- Email: [email protected]
Electrostatic effects on modification of charged groups in the active site cleft of subtilisin by protein engineering
J. Mol. Biol.
(1987)193,803-813
Electrostatic Effects on Modification of Charged Groups in the Active Site Cleft of Subtilisin by Protein Engineering Alan J. Russell, Paul G. Thomas and Alan R. Fersht Imperial
Department of Chemistry College of Science and Technology London, S W7 2A Y, England (Received 30 July
1986)
The dielectric constant in the active site cleft of subtilisin from Bacillus amyloliquefaciens has been probed by mutating charged residues on the rim and measuring the effect on the pK, value of the active site histidine (His64) by kinetics. Mutation of a negatively charged surface residue, which is 12 to 13 a from His64, to an uncharged one (Asp + Ser99) lowers the pK, of the histidine by up to 0.4 unit at low ionic strength (0.005 to 0.01 M). This corresponds to an apparent dielectric constant of about 40 to 50 between Asp99 and His64. The tnutation is in an external loop that is known to tolerate a serine at position 99 from homologies with subtilisins from other bacilli. The environment between His64 and Asp99 is predominantly protein. Another charged residue that is at a similar distance from His64 (14 t,o 15 8) and is also in an external loop that is known to tolerate a serine residue is Glul56, at the opposite side of the active site. There is only water in a direct line between His64 and Glu156. Mutation of Glu + Ser156 also lowers the pK, of His64 by up to O-4 unit at low ionic strength. This change again corresponds to an apparent dielectric constant of about’ 40 to 50. The pK, values were determined from the pH dependence of ko,/KMt for the hydrolysis of peptide substrates, with a precision of typically kO.02 unit. The following suggests that the changes in pK, are real and not artefacts of experimental conditions: Hill plots of the data for pK, determination have gradients (h) of - l.OO( +0.02), showing that there are negligible systematic deviations from theoretical ionization curves involving a monobasic acid: the pH dependence for the hydrolysis of two different substrates (succinyl-L-alanyl-Lalanyl-L-prolyl-L-phenylalanyl p-nitroanilide and benzoyl-L-valyl-L-glycyl-L-arginyl p-nitroanilide) gives identical results so that the pK, is independent of substrate; the pH dependence is unaffected by changing the concentration of enzyme, so that aggregation is not affecting the results; the shift in pK, is masked by high ionic strength, as expected qualitat,ivel.v for ionic shielding of electrostatic interactions.
1. Introduction
transmission of electrost,atic effects from such surface groups through a prot)ein is especially difficult to calculate using current, theoretical procedures, because the charges are at, an interface between the high dielectric constant of water and the lower, heterogeneous values of the protein. The standard computational procedures are at their weakest, in this situation (Rogers, 1984: Rogers et al., 1985). It’ is well-documented that altering overall surface charge on an enzyme by ext,ensive chemical modificat,ion can lead to significant changes in pH dependence of catalysis, as for example in t’he serine proteases (Valenzuela & Bender. 197 1). The extensive modifications in those studies can lead to structural changes in the protein and expansion because of the high charge density. The effects of a
Electrostatic elects are thought to play a key role in enzyme catalysis by stabilizing charged transition st)ates (Perutz, 1978; Warshel et al., 1984; LMatthew, 1985). The magnitudes of such effects are, unfortunately, extremely difficult to calculate because of the microheterogeneity of the dielectric constant of a protein. The electrostatic effects of surface charges on the ionization constants of catalytically active groups in enzymes are also interesting from a practical point of view, since modification of charge may provide a general means of tailoring the pH dependence of catalysis using protein engineering (Thomas et al., 1985). The t Symbol ustxl: KM. Michaelis constant. 803
0 1987 Academic Press Inc*. (London) Ltd.
A. J. Russell et al.
804
single substitution are little known. There has been a pioneering study on the effects of selective chemical modification of the surface lysine residues of cytochrome c on its redox potential (Rees, 1980). The consequent results have been criticized by Rogers et al. (1985) because the bulky chemical reagent used for modification could perturb the structure of the protein. The newer technique of site-directed mutagenesis avoids the problem of adding bulky groups to the side-chains of amino acids: a side-chain may be replaced by an isoteric or smaller one. We have begun investigating electrostatic effects in proteins by replacing surface residues of subtilisin from Bacillus amyloliquefaciens (subtilisin Novo or BPN’, EC 3.4.4.16.) by protein engineering (Thomas et al., 1985). The subtilisins are a family of extracellular serine proteases secreted by species of bacilli before sporulation. The crystal structure of the Novo or BPN’ enzyme has been solved (Wright et al., 1969; Drenth & Hol, 1967) and its kinetics wellcharacterized (Philipp & Bender, 1983). His64 acts as a general base during catalysis, accepting a proton from residue Ser221 as it forms a bond with the substrate carbonyl carbon. The enzyme is active only when His64 is unprotonated, at alkaline pH, and catalytic activity varies with the ionization of this residue. The pK, of His64 may be accurately determined by kinetics. The residue we have chosen to modify first is Asp99. This is in an external loop of the enzyme, on the rim of the active site cleft, and is not conserved in subtilisins from other species of bacilli (Table I). In subtilisin Carlsberg (from BuciEZus licheniformis), for example, it is replaced by a serine. The structures of the Novo and Carlsberg enzymes are remarkably similar (McPhalen et al.. 1985a; Bode et al., 1986): a least-squares superof the 274 structurally equivalent position a-carbons of the two enzymes gives a root-meansquare deviation of only 0.53 A (McPhalen et al., 1985a). Further, the conformations of the loop containing either Asp99 or Ser99 are identical. Therefore, a priori, the mutation of Asp + Ser99
should minimize the possibility of any conformational changes that could affect t)he pK, of the active site histidine. We have shown elsewhere t.hat the mutation of Asp + Ser99 causes a shift of about 0.29 unit in the pK, of the enzyme at ionic strengt’h 0.1 M as determined by kinetics (Thomas ef al.. 1985). To show that the mutation of Asp -+ Ser99 is not an isolated result, we have now constructed and analysed a further mutation, Glu -+ Ser 156. which is on the opposite rim of the cleft,. at a similar distance from His64. This residue too is in an external loop, which is known by homology t,o t,olerate a replacement by serine (Table 2). The carboxylate oxygen atoms of Glnl56 are 14 to 15 A from the imidazole nitrogen at,oms of His64 (R. R. Bott, personal communication; M. N. G. tJames: personal communication). Tnterestingly. the environment between the two residues is predominantly aqueous: the carboxylat,e points towards the histidine and there is no protein in a direct line hetween the two. The carboxglate oxygen atoms of Asp99 are 12 to 13 .h from the imidazole nitrogen atoms of His64 (R. R. Kott, personal communication; M. N. G. James, personal communication) but the environment between the two consists largely of protein. Our philosophy for probing enzyme structure and activity by protein engineering is to make only small perturbations in enzyme st,ructure in order to minimize artefacts arising from gross changes in conformation. This approach must also be applied t,o analysing changes in pK, on mutation of surface charge. The pK, values of groups on enz;ymes are sensitive to the overall charge on the protem whicth, itself, changes with pH. It is, therefore, necessa,ry to produce only small perturbations in pK, so that’ the pK, of the wild-type and mutant enzymes may be determined in the same pH range, under identical conditions. We show in this study that WV (*an produce significant, but small, changes in t*he pK, 01 the active site histidine by t)he tlwo mutations and measure them with sufficient precision to allow t,he calculation of electrostatic interact,ions across t)hr active site. These data are used t,o est,ima,tc the dielectric constant, between His64 and the two carboxylate residues.
Table 1 Homology
,$pecies
of subtilisin sequences around aspartate 99
Table 2 Homology
Amino acid sequcncet 90 9.5 100 LYAVKVLGADGSGQYSW LYAVKVLNSSGSGTYSG LYAI KVLNSSGSGTYSA LYAVKVLDSTGSGQYSWI LYAVKVLDSTGSGQYSWI
of subtilisin
sequenws around
glutamate 756 105 I 1 I
t Standard 1 -letter code for amino acids. $ Predicted from the DNA sequence (Wells et al., 1983; Jacobs rt nl., 1985). 3 From Nedov et aI. (1983). 11Stahl & Ferrari (1984). l ’ From Kurihara et al. (19’71).
References as for Table I. t Standard 1-letter code for amino acids. 1 Predicted from the DNA sequence.
