Domain swapping continues to be the most successful area of protein engineering. Complementary structural and functional information is now beginning to emerge from studies that aim to redesign dehydrogenases, proteases and metal-binding sites. One important conclusion from this work is that residues that do not contact the substrates can be important determinants of substrate specificity. These studies suggest an intriguing possibility: new enzymes may be more successfully engineered by selecting first for catalysis, then optimizing substrate binding, instead of attempting to adapt binding sites
for catalysis.
Current Opinion in Structural
Introduction
The tantalizing prospect of creating proteins with any desired function drives the protein engineering field. The most successful engineering strategies involve rear- ranging entire functional domains of proteins. Indeed, such domain swapping experiments are now routine. In contrast, redesigning enzymes with novel activities is a far less exact proposition. This uncertainty arises fro/rff our current lack of understanding of the forces tha~ govern enzyme-substrate interactions and protein stability [1].
In principle, novel enzymes can be designed simply by tinkering with their substrate-binding sites. For example, enlarging a binding site should create an enzyme specific for larger substrates; reducing the binding site should fa- vor smaller substrates, and specificity for charged sub- strates should simply be dependent on charge comple- mentarity. In practice, these types of experiments often result in altered substrate specificity because the enzyme is crippled and its activity toward the original substrate is decreased, rather than its activity toward a new substrate enhanced. This difficulty arises because protein function and stability have evolved together, so that many, if not all, protein residues have both structural and functional roles. Therefore substrate specificity is rarely the product of a small cassette of structural changes at the point of enzyme-substrate contact.
The complexity of the redesign problem necessitates a multifaceted approach involving thorough kinetic and structural characterization of both parent and mutant proteins. Ideally, enzymatic activity toward a panel of possible substrates should be analyzed to permit detec- tion of unsuspected specificity changes. The structures of both free enzyme and enzyme-substrate/inhibitor com- plexes should also be determined to assess the struc- tural consequences of mutations and their effect on enzyme-substrate interactions. Mutations should be an- alyzed in several proteins that have similar structures in order to draw generalizations about the structural basis
Biology 1994, 4:608-611
of substrate specificity. Families of proteins with varying substrate specificities are therefore attractive systems for this type of analysis. Modeling based on such sequence homology has identified the structural determinants of many enzymes, including most recently cholinesterases [2°,3",4], cytochrome P450s [5 °] and lectins [6,7].
Domain swapping
Although recombining ligand-binding domains and other regulatory units has become routine methodology, two elegant and insightful experiments of the past year are worth mentioning. In t'fie first, a novel receptor was designed that lacked an extracellular ligand-binding do- main [8°°]. The intracellular signaling domain of the T cell antigen receptor was instead fused to several copies of FKBP12, a protein which binds the immunosup- pressant FK506. This intracellular receptor is tethered to the cytosolic side of the membrane by an amino- terminal myristoylation, and thus is only accessible to membrane-soluble ligands; no mechanism exists for its activation at the cell surface. FKS06 does not activate the signaling pathway, presumably because FK506 bind- ing can not oligomerize the receptors. In contrast, the receptor is activated by a bidentate ligand consisting of two crosshnked FK506 molecules. This experiment supports the idea that oligomerization of the intracellu- lar domain of the receptor is required for signaling, and indicates that ohgomerization is a major mechanism of signal transduction from the cell surface. More impor- tantly, this methodology promises to be a powerful tool for the manipulation of signaling pathways in vivo.
The second experiment involved the coupling of the DNA-binding domain of the yeast regulatory protein GCN4 to a metal ion complex. Cuenoud and Schepartz [9°°,10] created novel DNA-binding proteins by using bisterpyridyliron(II) complexes to replace the dimer-
ization domain of GCN4. These engineered proteins recognize different DNA sequences, even though the DNA-binding domain is unchanged. This result shows that DNA recognition can also be modulated by a frame- work which does not directly contact the substrate.
