- Email: [email protected]
Engineering of tyrosyl tRNA synthetase
BIOCHIMIE, 1985, 67, 737-743
Engineering of tyrosyl tRNA synthetase. Hugues BEDOUELLE*, Paul CARTER*, Mary M.Y. WAYE*, Greg WINTER*, Denise M. LOWE**, Anthony J. W I L K I N S O N * * & Alan R. FERSHT**.
* Laboratory of Molecular Biology, Hills Rd, Cambridge CB2 2QH. ** Department of Chemistry, Imperial College, London SW7 2AY. (Refu le 19-3-1985, accept~ le 17-4-1985).
R6sum6 - - Le gkne codant pour la tyrosyl tRNA synthdtase de Bacillus stdarothermophilus a dtd systdmatiquement altdrd en utilisant des oligonucldotides synthdtiques comme mutag~nes. La construction des mutations a dtd facilitde par l'utilisation de souches bactdriennes ddfectives pour la rdparation des misappariements et aussi en utilisant une souche de Ml3 possddant un marqueur gdndtique (tel qu'une mutation ambre, ou un site EcoK ou EcoB) qui permet de sdlectionner la progdniture du phage issue du brin moins (mutagdnisd) lors de la triplication. Plusieurs mutations ont ~td construites dans le site de fixation de l'ATP pour dlucider le r61e de rdsidus particuliers dans la catalyse et la fixation du substrat et il a m~me dtd possible de construire des mutants qui ont une plus grande affinitd pour I'ATP. Des mutations changeant des rdsidus lysine ou arginine situds h la surface de l'enzyme nous ont permis d'identifier des pohlts de contact potentiels avec le tARN; ces mutations indiquent qu'un groupe de rdsidus basiques, pros de l'extrdmitd C-terminale de i'enzyme, interagit de fafon importante avec le tARN. Mots-cl~s : g~nie g~n~tlque / mutag~n~se dirig~e / tRNA synth~tase / specificitY.
S u m m a r y - - The gene encoding the enzyme tyrosyl tRNA synthetase from Bacillus stearothermophilus has been systematically altered using synthetic oligonucleotides as mutagens. The construction of mutations has been facilitated by using strains of bacteria defective in mismatch repair and also by utilising a genetic marker in the MI3 strain (such as an amber mutation, or an EcoK or EcoB site) which allows selection for the progeny of M13 replication derived from the minus (mutagenized} strand. Several mutations have.been constructed in the ATP binding site to elucidate the roles of individual residues in catalysis and substrate binding and it has even been possible to construct mutants which have improved affinity for ATP. Mutations in various surface lysine and arginine residues have allowed us to identify potential contacts with the tRNA, and indicate that a cluster of basic residues close to the C-terminus of the enzyme probably makes important interactions with the tRNA. Key-words : genetic engineering / site-directed mutagenesis / tRNA synthetase / specificity.
Introduction
The tyrosyl t R N A synthetase from the thermophile Bacillus stearothermophilus has three substrates : tyrosine, ATP and tRNA t~r. The enzyme catalyses the transfer o f tyrosine to tRNA in a two
stage reaction [1] in which the tyrosine is first activated at its carboxyl group with ATP, and then the tyrosyl adenylate is attacked by the 2' or 3' hydroxyl of the tRNA : Tyr + ATP = Tyr-AMP + PPi (1) Tyr-AMP + tRNA 'yr = Tyr-tRNA t~'~ + AMP (2)
H. Bedouelle and coll.
738
IMC38A~pl o&C~N ~H /
Il
H I
HN--...~ ,
H ,
H. 1
II
I
', H \+ co~
/
H,
I
~N~
NI
O II
..,,H / / ""
CH 2
..N
II~N~~
OI
' N ,'
_H--N-C-C--O--P--O--CH ' ' "
-
N
2
O
[~
--H--
co~
q /H /
/ / ~
o,..H
, ~
I
(~ II
H\
O
C
I IMC 36GIyJ
H\N I
IMC 192 GIy 1
FIG. 1. - - Potential hydrogen bonding contacts to the O'rosyl adenylate.
The gene for this enzyme has been cloned into the plasmid pBR322 [2] and subsequently into the single stranded phage M13 to facilitate mutagenesis [3]: the enzyme has been sequenced at the protein level as well as via the gene [4]. The enzyme is expressed well in the E. coli infected with the MI3 clone [3] and the X-ray crystallographic structure of an N-terminal domain has been solved : the C-terminal portion (one third of the molecule) appears to be disordered [5]. However the N-terminal domain contains all the residues essential for amino acid activation [6], and the binding site for tyrosyl adenylate could be identified directly by soaking the crystals with tyrosine and ATP [7]. Although the major contacts with the tRNA appear to be in the C-terminal domain, there are at least three potential contacts fringing the active site (Lys225, Lys230 and Lys233) which are partly protected from acetylation when tRNA binds [8]. The goal of site directed mutagenesis has been two-fold. Firstly, to evaluate the roles of each of the potential hydrogen bonds to the ATP or tyrosine, since the binding sites for these two substrates can be identified from the crystallographic map (Figure 1). Secondly, to map the path of the tRNA across the surface of the enzyme by altering surface lysine and arginine residues.