The Active Site Cleft of S&&in
2. Materials and Methods (a) Materials Reagents were obtained from Sigma (London). CM-52 ion-exchange resin was obtained from Whatman Ltd. (i) Preparation
and pur@cation
of subtilisin
The gene encoding the serine protease subtilisin BPN from B. amyloliquefaciens was cloned into the vector pUBll0 and expressed in a protease-deficient strain of subtilis (Thomas et al., 1985). Mutation Bacillus of Glu + Ser156 was performed as described for Asp -+ Ser99 the mutagenic primer using 5’-GAAGTGCCT&iGTTACCGG-3’. The entire sequence of each gene was verified by dideoxy sequencing. Wildtype and mutant enzymes were prepared from the supernatant of a culture grown for 36 h in 10 1 of L-broth containing kanamycin (25 pg ml-‘) and glucose (0.2%, w/v). The supernatant was concentrated using a Pellicon membrane filter (PTFCOO05, Millipore) before dialysis buffer 0.01 M-potassium/sodium phosphate against (pH 6.2) and purification on a CM-52 cellulose column as described (Estell et al., 1985; Thomas et al, 1985). All enzymes were purified to electrophoretic homogeneity, and autolytic digestion products were removed before kinetic analysis using a Sephadex G-25 gel filtration column and/or Centricon 10 microconcentrator (Amicon). (b) Methods All experiments were performed at 25°C. The concentration of each enzyme was determined by activesite titration with N-trans.cinnamoyl imidazole (Bender et aZ., 1966). Typical activity was 95% compared with the concentrations measured by A,,, (Eo.lo,o = 1.17; Matsubara et al., 1965). (i) Catalytic constants for the hydrolysis
of succinyl-
L-alanyl-L-alanyl-L-prolyl-L-phenylalanyl
p-nitroanilide (SWA-A-P-Ep-LNA)~ Stock solutions of substrate were prepared in 0.1 MTris. HCl (pH 8.6). The reaction was initiated by the addition of 15 ~1 of enzyme (typically 15 PM in 0.01 Mphosphate buffer, pH 6.2) to 3 ml of buffered substrate (0.01 to 080 mM, in 0.1 M-Tris . HCl (pH 8.6)) in a cuvette maintained at 25”C> in a Gilford 2600 spectrophotometer. Duplicate kinetic runs were performed for at least 10 different substrate concentrations. The increase in absorbance at 412 nm on the release of p-nitroaniline was monitored, and the initial rate calculated from the known extinction coefficient (s4i2 = 8480 M- ’ cm- ‘; DelMar et aZ., 1979). For determinations at pH 6.0, a stock solution of the substrate (69 mM) in DMSO was diluted 106fold to give a roncentration of DMSO of 1% in the cuvette. (ii) Catalytic constants for the hydrolysis of benzoylI,-valyl-L-glycyl-I,-argilzyl p-nitroanilide (bz- V-G-Rp-n;A) determined as for the kc,, were Khl and su-A-A-P-Fp-NA substrate, except that the range of substrate concentrations was 0.01 to 2.0 mM, in 0.1 MTris HCl buffer (pH 8.6), and the enzyme was typically 30 /LM in 0.01 .v-phosphate buffer (pH 6.2).
t Abbreviations used: su-A-A-P-Fp-NA. succinyl-nananyl-rd-alanyl-I,-prolyl-L-phenylalanyl p-nitroanilide: bz-V-G-Rp-NA, benzopl-I,-valyl-L-glycyl-L-arginyl p-nitroanilide; DMSO, dimet,hylsulfoxide.
805
(iii) pH dependence of kccrr/KMfor the hydrolysis of su-A-A-P-Fp-NA and bz- V-G-Rp-NA Phosphate buffers of ionic strength 0.1 M were used and the ionic strength adjusted by the addition of KCl, except for ionic strength 0.025 M and lower, where the stock buffers were diluted. pH was measured before and after addition of the enzyme using a PHM64 research pH meter and electrode (Radiometer, Copenhagen), and the results were discarded if there was significant change during reaction. The meter was calibrated with standard buffer solutions (Radiometer, Copenhagen): 15 ~1 of enzyme (15 pM in 0.01 M-phosphate buffer, pH 6.2) was added to 3 ml of buffered substrate (0.01 mM), the initial substrate concentration being at least 15 times lower than KM. At very low ionic strengths (<0.025 M), the buffering capacity of the phosphate buffer is so low that small changes in pH occur on addition of buffered substrate. Substrate was added for these experiments from a concentrated solution (69 mM) in DMSO to give a final concentration of 0.03% (v/v) solvent. The increase in absorbance at 412 nm was followed to completion spectrophotometrically and found to obey first-order kinetics with high precision. The value of k,,,lK, at each pH was determined from the first order plots exactly as described for the hydrolysis of acetyl-I,-tyrosine p-acetylanilide by chymotrypsin (Fersht & Renard, 1974). High pH values were obtained using both Tris . HCl and bicine-NaOH buffers. Control experiments showed that the rate of reaction with these buffers was lower than in phosphate buffers and, although corrected, these data were not used in the analysis. It was necessary to measure very small changes in absorbance for these determinations. The digital output from the spectrophotometer was fed directly int’o a BBC microcomputer using a data storage program written by Dr R. J. Leatherbarrow of this department. The data were fitted to theoretical curves by non-linear regression on the BBC microcomputer (R. J. Leatherbarrow. unpublished results; Marquardt, 1963). Excellent’ results were obtained for changes as low as 0.03 absorbance unit, which corresponds to a substrate concentration of 3.5 PM, over a time-course of 60 to 300 s.
3. Results (a) Measurement
of pK,
values and controls
a i n of ca tita y ic constants (i)iCZ a cu Et0
(Table
3)
of the synthetic substrates The hydrolysis succinyl-~-alanyl-~-alanyl-~-prolyl-~-phenylalanyl p-nitroanilide or benzoyl-L-valyl-L-glycyl-I,-arginyl p-nitroanilide by subtilisin is described by: E+&E.S.
in
which
2 EA2E+A
the
relevant
(kc,, = k,kd(k2 +kd;
(Scheme
constants
are
1)
defined
41 = K&/(-h + k3)).
For
nitroanilide substrates k, x k, (Philipp & Bender, 1983): the ratio of k, (deacylation) to-i* (acylation) has been report’ed to be 33 (Wells et al., 1986), thus steady-state measurements of k,,, and K, give directly k, and KS, respectively (Fersht, 1985). The activity of wild-type subtilisin towards su-A-A-P-Fp-NA (0.1 M-Tris . HCl, pH 8.6) is similar to that of the mutant (Asp -+ Ser99) under
A. J. Russell et al.
806
Table 3 Catalytic
Enzyme Wild-type
Glu --) Serl56
of p-nitroanilide at 25°C
substrates
by
enzyme
k cat (SC’)
KM
Buffer
PH
(mM)
Tris HCl (0.1 M) Tris . HCl (1% DMSO) K-phosphate (1% DMSO) Tris HCl (0.1 M)
8.6 8.6 6.0 8.6
0~17~0~01 0~09~0~01 0.44 * 0.04 0.32 + 0.05
57 42 11 1.3.5
3.8 x 4.7 x 2.5 x 4.2 x
10’ lo5 lo4 103
bz-V-G-Rp-NA
Tris HCl (0.1 M) Tris HCl (1% DMSO) K-phosphate (1% DMSO) Tris HCl (0.1 M)
8.6 8.6 6.0 8.6
0~13~0~01 0~10f0~01 0.54 + 0.05 0.92 +0.10
4.5 35 14 4.00
3.5 2.9 2.6 4.3
105 1O5 IO4 lo3
SWA-A-P-Fp-NA
Tris HCl (0.1
8.6
0.05 + OXtO
55
1.1 x lo6
Substrate su-A-A-P-Fp-NA bz-V-G-Rp-NA
Asp --t Ser99
constants for the hydrolysis wild-type and mutant
su-A-A-P-Fp-NA
M)
the same conditions: k,,, for the wild-type enzyme is 57 s-l and KM is 0.15 mM, whilst the corresponding values for the mutant enzyme are 45 s- ’ and 0.13 mM. The wild-type enzyme has activity identical with that produced from an independently cloned gene (Wells et al., 1983). The mutant Glu + Ser156 has a similar value for k,,, (MS-‘) but the value of KM is distinctly lower at 0.05 mM. The higher value of KM with Asp + Ser99 allows easier kinetic studies with this mutant, since experiments on the determination of pK, must be performed at [S] << KM. Consequently, Asp + Ser99 was chosen for more extended study. su-A-A-P-Fp-NA is not soluble in buffer at low pH values ( < 8.0), and so was dissolved in DMSO in order to measure values of k,,, and KM at pH 6.0. A control experiment at pH 8.6 on the effect of DMSO added to 1% (v/v) showed that k,,, is lowered by 150/b, whilst KM is unchanged. There is a threefold and mutant increase in KM for both wild-type enzyme on lowering the pH to 6.0. We have made no attempt to analyse the effects of mutation on k,,, and K,, since observed values of KM may be obscured by alternative modes of binding that are known to occur with subtilisin (Poulos et al., 1976). Observed values of k,,, may also reflect the relative occupancy of productive and non-productive modes of binding (Fersht, 1985). More extensive studies on the effects of electrostatic changes on values of KM are being performed elsewhere (J. A. Wells, T. P. Graycar. R. R. Bott & 1). Estell, personal communication). of pK, of the active site (ii) Determination Monitoring the pH dependence of k,,,lK, gives the ionization constants of free enzyme and free substrate (Fersht, 1985). Thus, pK, values determined from such plots should be independent’ of substrate in regions where the substrate does not ionize. The pH range investigated was restricted to the region around neutrality. This minimizes perturbations caused by the ionization of other residues, e.g. Asp and Glu at, low pH values and Lys: Tyr. Cys and, eventually, Arg at high pH values. The region of pH 6 to 8 spans nicely the expected pK,, about 7. The measured values of
(s-1 6’)
x x x x
katlK, could be fitted with high precision t.o theoretical ionization curves of a single base. The value of kca,/KM at each pH value was determined from measurements of the release of products at substrate concentrations much lower than KM as described by Fersht & Renard (1974). Under t’hese conditions, first-order kinetics should be obeyed, which are independent of initial substrat’e concentration. thus avoiding one source of error. Reactions were monitored for at least ten half-lives and excellent> first-order plots were obtained. The value of KM for t)hr hydrolysis of su-A-A-P-FpNA by wild-type enzyme is 0.44 T?IM Rt pH 6 and 0.15 mM at pH 8.6. The initial substrate concentrat~ion of O*Ol mM is at least, 15 times lower than K, at all values of pH. Further, since the firstorder plots were analysed over several half-lives, even lower concentrations of subst,rates were actually used for the determination of k,,,/K,. ‘l’he low concentration of substrate used in this method also avoids product inhibition. Experimental corlditions are summarized in Table 4. Measurement’s of wild-type and mutant enzymes were performed in parallel. Any errors in standard solut*ions for the calibration of the pH meter should thus cancel out, when taking the difference between wild-type and mutant. The reproducibility of pK, values is higher than that for kinetic constant,s, since only the relative values of k,,,/K, at various values of pH are required for the calculation and not the absolute values. (iii) pH dependerxe su-A-A-P-Fp-NA
of hydrolysis
(Table
of
5, Fig. 7)
The only ionizing group on the substrate is the succinyl group and its pK, is below the range of this study. At ionic strength 0.1 M, His64 in the wildtype enzyme has a pK, of 7*17( + 0.02); this is decreased to a value of 6.88( kO.02) in the Asp -+ Ser99 mutant, a shift of 0.29( &O-03) unit (Fig. 1). k,lKM is decreased by 2094 on mutation. The pK, of the wild-type enzyme is in excellent agreement with that derived from the hydrolysis of ester substrates under identical conditions and at a wide range of enzyme concentrations (Philipp et al.. 1979). Lowering the concentration of mutant
The Active
Site Cleft of Subtilisin
807
Table 4 Experimental Enzyme
Expt no.