Structural determinants of NAD and NADP specificity
The structural features of enzymes that discriminate be- tween NAD and NADP are among the best understood examples of specificity determinants that have been ma- nipulated by protein engineering. Coenzyme specificity seems to be controlled by a relatively small set of amino acid residues. Many NAD- and NADP-binding domains have a common ~0t~ct~ structural motif. The loop be- tween the first ~-sheet and the first 0t-helix contains a se- quence known as the fingerprint: Gly-X-Gly-X-X-Gly (where X is any amino acid) in NAD-specific enzymes; and Gly-X-Gly-X-X-Ala in NADP-specific enzymes. In NAD-specific enzymes, the end of the second [~-sheet contains an acidic residue, which forms hydrogen bonds to the 2' and 3' hydroxyls of the adenosine moiety of NAD. The first Gly-X peptide bond of the fingerprint region forms hydrogen bonds with this carboxylate, cre- ating a network of interactions that recognize NAD. This acidic residue is absent in NADP-specific enzymes. In- stead, the connecting loop between the second ~-sheet and the second 0t-helix contains Lys and Arg residues, which form a binding site for the 2' phosphate of NADE Glutathione reductase, an NADP-dependent enzyme, is converted to an NAD-dependent enzyme by substitu- tions that remove the positively charged residues, replace the Ala with Gly in the fingerprint region and provide a Glu at the end of the second ~-sheet [11"]. This mutant enzyme has an 8-fold higher preference for NAD, com- pared with the 103-fold higher preference for NADP in wild type glutathione reductase. However, this mutant has only 3 % of the activity of the wild type. The X-ray crystal structure of this mutant NAD-dependent enzyme has recently been solved in the presence and absence of NADH [12"]. The largest structural perturbation in the mutant compared with the wild type enzyme is in the fingerprint region. Substitution of Ala with Gly caused rotation of the first Gly-X bond, permitting formation of the hydrogen-bonding network that recognizes NADH. The structure confirms the role of the fingerprint re- gion in stabilizing the hydrogen-bonding interactions between the enzyme and the ribose of NADH.
Two groups have recently reported the complementary conversion of an NAD-utilizing enzyme to one with a preference for NADP while maintaining wild type lev- els of activity [13°°,14°]. These experiments confirm that coenzyme specificity is determined by the loop between the second ~-sheet and the second 0t-hehx. In the first, substitution of seven amino acids in this loop in NADH- specific malate dehydrogenase is suflqcient to change
coenzyme specificity from a 20-fold higher preference for NADH in wild type enzyme to a 20-fold higher preference for NADPH in the mutant [13°°]. Charac- terization of this mutant revealed that the largest effect of these substitutions was not on the Km for NADH and NADPH, but on the K m for oxaloacetate, although the oxaloacetate-binding site is removed from alterations of the coenzyme-binding site.
In the second experiment, Perham and colleagues [14 °] converted NAD-dependent lipoamide dehydrogenase to a NADP-dependent enzyme, also by making substitu- tions in this loop region. The NADP-dependent lipo- amide dehydrogenase has a 200-fold higher preference for NADP over NAD, compared with a 103-fold pref- erence for NAD for the wild type enzyme. Interestingly, no improvement was noted when the Gly--)Ala substi- tutions in the fingerprint region were included with the loop substitution, which suggests that the primary func- tion of the fingerprint region is the recognition of NAD.
Although these results substantiate the role of the fin- gerprint and loop regions in determining NAD/NADP specificity, it is important to note that other regions also play a role in substrate discrimination. NAD-dependent glyceraldehyde 3-phosphate is converted into an enzyme which recognizes both coenzymes by substitutions out- side the ~[~ct[~ region [15]. This result shows that sev- eral solutions often exist for a given protein engineering problem.
Protease specificity
The creation of proteases with restriction enzyme-like specificity has long been a goal of protein engineering. Protease engineering is particularly challenging, because alterations of one binding subsite often have unsuspected effects at adjacent subsites. For example, the specificity ofsubtilisin is changed dramatically by increasing the size of the $4 binding site, either by mutation of' Ile107 to Gly or by mutation of Tyrl04 to Ala [16°°,17]. Either of these single mutations selectively decreases enzymatic hydrolysis ofsubstrates containing Ala at the P4 position, while maintaining high activity on substrates with Phe at P4. Both of these mutations, however, destabilize the $1 binding pocket, as shown by a decrease in activity on single amino acid substrates. These results demonstrate that the $4 and $1 subsites are functionally linked, but the effect of these mutations on the specificity of the $1 binding site is unclear. Moreover, combinations of these two mutations are not additive.