Construction of mutants
The only general way of making precise changes at the nucleotide level is to use synthetic oligonucleotides [9]. Firstly a single stranded copy of the gene is prepared by cloning the gene into the replicative form of the bacteriophage MI3 • phage excreted from the transfected cell carry a single circular strand of the recombinant DNA. Secondly a short (15-20 bases) oligonucleotide is synthesised such that it is complementary to the region to be changed, apart from a limited internal mismatch which directs the mutation. The oligonucleotide is hybridised to the single stranded template and is extended with the Klenow subfragment of DNA polymerase in the presence of T4 DNA ligase. Several variations on this technique are then possible. In the "all the way round" method, the oligonucleotide is extended right round the template and the closed circular coil purified on an alkaline sucrose gradient [10]. In the "double primer" method, a second primer to the 5' side of the mutagenic oligonucleotide is extended concurrently to join up with the 5' end of the mutagenic oligonucleotide [11]. In the "gapped duplex" method, the oligonucleotide is extended within a "window" of exposed single strand • the gapped duplex is made by hybridising
Engineering of tyrosyl tRNA synthetase MI3 recombinant template with the linearised replicative form of the parent vector [12]. The mismatched heteroduplex prepared as above is then transfected into a male E. coli strain made competent for DNA uptake. In the process of replication both plus (template) and minus (in vitro synthesised) strands of the heteroduplex make a contribution to the progeny phage, but the contribution from the minus strand is the major one [13, 14]. Each plaque may be mixed (contains wild type and mutant phage), although one type usually predominates, indicating mismatch repair at a very early stage in replication [14]. The frequency of mutant plaques is rather variable, and for example depends on any displacement of the mutagenic oligonucleotide from the template by the polymerase, and also on the extent of mismatch repair in the cell [14]. We find that transfection of the heteroduplex into cells deficient in mismatch repair (such as mutL E. coh) can give dramatically enhanced frequencies of mutants for some mismatches. This applies to heteroduplex constructed by the "all the way round", "double primer" or "gapped duplex" methods [151. To improve mutant frequencies further we have attempted to eliminate the contribution of the plus strand to progeny phage, by constructing a selectable marker on the minus strand. We have used three markers successfully • an amber mutation in an essential gene of the virus, and an EcoK site or an EcoB site in the MI3 polylinker sequence. For example, an amber mutation was constructed in M13 gene IV by site directed mutagenesis, converting Ml3mpl8 and Ml3mpl9 to am IV versions. By priming on the single stranded template with the mutagenic primer in the insert and with another primer (the selection oligonucleotide) which removes the amber mutation in gene IV, a partial duplex is constructed as mutagenic p r i m e r
5" M I 3 template
t
amber | 5"
3" selection
FIG. 2. - -
primer
Coupledpriming.
A mutagenie primer and a selection oligonucleotide are extended with the Klenow subfragment of D N A polymerase in the presence of T4 D N A ligase. The selection oligonucleotide is designed to remove an a m b e r mutation in M I 3 gene IV.
739
in the "double primer" approach. We call this approach "coupled priming" (Figure 2). Transfection into repair su cells then selects against those phage progeny derived from plus strand replication. In preliminary experiments we find an increased frequency of mutants compared with transfection into repair-su ÷ cells [15].
Removing hydrogen bonds in the adenylate binding site There are several side and main chain atoms which could make hydrogen bonding contacts to the tyrosyl adenylate (Figure,l). For example there are three side chain contacts from the enzyme (Cys35, His48 and Thr51) to the ribose moiety of the adenylate. Two of the side chains, Cys35 and His48, are conserved in the sequences of the tyrosyl tRNA and methionyl tRNA synthetases (E. coli), and this suggested an important role for these residues [16]. However it is now clear that His48 can be replaced by Asn (in another strain of Bacillus stearothermophilus [17] and in yeast methionyl tRNA synthetase [18]) and Cys35 by Ser (in yeast methionyl tRNA synthetase [18]). Thr51 is not conserved, and is replaced by Pro in the tyrosyl tRNA synthetase from E. coli [4]. To evaluate the precise role of these residues, several mutants have now been made at each site and the k~,t and Km values for variation in [ATP] measured (Table 1 & [3, 19-23]). The fact that the steady state kinetics follows the Michaelis-Menten equation, also allows calculation of the change in binding of the transition state to the mutant enzyme [19]. Cys35 and His48 have each been altered to Gly, so removing their hydrogen bonding potential, and this results in a lower k~t and increased K m , and overall a loss in affinity for the transition state of about I kcal/mole [19, 21]. Does altering side chain such as Cys35 purely affect the local contact with the substrate, or does it cause more extensive changes around the active site ? To see whether the conversion of Cys35 to Gly alters the interaction at His48, we have constructed a double mutant Gly35 Gly48 (Figure 3). The kinetics reveal that the interaction of the imidazole group with substrate is identical in the Cys35 or Gly35 enzymes. Likewise the interaction of the sulphydryl group with substrate is identical in the His48 or Gly48 enzymes. Thus for these two mutants at least, the loss in energy of interaction can be ascribed to loss in the local contact [21].
H. Bedouelle and coll.
740
~.,kc/ ICy555 Hi548I~.2 kca~ (Cys35 GIy481
I GIy35 His481
'2kc~ IGIy35GIy481/' k~a' FIG. 3. - - Energetics of Gly35 Gly48 double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/mol [20]. Calculations were from k,~,/K~ terms as described in [19].
Does the energy calculated actually correspond to the direct interaction energy of side chain and substrate ? This seems unlikely as the absolute strengths of H-bonds are probably higher than the 1 kcal/mol values calculated above. For example, the enthalpy for CH3SH/water is - 3 . 2 kcal/mol, and for imidazolium/water is - 14 kcal/mol [24]. The binding of substrate to enzyme is in fact an exchange reaction as water must first be displaced : E--H...OH2 + H O - - H . . . A T P = [E--H...ATP] + HO--H...OH2 wild type (l) E
OH2 + HO--H...ATP = [E ATP] + HO--H...OH2 mutant
(2)
In the wild type enzyme (I), water bound to a hydrogen bonding group is displaced by the substrate. Since there are two H-bonds on the left of the equation and two similar on the right, the enthalpy is almost unchanged. However water bound to the enzyme is more ordered than bulk water, and the release of an ordered water molecule from the hydrogen bonding group helps drive the enzyme/substrate binding. In the mutant (2), the exchange is also isoenthalpic, but the ordered water molecule is no longer present. Furthermore, water molecules can now make van der Waals interactions with the portion o f the enzyme exposed on deleting the side chain, whereas there are steric constraints on the substrate doing this. These two factors may account for the net interaction energies of about 1 kcal/mole for His48 and Cys35 [22]. To what extent can other hydrogen bonding side chains substitute for Cys35 or His48 ? Taking a hint from evolution (see above), we can convert Cys35 (-CH2-SH) to Ser (-CH2-OH) [3]. We find that this smaller side chain contributes no more to transition state binding than the Gly35 substitution [19]. We attribute this to geometrical constraints against good hydrogen bonding in the
enzyme/substrate complex : if the length of the bond from Cys35 to the 3' hydroxyl of the ribose is optimal, then the bond from Ser 35 must be too long by at least 0.5A (see below, Equations 3, 4). On the other hand, if His48 is converted to Asn, the activity is relatively unchanged (Table 1 & [231). This implies that a hydrogen bond from Asn48 can effectively substitute for one from His48, and also that the hydrogen bond from His48 is probably made from the delta nitrogen atom o f the imidazole ring. Furthermore it indicates that His48 does not make any important electrostatic interaction with the phosphate of the adenylate. Thr51 has been altered to Ala, removing a potential hydrogen bond, and this results in an increased k~a~ and decreased K i n , and overall a gain in the affinity for the transition state of about 0.3 kcal/mole [20]. At first sight it may seem paradoxical that removing a hydrogen bonding contact can improve the affinity o f the enzyme for substrate. What is the explanation ? The X ray crystallography had indicated that the bond from Thr51 to the substrate was long, at 3.3 to 3.5,/t rather than the optimal 2.8A. This can be checked by construction of a Cys51 mutant : if the bond from Thr51 to substrate is too long, then the substitution of Cys should now permit an improved interaction with the ribose. Indeed there is an increase in affinity of 0.9 kcai/moI over Thr51 (0.5 kcal/mol over Ala51) (Table 1 & [22]). Thus in the wild type enzyme (Thr51) there must be geometrical constraints against good H-bonding in the enzyme/substrate complex. This can introduce an unfavourable term in the exchange reaction • E--H...OH2 + H O - - H . . . A T P = [ E - - H ATP] + HO--H...OH2 wild type (3) E
OHz + H O - - H . . . A T P = [E ATP] + HO--H...OH2 mutant
(4)
In the wild type enzyme there are two H-bonds on the left hand side of the equation and only one on the right • in the mutant there is one H-bond on each side of the equation. Thus in the wild type enzyme, substrate binding requires a water molecule to be displaced with no compensating hydrogen bond formed to the substrate. In the mutant enzyme, it is not necessary to displace a water molecule and the enzyme therefore has a higher affinity for the substrate [19, 22].