Wild-type
conditions
No. of pH points
x0. of kinetic runs
6.1-8.0 6.1-8.0 6.1-8.0 5.5-8.9 6.1-8.2 6.0-7.9 6.0-7.9 6.1-7.8 5.5-8.9 5.5-7.8
18 18 18 25 16 15 15 20 25 15
54 54 54 75 32 30 30 40 75 30
0@05 0.01 0.025 0.10 0.10 0.10 0.10 0.50 1.00 l@O
6.1-8.00 6.1-8.0 6.1-8.0 5.5-8.9 6.1-8.2 6.0-7.9 567.9 6.1-7.8 5.5-8.9 5.5-7.8
18 18 18 25 16 15 13 20 25 15
54 54 54 75 32 30 26 40 75 30
0@05 0.01 0.02.5 0.10 0.10 0.10 0.10 0.50 1X@ lxm
6.1-8.0 6.1-8.0 6.1-8.0 5.6-8.0 6.0-8.0
18 18 18 18 18
108 54 54 54 54
o+M15 0.01 0.025 0.10 1.00
1t 3t 4 5
61 7 8 9
10 11t 12t 1st
14 15
161 17§ 18 19 20 (Glu + Ser156
determinations
pH range
2t
Asp -+ Ser99
for pK,
21t w w
24 25
Ionic strength (M)
All experiments were done at 25°C; initial substrate concn,
rather than a
Table 5 The pH dependence of hydrolysis by wild-type succinyl-~-alanyl-~-alanyl-~-prolyl-~-phenylalanyl
Enzyme Ionic strength 0.005 .nt \I’ild-type .4sp + Her99 (ilu + Her156 Ionic strrngth 0.01 .vt \Vild-type Asp + Ser99 (:lu + Serl56 Ionic
and mutant
subtilisin
of
p-nitroanilide Limiting value of k,,,/K, at high pH
Expt no.
pK, of active site from kinetics (*s E.)
1 11 21
6.99+0.01 6.61k 0.01 6.67f0.03
4.03 x lo5 4.02 x 10’ 9.21 x lo5
2 12 22
7.08 * 0.02 6.66 + 0.02 6.66 * 0.03
4.06 x lo5 4.01 x 105 9.28 x lo5
3 13 23
7.0410.01 6.68 f 0.02 6.63kO.02
4.90 x 105 4.04 x 105 9.28 x lo5
4 14 24
7.17f0.02 6~88+0~02 6.92 kO.03
4.89 x IO5 4.08 x 1OS 9.09 x lo5
8 18
7.10f0.02 7.00 + 0.02
4.66 x lo5 4.08 x IO5
9 19 25
7.11 kO.02 7.09 f 0.02 7.17kO.03
4.48 x lo5 4.16 x lo5 1.64 x lo6
? DMSO present at a concn of 0.03%
(S-l
M-‘)
808
A. J. Russell et al.
5
6
7
0
9
IO
PH
Figure 1. pH dependence of Ic,,,/K, for the hydrolysis of su-A-A-P-Fp-NA by wild-type subtilisin (0) and the mutant Asp -+ Ser99 (0) at ionic strength 0.1 M. Data are normalized to loo:/, for the limiting value at high pH for each mut,ant.
enzyme Asp -+ Ser99 tenfold caused no significant change in its measured pK, (6.81( +0.04)). Aggregation therefore does not appear to he obscuring the results for either enzyme. (iv) $
rj;ep;yence a e
of hydrolysis
of bz-V-G-Rp-l\r,4
The pH dependence of kca,/KM should be independent of substrate in regions where the substrate does not ionize and so a second substrate was used to check that the pK, difference between wild-type and Asp + Ser99 is not an artefact of using su-A-A-P-Fp-NA. The pK, at ionic strength 0.1 M of wild-type subtilisin determined from the second substrate is 7.15(&0=02), whilst that of Asp -+ Ser99 is 6.92( +0.02); a pK, shift of 0.23( kO.03) pH unit, that is a 1.7.fold shift in K,. This is in good agreement with the results from the experiment using su-A-A-P-Fp-EA. (v) Effect of ionic strength on pK,
(Tables
intrinsic value in t.he absence of surface charge. Accordingly, we found from measurements on the hydrolysis of su-A-A-P-Fp-NA that,. at ionic strength 1.0 M, the pK, values of the active sit,es of wild-type and both mutant enzymes converge. At an intermediat,e ionic strength of 0.5 M, t,here is a shift in pK, of 0.1 unit between wild-type and Asp -+ Ser99. At the lowest ionic strengths, 0.005 t’o 0,025 M, the difference in pK, increases to about 0.4 unit,. The effect of high ionic strength on the pH dependence of t*hr second su bstrat,e. bz-V-G-Rp-NA, was identical within experimental error. The value of k,,,/K, for hydrolysis of bz-V-G-Rp-NA by subtilisin is much lower fhr t,his substrat’e and caused slightlv more error berause of longer time-courses (error -in f)K, = +043). ‘I%~ values for pK, quoted in Table 6 are the average values of both substrates and other variables. The Linderstrclm -Lang equation (set, IGdsali CI Wyman, 1958) predicts that the pK, of a group on will change with ionic strength an enzyme according to the net surface chargcx. Below thr isoelect’ric point. the pli, will incre,se and above it will decrease the magnitude of the change drprntling on net surface charge. The very small incsrrase in pK, of wild-type enzyme, 6.99 at ionic strength O-005 M t,o 7.13 at ionic strength 1.0 M, indic:iitrs ii small net positive charge on the enzyme, and is consistent, with the known isoelectric point of 7.X for subtilisin BPN’ (Matsubara et al.. 1965).
A further possible artefact is that c~alcium ions may bind to t,he enzyme and affect, the results (subtilisin is known to have calcium binding sites but Asp99 and Glu156 are not, implicated in these). Care was taken throughout t,he growing and harvesting of (aells and the subsequent steps t,o minimize the presence of calcium ions. Cont,rol experiments showed t,hat t#he presence of the calcium-chelating agent) EGTA has no effect on the pK, values of wild-type or mutant, enzymes at ionic strength 0.1 >I.
5 and 6)
The presence of high concentrations of ions should mask electrostatic interactions. At high ionic strength, the observed pK, should tend t’o its
(vii) Hill
plots of titration
data
A good indication of the quality of the fit’ of data to theoretical titration curves is the equiva,lent of
Table 6 The pH dependence of hydrolysis of N-benzoyl-~~-~~alyl-~~-gly~yl-~,-arg~~ni~~e p-r&roan&de by wild-type and m,utnnt subtilisin
Enzyme
All experiments
Expt no.
ph’, of active site from kin&w (&SE.)
were done at %5”(‘; buffered substrate wnm,
042 mM.
Limiting vitlue of Icca,/KM at high pH (s-l arc’)
The Active
809
Site Cleft of Subtilisin
Table 7 Effect of EGTA on pH dependence of hydrolysis by a wild-type and mutant subtilisin at 0.1 M ionic strength Enzyme
Expt no.
LVild-t,ype Asp -+ SW99
pK, of active site (from kinetics)
6 16
No EGTA
0.1 mM-EGTA. 0.03% DMSO
7~17+0.02 6+48+0~02
7.18f0.01 692f0.02
Experiments were done at 25°C. with 0.01 rnM buffered potassium phosphate buffer)
the Hill plot of 1% {I (kcatl&)max - (k,,,/li’M)ob,l/[(kcat/KM)obsl) against pH, where (k,,,/K,),,, is the limiting maximum value of k,,,/K, at high pH values and (katlKdo,s is the value of kca,/KM at the particular pH value. For a single ionization, this should be a straight line of slope h = - 1. This is found with high precision over all ionic strengths (Table 8, Fig. 2). The fit to such plots is generally h = - l.OO( kO.02).