Trypsin is the patriarch of a family that includes proteases of every imaginable substrate specificity. The specificity of trypsin for positively charged residues is apparently determined by an Asp at the bottom of the S1 binding pocket [18]; however, substitution of this Asp, in addition to the remaining residues of the $1 binding pocket, with the analogous residues of chymotrypsin does not transfer
610 Engineering and design
the specificity of chymotrypsin for Phe-containing sub- strates [19"°]. Trypsin is converted to an enzyme with chymotrypsin-like specificity by the exchange of two surface loops which connect to the walls of the $1 bind- ing pocket, but do not contact the substrate [19°']. This mutant, Tr--~Ch[SI+LI+L2], is defective in substrate- binding but processes enzyme-bound substrate at wild type rates. This defect in substrate-binding is a con- sequence of a deformed $1 binding site (L Hedstrom, JJ Perona, 1KJ Fetterick, unpublished data). This result suggests that it may be easier to design new enzymes by optimizing substrate-binding after catalytic activity is achieved, rather than the converse. Further characteriza- tion of this mutant enzyme reveals that the specificities of the adjacent subsites are also altered, although this site is far from the structural alterations [20"].
The unpredictability of these results suggests that random mutagenesis can provide important information about protease specificity. Two such experiments suggest that specificity for charged residues may be achieved with- out engineering a complementary charge. In the first, random mutagenesis of the $1 binding site of car- boxypeptidase Y produced an enzyme with increased activity for Lys-containing substrates, decreased activity for Ala-containing substrates and little effect on the ac- tivity of Phe-, Ser- and Glu-containing substrates [21]. A single mutation was responsible for this altered speci- ficity: Leu178--+Ser. In the second experiment, random mutagenesis of the $1 binding pocket of 0t lytic protease produced an enzyme with specificity for substrates con- taining His or Met at the P1 position [22"]; the key mu- tation appears to be the substitution of His for Met in the $1 binding pocket.
Protease engineering carries an extra burden because of the enormous number of substrates and the potential for context dependence. Specificity changes require, there- fore, analysis of large numbers ofsubstrates to obtain re- liable conclusions. One important recent achievement is the development of a statistical method for determining relative values of kcat/Km from mixtures of competing substrates [23"]. This method requires only that concen- trations of substrates can be monitored and is applicable to any enzyme system.
Construction of metal-binding sites
The design of low affinity metal-binding sites has be- come routine, requiring the introduction of two or three ligands (usually His) with the appropriate orien- tation. Zinc sites have been engineered into antibodies [24], changed into blue copper sites [25",26], and created from magnesium-binding sites [27]. The structural con- sequences of altering the ligands of carbonic anhydrase, the prototypical zinc enzyme, are currently being ex- plored [28]. In addition, the structure of an engineered,
metal-activated trypsin has recently been reported [29]. These types of studies should provide information that will allow the construction of high affinity metal-binding sites with new activities.
Conclusions
Enzymes catalyze reactions with amazing efficiency and incredible specificity. These two properties are often considered separately, which has spawned the idea that novel enzymes can be designed simply by tinkering with substrate-binding sites. Instead, networks of in- teractions determine comparatively subtle structural al- terations which account for substrate discrimination.
The challenge of protein engineering is to alter substrate specificity while maintaining wild type levels of func- tional activity. Not surprisingly, this goal can be accom- plished most easily by swapping independent domains to construct proteins with new combinations of functions. The next most successful arena has been the manipu- lation of recurring structural motit~ such as coenzyme- binding sites. However, the rational redesign ofproteases remains difficult to achieve.
Acknowledgements
L Hedstrom is supported by a grant from the Lucille P Markey Charitable Trust for the Structural Biology and Biochemistry Program at Brandeis University.
References and recommended reading
Papers of particular interest, published within the annual period of review, have been highlighted as: • of special interest • - of outstanding interest
1. Bone R, Agard DA: Mutational Remodeling of Enzyme Speci- ficity. Methods Enzymol 1991, 202:643-671.
2. Harel M, Sussman JL, Krejci E, 8on S, Chanal P, Massoulie • J, Silman I: Conversion of Acetylcholinesterase to Bulyryl-
cholinesterase: Modeling and Mutagenesis. Proc Natl Acad Sci USA 1992, 89:10827-10831.
The first of three independent studies that delineate the structural deter- minants of cholinesterase specificity using modeling based on sequence homology.