Engineering of tyrosyl tRNA synthetase
Distorting the active site
~.,kc/{ His48 Thr51 1 ~
The changes in side chain contacts discussed above, do not appear to evoke major changes in the disposition o f the polypeptide backbone. We had noticed that Thr51 of the Bacillus enzyme was replaced by Pro51 in the homologous E. coil enzyme, a little surprising as this residue lies toward the N-terminus o f the alpha-helix forming part of the ATP binding site. We decided to convert Thr51 to Pro as it might be expected to distort the active site [19]. Measurement of k~t and Km for variation in [ATP] shows an increase in catalytic rate and dramatic increase in affinity of the Pro51 enzyme for ATP (Table 1), well over TABLEI Activation of tyrosine by tyrosyl tRNA synthetase and its mutants. Mutant
Ref. k~,(s-') K,,(mM) k,,/K,,(s-~M -~)
wild type Gly35 Ser35 Gly48 Asn48 Ala51 Cys51 Pro51
122] [22] [22] [22] [23] [22] [22] [21]
8.4 2.95 2.52 2.0 7.90 8.75 12.4 12.4
1.1 2.6 2.4 1.3 1.4 0.54 0.35 0.058
7,640 1,130 1,050 1,540 5,640 16,200 35,400 213,790
Gly35 Gly35 Gly48 Ash48
[21] 1.32 [211 10.3 [21] 2.9 [23] 11.2
8.0 0.181 2.1 0.41
177 56,900 1,380 27,320
Gly48 Pro51 Pro51 Pro51
741
The k,, and K,, values for variation in [ATP] in the activation of tyrosine, determined as described in [21-23].
that for the Ala51 or Cys51 enzymes. The reasons for this could be three fold. The substrate might dock directly against the pyrollidine ring of Pro51 (much as the improved hydrogen bond is made from Cys51 to the ribose); the distortion introduced by Pro51 into the backbone might forge entirely new contacts with the enzyme; the distortion might improve existing contacts. In the Thr51, Ala51 or Cys51 enzymes, there is probably a hydrogen bond from the main chain amide of residue 51 back to the main chain carbonyl of Gly47 (alpha-helix) or His48 (3t0 helix). This is not present in the Pro51 enzyme as the imino group cannot make the bond. Whatever else the
IGIy4B Thr511
9kcal
IHis48Pro51 1
° kc~ [ Gly48 Pro51 11~3-°kca' FiG. 4. -- Energetics of Gly48 ProSZ double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/rnole [20]. Calculations were from k,,/Km terms as described in [19]. effect of Pro51, it seemed likely to affect the disposition of His48. We therefore decided to test whether Pro51 had improved the contacts at His48 or at Cys35 at the other side of the active site. To acheive this we made two double mutants, Gly48 Pro51 (Figure 4) and Gly35 Pro5l. The results indicate that in the wild type enzyme the imidazole group at position 48 is worth 1.1 kcal/mol but that in the Pro51 enzyme it is worth 3.0 kcal/mol. By comparison, the effect of Pro51 on Cys35 (across the other side o f the active site) is m i n o r : thus the major effect o f Pro51 is to improve the interaction at His48. At the same time these results rule out the possibility that the ribose makes a good docking interaction with the pyrrolidine ring, as the affinity of the Gly48 Pro51 enzyme is almost identical to that of the Gly48 Thr51 enzyme [21]. Is the improved interaction of His48 due entirely to an improved hydrogen bond with the substrate ? As discussed earlier, Asn48 is able to substitute effectively for His48, presumably by making the same hydrogen bond to the ribose. However the energetics of the double mutant Asn48 Pro51 (Table 1 & [23]) indicate that the proline residue improves affinity, but not as much as with His48 (Figure 5). Thus the effect o f Pro51
o2. l .,s4 T,rs, o.., (Aso48Thrs' / I,,so pros, l og c /A o40 FIG. 5. -- Energetics of ASh48 Pro51 double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/mol [231.Calculations were from kc,,/Kmterms as described in I19].
742
H. Bedouelle and coil
probably has two components : firstly an improved hydrogen bond to the ribose, and secondly an enhanced electrostatic interaction from the imidazole ring to a negative charge in the substrate.