(b) Estimation
of eSfective dielectric constants
(i) Calculation of dielectric constant, D Substitution of the known values of the fundamental physical constants into the standard formula for electrostatic interactions gives D = 332z,z,lrAC, where z1 and z2 are the two charges, measured in units of the charge of an electron, separated by a distance of r a and which wit,h an energy of AG kcal/mol interact ( 1 kcal = 4.184 kJ). Alternatively, for a measured The shift in pK, of ApK,, D = 244z,z2/rApK,. distances between the necessary interatomic carboxylate oxygen atoms of Asp99 or Glu156 and the 6-N and E-N atoms of the imidazole ring of His64 are known with some precision: unpublished high-resolution co-ordinates (1.8 A) from R. R. Bott (personal communication) on subtilisin Novo and similar data (2.1 A resolution) from McPhalen et aZ.
so-A-A-P-Fp-NA
(0.01 M ionic stren@h
(1985b; M. Iv. G. James, personal communication) on the complex with the barley (‘I-2 inhibitor give mean 0 to N distances that agree within 0.3 to 0.5 8. A value of 12.4 A was used for Asp99 carboxylate oxygen atoms to His64 imidazole nitrogen atoms, and 14.5 A for Glu156 to His64. Dielectric constant varies with ionic strength because of shielding by the ionic atmosphere (Table 9). The observed value at an ionic strength greater than zero is usually termed the effective dielectric constant (Hill, 1956). Denoting t$his by D,, and t’he value at zero ionic strength as D,, then, for simple point charges in solution separated by distance r: De, = D, exp (Kr),
where IC is the conventional term in the DebyeHiickel equation, i.e. the reciprocal of the Debye length. Since ICis proportional to a. where Z is the ionic strength: In D,, = In D, + afi,
(2)
where c( is a constant. Hill (1956) has analysed the effective dielectric constant between two charges embedded in idealized proteins consisting of waterimpenetrable spheres, but there is no simple theory to account for charges located on the surface of real proteins. The correct equation for the description of the effect of ionic strength will certainly be more than complex equation (2). Nevertheless, equation (2) fits the experimental da.ta (Fig. 3) and was used to extrapolate D,, to zero ionic strength
Table 8 Gradient
(h) of Hill plots for pK, -h
Substrate WA-4.P-Fp-NA
bz-V-GKp-NA
Ionic strength (M) 0.005 0.01 04)25 0.10 0.50 1.00 0.10 1.00
(1)
Wild-type 140+0~01 1~00~041 1.00 & 0.0% 0.99 * 0.02 1~00~0~01 1.03 + 0.02 140~0~02 0.99 f 042
Asp --t SW99
Glu - Ser156
0.99 * 092 1.01*0@2 0.99 f 0.03 140,042 0.99 + 0.01 1.02 f 042 l~oo+o~Ol 0.99fOa2
1.03 * 0.03 0.99f04)l 0.98 & 0.03 0.99 + 043 1.02 + 0.02 140 * 0.02
The Hill plot is log [(k,,,/KM),,,(k,,,/K~),,,]/[(k,,,/K,),,,1 wrsus pH, where (k,,,/KM),,, limiting value at, high pH and (kca,/KM)obr is the value at the particular pH.
is the
A. J. Russell et al.
810
'6
7
'6
8
7 PH
PH
1=0.025~
'6
I=
o-5
I=
M
-I
'6
7
7 PH
PH
8
PH
6
7
l.OOM
0
PH
Figure 2. Hill plots (log [(k,,,/K,),,,(k,,,/h’,),,,]/[(k,,,/K,),,,] versus pH, where (k,,,/K,),,, is the limiting value at is the value at the particular pH) for the hydrolysis of su-A-A-P-Fp-NA by wild-type and high PH and (kJ&)Obs mutant Asp -+ 8er99 subtilisin. For ionic strengths 0405, 04lO1, 0.025, 0.1 and 1.0 M, the average values of the data for each pH value are plotted; for ionic strength 0.5 M, the individual values (2 for each pH value) are shown to illustrate t,he spread. (Fig. 3). The extrapolation is, in any case, small were made down to measurements since 1 = O-005 M. Figure 3 also nicely illustrates the effect of ionic atmosphere on the effective dielectric constant. The effective dielectric constant between
Asp99 and His64 extrapolated to I = 0 is 42( +4). The lowest observed values are 47 t,o 55 and are found in the range I = O-005 t,o 0.025 M. The observed values between Glu156 and His64 are between 40 and 50 in this range.
The Active
811
Site Cleft of Subtilisin
Table 9 Effective dielectric constant (D,,,) at varying ionic strength Ionic strength (M)
PK Wild-type
Asp + Ser99
04)05t 0.01t
6.99 + 0.01
6.61 kO.01
0.38+0+2
0425t
7.08k0.02 7.04 k 0.01
6.66kO.02
0.42f0.03
0.10
7.17f0.021
6.68 + 0.02 690+0.02$
0.36 f 0.03 0.27kO.03
0.50
7.1OkO.02 7.13kO.02:
7GO~O.02 7.08+0.02$
O.lOf0.03 0.05+0.03
I .oo
LVild-type 0~005t 0.01t
o-025t
0.10
I7eff
APK,
52+3 47+3 55f5 73k8 200+60 400&-240
Glu + Ser156
6.99 k 0.01
7.08 k 0.02 7.04+0.01 7.17 + 0+2$
t DMSO conm. O.O3”i,. data $ Averaged for bz-V-G-Rp-NA.
3’
6.67 k 0.03
0.32 kO.04
6.66 + 0.02 6.63 k 0.01
0.42 f 0.03
52k7 40*3
6.92 + 0.03
0.41 f0.02 0.25 + 0.04
4152 67k9
both
su-A-A-P-Fp-NA
I 0.1
I 0.2
I 0.3
1 0.4
I 0.5
I 0.6
I 0.7
I 0.0
rh
and
Figure 3. The effect of ionic strength on effective dielectric constant between Asp99 and His64. The experimental results are plotted according to eqn (2), although it is expected that a more complex equation will be required to fit’ data of higher accuracy (see t,he text for details).
4. Discussion (a) Precision
of kinetic determination pK, values
of
Kinetics provides a very sensitive and convenient, means of determining the pK, values of active sites. Nuclear magnetic resonance spectroscopy is, of course, a far more general procedure for determining the pK, values of groups in proteins, especially histidine residues. But, for subtilisin, kinetics has two significant advantages. First, the activity of the enzyme tends to zero at low pH values and so the rate data fit an ionization equation of just two unknowns; the pK, and limiting value of rate constant at high pH. Nuclear magnetic resonance spectroscopy requires estimating three unknowns; the pK, and the limiting values of the chemical shift at both high and low pH values. Second, the experiments are performed at low concentration of enzyme and so autolysis is minimized. We find that the pH dependence of k,,,lK, between pH 6 and 8 fits with high precision the expected equation for the ionization of a single base, both to direct titration curves and derivative plots such as the Hill plot (Figs 1 and 2). Provided there is no change of ratedetermining step with pH, the pH dependence of kcaJKM gives the pK, of the unligated enzyme or the free substrate. The substrates employed here do not ionize in the region of the measurements and so t,he observed pK, values should be those of the enzyme. The reproducibility of the data is indicated from three different measurements at ionic strength enzyme: 7.15( -&0.02), 0.1 M of the pK, of wild-type ‘i.l’i(f0.02) and 7.18( kO.01). These values were determined for the positively charged substrate bz-V-G-Ap-NA and the negatively charged substrate su-A-A-P-Fp-NA in the absence and
presence of EGTA, respectively. (A wide range of ester substrates has given a value of 7.15 for the pK, of wild-type enzyme at ionic strength 0.1 M (Philipp et al., 1979).) The corresponding values for mutant Asp -+ Ser99 are: 6.92( + 0.02)) the 6.88( kO.02) and 6.91( kO.02). The changes in pK, respond to changes in ionic strength in the predicted manner: the differences are suppressed at high ionic strength and gradually increase with decreasing ionic strength. The perturbation of pK, on the mutation Asp + Ser99 is not a unique event. The mutation Glu --+ Ser156 also causes a similar change in pK,. We are therefore confident that we are measuring real and statistically significant effects on the pK, of the active site of the enzyme. (b) Magnitude
of pK,
shift
The results of previous experiments using extensive chemical modification predicted that we would find negligible changes in pK, on the modification of single charges remote from the active site histidine. Valenzuela & Bender (1971) reported a change of 1-O unit in the pK, of chymotrypsin (His57) on altering the overall surface charge by 28 units (i.e. changing 14 positively charged lysine residues to 14 negatively charged succinate half-amides by succinylation with succinic anhydride). A slightly lower effect was noted for the conversion of 13 surface carboxylates to positively charged residues by coupling with ethylenediamine. Acetylation of all surface lysine residues of trypsin lowers the pK, of its His57 by only 0.2 unit (Spomer & Wootton, 1971). Such drastic changes in surface charge could well alter the properties of the enzyme and/or surrounding solvent. Reorientation of solvent and surrounding
