction of CRP with DNA whereas Arg-185 and Lys-188 are not involved in specific complex formation. Substitutions of A at position 2 and G at position 5 in the DNA recognition sequence T 1 G2T 3 G 4 A5 both abolish specific binding of a wild-type CRP protein. The evolution of proteins represents (in part) natural selection engineering a protein to be suitable for different environments, and (in part) the non-functional drift of the protein's structure. Thus, some of the information drawn from the recent evolutionary history of a protein, deduced by sequence comparisons of homologous proteins, may assist the biochemist seeking to alter the protein's structure to achieve desired properties. This approach is being tested on two proteins: yeast alcohol dehydrogenase (ADH) and mammalian ribonuclease (RNase). Several mutants in ADH have been chosen by comparison of homologous dehydrogenases yeast and other organisms. A specific example is the change of Arg-211 to Thr. The residue, in a helix remote from the active site, would not be selected for mutagenesis based on a simple inspection of a crystal structure. Yet its alteration appears to effect cofactor binding predictably. In ribonuclease, alterations have been introduced into a synthetic gene expressed in Escherichia coli. Changes at positions 35 (Leu -Met) and 19 (Ala -Ser) recreate a ribonuclease from a now-extinct organism, the precursor of several modern bovoids. The many natural variants of chloramphenicol acetyltransferase (CAT) in diverse genera of chloramphenicol (CM)-resistant bacteria constitute a family of homologous proteins which inactivate CM by O-acetylation, using acetyl CoA as the acyl donor and a ternary complex mechanism with a general base role proposed for the imidazole of His-195. The plasmid-encoded type HI enzyme is a stable trimer (3 X 25 kd) with identical subunits, each of which contains a conserved acidic residue (Asp-199) which may contribute to catalysis as well as to the stability of CAT protomers. Site-directed mutagenesis studies, begun in advance of a structure from crystallographic data, aimed to demonstrate an interaction between His-195 and Asp-199 which might be important for catalysis and have been extended to studies of substrate binding and polypeptide folding, and protein stability. The CAT system may represent a particularly useful vehicle for protein engineering because of the availability of DNA sequences for nine naturally occurring CAT variants and a high-resolution (1.75 A) structure from X-ray crystallography. The latter locates the CM binding site with great precision and confirms in structural terms the importance of His-195 in the proposed catalytic mechanism.

42. Receptor chimeras: a new approach to study mitogenic and oncogenic transmembrane signalling H.Riedel, T.J.Dull, S.Massoglia, J. Schlessinger and A.Ullrich Genentech, Inc., 460 Port San Bruno Boulevard, San Francisco, CA 94080, USA Several growth factor receptors share general structural homology within their cytoplasmic tyrosine kinase and extracellular ligand binding domains. Distinct structural features define three receptor classes in this family that share close amino acid sequence homology with specific oncogene products [indicated below in brackets], (i) The human epidermal growth factor receptor (HER1) [\-erbB, AEV] and the related putative receptor HER2 [rat neu] are monomers with two cys-rich extracellular sequence repeats, (ii) The human receptors for insulin-like growth factor I and insulin [v-ros, ASV-UR2] appear to be heterotetrameric structures (fybi) &n<^ contain one cys-rich cluster per extracellular a-subunit. (lii) The receptors for platelet-derived growth factor and colony-stimulating factor I [v-jms, SM-FeSV] are monomers with 10 non-clustered extracellular cys-residues and a split tyrosine kinase domain. We have generated cDNA chimeras by combining structural domains of several receptors and oncogene products of the different classes described above for stable expression in mouse or rat fibroblasts. Expressed receptor-receptor and receptor-oncogene chimeras were then tested with regard to structural and functional characteristics. Our results demonstrate that these chimeric receptors are fully functional with respect to transmembrane activation of the tyrosine kinase and generation of biological signals. Furthermore, we find that various chimeras stimulate anchorage-independent cell growth in RatI and NIH-3T3 cells when activated with the appropriate ligand. The biological significance of these findings will be discussed. There are now many classes of proteins which are being synthesized in Escherichia coli, yeast or mammalian cells from cloned DNA including hormones, blood proteins, enzymes, inhibitors and antibodies. Using techniques of site-directed mutagenesis it is possible to change single amino acids in these proteins to make novel molecules with altered characteristics. Examples are provided for interleukin-2 and al-antiproteinase. The domain structure of many large eukaryotic proteins is often reflected in the genome by the arrangement of the exons and introns making up the gene. As the domains probably fold independently during synthesis it has been possible to change the domain structure of several proteins without destroying specificity of enzymatic activity and conferring novel functions to the hybrid. This is done by manipulating cloned genes or cDNA using naturally occurring or introduced restriction sites and oligonucleotides and expressing the chimeras in mammalian cells. Interferons and the plasminogen activators have been altered in this way but antibodies probably provide the best paradigm. The application of protein engineering in the pharmaceutical industry is now important. We have chosen here to describe two examples which clearly demonstrate the high potential of this new technology, (i) Human ai-antitrypsin analogues: the primary function of a-antitrypsin is the inhibition of neutrophil elastase. An ai-antitrypsin deficiency results in lung emphysema. Moreover, the a r antitrypsin from lungs of heavy smokers is inactivated due to the oxidation of the Met residues at the active site. A (Met-358 -Val-358) a|-antitrypsin analogue which remains fully active as an elastase inhibitor but which is also resistant to oxidative inactivation has been constructed and characterized.

The inhibition of the clotting pathway by other analogues with modified residues in the active site has also been studied, (u) The HTV env protein: the envelope glycoprotein (env) represents the major antigen at the surface of HTV viral particles and efforts to produce a vaccine against AIDS have centred on this protein.

We have constructed a recombinant vaccinia virus expressing the env protein. In infected tissue cultures cells the env precursor gpl60 is efficiently cleaved to two components on extracellular one (gpl20) and a transmembrane segment (gp41). We have observed that gpl20 is rapidly shed from the cell surface thus offering an explanation for the low immunogenicity of the env protein. In order to enhance immunogenicity of env, we have genetically engineered new variants and the results obtained will be discussed. It has been long appreciated that the crystal structure of an enzyme or protein of clinical interest can provide information which can serve as the basis for the rational design of drugs. Close examination of such a structure-based drug design strategy reveals the need for a fuller understanding of the contributions made to binding energy by specific interactions in a putative drug-receptor complex. We have investigated these interactions in the dihydrofolate reductase (DHFR)-methotrexate (MTX) complex by protein engineering. Specific amino acid substitutions in the active site of DHFR, made by site-directed mutagenesis, have allowed a detailed analysis of changes in binding interactions using the high-resolution (1.9 A) crystal structures of several mutant DHFRs in the binary complex with methotrexate. The analysis is facilitated by use of the thermodynamic cycle where localized changes in the system are considered rather than binding processes. This approach can aid in drug design by mapping, both qualitatively and quantitatively, the active site of a protein with respect to residues which are important to drug binding. Tissue plasminogen activator (t-PA) is a senne protease that converts the inactive zymogen plasminogen to plasmin, the major fibrin degrading enzyme in the fibrinolytic system. Tissue-PA is a multi-domain protein consisting of five putative domains: the fibronectin finger domain (F), the epidermal growth factor domain (G), the kringle 1 domain (K,), the kringle 2 domain (K2) and the protease domain (P). We have chemically synthesized a portion of the t-PA gene to allow the incorporation of convenient restriction sites between each of the domains, thus simplifying their manipulation. This system not only provides for the construction and expression of t-PA mutants having single domain deletions but also facilitates the site-directed mutagenesis and expression of individual domains. By using this approach, 238 we have constructed and expressed the prototype, full-length molecule (GFKiK 2 P) as well as the deletion mutants FK,K 2 P, GFK 2 P, GFK,P and FK 2 P. In addition, a number of mutants with multiple domain deletions, additions and transpositions have been made. We have purified each of these mutants on immunoaffinity columns and have assessed mol. wt by fibrin autography, amidolytic activity in the presence and absence of CNBr-digested fibrinogen fragments and specific activity via [l,3-3 H]di-isopropyl phosphorofluondate active site titration. Our analyses demonstrate that most gross domain deletions do not appear to have dramatic effects on specific enzymatic activity. These data suggest further that the interdomain amino acid modifications made as a result of inserting unique restriction sites do not appear to have any deleterious effect on the biochemical properties of t-PA we have examined thus far. Insulin associates above physiological concentrations (10~'° M) into dimers and hexamers. This strong tendency to self-assembly has numerous advantages in relation to the events following biosynthesis in the /3-cell of the pancreas, but does not necessarily represent useful properties for the therapeutic formulation and clinical use of insulin. We therefore applied protein engineering to the hormone with the aim of counteracting self-association and changing its physico-chemical properties. Substitutions were performed at five different residues (B9, B12, B26, B27 and B28) by oligonucleotide-directed mutagenesis or by total gene synthesis, and the insulin analogues were produced by fermentation of single chain precursors in yeast followed by semi-synthetic conversion into the respective mutants. As assessed by gel filtration, measurements of osmotic pressure and CD spectroscopy, mutation of B12-Val -lie, B26-Tyr -Glu or B27-Thr -Glu gave a partial reduction of the association tendency, whereas single mutation of B9-Ser -Asp, B12-Val -Glu or B28-Pro -Glu resulted in analogues which remained essentially monomeric even above millimolar concentrations Molecular modelling has revealed that electrostatic effects probably are the main explanation for the observed prevention of the association, but steric hindrance also accounts in some cases for the weakened self-assembly. In conclusion, monomeric insulins can be obtained by a single mutation in the monomer-monomer interface of the insulin dimer. Insulin dimerizes to significant levels at low concentrations. At the centre of the dimer-forming surface is Val-B12, an invariant side-chain, filling space between the aromatic groups of Phe-B24 and Tyr-B26 and making some close contacts to its equivalent in the dimer itself. Substitutions at B12 by site-directed mutagenesis have been conducted in an effort to produce a stable monomer for physical, chemical and biological studies. The substitution Val -De at B12 should not have altered internal contacts that appear to be important for the monomer folding, but the extra methyl group protrudes out from the dimer-forming surface and was expected to disturb the molecule's dimerizing behaviour. In the presence of zinc ions, however, B12 De insulin still forms hexamers. The analysis of the rhombohedral human 2Zn insulin B12 De crystals has been earned out with 2.0 A resolution data and the atomic model was refined to an agreement factor R = 0.14. The adjustments made by the residues surrounding the B12 De are rather small, extending across the dimerforming surface. Detailed comparison with the native human 2Zn insulin will be reported. The mature form of diphtheria toxin (mol. wt 58 348) is cleaved by trypsin into fragment A (mol. wt 21 167) and fragment B (mol wt 37 199). The catalytic centre for ADP-ribosyl transferase activity is carried by fragment A, while the function of fragment B is to bind the toxin molecule to the cell surface and facilitate the membrane translocation of fragment A into the cytosol. Receptor binding domain substitution should result in the formation of new recombinant toxin molecules whose target specificity is determined by the ligand component of the chimenc protein. We have genetically fused a cDNA insert encoding ammo acids 2-133 of interleukin-2 (IL2) to amino acid 465 of diphtheria toxin. The resulting chimeric toxin has a mol. wt of 68 086 and has antigenic determinants intrinsic to both diphtheria toxin and IL2. The toxin-related IL2 fusion protein (IL2-toxin) is expressed in recombinant Escherichia coli K12. Following purification from the periplasmic compartment, IL2-toxin is selectively cytotoxic for eukaryotic cells which bear high-affinity IL2 receptors (IC50 ^ 1 nM) and non-toxic toward cells which are devoid of the IL2 receptor (IC50 2: 5 X 10" 7 M). Cytotoxicity is blocked by excess free IL2, monoclonal antibodies to the IL2 receptor, and lysozomotrophic reagents. The method of Bishop and Thompson (1986) for the maximum Posters likelihood alignment of DNA sequences has been extended to the alignment of protein sequences. This has been done by considering each amino acid symbol in terms of the uncertain DNA triplet which represents it. Codon usage may be taken into account if it is so desired. The method is illustrated with some examples. Protein conformation depends on interactions of side-chain residues with solvent and with sterical neighbouring groups in the polypeptide chain. When these latter interactions occur water molecules are, with a few exceptions, excluded from the region between the interacting groups. Various methods have been employed to evaluate the absolute affinity of amino acids for an aqueous environment, leading to a number of hydrophobicity scales. The relative position of amino acids in such scales depends on the organic phase (sometimes the vapour phase) selected for the determination of the distribution coefficients and, consequently, discrepancies are found. We have investigated the conditions affecting the retardation of various apolar compounds and of amino acids when chromtographed in gel filtration media such as Sephadex G-25 and the polyacrylamide gel Bio Gel P-2. Such retardation (expressed as In K^) is a linear function of the ionic concentration in the eluting buffer and depends on the salt species, being related to their position in the Hofmeister series. In particular, salts exhibiting good salting-out properties, such as potassium phosphate and Na 2 SO 4 , enhanced the absorption onto the gel. On the contrary, salts known to exhibit salting-in properties, such as KI or KSCN, reduced the adsorption of aromatic compounds to the matrix. The behaviour of amino acids when chromatographed at high phosphate concentration on both media strikingly paralleled the features and ranking common to the above-mentioned hydrophobicity scales. The experimental In K4 values may thus be used for a quantitative estimation of the interaction of amino acids with water independent of the solvent. These results also throw light on the structure of the water coordinated by polymeric gel phases such as the media used in gel filtration. 52. Engineering of human lysozyme for elucidating its secretion mechanism by Saccharomyces cerevisiae M.Kikuchi, Y.Taniyama, Y.Yamamoto, K.Yoshimura 1 , K.Nakahama 1 and M.Dcehara Protein Engineering Research Institute, Tokyo 103, and 'Takeda Chemical Industries Ltd, Juso, Osaka 532, Japan We have investigated the effects of amino acids or peptide substitutions in the signal peptide region and mature protein on the secretion of human lysozyme (hLZM) by yeast. The signal peptides tested were the chicken lysozyme (CL) and the AspergiUus Posters glucoamylase signal (AG), and the two interspecies hybrids. Each hybrid was composed of one half from CL and the other half from AG, designated CL-AG or AG-CL. The mature hLZM was obtained with CL and CL-AG; however, no hLZM was secreted with AG and AG-CL, where hLZM synthesis was not detected. The hLZM secreted by the system was purified and crystallized. Substitution of the serine for the cysteine residue at 128 (S 128 ) completely inhibited the secretion of hLZM. The synthesized S 128 was located in the insoluble fraction and found to be processed and biologically inactive. The secretion mechanism of hLZM by yeast will be discussed on the basis of these results and those obtained from other substitutions of cysteine and die C-tenrunally truncated molecule. 53. Structure and prediction of calcium binding sites in globular proteins W.E.Hohne Humboldt University of Berlin, Institute of Biochemistry, Hessische Strasse 3-4, Berlin, DDR-1040, GDR Calcium binding behaviour and the structure of calcium binding sites of globular proteins are investigated in terms of the predictability of such sites. A computerized search program is presented which allows potential calcium binding loops in proteins to be detected on the basis of sequence data by means of a specially designed 'window'. This method is marked improved by taking into consideration the degree of accessibility of negatively charged non-salt-bridged amino acid residues if the spatial structure is known. The aim of this prediction is to select structural regions of a protein suitable for the generation of additional calcium binding by site-directed mutagenesis to increase the stability of the structure against the thermal denaturation and/or proteolytic degradation. Calcium binding serine proteinases are used as a model to verify the prediction methods. 54. Proteins, genes, hierarchies -a new strategy for protein design R.L.Somorjai and S.A.Narang