Contacts with the tRNA In the absence of a co-crystal of synthetase and tRNA, we are attempting to identify the surface patches of the synthetase which contact the tRNA. The tyrosyl tRNA synthetase binds one tRNA tyr molecule [8, 25]. The size of the tRNA is comparable to that of th e synthetase and it seems likely that the docking of these two macromolecules will involve multiple contacts, some important for binding and others for discrimination between related tRNAs. The tRNA presents an external surface of negatively charged phosphate groups, and these will likely form salt bridges with some of the lysine and arginine residues on the synthetase. For example, at least two lysine residues in the vicinity of the active site are protected from acetylation when the synthetase binds to tRNA. In addition to the interactions of the t R N A at the active site, is there a well defined tRNA binding domain, or does the tRNA straddle the entire surface of the synthetase ? The crystallographic structure of the enzyme shows an N-terminal domain of alpha helix and beta sheet (residues 1-225) which appears to make all the interactions with the tyrosyl adenylate and to form the subunit interface [5]. This is followed by an alpha helical domain (residues 226-319) and then a disordered C-terminus (residues 320-419). Removing the disordered C-terminus at the level of the gene yields a truncated dimer (residues 1-319), which catalyses the amino acid activation but not the charging of tRNA. Furthermore tRNA does not bind to the truncated enzyme, suggesting that the disordered C-terminus is a tRNA binding domain [6]. To identify the individual residues involved in interactions with the tRNA, we have developed an initial complementation assay for mutant tyrosyl tRNA synthetase [26]. Each mutant (present as recombinant MI3 phage) is carefully plaque purified and serial dilutions made. A drop of phage is placed on a freshly poured plate of male recA E. coli carrying a thermosensitive mutation in the essential, endogenous tyrS gene. At the non permissive temperature (42°C), only those cells infected with phage carrying viable tyrosyl tRNA synthetase will grow up. From the
presence or absence of bacterial colonies, each synthetase mutant can be scored as to whether it can charge tRNA at 42°C. From the size of the colonies, it is even possible to discern the three fold change in synthetase levels on constructing a promoter mutant. Only those mutants which score poorly in this assay need be further characterised to check whether the lesion is in amino acid activation, tRNA binding or tRNA charging. Inspection of the sequence of the C-terminal disordered domain and its comparison with that of E. coli reveals a conserved cluster of basic charges near the C-terminus, viz -Arg4°7-Arg4°sGly~-Lys41°-Lys TM [4]. Making single mutations in which Arg ~7 was converted to Gin, Arg4°8 to Gin, Lys41° to Asn, or Lys4~ to Asn (so as to remove the charge but not the hydrophilic character of the side chain) immediately abolished complementation [26]. Since the domain responsible for the formation of tyrosyl adenylate is unaltered, and SDS protein gels show that the tyrosyl t R N A synthetase is produced in good yield in the infected cells, it seems plausible that the residues of this basic cluster may each interact with the tRNA. Mutation of other basic residues in this domain, such as Lys367 or Arg385, do not affect complementation. Although we envisage that the C-terminal domain contains the major binding determinants for tRNA, the N-terminal domain probably contains contacts which dock the tRNA in the right orientation at the active site. The active site is fringed with two lengths of polypeptide chain carrying basic residues, viz : LysS2.LysS3.Ser~4_GluSS.ArgS6 and Lys~2S-...Lys23°-Phe231-Gly232-Lys233. Making single mutations as above, in which Lys was converted to Ash and Arg to Gin, virtually abolished complementation in Asn s2, Gln 86, Asn 23° or Asn TM enzymes, but not in the Asn 83 or Asn2:5 enzymes [26]. Since lysine residues in the region LysZ25...Lys233 were protected from acetylation by tRNA [8], it seems very likely that at least Lys23° and Lys233 are genuine contacts to the tRNA. These preliminary results suggest that surface lysine and arginine residues fringing the active site are required for docking of the tRNA.
REFERENCES 1. Fersht, A.R. & Jakes, R. (I975) Biochemistry, 14, 3350-3356.
Engineering o f tyrosyl tRNA synthetase
2. Barker, D.G. (1982) Eur. J. Biochem, 125, 357-360. 3. Winter, G., Fersht, A.R., Wilkinson, A.J., Zoller, M.J. & Smith, M. (1982) Nature, 299, 756-758. 4. Winter, G., Koch, G.L.E., Hartley, B.S. & Barker, D.G. (1983) Fur. J. Biochem., 132, 383-387. 5. Bhat, T.N., Blow, D.M., Brick, P. & Nyborg, J. (1982) J. Mol. BioL, 158, 699-709. 6. Waye, M.M.Y., Winter, G., Wilkinson, A.J. & Fersht, A.R. (1983) EMBO J., 2, 1827-1829. 7. Rubin, J. & Blow, D.M. (1981) J. MoL BioL, 145, 489-500. 8. Bosshard, H.R., Koch, G.L.E. & Hartley, B.S. (1978) J. Mol. BioL, 119, 377-389. 9. Hutchinson, C.A., Phillips, S., Edgell, M.H., Gillam, S., Jahnke, P. & Smith, M. (1978) J. Biol. Chem., 253, 6551-6560. 10. Zoller, M.J. & Smith, M. (1982) Nucleic Acids Res., 10, 6487-6500. 11. Norris, K., Norris, F., Christiansen, L. & Fiil, N. (1983) Nucleic Acids Res., 11, 5103-5112. 12. Kramer, W., Schughart, K. & Fritz, H.-J. (1982) Nucleic Acids Res., 10, 6475-6485. 13. Enea, V., Vovis, G.F. & Zinder, N.D. (1975) J. MoL BioL, 96, 495-509. 14. Kramer, B., Kramer, W. & Fritz, H.-J. (1984) Ce'll, 38, 879-887.
743
15. Carter, P., Bedouelle, H. & Winter, G. (1985) Nucleic Acids Res., 13, 4431-4443. 16. Barker, D.G. & Winter, G. (1982) FEBS Lett., 145, 191-193. 17. Jones, M.D., Lowe, D.M. & Fersht, A.R. unpublished. 18. Walter, P., Gangloff, J., Bonnet, J., Boulanger, P., Ebel, J.-P. & Fasiolo, F. (1983) Proc. Nat. Acad. Sci (U.S~4.) 80, 2437-2441. 19. Wilkinson, A.J., Fersht, A.R., Blow, D.M. &' Winter, G. (1983) Biochemistry, 22, 3581-3586. 20. Wilkinson, A.J., Fersht, A.R., Blow, D.M., Cai'ier, P. & Winter, G. (1984) Nature, 307, 187-188. 21. Carter, P., Winter, G., Wilkinson, A.J. & Fersht, A.R. (1984), Cell, 38, 835-840. 22. Fersht, A.R., Shi, J.P., Kniii-Jones, J., Lowe, D.M., Wilkinson, A.J., Blow, D.M., Brick, P., Carter, P., Waye, M.M.Y. & Winter, G. (1985) Nature, 314, 235-238. 23. Lowe, D.M., Fersht, A.R., Wilkinson, A.J., Carter P. & Winter, G. (1985) Biochemistry in the press. 24. Weiner, S.J., Kollman, P.A., Dase, D.A., Singh, U.C., Ghio, C., Alagona, G., Profeta, S. & Weiner, P.J. (1984) J. Am. Chem. Soc., 10, 765-784. 25. Dessen, P., Zaccai, G. & Blanquet, S. (1982) J. Mol. BioL, 159, 651-664. 26. Bedouelle, H. unpublished.