812
A. J. Russell et al.
counterions or an expansion of the protein caused by electrostatic repulsion of surface residues could affect’ the pK, of the active site by structural changes in addition to direct electrostatic interactions. (c) Dielectric constant across the active site cleft It is debatable whether the term dielectric constant,, which represents a macroscopic phenomenon, should be applied to proteins which are very heterogeneous in their dielectric properties. Nevertheless, it is important to know or calculate the transmission of electrostatic effects in proteins and hence know values of effective dielectric constants. The measured dielectric constant’ between Asp99 and His64 tends to about, 42 at zero ionic strength. Values of 47 to 55 are observed in the range of ionic strengths of 0.005 to 0.025 M. The measured dielectric constant between Glu156 and His64 is in the range 40 to 50 in this range of ionic strength. Rogers et aE. (1985) report a value of 27 between the iron of cytochrome cssl and a propionate 8.2 L%distant, measured from the change in redox potential on ionization of the propionate. But, both ions are buried in the protein by at least 5 A in this example. A change of ionic strength from 0.007 to 0.1 M also does not effect the change in redox potential on ionization of the propionate. The crystal structure of subtilisin shows predominantly protein between Asp99 and His64. There is no protein whatsoever in a direct line between Glu156 and His64. The observed values of the dielectric constant. are about half the value of 78.5 expected for water. There could be several reasons for this. First, there is evidence that ions may lower the dielectric constant of solvent water by a combination of solvent orientation and inductive polarization (e.g. see Mehler & Eichele. 1984). A tabulation of values by Conway et al. (1951) gives a value of dielectric constant of 65 at 12.4 A from a singly charged ion and 72 at 14.6 8. Second, some of the electrostatic field of the ion will be transmitted via the low dielectric constant of the protein. Third, the water molecules in the active site may differ from bulk water because of the orienting effects of the cavity and its amino acid side-chains. The observed increase in effective dielectric constant with increasing ionic strength is in accord with the classical Debye-Hiickel theory (Hill, 1956). It has been observed only rarely that ionic solutions have greater dielectric constants than do the pure solvents (Pethig, 1979). (d) Apparent dielectric constant The term effective dielectric constant is generally used to denote the observed values of dielectric constant when the dielectric is inhomogeneous. We suggest the use of the term apparent dielectric constant to describe the observed values of dielectric constant that are measured by site-
directed mutagenesis since, as discussed next. the observed value of the dielectric constant may have additional components from changes in wat,er structure. Mut’ation of residues that int’eract directly with a substrate may have complications caused by local changes in protein conformation that affect other interactions. Small local changes in conformabion are less likely to cause problems in experiments such as those described here, where a remote surface residue is mutated and long-range electrostatic interactions are monitored. But there could be effects transmit.ted by solvent’. The static dielectric constant of a polar liquid has contributions from two components: polarization bv induction and orientation of dipoles. For water. orientation of dipoles is by far the overwhelming contribution, accounting for about, 98Y; of the observed dielectric constant. (The cont~ribution from polariza,tion is calculated to he 1.79 from t)hfl refractive index, n, using Maxwell’s equation. U = n’.) hlutation of a charged side-chain of an amino acid may result in changes in the structure of water around the protein and. hence, the orientation of wat,er. Any change in dielet:t,ric consta,nt brought) about by reorientat)ion of solvent by t,he presence of an ion is, of course, an inhrrent component of t’he effect’ive dielectric constant of the medium surrounding that charge. In the abxencc of other charges affecting the pK, value of the a#ctivc site, the measured value of the dielectric constant would be identical with t,he classical effective dielectric constant. But if there are ot,her cha.rges present. t,hey could be affected indirectIT 1)~ changes in solvent structure. Mutat,ion of one residue could alter the effective dielectric constant between other charged residues and the ionizing group at the active site and, consequently. influence the magnitudes of other electrostatic interact’ions. These changes will be reflected in the measured effective dielectric constant. Because of this, we suggest the use of the’ term apparent dielectric constant, to acknowledge that bhe measured value of t,he dielectric constant contains additional components. Other interactions measured by sit~edirected tnutagenesis have component,s from changes in water stru&ure (e.g. hydrogen honds. Fersht et al., 1985) in addition to changes in protein conformation, and the term apparent is generally useful, since it describes experimentally observable results (Wells &, Fersht, 1986). We art’ condu&ng an extensive survey of the effects of mutation of surface charges on the ph’, of histidine residues in subtilisin t,o t,est, for effects of water strut:t.ure anti also to map the microhet,erogeneit’y of dielectric constant within other regions. Wr thank Dr M. N. (2. tJames for unpublished CC)ordinates of the complex between subtilisin Novo and t,hr barley CT-2 inhibitor and Dr R. Bott for measurements of interatomic distances in subtilisin Xovo. We thank also Drs D. Estell, ,J. Wells a.nd T. Graycar for exchange of’ unpublished information.
The Active Site Cleft of Subtilisin
References Bender, M. I,.. Begoue-Canton, M. L., Blakeley, R. L., Brubacher, I,. ,J., Feder, J., Gunter, C. R., Kezdy, F. J., Killheffer, J. V., Marshall, T. H., Miller, C. J.. Roeske, R. W. & Stoops, J. K. (1966). J. Amer. Chem,. Sot. 88. 5890-5914. Rode. W., Papamokos, E., Musil. D.. Seemueller, C. 8r Fritz, H. (1986). EMBO J. 5, 813-818. (lonway. B. E.. Bockris, J. O’M. & Ammar, I. (1951). Trans.
Faraday
Sot. 47. 756-766.
DrlMar, E. G.. Largman, C., Brodrick, J. W. & Goekas, M. C. (1979). Anal. Biochem. 99, 316-320. Drenth, ,J. & Hol. W. G. ,J. (1967). J. Mol. Biol. 28, 543551. Chemistry. Edsall, J. T. & Wyman, ?J. (1958). Biophysical p. 510. Academic Press. Xew York. Estell. I). A., Graycar. T. P. & Wells, J. A. (1985). J. Biol. (‘hem. 260, 6518-6521. i-\. R. ( 1985). In Enzyme Structure and Fersht. Mechanism, 2nd edit., chapt. 5, W. H. Freeman. Xew York and Oxford. Fersht. A. R. & Renard, M. (1974). Biochemistry, 13. 1416-1426. Fersht. A. R,.. Shi, J. I’.. Knill-Jones, J., Lowe, D. M.. Wilkinson. A. J.. Blow. D. M., Brick, I’., Carter. P.. Waye. M, M. Y. & Winter. G. (1985). Nature (London), 3 14, 235-238. Hill, T. L. (1956). ,J. Phys. Ch,em. 60, 253-255. *Jacobs. M.. Eliasson. M.. Uhlen, M. & Flock. J. (1985). LVucl. Acids ties. 13. 8913-8927. Kurihara. M.. Markland, F. & Smith. E. (1972). .I. BioZ. (‘hem,. 247, 5619-5631. Maryuardt. 1). W. (1963). J. Sot. Jndust. Appl. Math. 11, 431-441. Matsubara, H.. Kasper. C. B.. Brown, D. M. & Smith. E. L. (1965). J. Biol. Chem. 240, 1125-1130. Matthew, J. H. (1985). Annu. Rev. Biophys. Chem. 14, 387-417. McPhalen. (‘. ,4.. Schnebli. H. P. & James. M. I$, G. (19850). FEBS Letters, 188, 55-58.
813
McPhalen, C. A., Svendsen, I., Jonassen, I. & James, M. N. G. (1985b). Proc. Nat. Acad. Ski.. V.S. A. 82, 7242-7246. 23. Mehler, E. L. & Eichele, G. (1984). Biochemistry, 3887-3891. Xedov, P., Oberthur, W. & Braunitzer, G. (1983). HoppeSeyler’s 2. Physiol. Chem. 364, 1537-1540. Perutz, M. (1978). Science, 201, 1187-1191. Pethig, R. (1979). In Dielectric and Electronic Properties qf Biological Materials, p. 137, Wiley, Chichester. Philipp, M. & Bender, M. L. (1983). Mol. (‘ell. Biochem. 51. 5-32. Philipp. 11,. Tsai. I. H. & Bender. M. L. (1979). Biochemistry, 18, 3769-3773. I’oulos, T. I,., Alden, R. A., Freer. S. T., Birktoff, J. J. $ Kraut, J. (1976). J. Biol. Chem. 251. 1097-1103. Rees, D. C. (1980). J. Mol. Biol. 173. 323.-326. Rogers. h‘. K. (1984). D.Phil. thesis, Oxford University. Rogers. h’. K., Moore, G. R. & Sternberg, M. ,J. E. (1985). J. Mol. Biol. 182, 613-616. Spomer, W. E. & Wootton, ,J. F. (1971). Biochim. Biophys. Acta, 235, 164-171. Stahl. M. L. & Ferrari. E. (1984). .I. Bacterial. 158, 411L 418. Thomas, P. G., Russell. A. J. CGFersht. A. R. (1985). Xatuw (London), 318, 375-376. Valenzuela. P. & Bender, M. T,. (1971). Biochim. Hiophys. Acta, 250, 538-548.
Warshel, A., Russell, S. T. & Churg; A. K. (1984). I’roc. Nat. Acad. Ski., U.S.A. 81. 4785-4789. Wells. *J. A., Ferrari, E., Henner. D. J.. Estell. D. A. & Chen. E. Y. (1983). Nucl. Acids Res. 11. 7911-7925. Wells. J. A., Cunningham, B. C.. Graycar, T. I’. & Estell. I). A. (I 986). Phil. Tmnx Roy. Sot. ser. A. 317. 415.. 423. Wells. T. ?rj. C. 8: Fersht. A. R. (1986). Biochemistry, 25, 1881-1886. Wright, C. S.. Alden, R. A. & Kraut. ,J. (1969). Nature (Lon,don) , 221, 235-242.