Our strategy is based on the hypothesis that understanding the protein-folding process requires the elucidation of constraints. This hypothesis is supported by experimental facts and theoretical arguments. The crucial core of the hypothesis is that the constraints are manifested through the hierarchical nature of folding. The protein structure is demonstrably hierarchical. Our design method aims to delineate the critical constraints. The feasibility of the general strategy is tested on an enzyme of known structure, T4 lysozyme (see abstract, this meeting, Rose et al.). We first determine its hierarchical structure as the best consensus of a variety of clustering methods. This identifies the building blocks, e.g. secondary structural elements and hydrophobic/aromatic clusters at the lowest level. By following the fate of each of these as we ascend the hierarchy, a possible folding pathway can be traced. (Alternatively, we can descend the hierarchy.) The critically important experimental link-up is the existence of a regionspecific mutagenetic method. This enables us to change arbitrarily any protein at any level of its structural hierarchy by modifying, replacing, excising, shuffling any of its subunits. The new protein can then be subjected to a battery of physical-chemical -enzymatic tests to probe structure -function -stability relationships. 55. The double prediction method: an algorithm for improving the accuracy of secondary structure prediction of proteins from class prediction G.Deldage and B.Roux Laboratoire de Physico-Chimie Biologique, Universite Claude Bernard, 43 bd du 11 novembre 1918, Lyon I, 69622 Villeurbanne Cedex, France

The prediction of secondary structure of proteins from amino acid sequence has been extensively studied and applied to many proteins. Unfortunately, the accuracy of statistical methods fails to improve with the increase in data and currently achieves <56% for a three-state description of secondary structure (helix, sheet, coil) (Kabsch and Sander, 1983a ). An algorithm has been developed which takes into account the predicted class of proteins to increase the success in secondary structure prediction. This method has been called the 'double prediction method' and consists of a first estimation of the secondary structure from a new algorithm which uses Chou-Fasman (Chou and Fasman, 1978) type parameters, and the prediction of the class of the proteins from amino acid composition (Nakashtma et al., 1986) . These two predictions allow optimization of the parameters to provide the final secondary structure prediction. The 'double prediction method' has been tested on 59 proteins (i.e. 10 323 amino acids) of a secondary structure database (Kabsch and Sander, 1983b) . It yields a 72% success in class prediction and 61 % of amino acids are correctly predicted for three states (Deldage and Roux, 1987) . Typical applications of this method will be shown.

Packing of a-helices in a globule with a single core has been considered. It has been shown that such globules can be described by an idealized model of 'quasi-spherical' polyhedrons whose apexes stand for the ends of helices and the ribs are their axes or lines of contact between dieir ends. For example, all the threehelical globules can be represented by a single Platonian octahedron (six apexes stand for six ends of these helices, and the 'axes' ribs form either a clockwise or anti-clockwise helical bundle, etc.). The screening of a protein databank shows only little deviations from this model for the real architecture of helical complexes (20° in the direction of helix axes, 3 A in distance). 57. Classification of spatial motifs of the antiparallel 0-structure in globular proteins Y.N.Chirgadze

A systematization of the complex topological types for the polypeptide chain folding in the antiparallel /3-form is proposed. Three well-known simple topological types, /3, m and g, which are single /3-strand, a hairpin of two strands and a simple Greek key-type of four strands, were chosen as the basis of the system. The new topologically allowed motifs are created by a combination of these three basic motifs. All possible spatial motifs were considered for the more complicated double Greek key topology. This was done on the basis of a complete set of 14 deduced basic spatial motifs of simple Greek key topology. The analysis of -20 globular proteins shows that some spatial motifs appeared to be realized as the main part of the chain fold of the molecule. Therefore, chain folds of antiparallel (3-proteins are suggested to be necessarily determined by the simple topological requirements. The Bacillus Ucheniformis a-amylase is the liquefying enzyme most widely used in the starch industry. This enzyme is stable in a large range of pH, and retains >90% of its original activity after incubation at 85°C for 15 min. In order to study the thermostability of this enzyme, we decided to isolate mutants expressing a thermolabile enzyme. Two different mutant enzymes have been obtained. Both are thermolabile only at low concentrations of Ca 2+ . One was found to be affected in one of the conserved regions over all amylases. The second is under investigation. Several enzymes have been shown to be active versus small synthetic substrates in reversed micelles, but essentially nothing is known about the activity versus large biopolymers. Micelles hosted by enzymes have been shown to increase their size and change their structure, but the interactions between these filled micelles have not been studied. The investigation of the interactions of these large filled micelles has been conducted by measuring the rate of proteolysis of several proteins as a function of the pH, water versus surfactant concentration ratio and structureof the biopolymer to be hydrolysed. To this end, proteins known to be in their native or denatured form in the micelles and proteins whose surface hydrophobicity has been changed by controlled chemical modification have been used as substrates. The study is conducted in the sodium-2-ethylhexylsulfosuccinate (AOT)-isooctane-water system with trypsin and chymotrypsin as proteases. A comparison of the stability versus proteolysis of the various proteins with their behaviour in water, allows interesting conclusions to be drawn about the conformation and solvation of proteins in reversed micelles and the relation between their location and structure. These observations are relevant to the thermodynamics of protein denaturation. We have started studies to find out some structure -function relationships in thermostable proteins. a-Amylase from Bacillus stearothermophilus was chosen as a model enzyme since the protein possesses some interesting characteristics. It is rather thermostable, pH-stable and it is secreted in the culture medium. It also has a specific activity higher than the corresponding B. Ucheniformis enzyme. We wanted to see if some of the differences are due to an extra 32 amino acid stretch in B. stearothermophilus enzyme. A series of truncated B. stearothermophilus a-amylaselike enzymes were constructed by using site-directed mutagenesis. Indeed, by inserting a translation stop codon 32 amino acids from the carboxy terminus an a-amylase with novel features was created. As predicted, specific activity was lowered (~ 50%) but thermostability was enhanced. As hoped, the pH-stability profile was similar to the wild-type protein. The charactenstics of other mutant enzymes in this series will also be discussed. We have determined the crystal structure of 2Fe-2S ferredoxin from Halobacterium marismortui at 3.2 A resolution. The bulk of the structure is similar to a homologous 2Fe-2S ferredoxin from Spiruhna platensis, a non-halophilic organism, including conservation of the active site topology. The structure is distinguished by two characteristics: a third of all the residues are carboxyls, and there is an extra 22-residue segment at the N terminus. This headpiece, which is heavily loaded with carboxyls, folds into a separate domain connected to the rest of the molecule by two anchoring points. The apparent function of this domain is to solvate the protein at practically saturated salt concentrations. The insertion of a carboxyl-rich solvation domain may constitute a general mechanism for the stabilization of halophilic proteins. In order to learn about the structure -function relationship of proteins involved in supramolecular assembly, it is important to know the contributions of individual amino acid residues toward overall protein stability and in protein-protein interactions. Genetic studies have revealed three classes of mutants (tsB, tsC and tsBQ which synthesize a thermosensitive (ts) VP1, the major capsid protein of SV40. We have determined the nucleotide sequences of the DNA fragments containing the tsB/C mutations and will report the corresponding amino acid substitutions in 21 ts SV40 mutants. In 19 mutants a single amino acid substitution results in a protein which is thermosensitive over a narrow temperature range (33-40°C). The distribution pattern of these mutations suggests that there are two major domains in VP1. Mutations in one domain occur within a region of VP1 that is conserved among related papova viruses. Previous work has shown that mutants in this class (tsC) fail to initiate shell assembly at 40°C. The other domain contains the tsB mutations in a region of VP1 which is less conserved among related viruses. At 40°C, altered protein-protein interactions may result from these mutations since previous work has shown that tsB mutants fail to form a complete capsid around the SV40 chromatin at this temperature. Our results do not reveal an independent domain for the tsBC mutants. The glycolytic enzyme phosphoglycerate kinase (PGK) has been well characterized enzymologically and structurally. Comparison of yeast PGK with PGKs from various thermophilic sources provides a test-bed system to study the molecular basis of thermal stability in proteins. As part of this study PGK from the extreme thermophile Thermus thermophilus has been purified and crystallized. Synchrotron X-ray data have been collected to 3.5 A and solution of the structure, facilitated by molecular replacement methods, is in progress. Amino acid sequence analysis has made possible the construction of an oligonucleotide probe which has enabled the isolation of the T. thermophilus PGK gene; the primary sequence of the gene has been determined. Comparison of the mesophilic yeast and thermophilic enzymes will be discussed, together with proposals to engineer thermal stability into the yeast enzyme by site-directed mutagenesis. The most prominent l.r. band to study protein structure is the Amide I that is the 80% C=O stretching vibration and is located between 1610 and 1690 cm" 1 . Amide n is not so useful in assessing secondary structure, but its degree of disappearance after substituting the H 2 O medium for D 2 O gives an idea of tertiary structure. Sarcoplasmic reticulum exhibits a band with a maximum at 1652 cm" 1 and shoulders at 1634 and 1690 cm" 1 , indicating that the major structure is an a-helix with a contribution of /3-sheet. By changing the medium from H 2 O to D 2 O there is a shift of the Amide I band to 1645 cm" 1 , indicating a high accessibility of the amidic protons to deuterium and showing a low packing density. In the presence of the natural substrate ATP, an increase in the 1652 cm" 1 band is seen by l.r. difference spectroscopy compatible with an increase in a-helix content. When calcium is removed, an increase in unordered structure is seen. Mitochondnal Complex ID also shows structural changes that can be monitored by i.r. spectroscopy when the membrane is dehydrated. An increase in the 1630 cm" 1 band, which can be attributed to extended interconnected chains, is shown after the decrease in water content. Examples corresponding to data reduction techniques that allow the resolution of the broad Amide I band, such as Fourier deconvolution or Fourier derivation, will also be presented. Tubulin is the major constituent of microtubules and die cellular target for antimitotic drugs. Tubulin crystals suitable for X-ray analysis have not been reported and its tertiary structure and the interactions involved in microtubule assembly and ligand binding are mostly unknown. Hydrophobicity and secondary structure algorithms detect a total of -18 possibly exposed regions in each of the a-and /3-tubuhn sequences. Among these potential epitopes we have selected five tubulin fragments, which might be functionally relevant. These correspond to the positions a (241 -226), a (415-443), a (430-443), /3 (241-253) and /3 (412-431) and were prepared by solid-phase synthesis, purified to homogeneity by h.p.l.c, coupled to carrier proteins and employed to raise antibodies which were fully specific for the corresponding regions of tubulin. The peptides a (415-443) and /3 (412-431) were also recognized by the antitubulin monoclonal antibodies DM1A and DM IB respectively. We have employed the monospecific antibodies to: (i) identify the fragments produced by controlled proteolysis of tubulin with six endopeptidases, resulting in a new proteolytic map of this protein; and (ii) investigate the involvement of each of the corresponding tubulin regions in microtubule assembly. Ferredoxin and cytochrome c$ are specific partners in the sulphate-reducing bacteria electron transfer chain and might be exemplary for electron exchange mechanism studies between hemes and Fe-S clusters. The major feature of the cytochrome c 3 (M r 13 000) is the presence of four hemes of low redox potential. Ferredoxin I is a typi-al 4Fe-4S cluster ferredoxin (M r 6000). Thermodynamic parameters of the complex are: a stoichiometry of 1:1 and an association constant K a = 1.2 X l^M" 1 (pH 7.7, Tris 10~2 M). The main characteristics of the complex formation are an electrostatic recognition process associated with water molecule release. Involvement of the two highest potential hemes in the interaction has been demonstrated by 'H n.m.r. data. Primary and tertiary structure determinations of both partners allow a better understanding of the architectural basis of the interaction. Chemical modifications on tryosine, methionine or arginine and covalent cross-Unking of the two proteins should localize the cytochrome c 3 -ferredoxin interaction site. The solution conformation of azurin, the blue copper protein from Pseudomonas aeruginosa, as well as its electron transfer kinetics, appear to be influenced by pH. Modification of residue His-35, the copper ligands and the residues in the hydrophobic patch of the protein may provide insight into the mechanism of these pH effects. As a first step towards the site-directed mutagenesis of these residues the azunn gene was cloned and sequenced. The results are reported on the poster. The gene appears to code for a preprotein with a 19-amino acid long signal sequence. This sequence probably serves the transport of the azurin over the periplasmic membrane. The use of High Mass Fast Atom Bombardment Mass Spectrometry in conjunction with classical protein chemistry has proved to be a powerful and cost-effective method for structural determination of genetically engineered proteins (Morris et al., 1983) . We have developed specific methods for determining the identity of blocked N-terminal residues and for detecting the presence or absence of C-terminal fragments or 'ragged-ends', two areas which pose extreme problems for conventional classical techniques. In addition, characterization of recombinant proteins for both research and regulatory body approval purposes often requires determination of aspects of post-translational modification such as glycosylation or S-S bridge assignment (Morris and Pucci, 1985) . FAB mapping procedures have been developed to solve these problems and examples of the application of these new methods to recombinant and other proteins and glycoproteins will be given. The novel strategies described constitute a powerful new tool for protein engineers and chemists. The caseins have been described as one of the most rapidly diverging families of proteins on the basis of comparative studies of the sequences of codons and amino acids in homologous gene products from ruminant species, man, rat, mouse and guineapig. Solution measurements on the individual bovine caseins have indicated flexible polypeptide chains with little or no a-helix. Indeed, the caseins have been described as random coil-type proteins. Nevertheless, when secondary structure prediction methods were applied to bovine crsr, /3-and x-caseins a considerable fraction of each sequence was assigned to regular secondary structures, particularly the extended ^-configuration and jS-turns. We report secondary structure predictions for all the published (unphosphorylated and unglycosylated) casein sequences, including some for non-ruminant species. In spite of considerable sequence divergence, there are clear similarities in the predicted secondary structures within the family of x-casein homologues and also within the family of /3-casein homologues. Moreover, in the caseins that have major sites of phosphorylation there is a frequently predicted configuration in which the major sites are in a turn region close to the N terminus of an a-helix. The structure of bacteriorhodopsin by electron diffraction gives a picture of seven a-helical rods oriented roughly perpendicular to the membrane. I.r. spectroscopy shows anomalously high Amide I frequencies that can be attributed to distorted a-helices or to an-type helices. Phospholipase D cleaves the headgroups of the lipids and can be used to modify membranes. In a H 2 O medium, after water subtraction bacteriorhodopsin exhibits a maximum at 1659 cm" 1 and two shoulders at 1637 and 1684 cm" 1 supporting the idea of an anomalous helix and some j3sheet structure. In D 2 O the maximum is at 1660 cm" 1 , indicating a high packing density and a small contribution of unordered structure. Phospholipase D treatment reveals a shift in the Amide I to 1657 cm" 1 both in H 2 O and D 2 O media. However, no changes in the shoulders corresponding to /3-sheet are seen. The parallel shift in both media corroborates the small contribution of unordered structures to the Amide I band and indicates that no new structures are produced after phospholipase D treatment but relaxation of the distorted helical structures has taken place. Phospholipase D treatment of the purple membrane of Halobacterium halobium indicates, then, that the anomalous Amide I frequency is due to a distorted helix induced by membrane lipids rather than to a different helical structure induced by the amino acid backbone. The crystal structure of ROP, a dimenc protein which acts in the control of ColEl plasmid replication, has been determined at 1.7 A resolution. The 63 amino acid monomer consists of two helices joined by a very tight hairpin bend. In the dimer the helices pack -with each antiparallel to its neighbours -to form a highly regular four a-hehx bundle. The packing constraints for this coiled-cell structure can be analysed in terms of a repeating heptapeptide sequence a-b-c-d-e-f-g with positions a and d forming the hydrophobic core of the dimer. The stability and regularity of its structure makes ROP an attractive molecular scaffold for the attachment of, for example, catalytic sites. To assist in the design of such artificial enzymes, a data base is being created which documents in terms of structural changes the effects resulting from mutations of the ROP protein. Crystals of the hydrophobic protein crambin (mol. wt 5000) diffract to better than 1.0 A resolution and have >80% of the solvent in the crystal ordered. This protein provides an excellent opportunity to study the distribution of water molecules at a protein surface in atomic detail that is not generally available for other protein crystal structures. Water is important not only in stabilizing a protein in its folded conformation but also can contribute to an enzyme's activity through binding at its active site. The function of the plant protein crambin is not yet known; however, it is homologous to the membrane-active plant toxins purothionin (from wheat germ) and viscotoxin (from mistletoe), which have been shown to bind calmodulin and appear to bind lipids. Crambin's X-ray diffraction data to 0.945 A at 300 K and to 0.83 A at 140 K have been collected and the model of the structure has been refined. Two types of water networks are formed: chains at the hydrophilic protein surface and rings, primarily pentagons, at the hydrophobic protein surface. Insight into the role of water in stabilizing the protein is gained which applies to stabilization of proteins in general and to modelling the effect of changes on protein conformation on this water structure. We are in the process of constructing mutants of crambin and will explore how these mutants influence the water structure. Thermodynamic perturbation theory implemented with molecular dynamics is described together with its application to problems in drug and protein design. The perturbation method allows one to transform or 'mutate' a system between any two states that can be suitably parameterized. This method is applied to the protein thermolysin complexed with phosphonamidate and phosphonate ester inhibitors. The atomic structure of these proteininhibitor complexes has been determined and they are nearly identical, yet their binding constants differ by three orders of magnitude. We have calculated the difference between the free energy of binding of these inhibitors with a result of 4.2 ± 0.54 kcal/mol which compares remarkably well with the experimental value of 4.1. In preparation for the use of this method in protein design, we simulated the free energies of solvabon for all classes of amino acids. Calculated values are all within ~ 10% of relevant experimental data and provide the necessary experience to apply these methods to protein design. These methods have been applied to a site-specific mutant in dihydrofolate reductase. Asn-27 has been changed into a Ser with a calculated relative free energy of binding to methotrexate of 2.21 ± 0.76, which compares well with the experimental value of 2.6 kcal/mol. The resultant close correspondence between calculated and experimental values are very encouraging and lead us to believe that this technique will provide a significant advance towards a rational approach to both drug and protein design.