3. Vellom DC, Radic Z, Li Y, Pickering NA, Camp S, Taylor P: • Amino Acid Residues Controlling Acetylcholinesterase and Bu-
tyrylcholinesterase Specificity. Biochemistry 1993, 32:] 2-17. A thorough functional analysis of mutant acetylcholinesterases that delin- eates the roles of several residues in determining specificity for substrates and inhibitors.
4. Ordentlich A, Barak D, Kronman C, Flashner Y, Leitner M, Segall Y, Ariel N, Cohen S, Velan B, Shafferman A: Dissec- tion of the Human Acetylcholinesterase Active Center Deter- minants of Substrate Specificity. Identification of Residues Con- stituting the Anionic Site, the Hydrophobic Site and the Acyl Pocket. J Biol Chem 1993, 268:1 7083-1 7095.
5. Halpert JR, He Y: Engineering of Cytochrome P450 2B1 Speci- e ficity. Conversion of an Androgen 16~Hydroxylase to a 15cc-
Hydroxylase. J Biol Chem 1993, 268:44534457. The hydroxylation of alkanes is a very difficult synthetic reaction, yet cytochrome P450s can hydroxylate their substrates with great specificity under very mild conditions. Novel P450s are therefore greatly desired, and this paper is the latest example of P450 engineering.
6. Drickamer K: Engineering Galactose-Binding Activity into C- type Mannose Binding Protein. Nature ]992, 360:183-186.
7. Weis WI, Drickamer K, Hendrickson WA: Structure of a C-type Mannose Binding Protein Complexed with an Oligosaccharide. Nature 1992, 360:127-134.
8. Spencer DM, Wandless TJ, Schreiber SL, Crabtree GR: Control- o. ling Signal Transduction with Synthetic tigands. Science 1993,
262:1019-1024. An truly elegant example of protein engineering where a novel intracellu- lar receptor is constructed from the signaling domain of the T cell antigen receptor and the FKS06-binding domain.
9. Cuenoud B, Schepartz A: Altered Specificity of DNA-Binding • - Proteins with Transition Metal Dimerization Domains. Science
1993, 259:510-513. Replacing the dimerization domain of a bZIP DNA-binding protein with a metal ion complex creates proteins with different DNA-binding prop- erties, even though the DNA-binding domain is unchanged.
10. Cuenoud B, Schepartz A: Design of a Metallo-bZIP Protein that Discriminates between CRE and APt Target Sites: Selection against AP1. Proc Natl Acad Sci USA 1993, 90:1154-1159.
11. Scrutton NS, Berry A, Perham RN: Redesign of the Coenzyme • Binding Specificity of a Dehydrogenase by Protein Engineering.
Nature 1990, 343:38-43. This conversion of NADP-dependent glutathione reductase to an NAD- dependent enzyme is the first successful redesign of a coenzyme-binding site.
12. Mittl PRE, Berry A, Scrutton NS, Perham RN, Schulz GE: Struc- • tural Differences between Wild-type NADP-dependent Glu-
tathione Reductase from Escherichfa coil and a Redesigned NAD-Dependent Mutant. J Mol Biol 1993, 231:191-195.
The structure of an NAD-dependent mutant giutathione reductase in the presence and absence of NADH reveals that NAD specificity is deter- mined by a network of hydrogen bonds that extends at least 10 A from the NAD-binding site.
13. Nishiyama M, Birktoff JJ, Beppu T: Alteration of the Coenzyme • • Specificity of Malate Dehydrogenase from Thermus flavus by
Site Directed Mutagenesis. J Biol Chem 1993, 268:4656-4660. The conversion of an NAD-dependent dehydrogenase to an NADP-de- pendent dehydrogenase is described. The mutant enzyme has wild type levels of activity and substrate discrimination. Importantly, substrate dis- crimination is the consequence of improvements in the K m of the second substrate, oxaloacetate.
14. Bocanegra JA, Scrutton NS, Perham RN: Creation of an NADP- • Dependent Pyruvate Dehydrogenase Multienzyme Complex by
Protein Engineering. Biochemistry 1993, 32:2738-2740. This paper reports the more successful complementary transformation of lipoamide dehydrogenase, an NAD-dependent relative of glutathione re- ductase, to an NADP-dependent enzyme with wild type levels of activity and substrate discrimination.
15. Corbier C, Clermont S, BiHard P, Skarzynski T, Branlant C, Wonacott A, Branlant G: Probing the Coenzyme Specificity of Glyceraldehyde-3-phosphate Dehydrogenases by Site Directed Mutagenesis. Biochemistry 1990, 29:7101-7106.