Edited by G. A. Gilbert
(1987)193,803-813
Electrostatic Effects on Modification of Charged Groups in the Active Site Cleft of Subtilisin by Protein Engineering Alan J. Russell, Paul G. Thomas and Alan R. Fersht Imperial
Department of Chemistry College of Science and Technology London, S W7 2A Y, England (Received 30 July
1986)
The dielectric constant in the active site cleft of subtilisin from Bacillus amyloliquefaciens has been probed by mutating charged residues on the rim and measuring the effect on the pK, value of the active site histidine (His64) by kinetics. Mutation of a negatively charged surface residue, which is 12 to 13 a from His64, to an uncharged one (Asp + Ser99) lowers the pK, of the histidine by up to 0.4 unit at low ionic strength (0.005 to 0.01 M). This corresponds to an apparent dielectric constant of about 40 to 50 between Asp99 and His64. The tnutation is in an external loop that is known to tolerate a serine at position 99 from homologies with subtilisins from other bacilli. The environment between His64 and Asp99 is predominantly protein. Another charged residue that is at a similar distance from His64 (14 t,o 15 8) and is also in an external loop that is known to tolerate a serine residue is Glul56, at the opposite side of the active site. There is only water in a direct line between His64 and Glu156. Mutation of Glu + Ser156 also lowers the pK, of His64 by up to O-4 unit at low ionic strength. This change again corresponds to an apparent dielectric constant of about’ 40 to 50. The pK, values were determined from the pH dependence of ko,/KMt for the hydrolysis of peptide substrates, with a precision of typically kO.02 unit. The following suggests that the changes in pK, are real and not artefacts of experimental conditions: Hill plots of the data for pK, determination have gradients (h) of - l.OO( +0.02), showing that there are negligible systematic deviations from theoretical ionization curves involving a monobasic acid: the pH dependence for the hydrolysis of two different substrates (succinyl-L-alanyl-Lalanyl-L-prolyl-L-phenylalanyl p-nitroanilide and benzoyl-L-valyl-L-glycyl-L-arginyl p-nitroanilide) gives identical results so that the pK, is independent of substrate; the pH dependence is unaffected by changing the concentration of enzyme, so that aggregation is not affecting the results; the shift in pK, is masked by high ionic strength, as expected qualitat,ivel.v for ionic shielding of electrostatic interactions.
1. Introduction
transmission of electrost,atic effects from such surface groups through a prot)ein is especially difficult to calculate using current, theoretical procedures, because the charges are at, an interface between the high dielectric constant of water and the lower, heterogeneous values of the protein. The standard computational procedures are at their weakest, in this situation (Rogers, 1984: Rogers et al., 1985). It’ is well-documented that altering overall surface charge on an enzyme by ext,ensive chemical modificat,ion can lead to significant changes in pH dependence of catalysis, as for example in t’he serine proteases (Valenzuela & Bender. 197 1). The extensive modifications in those studies can lead to structural changes in the protein and expansion because of the high charge density. The effects of a
Electrostatic elects are thought to play a key role in enzyme catalysis by stabilizing charged transition st)ates (Perutz, 1978; Warshel et al., 1984; LMatthew, 1985). The magnitudes of such effects are, unfortunately, extremely difficult to calculate because of the microheterogeneity of the dielectric constant of a protein. The electrostatic effects of surface charges on the ionization constants of catalytically active groups in enzymes are also interesting from a practical point of view, since modification of charge may provide a general means of tailoring the pH dependence of catalysis using protein engineering (Thomas et al., 1985). The t Symbol ustxl: KM. Michaelis constant. 803
0 1987 Academic Press Inc*. (London) Ltd.
A. J. Russell et al.
804
single substitution are little known. There has been a pioneering study on the effects of selective chemical modification of the surface lysine residues of cytochrome c on its redox potential (Rees, 1980). The consequent results have been criticized by Rogers et al. (1985) because the bulky chemical reagent used for modification could perturb the structure of the protein. The newer technique of site-directed mutagenesis avoids the problem of adding bulky groups to the side-chains of amino acids: a side-chain may be replaced by an isoteric or smaller one. We have begun investigating electrostatic effects in proteins by replacing surface residues of subtilisin from Bacillus amyloliquefaciens (subtilisin Novo or BPN’, EC 3.4.4.16.) by protein engineering (Thomas et al., 1985). The subtilisins are a family of extracellular serine proteases secreted by species of bacilli before sporulation. The crystal structure of the Novo or BPN’ enzyme has been solved (Wright et al., 1969; Drenth & Hol, 1967) and its kinetics wellcharacterized (Philipp & Bender, 1983). His64 acts as a general base during catalysis, accepting a proton from residue Ser221 as it forms a bond with the substrate carbonyl carbon. The enzyme is active only when His64 is unprotonated, at alkaline pH, and catalytic activity varies with the ionization of this residue. The pK, of His64 may be accurately determined by kinetics. The residue we have chosen to modify first is Asp99. This is in an external loop of the enzyme, on the rim of the active site cleft, and is not conserved in subtilisins from other species of bacilli (Table I). In subtilisin Carlsberg (from BuciEZus licheniformis), for example, it is replaced by a serine. The structures of the Novo and Carlsberg enzymes are remarkably similar (McPhalen et al.. 1985a; Bode et al., 1986): a least-squares superof the 274 structurally equivalent position a-carbons of the two enzymes gives a root-meansquare deviation of only 0.53 A (McPhalen et al., 1985a). Further, the conformations of the loop containing either Asp99 or Ser99 are identical. Therefore, a priori, the mutation of Asp + Ser99
should minimize the possibility of any conformational changes that could affect t)he pK, of the active site histidine. We have shown elsewhere t.hat the mutation of Asp + Ser99 causes a shift of about 0.29 unit in the pK, of the enzyme at ionic strengt’h 0.1 M as determined by kinetics (Thomas ef al.. 1985). To show that the mutation of Asp -+ Ser99 is not an isolated result, we have now constructed and analysed a further mutation, Glu -+ Ser 156. which is on the opposite rim of the cleft,. at a similar distance from His64. This residue too is in an external loop, which is known by homology t,o t,olerate a replacement by serine (Table 2). The carboxylate oxygen atoms of Glnl56 are 14 to 15 A from the imidazole nitrogen at,oms of His64 (R. R. Bott, personal communication; M. N. G. tJames: personal communication). Tnterestingly. the environment between the two residues is predominantly aqueous: the carboxylat,e points towards the histidine and there is no protein in a direct line hetween the two. The carboxglate oxygen atoms of Asp99 are 12 to 13 .h from the imidazole nitrogen atoms of His64 (R. R. Kott, personal communication; M. N. G. James, personal communication) but the environment between the two consists largely of protein. Our philosophy for probing enzyme structure and activity by protein engineering is to make only small perturbations in enzyme st,ructure in order to minimize artefacts arising from gross changes in conformation. This approach must also be applied t,o analysing changes in pK, on mutation of surface charge. The pK, values of groups on enz;ymes are sensitive to the overall charge on the protem whicth, itself, changes with pH. It is, therefore, necessa,ry to produce only small perturbations in pK, so that’ the pK, of the wild-type and mutant enzymes may be determined in the same pH range, under identical conditions. We show in this study that WV (*an produce significant, but small, changes in t*he pK, 01 the active site histidine by t)he tlwo mutations and measure them with sufficient precision to allow t,he calculation of electrostatic interact,ions across t)hr active site. These data are used t,o est,ima,tc the dielectric constant, between His64 and the two carboxylate residues.
Table 1 Homology
,$pecies
of subtilisin sequences around aspartate 99
Table 2 Homology
Amino acid sequcncet 90 9.5 100 LYAVKVLGADGSGQYSW LYAVKVLNSSGSGTYSG LYAI KVLNSSGSGTYSA LYAVKVLDSTGSGQYSWI LYAVKVLDSTGSGQYSWI
of subtilisin
sequenws around
glutamate 756 105 I 1 I
t Standard 1 -letter code for amino acids. $ Predicted from the DNA sequence (Wells et al., 1983; Jacobs rt nl., 1985). 3 From Nedov et aI. (1983). 11Stahl & Ferrari (1984). l ’ From Kurihara et al. (19’71).
References as for Table I. t Standard 1-letter code for amino acids. 1 Predicted from the DNA sequence.
The Active Site Cleft of S&&in
2. Materials and Methods (a) Materials Reagents were obtained from Sigma (London). CM-52 ion-exchange resin was obtained from Whatman Ltd. (i) Preparation
and pur@cation
of subtilisin
The gene encoding the serine protease subtilisin BPN from B. amyloliquefaciens was cloned into the vector pUBll0 and expressed in a protease-deficient strain of subtilis (Thomas et al., 1985). Mutation Bacillus of Glu + Ser156 was performed as described for Asp -+ Ser99 the mutagenic primer using 5’-GAAGTGCCT&iGTTACCGG-3’. The entire sequence of each gene was verified by dideoxy sequencing. Wildtype and mutant enzymes were prepared from the supernatant of a culture grown for 36 h in 10 1 of L-broth containing kanamycin (25 pg ml-‘) and glucose (0.2%, w/v). The supernatant was concentrated using a Pellicon membrane filter (PTFCOO05, Millipore) before dialysis buffer 0.01 M-potassium/sodium phosphate against (pH 6.2) and purification on a CM-52 cellulose column as described (Estell et al., 1985; Thomas et al, 1985). All enzymes were purified to electrophoretic homogeneity, and autolytic digestion products were removed before kinetic analysis using a Sephadex G-25 gel filtration column and/or Centricon 10 microconcentrator (Amicon). (b) Methods All experiments were performed at 25°C. The concentration of each enzyme was determined by activesite titration with N-trans.cinnamoyl imidazole (Bender et aZ., 1966). Typical activity was 95% compared with the concentrations measured by A,,, (Eo.lo,o = 1.17; Matsubara et al., 1965). (i) Catalytic constants for the hydrolysis
of succinyl-
L-alanyl-L-alanyl-L-prolyl-L-phenylalanyl
p-nitroanilide (SWA-A-P-Ep-LNA)~ Stock solutions of substrate were prepared in 0.1 MTris. HCl (pH 8.6). The reaction was initiated by the addition of 15 ~1 of enzyme (typically 15 PM in 0.01 Mphosphate buffer, pH 6.2) to 3 ml of buffered substrate (0.01 to 080 mM, in 0.1 M-Tris . HCl (pH 8.6)) in a cuvette maintained at 25”C> in a Gilford 2600 spectrophotometer. Duplicate kinetic runs were performed for at least 10 different substrate concentrations. The increase in absorbance at 412 nm on the release of p-nitroaniline was monitored, and the initial rate calculated from the known extinction coefficient (s4i2 = 8480 M- ’ cm- ‘; DelMar et aZ., 1979). For determinations at pH 6.0, a stock solution of the substrate (69 mM) in DMSO was diluted 106fold to give a roncentration of DMSO of 1% in the cuvette. (ii) Catalytic constants for the hydrolysis of benzoylI,-valyl-L-glycyl-I,-argilzyl p-nitroanilide (bz- V-G-Rp-n;A) determined as for the kc,, were Khl and su-A-A-P-Fp-NA substrate, except that the range of substrate concentrations was 0.01 to 2.0 mM, in 0.1 MTris HCl buffer (pH 8.6), and the enzyme was typically 30 /LM in 0.01 .v-phosphate buffer (pH 6.2).
t Abbreviations used: su-A-A-P-Fp-NA. succinyl-nananyl-rd-alanyl-I,-prolyl-L-phenylalanyl p-nitroanilide: bz-V-G-Rp-NA, benzopl-I,-valyl-L-glycyl-L-arginyl p-nitroanilide; DMSO, dimet,hylsulfoxide.