76. Fragmentation of thermolysin as a tool to detect features of its structure, stability and dynamics

Thermolysin is an exceptionally stable metalloendopeptidase isolated from Bacillus thermoproteolyticus which contains a functional zinc ion and four calcium ions bound to the polypeptide chain. Since its amino acid sequence, three-dimensional structure at 1.6 A resolution and mode of binding of a number of inhibitors are known, thermolysin can be listed among the most suitable protein models to investigate structure-function relationships of proteins and in particular to study at the molecular level the key interactions responsible for the unusual stability of thermophilic enzymes. In our laboratory, fragmentation of thermolysin has been used to address questions regarding structural and folding properties of this protein, as well as its stability and dynamic features (see reference list). The hierarchical organization of the three-dimensional structure of thermolysin in terms of domains and subdomains has been studied by investigating the conformational properties of isolated protein fragments corresponding to structural domains (and subdomains) in the intact native protein (Dalzoppo et al., 1985a,b; Fontana et al., 1983 Fontana et al., , 1985 Vita and Fontana, 1982; Vita et al., 1984 Vita et al., , 1985a . It has been found that these isolated fragments are able to acquire a stable, native-like conformation independently from the rest of the polypeptide chain, as given by far-u.v. circular dichroism measurements (Dalzoppo et al., 1985a; Fontana et al., 1983 Fontana et al., , 1985 and immunochemical properties (Dalzoppo et al., 1985a; Vita et al., 1985a) . Moreover, these fragments show conformational transitions mediated by heat, pH or guanidine hydrochloride characterized by cooperativity and thermodynamic parameters typical of low mol. wt globular proteins (Dalzoppo et al., 1985b; Vita and Fontana, 1982) . Limited proteolysis or autolysis of thermolysin under specific experimental conditions allowed us to isolate 'nicked' protein species constituted by two as well as three protein fragments associated in stable complexes (Fassina et al., 1986; Fontana et al., 1986; Vita et al., 1985b) . Subtilisin cleaves thermolysin at a few specific peptide bonds leading to an active protein species (thermolysin S) given by a strong and stable association of fragments 5-224 (Dalzoppo et al., 1985b) and 225 (Vita et al., 1985a )-316 (Vita etal., 1985b) . Autolysis of thermolysin in the presence of EDTA at neutral pH leads to a selective removal of calcium from the protein and concomitant peptide bond fissions at the chain loop 109-205 involved in the binding of Ca-4 in the native protein (Fassina et al., 1986) . This result helps understanding, at a molecular level, the differences in stability between thermolysin and the homologous neutral protease from B. subtilis. In fact, the mesophilic protease, which is much more heat labile than thermolysin and prone to autolysis even in the presence of calcium ions, shows high sequence homology and shares many structural, functional and ion-binding features with thermolysin, but lacks the binding sites for Ca-4

only. An examination of the three-dimensional model of the thermolysin molecule reveals that all sites of chain fragmentation by limited proteolysis or autolysis by EDTA or heat treatment occur at exposed loops, and not within chain segments of regular secondary structure (helices, sheets). Moreover, a striking correlation is observed between sites of fragmentation and sites of segmental mobility given by the temperature factors (B values) determined crystallographically (Fontana et al., 1986) . This indicates that the overall process of proteolysis involves an initial interaction of a peptide segment of the protein substrate at the active site of the protease, after which some conformational change (chain distortion) is required in order to make the idealized transition state of the cleavage reaction. Overall, these results indicate that exposed and flexible loops are the most favoured sites for protein-protein recognition (in this case between the proteolytic enzyme and the globular protein substrate) and appear to demonstrate that protein mobility seen in protein crystals is relevant to motion in solution. In this connection the question arises whether the ribosomal proteins can have the unique tertiary structure in the isolated state. Structural studies of the ribosomal proteins with the help of n.m.r. spectroscopy, CD measurements and limited proteolysis permit the following conclusions to be drawn: (i) most ribosomal proteins in solution have a pronounced secondary and tertiary structure;

(ii) after denaturation by urea and guanidine hydrochloride and subsequent renaturation, the proteins have identical physical properties to the same proteins isolated in mild conditions; (iii) broad studies of some ribosomal proteins permit the role of their different amino acid sequences in the tertiary and quaternary structure formation to be identified; (iv) at the same time there are some proteins with a low level of secondary and tertiary structure. It can be shown that in such proteins the fixed unique structure in the ribosome can be achieved as a result of interactions with the other proteins and/or ribosomal RNA. A survey of 50 proteins indicates that intrahelical ion pairs influence the a-helix stability. It was found that on a-helices the i, i ± 3 and i, i ± 4 types of ion pairs are the most predominant. The observed frequencies of these ion pairs are significantly greater than the expected frequencies. Such a preference was not seen in the /3-sheets and non-helical regions. As a control, like charged pairs were also considered and they showed no similar preference. Importantly, the normalized frequencies of the i ± 3/4 type of ion pairs were alone found to increase with the helix length. The ion pairs were found on the solvent side of the ahelices. An analysis of the distances between the ion pairs suggests that they are stabilized by direct hydrogen bonding, electrostatic interactions and water bridges. The intrahelical ion pairs, by stabilizing the secondary structure, may also stabilize the protein tertiary structure. The fibrous and globular proteins are distinguished by the density of ion pairs. Research supported by NIH grants AM-34139 and GM-18455. Resonance Raman spectroscopy currently provides detailed information on the structures and conformations of several types of prosthetic groups of proteins, as well as on their molecular interactions with this host protein. Some recent applications of this method to bacterial photosynthetic reaction centres will be presented. Resonance Raman spectra permitted: (i) detailed studies of binding interactions assumed by each of the six chlorophyllic prosthetic groups present in this protein; (ii) investigations of the interspecific variabilities of these interactions. Resonance Raman spectroscopy also brought the first evidence of a local protein conformational change induced by the transfer of an electron from the primary donor to its early acceptor. Eukaryote cytochrome c consists of a haem group largely shielded from solvent by a single polypeptide chain that is economically wrapped around it in a series of loops, and to which it is covalently attached. Four of these structural elements conform to the definition of an fi-loop, a novel category of non-regular secondary structure (Leszczynski and Rose, 1986). It has been suggested that such loops may be modules of evolutionary change and thus candidates for protein engineering. Semi-synthetic techniques have been used to excise one of them, to give des-40-55 cytochrome c, since this deletion would be difficult to achieve by genetic manipulation methods. The likely structure of this product resembles that of a bacterial cytochrome, the c 555 of Chlorobium species, which differs from mitochondria! cytochrome c in lacking this same loop (Korszun and Salemme, 1977) . The physical and functional properties of the product have been compared with those of horse and Chlorobium proteins. The comparisons permit new conclusions on structure-function relationships in cytochrome c and, in particular, on the role of the loop itself. Tl, 2'GMP-RNase Tl and 3'GMP-RNase Tl. Thermodynamic analysis indicated that 2'GMP affected the structure of the Trp-59 region more than 3'GMP. This greater influence by 2'GMP appears to be primarily due to interactions involving the phosphate moiety of 2'GMP with the active site of RNase Tl which are then transmitted through the protein matrix to the Trp-59 region. Theoretical studies via molecular dynamics were performed by removing 2'GMP from the 2'GMP-RNase Tl X-ray coordinates yielding an averaged structure and dynamics for free RNase Tl. RMS differences between the averaged RNase Tl structure and the 2'GMP-RNase Tl structure showed significant changes in loop regions (containing residues 44-55 and 95-98) which interact with the guanine moiety of 2'GMP. Results from the molecular dynamics also indicated a motion of Trp-59 similar to that observed via time-resolved fluorescence in free RNase T2. Combination of the experimental and theoretical results indicates that 2'GMP-induced structural changes leading to alterations of RNase Tl dynamics occur both in regions directly adjacent to and regions more removed from the bound 2'GMP. The bacterial elongation factor Tu belongs to the GTP hydrolysing group of proteins. The primary structure of Escherichia coli EF-Tu was determined in Aarhus and recently we have solved the structure of the G-domain at a resolution of 0.29 nm by X-ray crystallography. As an increasing number of amino acid sequences of GTP hydrolysing proteins have been determined it appears that many exhibit partial homology with the primary structure of the G-binding domain of E. colt EF-Tu. We have focused our interest on three such proteins, namely chloroplast EF-Tu, the E. coli initiation factor IF2 and the ras oncogene product p21. Based on the sequence homologies and using the crystallographic model of E. coli EF-Tu as a framework we have built computer models of the G-domains of the chloroplast EF-Tu, E. coli IF2 and p21. These structural models are valuable tools in the engineering of the proteins. Mutagenesis work is already in progress with these proteins. The enzymatic hydrolysis of cytidine 2':3'-cychc phosphate (CP) by bovine pancreatic RNase A follows non-hyperbolic kinetics when a large substrate concentration range (0-50 mM) is used. The transition found at concentrations of CP > -15 mM could be explained by an allostenc phenomenon in which several substrate molecules bind to different subsites in the enzyme thus causing a conformational change which results in a higher rate of hydrolysis. The simultaneous assay of the hydrolytic and synthetic activities of the enzyme by means of an h.p.I.e. technique can be described by a kinetic model based on the existence of several subsites, one of them having a high K a and the other ones with a low A^ value. At higher concentrations of substrate (-25 mM) a drop in activity is found. Although no direct evidence has been found, this phenomenon could be promoted by the formation of substrate-induced enzyme dimers known to occur at high substrate concentrations.