Engineering of tyrosyl tRNA synthetase. Hugues BEDOUELLE*, Paul CARTER*, Mary M.Y. WAYE*, Greg WINTER*, Denise M. LOWE**, Anthony J. W I L K I N S O N * * & Alan R. FERSHT**.
* Laboratory of Molecular Biology, Hills Rd, Cambridge CB2 2QH. ** Department of Chemistry, Imperial College, London SW7 2AY. (Refu le 19-3-1985, accept~ le 17-4-1985).
R6sum6 - - Le gkne codant pour la tyrosyl tRNA synthdtase de Bacillus stdarothermophilus a dtd systdmatiquement altdrd en utilisant des oligonucldotides synthdtiques comme mutag~nes. La construction des mutations a dtd facilitde par l'utilisation de souches bactdriennes ddfectives pour la rdparation des misappariements et aussi en utilisant une souche de Ml3 possddant un marqueur gdndtique (tel qu'une mutation ambre, ou un site EcoK ou EcoB) qui permet de sdlectionner la progdniture du phage issue du brin moins (mutagdnisd) lors de la triplication. Plusieurs mutations ont ~td construites dans le site de fixation de l'ATP pour dlucider le r61e de rdsidus particuliers dans la catalyse et la fixation du substrat et il a m~me dtd possible de construire des mutants qui ont une plus grande affinitd pour I'ATP. Des mutations changeant des rdsidus lysine ou arginine situds h la surface de l'enzyme nous ont permis d'identifier des pohlts de contact potentiels avec le tARN; ces mutations indiquent qu'un groupe de rdsidus basiques, pros de l'extrdmitd C-terminale de i'enzyme, interagit de fafon importante avec le tARN. Mots-cl~s : g~nie g~n~tlque / mutag~n~se dirig~e / tRNA synth~tase / specificitY.
S u m m a r y - - The gene encoding the enzyme tyrosyl tRNA synthetase from Bacillus stearothermophilus has been systematically altered using synthetic oligonucleotides as mutagens. The construction of mutations has been facilitated by using strains of bacteria defective in mismatch repair and also by utilising a genetic marker in the MI3 strain (such as an amber mutation, or an EcoK or EcoB site) which allows selection for the progeny of M13 replication derived from the minus (mutagenized} strand. Several mutations have.been constructed in the ATP binding site to elucidate the roles of individual residues in catalysis and substrate binding and it has even been possible to construct mutants which have improved affinity for ATP. Mutations in various surface lysine and arginine residues have allowed us to identify potential contacts with the tRNA, and indicate that a cluster of basic residues close to the C-terminus of the enzyme probably makes important interactions with the tRNA. Key-words : genetic engineering / site-directed mutagenesis / tRNA synthetase / specificity.
Introduction
The tyrosyl t R N A synthetase from the thermophile Bacillus stearothermophilus has three substrates : tyrosine, ATP and tRNA t~r. The enzyme catalyses the transfer o f tyrosine to tRNA in a two
stage reaction [1] in which the tyrosine is first activated at its carboxyl group with ATP, and then the tyrosyl adenylate is attacked by the 2' or 3' hydroxyl of the tRNA : Tyr + ATP = Tyr-AMP + PPi (1) Tyr-AMP + tRNA 'yr = Tyr-tRNA t~'~ + AMP (2)
H. Bedouelle and coll.
738
IMC38A~pl o&C~N ~H /
Il
H I
HN--...~ ,
H ,
H. 1
II
I
', H \+ co~
/
H,
I
~N~
NI
O II
..,,H / / ""
CH 2
..N
II~N~~
OI
' N ,'
_H--N-C-C--O--P--O--CH ' ' "
-
N
2
O
[~
--H--
co~
q /H /
/ / ~
o,..H
, ~
I
(~ II
H\
O
C
I IMC 36GIyJ
H\N I
IMC 192 GIy 1
FIG. 1. - - Potential hydrogen bonding contacts to the O'rosyl adenylate.
The gene for this enzyme has been cloned into the plasmid pBR322 [2] and subsequently into the single stranded phage M13 to facilitate mutagenesis [3]: the enzyme has been sequenced at the protein level as well as via the gene [4]. The enzyme is expressed well in the E. coli infected with the MI3 clone [3] and the X-ray crystallographic structure of an N-terminal domain has been solved : the C-terminal portion (one third of the molecule) appears to be disordered [5]. However the N-terminal domain contains all the residues essential for amino acid activation [6], and the binding site for tyrosyl adenylate could be identified directly by soaking the crystals with tyrosine and ATP [7]. Although the major contacts with the tRNA appear to be in the C-terminal domain, there are at least three potential contacts fringing the active site (Lys225, Lys230 and Lys233) which are partly protected from acetylation when tRNA binds [8]. The goal of site directed mutagenesis has been two-fold. Firstly, to evaluate the roles of each of the potential hydrogen bonds to the ATP or tyrosine, since the binding sites for these two substrates can be identified from the crystallographic map (Figure 1). Secondly, to map the path of the tRNA across the surface of the enzyme by altering surface lysine and arginine residues.
Construction of mutants
The only general way of making precise changes at the nucleotide level is to use synthetic oligonucleotides [9]. Firstly a single stranded copy of the gene is prepared by cloning the gene into the replicative form of the bacteriophage MI3 • phage excreted from the transfected cell carry a single circular strand of the recombinant DNA. Secondly a short (15-20 bases) oligonucleotide is synthesised such that it is complementary to the region to be changed, apart from a limited internal mismatch which directs the mutation. The oligonucleotide is hybridised to the single stranded template and is extended with the Klenow subfragment of DNA polymerase in the presence of T4 DNA ligase. Several variations on this technique are then possible. In the "all the way round" method, the oligonucleotide is extended right round the template and the closed circular coil purified on an alkaline sucrose gradient [10]. In the "double primer" method, a second primer to the 5' side of the mutagenic oligonucleotide is extended concurrently to join up with the 5' end of the mutagenic oligonucleotide [11]. In the "gapped duplex" method, the oligonucleotide is extended within a "window" of exposed single strand • the gapped duplex is made by hybridising
Engineering of tyrosyl tRNA synthetase MI3 recombinant template with the linearised replicative form of the parent vector [12]. The mismatched heteroduplex prepared as above is then transfected into a male E. coli strain made competent for DNA uptake. In the process of replication both plus (template) and minus (in vitro synthesised) strands of the heteroduplex make a contribution to the progeny phage, but the contribution from the minus strand is the major one [13, 14]. Each plaque may be mixed (contains wild type and mutant phage), although one type usually predominates, indicating mismatch repair at a very early stage in replication [14]. The frequency of mutant plaques is rather variable, and for example depends on any displacement of the mutagenic oligonucleotide from the template by the polymerase, and also on the extent of mismatch repair in the cell [14]. We find that transfection of the heteroduplex into cells deficient in mismatch repair (such as mutL E. coh) can give dramatically enhanced frequencies of mutants for some mismatches. This applies to heteroduplex constructed by the "all the way round", "double primer" or "gapped duplex" methods [151. To improve mutant frequencies further we have attempted to eliminate the contribution of the plus strand to progeny phage, by constructing a selectable marker on the minus strand. We have used three markers successfully • an amber mutation in an essential gene of the virus, and an EcoK site or an EcoB site in the MI3 polylinker sequence. For example, an amber mutation was constructed in M13 gene IV by site directed mutagenesis, converting Ml3mpl8 and Ml3mpl9 to am IV versions. By priming on the single stranded template with the mutagenic primer in the insert and with another primer (the selection oligonucleotide) which removes the amber mutation in gene IV, a partial duplex is constructed as mutagenic p r i m e r
5" M I 3 template
t
amber | 5"
3" selection
FIG. 2. - -
primer
Coupledpriming.