16. Rheinnecker M, Baker G, Eder J, Fersht AR: Engineering • • a Novel Specificity in Subtilisin BPN'. Biochemistry 1993,
32:1199-1203.
Engineering for redesign Hedst rom 611
This work illustrates the difficulty of protease engineering: subsites are functionally linked, so that mutations in one subsite trigger changes in adjacent subsites.
17. Rheinnecker M, Eder J, Pandey PS, Fersht AR: Variants of Subtilisin BPN' with Altered Specificity Profiles. Biochemistry 1994, 33:221-225.
18. Perona lJ, Tsu CA, McGrath ME, Craik CS, Fletterick RJ: Relo- cating the Negative Charge in the Binding Pocket of Trypsin. J Mol Biol 1993, 230:934-949.
19. Hedstrom L, Szilagyi L, Rutter WJ: Converting Trypsin to e. Chymotrypsin: the Role of Surface Loops. Science 1992,
255:1249-1253. Two surface loops which do not contact the substrate are critical deter- minants of substrate specificity in the trypsin family of serine proteases. Careful kinetic analysis reveals that the mutant proteases are defective in substrate binding, but can process enzyme bound substrates at rates comparable to wild type trypsin.
20. Schellenberger V, Turck CW, Hedstrom L, Rutter WJ: Map- • ping the S' Subsites of Serine Proteases using Acyl Trans-
fer to Mixtures of Peptide Nucleophiles. Biochemistry 1993, 32:4349-4353.
The $1' subsite specificity of trypsin is changed by mutations that alter the specificity of the $1 binding site.
21. Olesen K, Keilland-Brandt MC: Altering Substrate Preference of Carboxypeptidase Y by a Novel Strategy of Mutagenesis Elim- inating Wild Type Background. Protein Eng 1993, 6:409-415.
22. Graham LD, Haggett KD, Jennings PA, Le Brocque DS, Whit- • taker RG: Random Mutagenesis of the Substrate Binding Site
of a Serine Protease can Generate Enzymes wilh Increased Ac- tivities and Altered Primary Specifieities. Biochemistry 1993, 32:6250-6258.
The $1 binding pocket of ~ Lytic protease was randomly mutated and the resulting library of enzymes was screened with a wide range of substrates. The result was a protease with a specificity for substrates containing his- tidine or methionine at the P1 site.
23. Schellenberger V, Siegel RA, Rutter WJ: Analysis of Enzyme • Specificity by Multiple Substrate Kinetics, Biochemistry 1993,
32:4344-4348. A general method for extracting relative values of kcat/K m from the anal- ysis of substrate mixtures.
24. Wade WS, Koh JS, Han N, Hoekstra DM, Lerner RA: Engineer- ing Metal Coordination Sites into the Antibody Light Chain. J Am Chem Soc 1993, 115:4449-4456.
25. Canters GW, Gilardi G: Engineering Type 1 Copper Sites in • Proteins. FEBS Lett 1993, 325:39-48. A good general review of the field.
26. Lu Y, LaCroix LB, Lowery MD, Solomon El, Bender CJ, Peisach J, Roe JA, Gralla EB, Valentine JS: Construction of a 'Blue' Cop- per Site at the Native Zinc of Yeast Copper-Zinc Superoxide Dismutase. J Am Chem Soc 1993, 115:5907-5918.
27. Murphy JE, Xu X, Kantrowitz ER: Conversion of a Magnesium Binding Site into a Zinc Binding Site by a Single Amino Acid Substitution in Escherichia coil Alkaline Phosphatase. J Biol Chem 1993, 268:21497-21500.
28. Alexander RS, Kiefer LL, Fierke CA, Christianson DW: Engi- neering the Zinc Binding Site of Human Carbonic Anhydrase Ih Structure of the His94 -~Cys Apoenzyme in a New Crystal Form. Biochemistry 1993, 32:1510-1518.
29. McGrath ME, Haymore BL, Summers NL, Craik CS, Fletter- ick RJ: Structure of an Engineered, Metal-Activated Switch in Trypsin. Biochemistry 1993, 32:1914-1919.
L Hedstrom, Graduate Department of Biochemistry, Brandeis University, Waltham, Massachusetts 02254, U S A.