The lactose permease of Escherichia coli is an integral membrane protein catalysing active transport of galactosides. The conformational change which is necessary for transport was studied by transport measurements in the millisecond time-scale. The internal dynamics were investigated by time-resolved fluorescence anisotropy experiments, using tryptophane and pyrene bound to a distinct cysteine residue as probes. For all measurements, the permease was purified and reconstituted in dimyristoylphosphatidylcholine. When the lipids undergo the phase transition from the fluid to the ordered state, the transport rate decreased. Concomitantly, a 50-ns fluctuation was slowed down, whereas faster fluctuations were not influenced. The 50-ns fluctuation is interpreted as a wobbling motion of a membrane-spanning helix, and the faster fluctuations as motions of the probes relative to the polypeptide backbone. Application of the reaction theory of Kramers in the high-viscosity limit permits the 50-ns fluctuation to be related to the transport rate. Measured transport rates and their activation energies agree with this interpretation. Thus we propose that the wobbling motion of a membrane-spanning helix which occurs in the range of 10~7 s is the origin of a slow conformational change in the range of 10~7 s which represents the transport step. Glutamate dehydrogenase catalyses the oxidative deamination of L-glutamate to 2-oxoglutarate and ammonia with concomitant reduction of NAD(P) + . The crystal structure of an NAD-linked GDH from Clostridium symbiosium has been solved at 2.5 A using a combination of isomorphous replacement and molecular averaging. The structure is hexameric and each subunit is comprised of two domains separated by a deep cleft. Difference Fourier analysis has shown that one of the domains binds NAD in a manner closely related to that of other dehydrogenases. The second domain is important in the assembly of the molecule, forming important interactions across the 2-and 3-fold interfaces. Although no sequence is available for this GDH, a consensus sequence derived from alignment of GDH sequences from other species was used to provide an initial interpretation of the electron density map. The regions most highly conserved across the different sequences were seen to comprise the core of the molecular structure. Furthermore our attention was drawn to a pocket near the NAD binding site, deep in the cleft, largely comprised of residues conserved over all GDH sequences., We suspect this pocket forms the binding site for the second substrate and also contains the catalyic residues. Crystallographic data on a variety of enzyme/substrate complexes are currently being processed to provide independent evidence for the location of the active site. Tyrosyl-tRNA synthetase (TyrTS) from Bacillus stearothermophilus is composed of two identical subunits, each having a complete active site. The enzyme displays half-of-the-sites activity in that only one active site appears functional per dimer. Heterodimers of TyrTS were produced by engineering oppositely charged groups into the subunit interface so that they could form a complementary pair. In wild-type TyrTS, the two Phe-164 residues are on the axis of symmetry and interact in a hydrophobic region of the subunit interface. Phe-164 was mutated to Asp or Glu in TyrTS and to Lys or Arg in an active truncated enzyme. This size difference allowed subunit association to be studied by gel filtration. These changes induce reversible dissociation into inactive monomers at pH values which favour ioruzation at position 164. Mixing full-lenguh and truncated mutants near neutral pH generates active heterodimers, but no detectable homodimers. Kinetic analysis allowed estimation of dissociation constants for the dimers. Each of the four salt-bridges engineered into TyrTS is weak, but sufficiently strong to direct specificity in dimenz-ation. Mutation of the subunit interface is not required to form heterodimers. They were also produced by mixing full-length and truncated enzymes and reversibly denaturing in urea. Heterodimeric enzymes are useful in investigations of half-of-the-sites activity because different mutations can be introduced into each subunit of the dimer. Such studies have shown that, in any given dimer, it is always the same subunit which functions in the steady-state.

89. Probing structure-function relationships using site-directed mutants of tyrosyl-tRNA synthetase K.A.Brown and D.M.Blow Blacken Laboratory, Imperial College, London SW7 2BZ, UK

Crystallographic and kinetic studies have been made of several mutants of tyrosyl-tRNA synthetase which have been altered at the active site by site-directed mutagenesis techniques. This enzyme catalyses the adenylation of tyrosine with ATP, followed by the transfer of tyrosine to its cognate tRNA. Structure analysis of a mutant in which Thr-51 is changed to prohne shows that there is no change to the main chain conformation, while kinetic data show that this mutant is much more active in the adenylation step. The results suggest that the substitution of prohne at position 51 changes the enzyme surface so as to improve its affinity for the adenine moiety They also indicate a rearrangement of solvent over this part of the enzyme surface. Acylphosphatase (EC 3.6.1.1) is a highly conserved, 80-100 residue enzyme, which is found in several tissues and cleaves specifically the carboxyphosphate bonds in several metabolically important compounds (e.g. 1,3-diphosphoglycerate, carbamyl phosphate). Although the sequence has been determined, little is known about the overall structure. Our current research is on the three-dimensional structure of acylphosphatases from horse and turkey muscles using conventional two-dimensional n.m.r. techniques. Most of the spin-systems of the individual amino acids have been identified and some of them assigned to positions in the sequences. An antiparallel /3-sheet consisting of ~20 amino acids has been found. To elucidate the role of individual amino acid residues in stabiliz-ing the conformation of a protein, stabilities of a series of variant a-subunits of tryptophan synthase from Escherichia coli substituted by each of the amino acids at position 49, which is buried in the interior of the protein, were quantitatively examined. Gibbs energy of unfolding in water and activation Gibbs energy of unfolding in 3 M GuHCl for each protein were compared at pH 7 and 9. Our results indicated that: (i) the values of Gibbs energy of unfolding in water at pH 7 of the mutant proteins changed 0.72-to 1.92-fold of that of the wild-type protein, but those of activation Gibbs energy of unfolding in 3 M GuHCl changed only 0.95-to 1.03-fold of the wild-type protein; and (ii) the stabilities of the proteins substituted at a position buried in the interior of the molecule increase linearly with hydrophobicity of the substituting residue, if the residual volume of the substituting residue lies within a certain limit.

92. The mechanism of interaction of some high molecular weight polyethylene glycols with proteins Polyethylene glycols (PEGs) have been used as effective protein salting-out and crystallizing agents. However, the mechanism of their interaction with proteins still remains unclear. We have studied the interaction of a number of proteins of varying hydrophobicities and mol. wt with PEGs ranging in mol. wt from 2000 to 6000 using high-precision densimetry. This technique enables us to deduce the preferential interaction parameter for the interaction of various additives with proteins using multicomponent thermodynamic analysis. It has been found that in all the PEGs used the proteins remain preferentially hydrated and that there is preferential exclusion of the PEGs from the protein domain. The preferential exclusion increases with the increasing PEG size indicating that steric exclusion of PEGs leads to the preferential hydration of proteins. Also, the chemical potential of the proteins is increased by PEG addition and it increases with the increasing PEG size. This unfavourable situation leads to the salting-out of proteins. It is suggested that unfavourable interactions of solvents with the charged groups on protein surface, with the hydrophobic groups of proteins, and their steric exclusion from the protein domain are the key factors responsible for the salting-out of proteins. Thermostability is a valuable characteristic for industrial enzymes, but little is known about its molecular basis. To gain insight into the rules that govern protein thermostability we have performed the analysis of replacements between mesostable and thermostable counterparts of several isoenzymes whose sequences are known. The secondary structures, hydropathic profiles, hydrophobic moments and flexibility profiles have been predicted by means of a microcomputer program. From the analysis of proteins whose mesostable, thermostable and/or thermosensitive counterparts are known a table of replacements has been established. These replacements have been used to check for similar ones in other sequentially less related proteins. We have found that replacements leading to an enhancement of thermostability preferentially occur at external regions, mainly in /3-turns and a-helices (at the edges of amphipathic a-helices) and rarely in j3-strands. Replacements highly conserve the corresponding secondary structure and the hydrophobic moment of the helices. Regions in which replacements take place are conservative in flexibility/rigidity and mainly correspond to a flexible region. The determination of the three-dimensional structure of native and mutant proteins has become a well-established branch of protein engineering machinery. The crystallization of the proteins is still a time-consuming and labour-intensive step due to the great number of parameters that can be changed. As a first step towards automatizing the crystallization procedure we used an autosampler for our crystallization experiments. Our methods and a first successful result will be presented. In aqueous solution human insulin molecules (mol. wt 5808) associate, depending on concentration, pH, metal ions and salts, into dimers, tetramers and hexamers. The association state can be assessed by several methods, all with inherent problems and advantages. Using osmotic pressure only the average state of association is obtained but no dilution or dissociation is introduced during the measurements. This is not the case with gel filtration which, however, gives an indication of the distribution of oligomeric species. In neutral solution zinc-free human insulin is found as hexamers in concentrations s 1 mM. In contrast, at 1 mM concentrations, genetically engineered insulin analogues with mutations in the monomer-monomer interface all showed less tendency to self-association. Substitution of B12-Val -De or B27-Thr -Glu resulted in a mean association state corresponding to tetramers of insulin, B26-Tyr -Glu or B28-Pro -Asp gave only weak tendency to self-assembly corresponding on average approximately to dimers and B9-Ser -Asp, B12-Val -Glu or B27-Thr -Glu plus A21-Asn -Asp were fully dissociated into monomers. The analogue B27 -Glu plus B9 -Asp had no self-association even in the presence of Zn 2+ . It is concluded that osmotic pressure measurement in combination with gel filtration is a convenient method to characterize association behaviour of insulins in the concentration range 0. In order to express the mature human apolipoprotein (apo) AI in Escherichia coli, several plasmids with different promoters and Shine-Dalgarno sequences were constructed. None of them gave high levels of expression. By fusing fragments of the gene coding for apo AI at the 5' end of the /3-galactosidase gene, we could verify the poor efficiency of initiation of mRNA translation. With the help of synthetic oligonucleotides we have constructed several random mutants in the NH2-terminal coding sequence of apo AI, but without altering the amino acidic sequence. Some mutations led to high-level expression while others did not noticeably affect it. These results strengthen the importance of secondary structures in the 5' end of the mRNA which influence the efficiency of initiation of translation. The isolation and sequencing of the mutants has led to the identification of the molecular determinants of this parameter. We have in fact been able to express recombinant apo AI after modifying the NH2-terminal sequence. The glycoprotein hormones consist of two non-covalently associated subunits; the a-subunit is common to the hormones within a species whereas the /J-subunits are unique to each hormonal activity and confer receptor specificity. TSH, FSH, LH are secreted by the anterior pituitary whereas CG, an LH /3-gene readthrough, is produced by the human placenta. Specific target cell activity is associated with adenylate cyclase stimulation. These hormones are the most complicated known and their tertiary structure has defied elucidation. We have developed an analysis package to predict their tertiary structure, using Chou-Fasman, hydrophobicity and core residue propensities. Difference analysis of the /S-subunits has identified two major receptor-specific loops. The a-subunit shares a number of features, including an appropriate Ser, with the substrate binding pocket of serine proteases. Since two of the three His in the a-subunit are strictly conserved in evolution, we have evaluated whether they mediate a similar catalytic activity in the hormones. Specific and progressive acetylation of His residues increasingly attenuated HCG bioactivity, with agonist activity undetectable when all His were modified. Treatment of this analogue with alkali, breaking the highly substituted His rings, produced a molecule which antagonized the activity of intact HCG and, yet, was still devoid of agonist activity. Unusually, antagonism was appreciably greater for the hormone-stimulatable cytosolic enzyme pathways than for the plasma membrane-associated cyclase activity, thereby divorcing the two pathways. The tertiary designations of the receptor and two different bioactivity sites map to complementary, but independent, areas of both the a-and /3-subunits. We have developed a non-chemical DNA synthesis method in which assembling of base-pairs is based on repeated cloning of a modular DNA fragment into specialized vectors. Insertion and excision of the fragment occurs at positions different by one or two base-pairs, thus resulting in stepwise elongation of the fragment. In the 'two base-pairs stepwise-elongation system', 16 vectors contain the following sequence:

Xmnl N-N represent the 16 possible dinucleotide sequences. Xmnl and MnH yield blunt ends and cleave at (+) and (*) respectively. After cloning into the Xmnl site, the DNA module is excised again with MnR at two base-pairs from the original site of insertion. Similarly, vectors for single base-pair elongation were constructed. All vectors contain suitable biochemical and genetical properties to allow forced insertion of the fragment in the correct orientation and to avoid background. This method will be illustrated by several examples of site-specific mutagenesis.

101. Theoretical prediction of regions in interleukin-2 primary structure important for its activity The recently developed informational spectrum method for the analysis of informational content of macromolecules (Veljkovid et al., 1985) was used here to identify frequency values, i.e. amino acid residues that are supposed to be responsible for interleukm-2 (IL-2) binding to its receptor and mitogenic transmembrane signalling. So the characteristic frequence (F) ensuring IL-2 binding to the specific receptor is determined (F = 0.02734) which comprise defined amino acid residues. The frequence values of F = 0.30859 is supposed to characterize IL-2 activity responsible for information transduction. Site-directed mutageneses are found to influence such frequencies, namely the mutation of defined amino acids decrease or increase IL-2 activities. This approach allows us to design IL-2 molecules with the desired activity. The informational spectra method (Veljkovid et al., 1985) was applied for the analysis of structure-function relationships in the murine epidermal growth factor (EGF). A segment active both in receptor binding and biological stimulation activities was predicted. The results obtained in this way are in good agreement with those achieved experimentally for A, B and C segments of EGF (Williams et al., 1986) . Furthermore, a possible role for the N-and C-terminal portions of sequences was proposed as well as hot-spots for 'up' and 'down' functional mutations.

103. Preparation of an active fragment of human erythrocyte acetylcholinesterase R.Ravazzolo, R.Barresi and C.Garre" Department of Biology, University of Genova, ISMI, Genova, Italy

Proteolysis with papain of intact human erythrocytes releases in the incubation medium acetylcholinesterase (AChE; EC 3.1.1.7) activity from the external surface of the membrane where the enzyme is tightly bound as an integral protein. Intact erythrocytes were incubated with papain immobilized on Sepharose beads and the AChE activity released was further purified by affinity chromatography. Such enzymatic activity showed specificity for the substrate acetylthiochohne, was sensitive to the specific AChE inhibitor 1,5-^ts(4-allyldimethylammoniumphenyl)pentan-3-one, and had a K m for the substrate comparable with AChE punfied from the erythrocyte membranes. Monoclonal antibodies to human erythrocyte AChE reacted both with native membrane enzyme and with the proteolytic fragment. It was also possible to label the proteolytic fragment with the active site directed ligand [ 3 H]diisopropylfluorophosphate. By electrophoresis in nondenaturing polyacrylamide gel and gel filtration chromatography, both performed in the presence and absence of non-ionic detergents, the fragment behaved like a hydrophilic protein.

104. Synthesis and expression of the T4 endonuclease V gene designed for cassette mutageneses E.Ohtuka, T.Inaoka, M.Ishida and K.Miura Faculty of Pharmaceutical Sciences, Hokkaido University, Sapporo 060, Japan T4 endonuclease V (T4endV), which consists of 138 amino acids, is a pyrimidine-dimer-specific excision-repair enzyme. It has two distinct catalytic activities, a pyrimidine-dimer DNA glycosylase and an apurinic/apynmidinic endonuclease. In order to elucidate the structure-function relationship of this enzyme, we designed a gene coding T4endVm, which contained restriction sites suitable for cassette mutageneses. The gene was synthesized by joining of 26 chemically synthesized oligodeoxyribonucleotides (each of 22-36 nucleotides) and inserted downstream of the tryptophan promoter of an expression vector. Escherichia coli HB101 was transformed by this plasmid and the gene was expressed under the control of the tip promoter. A band corresponding to a product with the expected mol. wt of T4endV was detected by SDS-acrylamide gel electrophoresis. The amino acid sequence was analysed and the sequence up to 20 amino acid residues of the amino terminus was identical to T4endV. The endonuclease activity of the product was confirmed using a synthetic substrate. The product was also effective in xeroderma pigmentosum cells on DNA damaged by u.v. light. The problem of rapid degradation by endogenous proteases of heterologous polypeptides expressed in bacteria has been investigated. With the use of synthetic DNA fragments, a set of plasmid vectors has been obtained (Chakhmakhcheva et al., 1987) . The vectors provided high-level expression of peptides and small proteins in Escherichia coli as fusions with the leader sequences of different length. They were used to achieve expression of synthetic genes for human proinsuhn and its mutants and functionally active fragment of bactenorhodopsin. The yields of polypeptides were in the range 5-30% of the total amount of cellular protein. Zeins are a microheterogeneous mixture of proteins that constitute the major storage proteins in the corn endosperm. The zein proteins are notably deficient in Lys and some other amino acids.