A mutagenie primer and a selection oligonucleotide are extended with the Klenow subfragment of D N A polymerase in the presence of T4 D N A ligase. The selection oligonucleotide is designed to remove an a m b e r mutation in M I 3 gene IV.
739
in the "double primer" approach. We call this approach "coupled priming" (Figure 2). Transfection into repair su cells then selects against those phage progeny derived from plus strand replication. In preliminary experiments we find an increased frequency of mutants compared with transfection into repair-su ÷ cells [15].
Removing hydrogen bonds in the adenylate binding site There are several side and main chain atoms which could make hydrogen bonding contacts to the tyrosyl adenylate (Figure,l). For example there are three side chain contacts from the enzyme (Cys35, His48 and Thr51) to the ribose moiety of the adenylate. Two of the side chains, Cys35 and His48, are conserved in the sequences of the tyrosyl tRNA and methionyl tRNA synthetases (E. coli), and this suggested an important role for these residues [16]. However it is now clear that His48 can be replaced by Asn (in another strain of Bacillus stearothermophilus [17] and in yeast methionyl tRNA synthetase [18]) and Cys35 by Ser (in yeast methionyl tRNA synthetase [18]). Thr51 is not conserved, and is replaced by Pro in the tyrosyl tRNA synthetase from E. coli [4]. To evaluate the precise role of these residues, several mutants have now been made at each site and the k~,t and Km values for variation in [ATP] measured (Table 1 & [3, 19-23]). The fact that the steady state kinetics follows the Michaelis-Menten equation, also allows calculation of the change in binding of the transition state to the mutant enzyme [19]. Cys35 and His48 have each been altered to Gly, so removing their hydrogen bonding potential, and this results in a lower k~t and increased K m , and overall a loss in affinity for the transition state of about I kcal/mole [19, 21]. Does altering side chain such as Cys35 purely affect the local contact with the substrate, or does it cause more extensive changes around the active site ? To see whether the conversion of Cys35 to Gly alters the interaction at His48, we have constructed a double mutant Gly35 Gly48 (Figure 3). The kinetics reveal that the interaction of the imidazole group with substrate is identical in the Cys35 or Gly35 enzymes. Likewise the interaction of the sulphydryl group with substrate is identical in the His48 or Gly48 enzymes. Thus for these two mutants at least, the loss in energy of interaction can be ascribed to loss in the local contact [21].
H. Bedouelle and coll.
740
~.,kc/ ICy555 Hi548I~.2 kca~ (Cys35 GIy481
I GIy35 His481
'2kc~ IGIy35GIy481/' k~a' FIG. 3. - - Energetics of Gly35 Gly48 double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/mol [20]. Calculations were from k,~,/K~ terms as described in [19].
Does the energy calculated actually correspond to the direct interaction energy of side chain and substrate ? This seems unlikely as the absolute strengths of H-bonds are probably higher than the 1 kcal/mol values calculated above. For example, the enthalpy for CH3SH/water is - 3 . 2 kcal/mol, and for imidazolium/water is - 14 kcal/mol [24]. The binding of substrate to enzyme is in fact an exchange reaction as water must first be displaced : E--H...OH2 + H O - - H . . . A T P = [E--H...ATP] + HO--H...OH2 wild type (l) E
OH2 + HO--H...ATP = [E ATP] + HO--H...OH2 mutant
(2)
In the wild type enzyme (I), water bound to a hydrogen bonding group is displaced by the substrate. Since there are two H-bonds on the left of the equation and two similar on the right, the enthalpy is almost unchanged. However water bound to the enzyme is more ordered than bulk water, and the release of an ordered water molecule from the hydrogen bonding group helps drive the enzyme/substrate binding. In the mutant (2), the exchange is also isoenthalpic, but the ordered water molecule is no longer present. Furthermore, water molecules can now make van der Waals interactions with the portion o f the enzyme exposed on deleting the side chain, whereas there are steric constraints on the substrate doing this. These two factors may account for the net interaction energies of about 1 kcal/mole for His48 and Cys35 [22]. To what extent can other hydrogen bonding side chains substitute for Cys35 or His48 ? Taking a hint from evolution (see above), we can convert Cys35 (-CH2-SH) to Ser (-CH2-OH) [3]. We find that this smaller side chain contributes no more to transition state binding than the Gly35 substitution [19]. We attribute this to geometrical constraints against good hydrogen bonding in the
enzyme/substrate complex : if the length of the bond from Cys35 to the 3' hydroxyl of the ribose is optimal, then the bond from Ser 35 must be too long by at least 0.5A (see below, Equations 3, 4). On the other hand, if His48 is converted to Asn, the activity is relatively unchanged (Table 1 & [231). This implies that a hydrogen bond from Asn48 can effectively substitute for one from His48, and also that the hydrogen bond from His48 is probably made from the delta nitrogen atom o f the imidazole ring. Furthermore it indicates that His48 does not make any important electrostatic interaction with the phosphate of the adenylate. Thr51 has been altered to Ala, removing a potential hydrogen bond, and this results in an increased k~a~ and decreased K i n , and overall a gain in the affinity for the transition state of about 0.3 kcal/mole [20]. At first sight it may seem paradoxical that removing a hydrogen bonding contact can improve the affinity o f the enzyme for substrate. What is the explanation ? The X ray crystallography had indicated that the bond from Thr51 to the substrate was long, at 3.3 to 3.5,/t rather than the optimal 2.8A. This can be checked by construction of a Cys51 mutant : if the bond from Thr51 to substrate is too long, then the substitution of Cys should now permit an improved interaction with the ribose. Indeed there is an increase in affinity of 0.9 kcai/moI over Thr51 (0.5 kcal/mol over Ala51) (Table 1 & [22]). Thus in the wild type enzyme (Thr51) there must be geometrical constraints against good H-bonding in the enzyme/substrate complex. This can introduce an unfavourable term in the exchange reaction • E--H...OH2 + H O - - H . . . A T P = [ E - - H ATP] + HO--H...OH2 wild type (3) E
OHz + H O - - H . . . A T P = [E ATP] + HO--H...OH2 mutant
(4)
In the wild type enzyme there are two H-bonds on the left hand side of the equation and only one on the right • in the mutant there is one H-bond on each side of the equation. Thus in the wild type enzyme, substrate binding requires a water molecule to be displaced with no compensating hydrogen bond formed to the substrate. In the mutant enzyme, it is not necessary to displace a water molecule and the enzyme therefore has a higher affinity for the substrate [19, 22].