805
(iii) pH dependence of kccrr/KMfor the hydrolysis of su-A-A-P-Fp-NA and bz- V-G-Rp-NA Phosphate buffers of ionic strength 0.1 M were used and the ionic strength adjusted by the addition of KCl, except for ionic strength 0.025 M and lower, where the stock buffers were diluted. pH was measured before and after addition of the enzyme using a PHM64 research pH meter and electrode (Radiometer, Copenhagen), and the results were discarded if there was significant change during reaction. The meter was calibrated with standard buffer solutions (Radiometer, Copenhagen): 15 ~1 of enzyme (15 pM in 0.01 M-phosphate buffer, pH 6.2) was added to 3 ml of buffered substrate (0.01 mM), the initial substrate concentration being at least 15 times lower than KM. At very low ionic strengths (<0.025 M), the buffering capacity of the phosphate buffer is so low that small changes in pH occur on addition of buffered substrate. Substrate was added for these experiments from a concentrated solution (69 mM) in DMSO to give a final concentration of 0.03% (v/v) solvent. The increase in absorbance at 412 nm was followed to completion spectrophotometrically and found to obey first-order kinetics with high precision. The value of k,,,lK, at each pH was determined from the first order plots exactly as described for the hydrolysis of acetyl-I,-tyrosine p-acetylanilide by chymotrypsin (Fersht & Renard, 1974). High pH values were obtained using both Tris . HCl and bicine-NaOH buffers. Control experiments showed that the rate of reaction with these buffers was lower than in phosphate buffers and, although corrected, these data were not used in the analysis. It was necessary to measure very small changes in absorbance for these determinations. The digital output from the spectrophotometer was fed directly int’o a BBC microcomputer using a data storage program written by Dr R. J. Leatherbarrow of this department. The data were fitted to theoretical curves by non-linear regression on the BBC microcomputer (R. J. Leatherbarrow. unpublished results; Marquardt, 1963). Excellent’ results were obtained for changes as low as 0.03 absorbance unit, which corresponds to a substrate concentration of 3.5 PM, over a time-course of 60 to 300 s.
3. Results (a) Measurement
of pK,
values and controls
a i n of ca tita y ic constants (i)iCZ a cu Et0
(Table
3)
of the synthetic substrates The hydrolysis succinyl-~-alanyl-~-alanyl-~-prolyl-~-phenylalanyl p-nitroanilide or benzoyl-L-valyl-L-glycyl-I,-arginyl p-nitroanilide by subtilisin is described by: E+&E.S.
in
which
2 EA2E+A
the
relevant
(kc,, = k,kd(k2 +kd;
(Scheme
constants
are
1)
defined
41 = K&/(-h + k3)).
For
nitroanilide substrates k, x k, (Philipp & Bender, 1983): the ratio of k, (deacylation) to-i* (acylation) has been report’ed to be 33 (Wells et al., 1986), thus steady-state measurements of k,,, and K, give directly k, and KS, respectively (Fersht, 1985). The activity of wild-type subtilisin towards su-A-A-P-Fp-NA (0.1 M-Tris . HCl, pH 8.6) is similar to that of the mutant (Asp -+ Ser99) under
A. J. Russell et al.
806
Table 3 Catalytic
Enzyme Wild-type
Glu --) Serl56
of p-nitroanilide at 25°C
substrates
by
enzyme
k cat (SC’)
KM
Buffer
PH
(mM)
Tris HCl (0.1 M) Tris . HCl (1% DMSO) K-phosphate (1% DMSO) Tris HCl (0.1 M)
8.6 8.6 6.0 8.6
0~17~0~01 0~09~0~01 0.44 * 0.04 0.32 + 0.05
57 42 11 1.3.5
3.8 x 4.7 x 2.5 x 4.2 x
10’ lo5 lo4 103
bz-V-G-Rp-NA
Tris HCl (0.1 M) Tris HCl (1% DMSO) K-phosphate (1% DMSO) Tris HCl (0.1 M)
8.6 8.6 6.0 8.6
0~13~0~01 0~10f0~01 0.54 + 0.05 0.92 +0.10
4.5 35 14 4.00
3.5 2.9 2.6 4.3
105 1O5 IO4 lo3
SWA-A-P-Fp-NA
Tris HCl (0.1
8.6
0.05 + OXtO
55
1.1 x lo6
Substrate su-A-A-P-Fp-NA bz-V-G-Rp-NA
Asp --t Ser99
constants for the hydrolysis wild-type and mutant
su-A-A-P-Fp-NA
M)
the same conditions: k,,, for the wild-type enzyme is 57 s-l and KM is 0.15 mM, whilst the corresponding values for the mutant enzyme are 45 s- ’ and 0.13 mM. The wild-type enzyme has activity identical with that produced from an independently cloned gene (Wells et al., 1983). The mutant Glu + Ser156 has a similar value for k,,, (MS-‘) but the value of KM is distinctly lower at 0.05 mM. The higher value of KM with Asp + Ser99 allows easier kinetic studies with this mutant, since experiments on the determination of pK, must be performed at [S] << KM. Consequently, Asp + Ser99 was chosen for more extended study. su-A-A-P-Fp-NA is not soluble in buffer at low pH values ( < 8.0), and so was dissolved in DMSO in order to measure values of k,,, and KM at pH 6.0. A control experiment at pH 8.6 on the effect of DMSO added to 1% (v/v) showed that k,,, is lowered by 150/b, whilst KM is unchanged. There is a threefold and mutant increase in KM for both wild-type enzyme on lowering the pH to 6.0. We have made no attempt to analyse the effects of mutation on k,,, and K,, since observed values of KM may be obscured by alternative modes of binding that are known to occur with subtilisin (Poulos et al., 1976). Observed values of k,,, may also reflect the relative occupancy of productive and non-productive modes of binding (Fersht, 1985). More extensive studies on the effects of electrostatic changes on values of KM are being performed elsewhere (J. A. Wells, T. P. Graycar. R. R. Bott & 1). Estell, personal communication). of pK, of the active site (ii) Determination Monitoring the pH dependence of k,,,lK, gives the ionization constants of free enzyme and free substrate (Fersht, 1985). Thus, pK, values determined from such plots should be independent’ of substrate in regions where the substrate does not ionize. The pH range investigated was restricted to the region around neutrality. This minimizes perturbations caused by the ionization of other residues, e.g. Asp and Glu at, low pH values and Lys: Tyr. Cys and, eventually, Arg at high pH values. The region of pH 6 to 8 spans nicely the expected pK,, about 7. The measured values of
(s-1 6’)
x x x x
katlK, could be fitted with high precision t.o theoretical ionization curves of a single base. The value of kca,/KM at each pH value was determined from measurements of the release of products at substrate concentrations much lower than KM as described by Fersht & Renard (1974). Under t’hese conditions, first-order kinetics should be obeyed, which are independent of initial substrat’e concentration. thus avoiding one source of error. Reactions were monitored for at least ten half-lives and excellent> first-order plots were obtained. The value of KM for t)hr hydrolysis of su-A-A-P-FpNA by wild-type enzyme is 0.44 T?IM Rt pH 6 and 0.15 mM at pH 8.6. The initial substrate concentrat~ion of O*Ol mM is at least, 15 times lower than K, at all values of pH. Further, since the firstorder plots were analysed over several half-lives, even lower concentrations of subst,rates were actually used for the determination of k,,,/K,. ‘l’he low concentration of substrate used in this method also avoids product inhibition. Experimental corlditions are summarized in Table 4. Measurement’s of wild-type and mutant enzymes were performed in parallel. Any errors in standard solut*ions for the calibration of the pH meter should thus cancel out, when taking the difference between wild-type and mutant. The reproducibility of pK, values is higher than that for kinetic constant,s, since only the relative values of k,,,/K, at various values of pH are required for the calculation and not the absolute values. (iii) pH dependerxe su-A-A-P-Fp-NA
of hydrolysis
(Table
of
5, Fig. 7)
The only ionizing group on the substrate is the succinyl group and its pK, is below the range of this study. At ionic strength 0.1 M, His64 in the wildtype enzyme has a pK, of 7*17( + 0.02); this is decreased to a value of 6.88( kO.02) in the Asp -+ Ser99 mutant, a shift of 0.29( &O-03) unit (Fig. 1). k,lKM is decreased by 2094 on mutation. The pK, of the wild-type enzyme is in excellent agreement with that derived from the hydrolysis of ester substrates under identical conditions and at a wide range of enzyme concentrations (Philipp et al.. 1979). Lowering the concentration of mutant
The Active
Site Cleft of Subtilisin
807
Table 4 Experimental Enzyme
Expt no.