To develop an approach to the construction of mutant zeins of improved nutritional quantities, the chemical-enzymatic synthesis ofa gene for zein cZ22Bl (22 kd) (Marks et al., 1985) has been undertaken. The synthesis of this ~800 bp-long DNA was planned with the use of a general methodology for artificial gene construction developed by us earlier (Chakhmakhcheva et al., 1986) . The determination of appropriate sites for alterating the amino acid sequence and the possibility of obtaining modified zeins containing the Lys and Tip residues in specified positions of the polypeptide chain by directed mutagenesis of the corresponding gene will be discussed.

107. Selective alteration of protease inhibitory activity of Eglin c upon reaction with 1,2-cyclohexanedione N.J.Braun, Y.J.Chang and H.P.Schnebli

N-Ac-Eglin c is a small protein inhibitor of a number of serine proteases including human leucocyte elastase and Cathepsin G, chymotrypsin, pancreatic elastase and subtilisin. Eglin c contains four arginine residues, three of which are in close proximity to the proposed 'scissile' peptide bond. All of these residues react at a similar rate with the a-dicarbonyl reagent 1,2-cyclohexanedione, and the final extent of modification (-50%) is the same for all residues. Modified Eglin retains, to a large degree, inhibitory activity towards leucocyte elastase and chymotrypsin, but is much less active towards pancreatic elastase and subtilisin when compared with the native inhibitor. These studies help to define the role of arginine residues in the interaction of Eglin c with several proteases and can serve as a guide to constructing altered inhibitors through site-specific mutagenesis. The C terminus of the insulin B-chain plays a major role in current structure -function studies. The aim of the present work was to investigate the influence of biological activity by chain length variation and substituting the C terminus by amidation with different amines. The preparation of analogues shortened by 2-7 amino acids was accomplished by trypsin-assisted semi-synthesis with a-amino-protected des(B23-B30)-insuhn and a 10-to 15fold excess of peptide amide in aqueous-organic solvent mixtures. After deblocking, the analogues were purified by gel filtration and ion exchange chromatography at pH 8. The analogues were homogeneous in reversed-phase h.p.l.c. and were further characterized by electrophoresis (pH 2.2, 4.8 and 8.6) and amino acid analysis. The C termini in the positions B24 and B25 have been blocked by ammonia, hexylamine, benzylamine, phenethylamine and 3-phenylpropylamine. A pronounced enhancement of bioactivity (lipog enesis m isolated rat adipocytes) by amidation was observed in the case of des(B26-B30)-insulin, especially with ammonia and phenethylamine. Coronaviruses are a group of positive-stranded RNA viruses which infect vertebrates and are responsible for diseases of economic and clinical importance. The surface protein (S-protein) of the coronavirion is responsible for both cell and tissue tropism, recognition by the cell receptors, membrane fusion and is the structural protein primarily responsible for eliciting both the humoral-and cell-mediated response. It is therefore of central importance in defining the disease process during coronavirus infection. We have cloned and sequenced the murine hepatitis virus (MHV) S-protein and are producing vaccinia virus recombinants which express this protein in a variety of cell lines. Using the technique of 'gapped duplex' oligonucleotide-directed mutagenesis, we have produced S-gene mutations which are currently being tested for their functional significance with respect to immune recognition by B and T cell clones. A nucleotide sequence encoding Escherichia coli translational initiation factor IF1 was deduced from the primary structure (71 amino acid residues) of this protein. The codons in the sequence were selected on the basis of the preferred codon usage found for the structural genes of initiation factor IF2 of both E. coli and Bacillus stearothermophilus, which can be expressed to high levels in E. coli. To facilitate site-directed mutagenesis experiments, we designed the IF1 gene to have a modular structure (each module containing 3-10 amino acid residues) by selecting, at regular intervals, triplets which would introduce specific restriction sites in the sequence. The IF1 gene was assembled by shot-gun ligation of nine synthetic oligodeoxynucleotides ranging in size from 31 to 65 nucleotides and cloned into an expression vector under the control of an inducible promoter. Upon induction, E. coli cells transformed with this vector were found to express very efficiently the artificial gene producing large amounts (60 mg/100 g cells) of a protein indistinguishable from natural IF1 in both chemical and biological properties. We have begun a protein engineering approach to probe the basepair recognition mechanisms of DNA polymerases. A genetic screening method was developed to facilitate the generation of mutant forms of Escherichia coli DNA polymerase I (Klenow fragment) that less efficiently discriminate against non-Watson -Crick base-pairs during DNA synthesis. Plasmids encoding the Klenow enzyme are engineered in vitro, then introduced into a strain of E. coli (polA^, polC a , pcbA) in which the plasmidencoded Klenow enzyme substitutes for DNA polymerase III in DNA replication. This system permits a rapid genetic assessment of the functional properties of mutant forms of the Klenow enzyme, including mutator activity. An improved segmented method of oligonucleotide synthesis was developed and used to prepare 62 different mutant oligonucleotides in a single day, for use in oligonucleotide-directed mutagenesis of Asp and Glu residues that are conserved within the polymerase domain of Klenow and T7 DNA polymerases. A similar protein engineering approach is also being taken with eucaryotic DNA polymerase beta. The isolated heavy and light chain variable regions (Fv) of the anti-hen egg lysozyme (HEL) monoclonal antibody (MAB), Gloop2 (Darsley and Rees, 1985) are being expressed in Escherichia coli using a dual origin expression vector, in order to procure large quantities of the antigen binding fragments for binding and physical -X-ray crystallographic and n.m.r. -studies. Such studies will aid the verification of the computer-derived model of the Gloop2-HEL complex (de la Paz et al., 1986) . In addition, the structure of Fv proteins with altered antigen binding properties (via site-directed mutagenesis of the cloned cDNAs) can be determined and thus provide further insight into antibody-antigen interactions. The cloned Gloop2 heavy 0y2b) and light (K) chain cDNAs (Roberts and Rees, 1986) The human class I alcohol dehydrogenases differ at 35 positions, and some of these exchanges affect catalytically important residues. We have started to change residues in the j3i-suburut in order to test their relative importance for enzymatic activity. The site-directed mutagenesis has been performed in the M13 system with mismatching oligonucleotides. Thr-48 has been changed to Ser and Phe-93 to Trp. Thr-48 and Phe-93 are both located in the active site pocket, restricting the space available for the substrate. The residue at position 48 is also directly involved in the binding of the NMN-ribose of NAD. The Thr/Ser change makes the human /3-subunit more similar to the -y-subunit and the horse form, and the Phe/Thr change makes the human form identical to the yeast alcohol dehydrogenase at that position. One of the zinc-ligands, Cys-111, has been changed to Ser in a third mutant. ADH-expression plasmids were constructed from the mutated cDNA inserts and the vector pKK 223-3 containing the tac promoter. These constructions gave unfused human alcohol dehydrogenase expressed in Escherichia coli cells. The recombinant proteins showed both lmmunological and enzymatic activity.

114. Kinetic studies of engineered mutants of the dihydrofolate reductase gene from A kinetic scheme for wild-type Lactobacillus casei DHFR has been derived by measuring association and dissociation rate constants together with pre-steady-state transients, using stoppedflow fluorescence and absorbance spectroscopy. As has been seen for Escherichia coli DHFR (Fierke et al., 1987) , the binding kinetics suggest that during steady-state turnover, the major ratelimiting step in the conversion of dihydrofolate (H 2 F) to tetrahydrofolate (H4F) is the dissociation of the H4F product from an ENADPHH4F ternary complex. This seems to be the preferred pathway since H4F dissociation from either E • H4F or ENADP + H4F is much slower than the observed K^. The preferred complex is formed when NADPH replaces NADP + in the ternary complex and after dissociation of H4F, the binary complex binds H 2 F to repeat the cycle. Thus, during steadystate turnover, DHFR is only present as a binary or ternary complex and never as free enzyme. This scheme can now be used to study the effects of single amino acid substitutions on binding and catalysis. One such mutant (Trp-21 -Leu) appears to have had a large effect on cooperativity of substrate binding and also on the overall rate. In this case K^ is reduced from 17 to 0.4 s~' with hydride ion transfer being slower than product dissociation to become the rate-limiting step. The endopeptidase papain is the most thoroughly studied of the plant proteases that employ cysteine in the hydrolytic process. Its X-ray crystallographic structure has been refined to 1.65 A and it has been the subject of extensive QSAR (quantitative structure-activity relations) analysis. In preparation for sitespecific mutagenesis to genetically cause substitution of amino acids thought to be critical to binding and catalysis (e.g. Cys-25 to Ser, Gin-142 to Lys and Gly) and thereby to study further structure and function, we have cloned and sequenced cDNA corresponding to the entire enzyme. The sequence includes infor-mation for a 133 amino acid prosegment, suggesting that the enzyme is synthesized as an inactive zymogen. The downstream segment contained the sequence (ATV^GAA in the approximate position expected for the polyadenylation signal. The sequence AAATAATAAA, which is more characteristic of the polyadenylation signal, is found further downstream. Our sequence shows some differences from the published amino acid sequence: at positions 47, 118 and 135 the codon for glutamate is found instead of glutamine; at position 169 the codon for asparagine instead of glycine; at amino acid positions 86 -88, the codons for Tyr, Pro, Tyr instead of Pro, Tyr, Tyr. The cDNA has been subcloned into the Tabor T7 promoter regulated peptide system (pT7-7) for expression in Escherichia coli and into pT7/ T3-18 (BRL) for in vitro transcription and translation. The results of those experiments will be discussed.

enzymes. The sequence comparison reveals interesting points. There is a rather high degree of homology between these four enzymes over the whole sequences. More interestingly there are special hot-spots of conserved sequences, namely the His-De-Gly-His (HIGH) region as postulated ATP binding site, the Ser-Lys-Ser (SKS) region involved in the 3' end binding of the tRNA and a further region of still unknown function. The localization of these conserved regions is roughly the same in all the four synthetases. The HIGH region is at the N terminus, the SKS region in the centre of the sequences and the third regioñ 100-140 amino acids towards the N terminus. Hybrid constructions with exchanges in these regions together with sitedirected mutagenesis could be useful in clarifying their functions. Additionally, more general information concerning arrangement and functions of protein domains could be obtained. Thrombin initiates proliferation of fibroblasts and other cells by binding to specific receptors on the cell surface. Signals to the mitogenic machinery are generated by both the binding of thrombin to the receptor and the enzymatic activity of thrombin. A synthetic peptide representing residues 508-530 of the B-chain of human prothrombin binds to the receptor on mouse enzyme cells and inhibits binding of thrombin to these cells. Computer modelling studies have produced a plausible model for thrombin based on the crystal structures of the senne proteases. The region of the thrombin model represented by the 508-530 peptide is part of the active site of the serine proteases with a /3-turn region extending to the surface of the thrombin model. Since neither sequence nor structural data exist for the receptor, we are constructing model peptides which will fit the surface features of this region in the thrombin model and form the appropriate binding interactions. These peptides will provide information which may lead to identification of unique sequences in the receptor which can be utilized for antibodies, cloning or generation of additional synthetic agonists or antagonists of cellular events mediated through this receptor system. Among the 20 aminoacyl-tRNA synthetases of Escherichia coli no extensive homology could be shown until now. We sequenced the genes of two synthetases of the XUX tRNA family, namely the valine and the leucine enzyme, and compared their protein sequences with the sequences of the isoleucine and the methionine 118. Determination of transformation-associated domains of a viral transforming protein using in vitro mutagenic techniques William Markland Integrated Genetics, 31 New York Avenue, Framingham, MA 01701, USA The early region of the polyomavirus genome encodes three tumour (T) antigens named large-T, middle-T and small-T. Middle-T has the ability to transform established rodent cells in vitro in the absence of the other polyomavirus T antigens. The transforming event would appear to be triggered by the association of rruddle-T with pp6(f' src (the cellular homologue of the transforming protein of Rous sarcoma virus) which leads to an enhancement of the tyrosyl kinase activity of the proto-oncogene product. This is a unique example of a protein-protein interaction inducing transformation. A variety of in vitro mutagenic techniques have been utilized to analyse polyomavirus middle-T as a transforming protein and in an attempt to determine what region of the protein is required for its association with ppoXF' 1 " 7 . The majority of the gene encoding middle-T has now been mutated. The phenotypes of these mutants indicate that the protein can be considered as a molecule consisting of three domains: one important for cellular localization, one important for its association with pp6(f' src , and one as a potential modulator of the tyrosyl kinase activity.