Engineering of tyrosyl tRNA synthetase
Distorting the active site
~.,kc/{ His48 Thr51 1 ~
The changes in side chain contacts discussed above, do not appear to evoke major changes in the disposition o f the polypeptide backbone. We had noticed that Thr51 of the Bacillus enzyme was replaced by Pro51 in the homologous E. coil enzyme, a little surprising as this residue lies toward the N-terminus o f the alpha-helix forming part of the ATP binding site. We decided to convert Thr51 to Pro as it might be expected to distort the active site [19]. Measurement of k~t and Km for variation in [ATP] shows an increase in catalytic rate and dramatic increase in affinity of the Pro51 enzyme for ATP (Table 1), well over TABLEI Activation of tyrosine by tyrosyl tRNA synthetase and its mutants. Mutant
Ref. k~,(s-') K,,(mM) k,,/K,,(s-~M -~)
wild type Gly35 Ser35 Gly48 Asn48 Ala51 Cys51 Pro51
122] [22] [22] [22] [23] [22] [22] [21]
8.4 2.95 2.52 2.0 7.90 8.75 12.4 12.4
1.1 2.6 2.4 1.3 1.4 0.54 0.35 0.058
7,640 1,130 1,050 1,540 5,640 16,200 35,400 213,790
Gly35 Gly35 Gly48 Ash48
[21] 1.32 [211 10.3 [21] 2.9 [23] 11.2
8.0 0.181 2.1 0.41
177 56,900 1,380 27,320
Gly48 Pro51 Pro51 Pro51
741
The k,, and K,, values for variation in [ATP] in the activation of tyrosine, determined as described in [21-23].
that for the Ala51 or Cys51 enzymes. The reasons for this could be three fold. The substrate might dock directly against the pyrollidine ring of Pro51 (much as the improved hydrogen bond is made from Cys51 to the ribose); the distortion introduced by Pro51 into the backbone might forge entirely new contacts with the enzyme; the distortion might improve existing contacts. In the Thr51, Ala51 or Cys51 enzymes, there is probably a hydrogen bond from the main chain amide of residue 51 back to the main chain carbonyl of Gly47 (alpha-helix) or His48 (3t0 helix). This is not present in the Pro51 enzyme as the imino group cannot make the bond. Whatever else the
IGIy4B Thr511
9kcal
IHis48Pro51 1
° kc~ [ Gly48 Pro51 11~3-°kca' FiG. 4. -- Energetics of Gly48 ProSZ double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/rnole [20]. Calculations were from k,,/Km terms as described in [19]. effect of Pro51, it seemed likely to affect the disposition of His48. We therefore decided to test whether Pro51 had improved the contacts at His48 or at Cys35 at the other side of the active site. To acheive this we made two double mutants, Gly48 Pro51 (Figure 4) and Gly35 Pro5l. The results indicate that in the wild type enzyme the imidazole group at position 48 is worth 1.1 kcal/mol but that in the Pro51 enzyme it is worth 3.0 kcal/mol. By comparison, the effect of Pro51 on Cys35 (across the other side o f the active site) is m i n o r : thus the major effect o f Pro51 is to improve the interaction at His48. At the same time these results rule out the possibility that the ribose makes a good docking interaction with the pyrrolidine ring, as the affinity of the Gly48 Pro51 enzyme is almost identical to that of the Gly48 Thr51 enzyme [21]. Is the improved interaction of His48 due entirely to an improved hydrogen bond with the substrate ? As discussed earlier, Asn48 is able to substitute effectively for His48, presumably by making the same hydrogen bond to the ribose. However the energetics of the double mutant Asn48 Pro51 (Table 1 & [23]) indicate that the proline residue improves affinity, but not as much as with His48 (Figure 5). Thus the effect o f Pro51
o2. l .,s4 T,rs, o.., (Aso48Thrs' / I,,so pros, l og c /A o40 FIG. 5. -- Energetics of ASh48 Pro51 double mutant. The energy of interaction of each side chain with the transition state in the amino acid activation reaction in kcal/mol [231.Calculations were from kc,,/Kmterms as described in I19].
742
H. Bedouelle and coil
probably has two components : firstly an improved hydrogen bond to the ribose, and secondly an enhanced electrostatic interaction from the imidazole ring to a negative charge in the substrate.