Upon dilution to 10~6 M, human insulin dissociates into monomers. This process implies changes in the near-u.v. circular dichroism (c.d.) spectrum. During association the aromatic residues, especially those in the monomer-monomer interface, interact and enhance the c.d. band at 275 run arising from the tyrosines. Thus, the degree of asociation can be evaluated from the concentration dependency of the intensity of this band. The association properties of mutated insulins in the concentration range from 3 X 10~3 M to 3 X 10" 6 M in 5 X 10~2 M NaCl and 10 2 M Tris buffer at pH 7.5 were assessed. Compared with zinc-free human insulin the mutant B27-Thr -Glu showed a reduced association tendency which was further diminished when this mutant was deamidated in the C-terminal residue A21 (Asn -Asp). The spectra of the mutant B9-Ser -Asp were nearly concentration independent indicating that only one single mutation in the monomer-monomer interface is sufficient to impair association so that monomers predominate even at high insulin concentrations. In conclusion, the c.d. spectra indicate that the tendency of the insulin molecules to associate into dimers and hexamers can be reduced and even lost by substitution of one or two ammo acid residues in the interface. 120. Expression of the synthetic gene of an artificial DDT-binding polypeptide in Escherichia coli R.Moser Biochemisches Institut der University Zurich, Winterthurerstrasse 190, Switzerland A hydrophobic 24-residue polypeptide (DBP) with the specific function of binding the insecticide DDT was designed using secondary structure prediction methods and model building (Moser et al., 1983) . The synthetic DBP was indeed found to bind DDT (Moser et al., 1983) . To study the in vivo properties of DBP, the corresponding double-stranded DNA was synthesized, inserted between the BamYO. and Pstl cleavage sites of plasmid pUR 291 and the hybrid plasmid pUR 291-DBP cloned in Escherichia coli JM 109 (Moser et al., submitted) . The fusion construction in which DBP was linked to the C terminus of j3galactosidase made it possible to isolate and purify the hybrid protein after induction by isopropyl-/3-D-thiogalactopyranoside. DBP, which is stable to trypsin, was obtained by reversed-phase h.p.l.c. after tryptic digestion of the fusion protein. Recombinant and chemically synthesized DBP showed identical chromatographic properties, amino acid composition and chymotryptic digestion patterns. The in vivo properties of the /3-galactosidase -DBP fusion protein are under investigation. 121. Examination of the role of active site residues of ribulose bisphosphate carboxylase/oxygenase from Rhodospirillwn rubrum by site-directed mutagenesis Richard J.Mural, Robert S.Foote, Frank W.Larimer, Eva H.Lee, Richard Machanoff, Sankar Mitra, Salil K.Niyogi, Thomas S.Soper and Fred C.Hartman Protein Engineering Program, Biology Division, ORNL, Oak Ridge, TN 37831, USA Based on evolutionary conservation and chemical modification, His-291 (Igarashi etal., 1985; Paech, 1985) and Lys-166 (Hartman et al., 1985) of the title enzyme have been implicated as the essential base that initiates catalysis by the enolization of ribulose-P 2 . To distinguish between these two possibilities, site-directed mutagenesis has been used to replace His-291 (Niyogi et al., 1986) and Lys-166 (Hartman et al., 1987) with other amino acids. When Ala is substituted for His-291, the mutant enzyme retains 40% of the wild-type activity, thereby arguing against an essentiality of this histidyl residue. In contrast, substitution of Lys-166 with Gly, Ala, Ser, Gin, Arg, Cys or His yields mutant proteins with <0 2% of wild-type activity. Since some of the position 166 mutants undergo carbamylation (the obligatory activation process for the wild-type enzyme) and are competent in substrate binding, their severe deficiencies in catalytic activity confirm a crucial role for Lys-166 and are consistent with its functioning as the proton transfer group that promotes the enolization of ribulose-P 2 . a-Actinin, a dimer of 110-kd monomers, and spectrin are known to interact with F-actin. Spectrin contains 240-kd a-and 200-kd /3-subunits. Both subunits form repeated homologous units of 106 residues. Most of the units contain the consensus sequence SEDYG. We compared a 478-residue-long fragment of Dictyostelium a-actinin using the computer program DIAGON. There was a common repeated structure in both proteins corresponding to a double matching probability of 1 X 10~5 and 2 x 10~4 respectively. We also raised monoclonal antibody against synthetic peptide SEDYKDL(GGC) derived from chicken a-spectrin. This antibody reacted in Western blot with human red blood cell and chicken muscle spectrins. In addition antibody reacted with animal a-actinins. 123. A model for binding and oxidation of water in the active site of water splitting enzyme of thylakoid M.K.Raval and U.C.Biswal

A water-specific active site in the water splitting enzyme of thylakoid is proposed. The active site consists of two Mn 2+ and four histidine residues Oxygen atoms of two molecules of water bind to the Mn 2+ and four hydrogen atoms form hydrogen bonds with four nitrogen atoms of histidine residues. Light-induced oxidation of water to oxygen is catalysed by higher oxidation states of Mn 2+ . The transient species formed during oxidation are H2O2 and O^~ bound to the active site. The geometry of the active site with bound substrate and transients is calculated using contact criteria and hydrogen bonding for optimum orientation of binding species in the active site. Since the transient species are structurally different and need space adjustment in the active site, changes in the geometry of the active site and the conformation of the enzyme during oxidation of water are predicted which account for the correlation of thylakoid membrane fluidity and activity of water-splitting enzyme.

124. Protein engineering of ribonucleotide reductase from Escherichia coli Hans Eklund, Solveig Hahne, Ake Larsson, Tomas Lundqvist, Margareta Karlsson, Olle Nilsson, Par Nordlund, Ulla Uhlin and Agneta Ahgren Department of Molecular Biology, Swedish University of Agricultural Sciences, Uppsala Biomedical Center, Box 590, Sweden The ribonucleotide reductase system provides a balanced supply of deoxyribonucleotide precursors for DNA synthesis in all living cells. Ribonucleotide reductase from Escherichia coli consists of two non-identical subunits, denoted proteins Bl and B2. Protein Bl is allosterically regulated and contributes redox-active sulphydryls to the active site, which is formed at the interface of Bl and B2. Protein B2 contributes an essential tyrosyl radical. Both subunits have been crystallized separately and as a complex. Two catalytically defective Bl subunits were isolated by random point mutagenesis. The mutational changes were a Gly-410 -Ser substitution in one case and a Glu-636 -Lys substitution in the other. Further characterization of these mutant proteins were consistent with those residues being part of the active site of Bl. In another study, protein Bl was photoaffinity labelled by the allostenc effector dTTp. The labelled residue, Cys-292, was subsequently changed to an alanine by oligonucleotidedirected mutagenesis. A characterization of the mutant protein identified Cys-292 as part of the allosteric site of Bl. An amino acid sequence alignment of protein B2 with its homologous eukaryotic subunits was the basis for predicting the radical tyrosine residue. Tyr-122 was changed to phenylalanine and a characterization of the engineered B2 subunit unequivocally established that Tyr-122 harbours the radical. Protein B2 was spontaneously degraded in vivo to a truncated form lacking 30 C-terminal residues. Characterization of the truncated form and a heterodimeric form, consisting of one native and one truncated polypeptide, suggested that side-chains responsible for the subunit interaction between Bl and B2 are in B2 primarily localized in the C-terminal part. The gene for fructose-2,6-bisphosphatase (FbPase) was sequenced from both Saccharomyces cerevisiae and Schizosaccharomyces pombe and the amino acid sequence determined. These and pig kidney FbPase amino acid sequences were compared as a first attempt to define functional domains within the protein. FbPase shows complex regulation, especially in S. cerevisiae where a cAMP-dependent phosphorylation of Ser-11 has been implicated in glucose repression and proteolytic degradation. Amino acid substitutions were made at this locus using site-directed mutagenesis (Ser -Ala and Ser -Asp). These mutant enzymes and the wild-type enzyme were overexpressed and the enzymes purified to homogeneity. P-FbPase was prepared by in vitro phosphorylation. A comparison of the properties of phospho and dephospho wild-type FbPases and the mutant FbPases demonstrated that all enzymes had similar maximum activity. However, the P-enzyme was more sensitive to AMP and fructose-2,6-bisphosphate inhibition than the dephospho enzyme, suggesting that regulation operates in vivo by this mechanism, leading to decreased enzyme activity. The purified mutant enzymes Ala-11 and Asp-11 exhibited properties closely similar to those of dephospho and P-FbPase respectively. These results indicate that the functional group at residue 11 is an important factor in the regulation of FbPase activity and that P-Ser can be functionally substituted by Asp in this enzyme. The enzyme RuBP carboxylase/oxygenase ('RuBisCO') is responsible for the fixation of atmospheric CO2 in plants, but also catalyses the reaction of RuBP with oxygen. The enzyme has a mol. wt of 550 000 daltons and is composed of eight large and eight small subunits. We have expressed the genes for cyanobacterial RuBisCO in Escherichia coli to yield active enzyme. We have carried out site-directed mutagenesis on the gene for the large subunit to yield enzymes with altered residues at the active site. We have also constructed chimeric genes between cyanobacterial and higher plant RuBisCOs to investigate the role of regions of polypeptide on assembly and catalysis.

127. Site-directed mutagenesis of elongation factor Tu A.Parmeggiani, P.H.Anborgh, R.H.Cool, F.Gumusel, E.Jacquet and M.Jensen Laboratoire de Biochimie, Ecole Polytechnique, Laboratoire Associe du CNRS no. 240, F-91128 Palaiseau Cedex, France Elongation factor Tu (EF-Tu) from Escherichia coli is one of the best-studied guanine nucleotide binding proteins. Its primary structure as well as that of its two almost identical coding genes (tufA and tufB) is known and three-dimensional models at high resolution have been proposed (La Cour et al., 1985; Jurnak, 1985) . EF-Tu, the carrier of aa-tRNA to the ribosome during protein biosynthesis, displays a GTPase activity. The hydrolysis of GTP is a critical function, since the activity of EF-Tu is regu-Posters lated allosterically by GDP and GTP. In order to achieve a deeper insight into the molecular mechanisms of the EF-Tu functions, we have isolated EF-Tu mutants via site-directed mutagenesis, using synthetic ohgonucleotides as mutagenic primers and colony hybridization as a selection method. So far we have isolated in preparative amounts several mutated EF-Tu factors carrying substitutions in amino acid residues thought to be critical for the GTPase activity and for the binding of GDP/GTP. By deleting a large part of the tufA gene, we have isolated the N-terminal domain of EF-Tu containing the guanine nucleotide binding site and its catalytic centre (Parmeggiani et al., 1987) . The isolated N-terminal domain (mol. wt 21 000), which displays a pronounced homology with other nucleotide binding proteins including the P21 ras protein, conserved the property to interact with GDP/GTP and to express the GTPase activity. The antibody combining site (ACS) of Gloop2, one of five monoclonal antibodies (Gloopl -5) raised to the isolated 'loop' determinant of hen egg lysozyme (HEL) (Darsley and Rees, 1985) , has been modelled and 'docked' with HEL (de la Paz et al., 1986) . This preliminary complex and the response of Gloop2 to various lysozymes implicated several residues on or within the ACS possibly involved in the antibody-antigen interaction. These target residues were substituted both in the cDNA clones of the heavy and light chains of Gloop2 via site-directed mutagenesis and in the refined model of the Gloop2-HEL complex. Single and double mutations have been made and their effect upon the binding of antibody to a panel of variant lysozymes assessed and compared with the effects predicted by the computer models. The results have highlighted the role of electrostatic interactions in determining the conformation of the ACS and in the interaction of the antibody with its antigen. Cellulose is degraded by a family of enzymes acting together in a complex synergistic manner but the molecular mechanism of cellulose hydrolysis is not yet understood in detail. A similar mode of action by general acid catalysis of the cellulases and the well-studied lysozymes was suggested over 20 years ago. There are implications from both kinetic and chemical modification studies of the involvement of carboxyl groups in the hydrolysis of cellulose by these enzymes. Moreover, amino acid sequence comparisons have revealed interesting homologies in the primary structures of cellulases to the active site regions of the different lysozymes. Three-dimensional structures of the cellulases, which would greatly facilitate studies of the structurefunction relationships of these enzymes, are not as yet available. We have used interactive computer graphics for the molecular modelling of the proposed active site region of one of the cellulases, Trichoderma reesei cellobiohydrolase I. This region, homologous to the active sites of both T4 and hen egg white lysozymes, was fitted in the known crystal structure of the T4 lysozyme. Refinement of the model and the subsequent docking of the competitive inhibitor, cellobiose, are discussed and the usefulness of the model as the basis for site-directed mutagenesis experiments is evaluated. The three-dimensional structure of the methionyl-tRNA synthetase from Escherichia coli has been investigated using X-ray analysis at a resolution of 1.8 A. Ninety per cent of the molecule is now well defined and the zinc atom has been identified in a buried region of the molecule, close to the active site. At the same time, the refinement of the complex ATP-MetRS has been carried out at 2.5 A. A series of 60 modified enzymes truncated at the C terminus have been constructed in vitro and assayed for activity. In agreement with the graphics model, the results show that a minimum of 547 residues are necessary to sustain the aminoacylation reaction. A programme of site-directed mutagenesis is in progress: residues thought to be important for catalytic activity, metal coordination and interaction with tRNA are being modified. The effects of the resulting mutations on both ATP-PP, exchange and tRNA aminoacylation activities will be discussed in the light of crystallographic model and affinity labelling data.

131. Correlation of surface electrostatic potential and redox activity for thioredoxin variants altered in amino acid sequence Knut Langsetmo, James Fuchs and Clare Woodward Department of Biochemistry, University of Minnesota, St. Paul, MN 55108, USA Escherichia coli thioredoxin is a small, stable globular protein containing one disulphide bond. The reduced form of the protein (SH>2 acts as a hydrogen donor in the enzymatic reduction of ribonucleou'des to deoxyribonucleotides. The redox-active disulphide is formed by Cys-32-Cys-35, in the highly conserved active site sequence Cys-Gly-Pro-Cys located in a protrusion of the molecule. The hydrophobic surface below the active site pro-Posters trusion is important for protein-protein interactions. Above the active site protrusion the protein surface is highly charged, and contains the conserved residues Asp-26, Lys-36 and Lys-57. Using oligonucleotide-directed mutagenesis, we have developed thioredoxin variants with each of these charged residues replaced by a glycine residue. The electrostatic potential at the surface of the proteins for the wild-type and each of the variants was calculated using the UCSF Midas molecular graphics program. Changes in the redox activity of these thioredoxin variants are correlated to changes in the surface electrostatic potential in this region.

'pseudo EF-hand' structure -that contains two ammo acids more than the archetypical EF-hand. The studies made so far by *H, 43 Ca and " 3 Cd n.m.r., stopped-flow kinetics of calcium off rates, equilibrium constant measurements of Ca 2+ affinity and thermochemical studies of the Ca 2+ binding processes indicate that the mutations made cause no gross structural changes in the protein, and that mutations in the 'pseudo EF-hand' influence not only the properties of that site but also of the distant second Ca 2+ binding site (the EF-hand). This includes effects on kinetic properties as well as equilibrium properties as reflected in free energies associated with Ca 2+ binding. The affinity labelling of bovine pancreatic ribonuclease A with 6-chloropurine 5'-ribonucleotide (Part's etal.. 1980 ) allowed us to postulate the existence of subsites P2, R3 and B 3 on the active site cleft of the enzyme. For nomenclature of subsites see Richards and Wyckoff (1971) . The study of this reaction in greater detail together with the study of an enzyme-pentanucleotide complex by means of computer graphics indicates that at least five phosphate groups of the RNA molecule can interact with five positive regions of the enzyme. Present on each one is a lysine residue of the active site cleft (sequentially in the 5'-3' direction Lys-104j 66, 41, 7 and 37). The distance between each lysine is -7 -8 A, the same distance as that found between the phosphate groups on the RNA molecule.

133. Protein engineering and biophysical studies of structure-function relationships in bovine calbindin D 9K (ICaBP) S.Forsen, P.Brodin 1 , T.Drakenberg, T.Grundstrom 1 , S.Linse and E.Thulin Chemical Centre, Lund University, Lund, and 'Unit for Applied Cell and Molecular Biology, University of Umea, Umea, Sweden

The 9-kd mammalian intestinal calcium binding protein, calbindin D 9K (ICaBP), also found in kidney and placenta, is a very suitable system for studies of structure-function relationships in 'EFhand' calcium binding proteins because it is the smallest protein of this type with an established high-resolution crystal structure. We have synthesized genes for wild-type and several mutant bovine calbindins (14 at the time of writing) by the 'shot-gun' ligation technique, and produced the proteins in Escherichia coli in 20-100 mg quantities. In one set of 'muteins' we have modified the first Ca 2+ binding site of calbindin -the so-called 134. Pentameric structure of the murine IgM antibody The pentameric structure of IgM is presumed to depend on inter-u chain disulphide bonds involving cysteines at amino acid positions 337, 414 and 575. We have studied the importance of these cysteines by using site-directed mutagenesis to generate mutant u genes encoding serine rather than cysteine at various positions. The mutant and normal u genes are then expressed in appropriate cell lines and the structure and function of the secreted IgM of the mutants can then be compared with wild-type. Two mutants have been studied to date. Ser-575 is a Cys to Ser change at amino acid 575 in Cu-4 and Ser-414 is a similar substitution at amino acid 414 in Cu-3. Both mutants produce pentamer although less efficiently than the wild-type and neither pentamer contains J chain. The mutants differ in that Ser-575 is fully functional in complement-dependent cytolysis while Ser-414 is not. These results show that neither of these cysteines is an absolute requirement for pentamer formation. In addition, J chain is not necessary for the pentamer to mediate cytolysis, although the Cys-414 would appear to be critical for this function.

135. Investigation of the catalytic mechanism of glutamate dehydrogenase with site-directed mutants Glutamate dehydrogenase (GDH; EC 1.4.1.4) of Escherichia coli catalyses the NADP-dependent reductive amination of 2-oxoglutarate to glutamate during assimilation of ammonium. In the proposed general 'acid-base' mechanism of catalysis of bovine GDH two basic groups of pK 7.7 (probably lysines) and one carboxyl group are involved in dicarboxylate substrate binding and catalysis. GDH is unusual amongst dehydrogenases in not possessing a catalytic histidine. Critical residues identified from physico-chemical and in vivo mutation experiments are being investigated in parallel with the X-ray structure determination of two GDHs (E. coli at Leeds, Clostridium symbiosum at Sheffield). Lys-128 (E. coli) is conserved in all GDHs sequenced and is reactive to pyridoxal phosphate with a pK of 7.7. To deter-mine whether Lys-128 is involved in substrate binding or in catalysis, this residue has been altered to His, Arg or Glu by oligonucleotide-directed mutagenesis of the ghdA gene. While not affecting production, gross conformation or apparent stability of the enzyme, these replacements result in altered catalytic activity of GDH. The unexpectedly different properties of the His-128 and Arg-128 are presented. These properties suggest interactions of Lys-128 that are necessary for catalytic selectivity for L-glutamate as a substrate, possibly by transition state stabilization.

136. The use of interactive computer graphics in the simulation and design of site-specific mutations to dihydrofolate reductase Ian Wright Astbury Department of Biophysics, University of Leeds, Leeds LS2 9JT, UK Crystallographic studies of several species variants of dihydrofolate reductase have shown a high degree of structural homology despite the well-documented species-dependent variations in affinity for a selection of inhibitors, particular differences being noticed between vertebrate and invertebrate species. It has been suggested that the basis for these variations lies in differences in the shapes and relative sizes of active sites of the enzymes and the consequent availability of favourable hydrophobic interactions. The work described has involved the selection and computer modelling of specific mutations to key hydrophobic residues in the active site of dihydrofolate reductase from Lactobacillus casei, the object being to design a mutated bacterial enzyme that might display an inhibitor specificity normally characteristic of a vertebrate species variant of the enzyme.

137. The use of site-directed mutagenesis to study the glycolytic enzyme phosphoglycerate kinase P.Minard, P.Walker, D.Bowen, G.Davies, J.Littlechild, L.Hall and H.C.Watson Department of Biochemistry, University of Bristol, Bristol BS81TD, UK Site-directed mutagenesis is being used to study the catalytic mechanism and to investigate the proposed domain movement in the monomeric glycolytic enzyme phosphoglycerate kinase (PGK). The enzyme can be overexpressed using a multicopy yeast vector to yield 50-80% of the total cell protein as 'well-folded' yeast PGK. Several mutants have been constructed and expressed which include a change His-388 -Gin in the 'hinge' region of the protein (Wilson et al., 1987) , His-170 -Asp which forms part of the 'basic-patch' thought to be involved in the 3-PGA binding site and Arg-168 -Lys which is implied in transitionstate stabilization of the enzyme. The effects of these mutations will be discussed. which catalyses the phosphorylation of fructose 6-phosphate to fructose-l.o^bisphosphate. It is inhibited by phosphoenolpyruvate (PEP) and activated by ADP and GDP. Site-directed mutagenesis in the effector site was carried out systematically to alter the binding of effector molecules. We find that mutation of Tyr-55, which possibly interacts hydrophobically with the base of ADP, to phenylalanine does not significantly affect effector binding. However, a 5.5-fold increase in dissociation constant resulted when it was changed into a glycine. The binding of the or-and /3-phosphate groups in ADP is enhanced by a number of positively charged residues. Removal of the side-chains of the relevant residues Arg-21, Arg-25 and Lys-213 causes changes in the binding of both GDP and PEP. This provides evidence that both inhibitor and activator bind in the same site. Glu-187 binds the Mg 2 " 1 " ion coordinated to ADP. Mutation of Glu-Ala-187 leads to a dramatic inversion of allosteric inhibition. At low concentrations of F6P there is a 6-fold increase in activity on binding PEP even though wild-type enzyme is strongly inhibited under the same conditions: the mutant is 100 times more active than wild-type. For an enzyme to be a good catalyst, it must enhance both the rate and yield of the chemical reaction catalysed. The means of rate enhancement is sometimes very amenable to study by protein engineering. Amino acid side-chains believed to be involved in catalysis can be removed by protein engineering, and the resultant effects on the catalytic rate rationalized (see, for example, Trends Biochem. Sci., 11, 321 -326, 1986) . The mechanisms by which the enzyme prevents catalysis of unwanted side-reactions, thus increasing the product yield, present a more complicated problem to investigate. The hydrolysis of the enzyme-bound tyrosyladenylate in the activation reaction of tyrosyl -tRNA synthetase is one such reaction. Kinetic analyses of the hydrolysis reaction in the wild-type and 10 mutant proteins suggest that the hydrolysis transition state shares some structural similarities with the reaction transition state. Proteins mutated at position 48 have much higher rates of hydrolysis. The small size of the hydroxyl nucleophile means that it is a useful probe for structural changes in the active site. From the crystal structure, mechanisms fitting all these data may be proposed, and tested experimentally by l *O distribution studies. Site-specific mutagenesis has been used to produce a mutant form of yeast phosphoglycerate kinase (PGK) in which His-170 has been replaced by an aspartate residue. No difference in activity was observed between the mutant PGK and the wild-type yeast PGK. Comparison of the 500 MHz 'H n.m.r. spectrum of the mutant enzyme with that of the wild-type has enabled us to assign resonances to the C-2 proton of His-170 (with confidence), 167 and 62 (tentatively). All three histidines have been implicated in the fully developed 3-phosphoglycerate (3-PG) binding site. Anion binding has been studied in both the mutant and wild-type enzymes, and dissociation constants were determined for the substrate 3-PG and the paramagnetic shift reagent [Fe(CN) 6 ] 3~. The wild-type PGK binds 3-PG and [Fe(CN) 6 ] 3~ more tightly than the mutant, although only by a factor of two. The resonances assigned above were shifted downfield by addition of 3-PG to both the wild-type and mutant enzymes. This can be attributed to a pK a increase in the substrate (3-PG) binding area of the enzyme. The minimum antibody fragment that contains a complete binding site is designated Fv, comprising the variable heavy (VH) and variable light chain (VL) domains. While the Fv region can be released from antibodies under special circumstances, as in the case of MOPC 315 (Hochman et al., 1973) , it is not generally practical to isolate active Fv regions. In order to circumvent this problem, strategies for preparing biosynthetic Fv regions have been developed. These methods have been applied to the antidigoxin 26-10 monoclonal antibody (Mudgert-Hunter et al., 1982) . The genes corresponding to the VH and VL regions of 26-10 were synthesized with unique restriction sites bracketing the hypervariable segments. The VH and VL genes were expressed in bacteria; protein products were purified and characterized by immunoblotting and protein sequencing. The VH and VL products were utilized in reconstitution experiments to recover biosynthetic 26-10 Fv. The VH and VL genes were also expressed in mammalian cells wherein protein folding and association of polypeptide chains are favoured. Gene expression was assessed by Northern analysis and protein products were immunoprecipitated.

Supported in part by NIH grant No. 1 R43 CA39870-01, funded by NCI. the binding of the bisubstrate analog N-phosphonacetyl-Laspartate (PALA). In particular, the equatorial domain moves towards the polar domain along with a major displacement of the 240's loop, resulting in a closure of the active site. Sitedirected mutagenesis has been used to probe the functional importance of the major interactions which stabilize the two conformations of the enzyme which have low (T) and high (R) affinity for the substrates. Replacements involving the link between Tyr 240 and Asp 271 indicate that this link is essential for stabilization of the T state. Replacements involving the link between Glu 50 and both Arg 167 and Arg 234 indicate that this set of interactions is critical for both cooperativity and the configuration of the active site found in the R state. Analysis of data for a series of single and double mutants suggests that substrate binding to the catalytic chain results in a major reorganization of that chain. However, stenc constraints imposed by the quaternary structure prevent the 240's loop from attaining its final position without a restructuring of the entire molecule. Thus, substrate binding facilitates not only tertiary changes but promotes the concerted allosteric transition.

145. Probing the catalytic mechanism of glyceraldehyde-3 phosphate dehydrogenase by site-specific mutagenesis Improved techniques for site-directed mutagenesis have been developed over the last ten years. Until routine utilization of these techniques for industrial enzyme engineering, much has to be learned from model enzyme studies. We chose the glycolytic glyceraldehyde-3P-dehydrogenase (GAPDH) as a model enzyme. GAPDH is a tetrameric enzyme, with cooperative properties, requiring NAD + as a cofactor. Kinetic, chemical and highresolution X-ray crystallography studies have permitted us to propose a plausible molecular model, which could explain catalysis and substrate specificity. Using a site-specific mutagenesis approach on Bacillus stearothermophilus and Escherichia coh cloned GAPDH genes, we studied the role of the amino acids postulated to be implicated in the catalytic mechanism. The replacement of His 176 by Asn 176 seems to have no effect on the nucleophilic property of Cys 149. In contrast, the catalytic efficiency of the mutant is strongly reduced (l/50th) in accord with an acid-base catalytic role of the His 176 residue. The replacement of Cys 149 by a Ser or Gly residue has permitted us to confirm the implication of the Cys residue in the formation of the acyl intermediate. No activity was observed for the Gly 149 mutant. The Ser 149 mutant had a highly reduced k^ value (1/300th of the wild-type enzyme) which could be explained by a change in the limiting step of the reaction. Replacement of Asn 313 by Thr confirms the implication of this residue in the binding of NAD + . The mutant enzyme had a k^ similar to that of the wild-type enzyme, but its K M for NAD + increased by a factor of 10 (5.10" 4 M). Finally, replacement of Ser 148 by Ala strongly increased the K M for P, and only slighdy affected that for GAP, whereas the k^ is decreased by a factor of 30. This is in favor of the involvement of Ser 148 in P, binding. The potential for preparing an effective pregnancy vaccine based on immunization against the /3-subunit of human chorionic gonadotrophin (HCG) has been demonstrated in studies in primates. However, worldwide availability of such a vaccine will depend upon die ability to produce large amounts of the /3-HCG antigen simply and inexpensively. This may be achieved by protein engineering, which will also allow specific mutagenesis of the /3-HCG sequence in order to minimize immunological crossreaction with the related glycoprotein hormone, luteotrophin (HLH). /3-HCG complementary DNA (cDNA) was subcloned into the lacZ gene of a bacterial expression plasmid. The hybrid plasmid, designated placZMRW838, was constructed to produce a fusion protein between /3-galactosidase and a (3-HCG peptide that lacks the N-terminal signal peptide but includes a unique N-terminal tryptophan at which cleavage of the /3-HCG peptide from the fusion protein can be achieved. Partially purified fusion product, which was subjected to polyacrylamide gel electrophoresis followed by transfer to a nitrocellulose membrane, was detected by both anti-/3-galactosidase and anti-/3-HCG. Preparations of the fusion protein have been used as immunogens in rabbits and are being tested for their anti-HCG activity by in vitro bioassay. Analytical studies are currendy being carried out to optimize die chemical cleavage and separation of the /3-HCG peptide from /3-galactosidase.

147. Expression of a synthetic gene for interleukin-2 receptor in Escherichia coli and higher cells Joachim W.Engels, Petra Artelt 1 , Thomas Beckers, Hans-Jorg Hauser 1 , Dieter Hiisken and Waldemar Wetekam 2 Universitiit Frankfurt, Organische Chemie, Niederurseler Hang, D-6000 Frankjurt-am-Main 50, X GBF, 3300 Braunschweig, and 2 Hoechst AG, 6230 Frankfurt-am-Main 80, FRG By chemoenzymatic synthesis we constructed a des-Met gene for the human interleukin-2 (IL-2) receptor. For its expression two cellular systems were employed. The unmodified protein should be the result of a fusion expression wim /3-galactosidase in Escherichia coli, when cleaved with cyanogen bromide. On the other hand, with the help of a SV-40 shuttle vector die gene was incorporated into HeLa, BHK and mouse L-cells. Functionality of the synthetic gene was established by detecting the receptors on the surface of these cells in a cell sorter. Comparison of authentic cDNA and our synthetic gene in these different cell systems as well as substrate binding with IL-2 and mutants thereof are now underway.