Contacts with the tRNA In the absence of a co-crystal of synthetase and tRNA, we are attempting to identify the surface patches of the synthetase which contact the tRNA. The tyrosyl tRNA synthetase binds one tRNA tyr molecule [8, 25]. The size of the tRNA is comparable to that of th e synthetase and it seems likely that the docking of these two macromolecules will involve multiple contacts, some important for binding and others for discrimination between related tRNAs. The tRNA presents an external surface of negatively charged phosphate groups, and these will likely form salt bridges with some of the lysine and arginine residues on the synthetase. For example, at least two lysine residues in the vicinity of the active site are protected from acetylation when the synthetase binds to tRNA. In addition to the interactions of the t R N A at the active site, is there a well defined tRNA binding domain, or does the tRNA straddle the entire surface of the synthetase ? The crystallographic structure of the enzyme shows an N-terminal domain of alpha helix and beta sheet (residues 1-225) which appears to make all the interactions with the tyrosyl adenylate and to form the subunit interface [5]. This is followed by an alpha helical domain (residues 226-319) and then a disordered C-terminus (residues 320-419). Removing the disordered C-terminus at the level of the gene yields a truncated dimer (residues 1-319), which catalyses the amino acid activation but not the charging of tRNA. Furthermore tRNA does not bind to the truncated enzyme, suggesting that the disordered C-terminus is a tRNA binding domain [6]. To identify the individual residues involved in interactions with the tRNA, we have developed an initial complementation assay for mutant tyrosyl tRNA synthetase [26]. Each mutant (present as recombinant MI3 phage) is carefully plaque purified and serial dilutions made. A drop of phage is placed on a freshly poured plate of male recA E. coli carrying a thermosensitive mutation in the essential, endogenous tyrS gene. At the non permissive temperature (42°C), only those cells infected with phage carrying viable tyrosyl tRNA synthetase will grow up. From the
presence or absence of bacterial colonies, each synthetase mutant can be scored as to whether it can charge tRNA at 42°C. From the size of the colonies, it is even possible to discern the three fold change in synthetase levels on constructing a promoter mutant. Only those mutants which score poorly in this assay need be further characterised to check whether the lesion is in amino acid activation, tRNA binding or tRNA charging. Inspection of the sequence of the C-terminal disordered domain and its comparison with that of E. coli reveals a conserved cluster of basic charges near the C-terminus, viz -Arg4°7-Arg4°sGly~-Lys41°-Lys TM [4]. Making single mutations in which Arg ~7 was converted to Gin, Arg4°8 to Gin, Lys41° to Asn, or Lys4~ to Asn (so as to remove the charge but not the hydrophilic character of the side chain) immediately abolished complementation [26]. Since the domain responsible for the formation of tyrosyl adenylate is unaltered, and SDS protein gels show that the tyrosyl t R N A synthetase is produced in good yield in the infected cells, it seems plausible that the residues of this basic cluster may each interact with the tRNA. Mutation of other basic residues in this domain, such as Lys367 or Arg385, do not affect complementation. Although we envisage that the C-terminal domain contains the major binding determinants for tRNA, the N-terminal domain probably contains contacts which dock the tRNA in the right orientation at the active site. The active site is fringed with two lengths of polypeptide chain carrying basic residues, viz : LysS2.LysS3.Ser~4_GluSS.ArgS6 and Lys~2S-...Lys23°-Phe231-Gly232-Lys233. Making single mutations as above, in which Lys was converted to Ash and Arg to Gin, virtually abolished complementation in Asn s2, Gln 86, Asn 23° or Asn TM enzymes, but not in the Asn 83 or Asn2:5 enzymes [26]. Since lysine residues in the region LysZ25...Lys233 were protected from acetylation by tRNA [8], it seems very likely that at least Lys23° and Lys233 are genuine contacts to the tRNA. These preliminary results suggest that surface lysine and arginine residues fringing the active site are required for docking of the tRNA.
REFERENCES 1. Fersht, A.R. & Jakes, R. (I975) Biochemistry, 14, 3350-3356.
Engineering o f tyrosyl tRNA synthetase
2. Barker, D.G. (1982) Eur. J. Biochem, 125, 357-360. 3. Winter, G., Fersht, A.R., Wilkinson, A.J., Zoller, M.J. & Smith, M. (1982) Nature, 299, 756-758. 4. Winter, G., Koch, G.L.E., Hartley, B.S. & Barker, D.G. (1983) Fur. J. Biochem., 132, 383-387. 5. Bhat, T.N., Blow, D.M., Brick, P. & Nyborg, J. (1982) J. Mol. BioL, 158, 699-709. 6. Waye, M.M.Y., Winter, G., Wilkinson, A.J. & Fersht, A.R. (1983) EMBO J., 2, 1827-1829. 7. Rubin, J. & Blow, D.M. (1981) J. MoL BioL, 145, 489-500. 8. Bosshard, H.R., Koch, G.L.E. & Hartley, B.S. (1978) J. Mol. BioL, 119, 377-389. 9. Hutchinson, C.A., Phillips, S., Edgell, M.H., Gillam, S., Jahnke, P. & Smith, M. (1978) J. Biol. Chem., 253, 6551-6560. 10. Zoller, M.J. & Smith, M. (1982) Nucleic Acids Res., 10, 6487-6500. 11. Norris, K., Norris, F., Christiansen, L. & Fiil, N. (1983) Nucleic Acids Res., 11, 5103-5112. 12. Kramer, W., Schughart, K. & Fritz, H.-J. (1982) Nucleic Acids Res., 10, 6475-6485. 13. Enea, V., Vovis, G.F. & Zinder, N.D. (1975) J. MoL BioL, 96, 495-509. 14. Kramer, B., Kramer, W. & Fritz, H.-J. (1984) Ce'll, 38, 879-887.
743
15. Carter, P., Bedouelle, H. & Winter, G. (1985) Nucleic Acids Res., 13, 4431-4443. 16. Barker, D.G. & Winter, G. (1982) FEBS Lett., 145, 191-193. 17. Jones, M.D., Lowe, D.M. & Fersht, A.R. unpublished. 18. Walter, P., Gangloff, J., Bonnet, J., Boulanger, P., Ebel, J.-P. & Fasiolo, F. (1983) Proc. Nat. Acad. Sci (U.S~4.) 80, 2437-2441. 19. Wilkinson, A.J., Fersht, A.R., Blow, D.M. &' Winter, G. (1983) Biochemistry, 22, 3581-3586. 20. Wilkinson, A.J., Fersht, A.R., Blow, D.M., Cai'ier, P. & Winter, G. (1984) Nature, 307, 187-188. 21. Carter, P., Winter, G., Wilkinson, A.J. & Fersht, A.R. (1984), Cell, 38, 835-840. 22. Fersht, A.R., Shi, J.P., Kniii-Jones, J., Lowe, D.M., Wilkinson, A.J., Blow, D.M., Brick, P., Carter, P., Waye, M.M.Y. & Winter, G. (1985) Nature, 314, 235-238. 23. Lowe, D.M., Fersht, A.R., Wilkinson, A.J., Carter P. & Winter, G. (1985) Biochemistry in the press. 24. Weiner, S.J., Kollman, P.A., Dase, D.A., Singh, U.C., Ghio, C., Alagona, G., Profeta, S. & Weiner, P.J. (1984) J. Am. Chem. Soc., 10, 765-784. 25. Dessen, P., Zaccai, G. & Blanquet, S. (1982) J. Mol. BioL, 159, 651-664. 26. Bedouelle, H. unpublished.