148. Expression and structure-function relationship of nerve growth factor (NGF) R.Gotz, G.Dechant, Y.-A.Barde and H.Thoenen

Mouse nerve growth factor (NGF) is a basic polypeptide of 118 ammo acid residues that plays an important role in the embryonic development and maintenance of the nervous system. We want to produce NGF in order to study its functional and structural properties. We have achieved overproduction of NGF in Escherichia coli cells that had been transformed with plasmid pEN6. In this plasmid, the DNA fragment encoding NGF is inserted in the correct translational reading frame into the lacZ gene of E. coli. Upon induction of the hybrid gene promoter, E. coli accumulates large amounts of a hybrid protein of mol. wt 130 kd that cross-reacted with an anti-NGF antiserum. The NGF peptide released by cyanogen bromide cleavage from the hybrid protein had the same size as NGF isolated from murine submaxillary glands. Experiments dealing with its biological activity and its correct folding will be presented. Furthermore, we are currently raising and testing antibodies directed against different parts of NGF and data from these experiments will be discussed. Conventional methods of toxoid production, which employ heat or chemical inactivation of native toxin, can considerably alter the immunogenic properties of the protein by inducing major conformational changes. Ideally the toxoid structure should resemble that of the toxin as closely as possible. We have used genetic manipulation to inactivate the porcine Escherichia coh heat-labile enterotoxin (LT) without significantly altering the holotoxin configuration. A plasmid carrying the gene encoding the LT 'A' subumt was subjected to random chemical mutagenesis then used to transform E. coli HB101. Clones producing full-length but inactive LTA were sequenced to identify the inactivating mutations. One totally inactive mutant was found to differ from native LTA by only two amino acid residues and was capable of associating into an inactive, but otherwise normal, holotoxin structure with unmutated LT 'B' subunits. Intramuscular injections of crude and purified preparations of this toxoid into pigs have been shown to induce formation of neutralizing antibodies against native LT. Site-directed mutagenesis has been used to demonstrate that only one of the two amino acid changes found is necessary for abolition of activity. We are studying the expression of heterologous genes and se-cretion of their products in Escherichia coli by using human a-1-interferon as a model protein. We have inserted the interferon gene under the control of bacteriophage lambda pL-promoter. The gene was placed downstream from a synthetic SD sequence and initiation codon without changing its DNA sequence. With this construction it is possible to study the transcription and translation of foreign gene in E. coli, especially the effect of the 5' end of mRNA on the translation efficiency. Another construction was made by using the observation that Bacillus stearothermophilus a-amylase is produced by E. coli cells in certain conditions. By adding the a-amylase gene to the interferon vector we have made a construction with which we are able to study the localization of foreign proteins in E. coli.

151. Protein engineering of human interferon-7 W.C.Leung Departments of Medicine, Medical Microbiology and Infectious Diseases, University of Alberta, Edmonton, Alberta, Canada T6G 2G3

We previously have chemically synthesized the structural gene for human interferon-7 (HuIFN-7) with modified nucleotide sequences to accommodate additional restriction endonuclease sites. Site-specific mutants can then be generated by the fragment replacement method. The gene was expressed in Escherichia coli and served as a quick screen method. The mutants of interest were then ligated under a SV40 promoter and used to transfect the CHO cells. The intracellular as well as secreted form of expressed HuIFN-7 activity were assayed. Since the three-dimensional structure for HuIFN-7 was not available, we based the mutagenesis strategy on a secondary structure predicted from the primary sequence. The first objective is to localize regions on the protein molecule which affect its biological activities. We chose to generate mutants with minimal, but definite, conformational changes. Out of 30 mutant genes constructed, seven of them were able to re-express into biologically active molecules. These mutants include deletion of N-terminal Gin; deletion of amino acids 27-30, substitution of Gly with Glu at amino acid 31; and insertion of mutants at 121 and 122 to extend the a-helix, etc. The antiviral activities of these mutants exhibit significant changes, ranging from 4-fold reduction to 10-fold enhancement. This study served to illustrate the feasibility of protein engineering on HuIFN-7 which will result in construction of mutants with significant changes in biological activity.

152. Isolation and characterization of normal and activated human antihaemophilic Factor vm N.Bihoreau, P.Paolantonacci, V.Game and H. Van de Pol Centre National de Transfusion Sanguine, 91943 Les Ulis, France Normal and activated Factor VIE were purified from commercial concentrates by immunoaffinity chromatography with an anti-VTJI:CAg monoclonal antibody bound to Sepharose. Different fractions of Factor VTII were isolated by f .p.I.e. on Mono Q gel. We have identified, by SDS-PAGE, the presence of the light chain of mol. wt 78-80 kd in all active fractions, together with a different size distribution of the heavy chains, ranging from Posters mol. wt 200 to 110 kd. The smallest active Factor VHI complex, consisting of the 90-kd polypeptide associated with the 80-kd chain can be eluted in a separate peak from the anion exchanger. A 3-fold activation of Factor VIE in plasma concentrate starts with a generation of the 70-kd chain. This activation is correlated, on the chromatographic pattern, with the absence of the last peak (corresponding with the 90-80-kd active Factor VOT). We have separated different activated heterodimers and noticed that the 70-kd proteolysed polypeptides are present in all fractions, and were associated with the different, fragmented heavy chains. Furthermore, the presence of the 78 -80-kd doublet only with the higher mol. wt chains suggests that the B region protects this doublet from thrombin activation.

Roy A.Quinlan, Simon R.Clarke and Murray Stewart MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK Myosin, intermediate filament proteins and lamins are major constituents of their respective filament systems which are important for cell architecture, movement and nuclear envelope organization. All contain extensive a-helical-coiled coils and interaction between these is fundamental to filament integrity. The three-dimensional molecular packing has been inferred from sequence analyses and appears similar in each case. Highresolution structural studies require fragments which cannot always be made proteolytically. For intermediate filament proteins and lamins there is also the problem of producing sufficient quantities for biochemical and structural studies. Expression of genetically engineered fragments circumvents many of these problems. We have compared two expression vectors. One is based on the pLcII system, where fragments are expressed as a fusion protein with the X ell protein (Nagai and Thogerson, 1984) . The fragment is removed from the fusion protein with Factor X. Its cleavage site was engineered between the X ell protein and the fragment. The second system is based on lpp vectors where expression is under the control of the lac UV5 promoter (Masui et al., 1984) . The product is again a fusion product, and so a Factor X site was similarly incorporated. Various DNA fragments from myosins, lamins and glial filament protein have been inserted into these vectors. With lamins and the nematode pharyngeal myosin the fusion products represented up to 10% of total cell protein and were incorporated into inclusion bodies. Fusion proteins have been purified and these fragments are being used for biochemical and structural studies.

Proc Natl. Acad Set. USA

Submitted Kabsch.W and Sander.C (1983a) FEES Lett

Proc. Nail Acad Sa USA

press References Chakhmakhcheva

press Igarashi et al

A common, conserved structure between spectrin and a-actinin References Jumak

press References Paies,X , Uorens.R , Anis.C and Cuchillo.C M

A model for the concerted allosteric transition of aspartate transcarbamylase as deduced by site-directed mutagenesis

USA Profound tertiary and quaternary changes in the structure of the Escherichia coli aspartate transcarbamylase are observed upon References Masui et al

Using site-directed mutagenesis, Arg-171 at the substrate-binding site of Bacillus stearothermophilus LDH has been replaced by lysine. In the closely homologous eukaryotic LDHs this residue binds the carboxylate group of the substrate by forming a planar, bifurcated bond. The mutation diminishes the binding energy of pyruvate, a-ketobutyrate and a-ketovalerate (measured by k^J ATj n) by the same amount (~6 kcal/mol). For each additional methylene group on the substrate there is a loss of -1.5 kcal/mol of binding energy in both mutant and wild-type enzymes. From these parallel trends in the two forms of enzyme we infer that the mode of productive substrate binding is identical in each, the only difference being the loss of a strong carboxylate-guanidinium interaction in the mutant. In contrast to this simple pattern in kaJKm, the K m alone increases with substrate size in the wildtype enzyme but decreases in the mutant. These results can be most simply explained by the occurrence of tighter, unproductive enzyme-substrate complexes in the mutant enzyme as the substrate alkyl chain is extended. This does not occur in the wildtype enzyme because the strong orienting effect of Arg-171 will maximize the frequency of substrates binding in the correct alignment.139. Hydrogen bonding of the carboxyamide of NADH is not important for catalysis in lactate dehydrogenase D.B.Wigley, A.R.Clarke and J.J.Holbrook Department of Biochemistry, Bristol University Medical School, University Walk, Bristol BS8 1TD, UK Hydride transfer in dehydrogenases is stereospecific -occurring from either the A or B face of the nicotinamide ring of NADH. Crystal structures of dehydrogenase complexes with NADH show this to be due to orientation of the coenzyme such that hydride transfer is more favourable from one face of the nicotinamide ring. It has been proposed from the crystal structures of ternary complexes of lactate dehydrogenase that the orientation of the nicotinamide ring is imposed by a hydrogen bond between the carboxyamide of the coenzyme and the hydroxyl of Ser-163. By changing Ser-163 to alanine we have removed the possibility of forming this hydrogen bond Both k^ and affinity for NADH are unaltered in the mutant enzyme. Therefore Ser-163 does not appear to be as important as previously suggested.140. Site-directed mutagenesis in the effector site of Escherichia coli phosphofructokinase F.T.K.Lau and A.R.Fersht Department of Chemistry, Imperial College, London SW7 2AY, UK

Arthrobacter glucose isomerase is a good industrial enzyme but 264 was never commercialized because of EEC tariffs on high fructose corn syrups (HFCS). The enzyme requires only low Mg 2 " 1 " and is stable to 80°C. It has been crystallized and its tertiary structure has been determined at 2.7 A. It is a /3-barrel almost identical to the Streptomyces glucose isomerase. We have developed a host-vector and transformation system for the commercial organism that is as efficient as those for Escherichia coli, and are currently cloning the gene. Our protein engineering target is to replace surface acidic groups with Lys or Arg residues so as to lower the pH optimum, whilst retaining or improving thermostability with other replacements if necessary. Current HFCS production is at 50-60°C only, since unwanted byproducts are produced at higher temperatures at the optimum pH of the enzyme (7-8). There is also scientific interest in using this protein as a model to map changes in surface electrostatic fields consequent on the charge replacements. Hence we are using 'competitive labelling' techniques to determine p^T a s of individual amino and carboxyl groups. The leech protein, hirudin, is a highly specific tight-binding inhibitor of the procoagulant protease, a-thrombin (k^ = 20 fM). Native hirudin is a 65-amino acid residue protein containing a sulphated tyrosine at position 63. The production of recombinant hirudin for possible pharmaceutical use is currently an active area of research. A desulphated form of hirudin has been expressed at high yield in yeast. This form has a k^ for thrombin binding which is 10-fold higher than native hirudin. Preparations of hirudin from yeast culture also contain hirudin forms which lack one or two amino acids from their C-terminus. These forms show small decreases in their affinity for thrombin. Synthetic genes coding for artificial proteins with predefined and nutritionally valuable amino acid compositions have been constructed and cloned into Escherichia coli plasmid vector pKK233-2. The genes are constructed from three easily interchangeable cassettes encoding either essential (cassette A), nonessential (cassette B) or branched chain (cassette H) amino acid residues. Furthermore, an AF~B gene was constructed using cassette mutagenesis, which contains all the amino acids in optimal proportions but lacking phenylalanine. After recloning into the newly developed expression vectors pBI052 and pGFY221N they could be expressed in an in vitro transcription-translation system and some of them in a minicell-producing E. coli strain.

To force expression of the artificial polypeptides, the AHB gene (containing one copy of each cassette) and the A6 gene (consisting of six copies of concatemerised A cassette) were cloned into pUR300, a /3-Gal fusion vector and expressed as fusion proteins /3-Gal-AHB and 0-Gal-A6 in E. coli strain BMH71-18. Analysis of the protein synthesis in vivo realized in minicells showed differentiate product stability of engineered proteins in living bacterial cells. Our report is an example of expression of completely novel, not naturally occurring proteins which have not been optimized during the evolution process.

Fragmentation of thermolysin as a tool to detect features of its structure, stability and dynamics C.Vita, D. Dalzoppo 

Limited proteolysis by subtilisin or autolysis by EDTA or heat treatment of the metalloendopeptidase thermolysin isolated from B. thermoproteolyticus allowed isolation of 'nicked' protein species constituted of two as well as three protein fragments associated in stable complexes. In particular, autolysis of thermolysin in the presence of EDTA at neutral pH leads to peptide bond fissions at the chain loop 190-205 involved in the binding of Ca-4 in the native protein, suggesting an important stabilizing role for this calcium ion, which is missing in the homologous heat-labile neutral protease from B. subtilis. An examination of the three-dimensional model of the thermolysin molecule reveals that all sites of chain fragmentation occur at exposed loops, and not within chain segments of regular secondary structure (helices, sheets). Moreover, a striking correlation is observed between sites of fragmentation and sites of segmental mobility as given by the temperature factors (B values) determined crystallographically, thus demonstrating that protein mobility seen in protein crystals is relevant to motion in solution.residue would take place when, and only when, the accessible area of the lateral chain of the second amino acid is inferior to 147 Az. To probe the model, we performed systematic substitution of the second amino acid of methionyl-tRNA synthetase by site-directed mutagenesis. An original expression shuttle vector, based on protein fusion with /3-galactosidase, was designed to facilitate rapid mutant protein purification and Nterminal microsequencing. Now, several substitutions have been studied to completion and the experimental results are in full agreement with the proposed model. In particular, they confirm the predictions based on the model concerning amino acids not yet found in the second position in bacterial proteins. P.H. is supported by a predoctoral fellowship from Rhone-Poulenc Recherches S.A.Computer-assisted analyses of the secondary structure of viral and cellular receptor-like proteins Y.Becker

The availability of amino acid sequences in computer data banks and a program which utilizes the Chou-Fasman and Robson-Gamier prediction of the secondary structure of proteins made possible analyses of membrane-anchored viral and cellular proteins, some having oncogene-like properties. Using the software developed by the University of Wisconsin Genetic Computer Group and a program for the prediction of protein secondary structure (Jamieson and Wolf, 1986), we studied the properties of (i) an EB V latent membrane protein with oncogenelike properties and related viral glycoproteins; (ii) an insulin receptor which resembles tyrosine kinase oncogenes; (iii) an estrogen-receptor; (iv) an interleukin-2 receptor and (v) a fibronectin receptor. The computer program was also used for the prediction of antigenic domains in viral and cellular proteins.Methionyl-aminopeptidase: processing of the initiator methionine of Escherichia coli proteins Ph-Herv6 Hirel, Jean-Marie Schmitter, Philippe Dessen, Guy Fayat and Sylvain Blanquet Laboratoire de Biochimie, UA240 CNRS, Ecole Polytechnique, 91128 Palaiseau Cedex, France Less than 40% of mature E. coli proteins actually retain their initiator methionyl residue. Computer analyses of about 100 chemically determined amino-terminal sequences of E. coli proteins were performed. This allowed us to class protein Nterminal sequences into two categories, corresponding to processed and non-processed polypeptide chains. From this analysis, we could deduce a molecular enzymatic model, accounting for the occurrence or non-occurrence of the maturation process, and consistent with existence of a unique enzyme. According to our model, the excision of the initiator methionyl