Wild-type
conditions
No. of pH points
x0. of kinetic runs
6.1-8.0 6.1-8.0 6.1-8.0 5.5-8.9 6.1-8.2 6.0-7.9 6.0-7.9 6.1-7.8 5.5-8.9 5.5-7.8
18 18 18 25 16 15 15 20 25 15
54 54 54 75 32 30 30 40 75 30
0@05 0.01 0.025 0.10 0.10 0.10 0.10 0.50 1.00 l@O
6.1-8.00 6.1-8.0 6.1-8.0 5.5-8.9 6.1-8.2 6.0-7.9 567.9 6.1-7.8 5.5-8.9 5.5-7.8
18 18 18 25 16 15 13 20 25 15
54 54 54 75 32 30 26 40 75 30
0@05 0.01 0.02.5 0.10 0.10 0.10 0.10 0.50 1X@ lxm
6.1-8.0 6.1-8.0 6.1-8.0 5.6-8.0 6.0-8.0
18 18 18 18 18
108 54 54 54 54
o+M15 0.01 0.025 0.10 1.00
1t 3t 4 5
61 7 8 9
10 11t 12t 1st
14 15
161 17§ 18 19 20 (Glu + Ser156
determinations
pH range
2t
Asp -+ Ser99
for pK,
21t w w
24 25
Ionic strength (M)
All experiments were done at 25°C; initial substrate concn,
rather than a
Table 5 The pH dependence of hydrolysis by wild-type succinyl-~-alanyl-~-alanyl-~-prolyl-~-phenylalanyl
Enzyme Ionic strength 0.005 .nt \I’ild-type .4sp + Her99 (ilu + Her156 Ionic strrngth 0.01 .vt \Vild-type Asp + Ser99 (:lu + Serl56 Ionic
and mutant
subtilisin
of
p-nitroanilide Limiting value of k,,,/K, at high pH
Expt no.
pK, of active site from kinetics (*s E.)
1 11 21
6.99+0.01 6.61k 0.01 6.67f0.03
4.03 x lo5 4.02 x 10’ 9.21 x lo5
2 12 22
7.08 * 0.02 6.66 + 0.02 6.66 * 0.03
4.06 x lo5 4.01 x 105 9.28 x lo5
3 13 23
7.0410.01 6.68 f 0.02 6.63kO.02
4.90 x 105 4.04 x 105 9.28 x lo5
4 14 24
7.17f0.02 6~88+0~02 6.92 kO.03
4.89 x IO5 4.08 x 1OS 9.09 x lo5
8 18
7.10f0.02 7.00 + 0.02
4.66 x lo5 4.08 x IO5
9 19 25
7.11 kO.02 7.09 f 0.02 7.17kO.03
4.48 x lo5 4.16 x lo5 1.64 x lo6
? DMSO present at a concn of 0.03%
(S-l
M-‘)
808
A. J. Russell et al.
5
6
7
0
9
IO
PH
Figure 1. pH dependence of Ic,,,/K, for the hydrolysis of su-A-A-P-Fp-NA by wild-type subtilisin (0) and the mutant Asp -+ Ser99 (0) at ionic strength 0.1 M. Data are normalized to loo:/, for the limiting value at high pH for each mut,ant.
enzyme Asp -+ Ser99 tenfold caused no significant change in its measured pK, (6.81( +0.04)). Aggregation therefore does not appear to he obscuring the results for either enzyme. (iv) $
rj;ep;yence a e
of hydrolysis
of bz-V-G-Rp-l\r,4
The pH dependence of kca,/KM should be independent of substrate in regions where the substrate does not ionize and so a second substrate was used to check that the pK, difference between wild-type and Asp + Ser99 is not an artefact of using su-A-A-P-Fp-NA. The pK, at ionic strength 0.1 M of wild-type subtilisin determined from the second substrate is 7.15(&0=02), whilst that of Asp -+ Ser99 is 6.92( +0.02); a pK, shift of 0.23( kO.03) pH unit, that is a 1.7.fold shift in K,. This is in good agreement with the results from the experiment using su-A-A-P-Fp-EA. (v) Effect of ionic strength on pK,
(Tables
intrinsic value in t.he absence of surface charge. Accordingly, we found from measurements on the hydrolysis of su-A-A-P-Fp-NA that,. at ionic strength 1.0 M, the pK, values of the active sit,es of wild-type and both mutant enzymes converge. At an intermediat,e ionic strength of 0.5 M, t,here is a shift in pK, of 0.1 unit between wild-type and Asp -+ Ser99. At the lowest ionic strengths, 0.005 t’o 0,025 M, the difference in pK, increases to about 0.4 unit,. The effect of high ionic strength on the pH dependence of t*hr second su bstrat,e. bz-V-G-Rp-NA, was identical within experimental error. The value of k,,,/K, for hydrolysis of bz-V-G-Rp-NA by subtilisin is much lower fhr t,his substrat’e and caused slightlv more error berause of longer time-courses (error -in f)K, = +043). ‘I%~ values for pK, quoted in Table 6 are the average values of both substrates and other variables. The Linderstrclm -Lang equation (set, IGdsali CI Wyman, 1958) predicts that the pK, of a group on will change with ionic strength an enzyme according to the net surface chargcx. Below thr isoelect’ric point. the pli, will incre,se and above it will decrease the magnitude of the change drprntling on net surface charge. The very small incsrrase in pK, of wild-type enzyme, 6.99 at ionic strength O-005 M t,o 7.13 at ionic strength 1.0 M, indic:iitrs ii small net positive charge on the enzyme, and is consistent, with the known isoelectric point of 7.X for subtilisin BPN’ (Matsubara et al.. 1965).
A further possible artefact is that c~alcium ions may bind to t,he enzyme and affect, the results (subtilisin is known to have calcium binding sites but Asp99 and Glu156 are not, implicated in these). Care was taken throughout t,he growing and harvesting of (aells and the subsequent steps t,o minimize the presence of calcium ions. Cont,rol experiments showed t,hat t#he presence of the calcium-chelating agent) EGTA has no effect on the pK, values of wild-type or mutant, enzymes at ionic strength 0.1 >I.
5 and 6)
The presence of high concentrations of ions should mask electrostatic interactions. At high ionic strength, the observed pK, should tend t’o its
(vii) Hill
plots of titration
data
A good indication of the quality of the fit’ of data to theoretical titration curves is the equiva,lent of
Table 6 The pH dependence of hydrolysis of N-benzoyl-~~-~~alyl-~~-gly~yl-~,-arg~~ni~~e p-r&roan&de by wild-type and m,utnnt subtilisin
Enzyme
All experiments
Expt no.
ph’, of active site from kin&w (&SE.)
were done at %5”(‘; buffered substrate wnm,
042 mM.
Limiting vitlue of Icca,/KM at high pH (s-l arc’)
The Active
809
Site Cleft of Subtilisin
Table 7 Effect of EGTA on pH dependence of hydrolysis by a wild-type and mutant subtilisin at 0.1 M ionic strength Enzyme
Expt no.
LVild-t,ype Asp -+ SW99
pK, of active site (from kinetics)
6 16
No EGTA
0.1 mM-EGTA. 0.03% DMSO
7~17+0.02 6+48+0~02
7.18f0.01 692f0.02
Experiments were done at 25°C. with 0.01 rnM buffered potassium phosphate buffer)
the Hill plot of 1% {I (kcatl&)max - (k,,,/li’M)ob,l/[(kcat/KM)obsl) against pH, where (k,,,/K,),,, is the limiting maximum value of k,,,/K, at high pH values and (katlKdo,s is the value of kca,/KM at the particular pH value. For a single ionization, this should be a straight line of slope h = - 1. This is found with high precision over all ionic strengths (Table 8, Fig. 2). The fit to such plots is generally h = - l.OO( kO.02).
(b) Estimation
of eSfective dielectric constants
(i) Calculation of dielectric constant, D Substitution of the known values of the fundamental physical constants into the standard formula for electrostatic interactions gives D = 332z,z,lrAC, where z1 and z2 are the two charges, measured in units of the charge of an electron, separated by a distance of r a and which wit,h an energy of AG kcal/mol interact ( 1 kcal = 4.184 kJ). Alternatively, for a measured The shift in pK, of ApK,, D = 244z,z2/rApK,. distances between the necessary interatomic carboxylate oxygen atoms of Asp99 or Glu156 and the 6-N and E-N atoms of the imidazole ring of His64 are known with some precision: unpublished high-resolution co-ordinates (1.8 A) from R. R. Bott (personal communication) on subtilisin Novo and similar data (2.1 A resolution) from McPhalen et aZ.
so-A-A-P-Fp-NA
(0.01 M ionic stren@h
(1985b; M. Iv. G. James, personal communication) on the complex with the barley (‘I-2 inhibitor give mean 0 to N distances that agree within 0.3 to 0.5 8. A value of 12.4 A was used for Asp99 carboxylate oxygen atoms to His64 imidazole nitrogen atoms, and 14.5 A for Glu156 to His64. Dielectric constant varies with ionic strength because of shielding by the ionic atmosphere (Table 9). The observed value at an ionic strength greater than zero is usually termed the effective dielectric constant (Hill, 1956). Denoting t$his by D,, and t’he value at zero ionic strength as D,, then, for simple point charges in solution separated by distance r: De, = D, exp (Kr),
where IC is the conventional term in the DebyeHiickel equation, i.e. the reciprocal of the Debye length. Since ICis proportional to a. where Z is the ionic strength: In D,, = In D, + afi,
(2)
where c( is a constant. Hill (1956) has analysed the effective dielectric constant between two charges embedded in idealized proteins consisting of waterimpenetrable spheres, but there is no simple theory to account for charges located on the surface of real proteins. The correct equation for the description of the effect of ionic strength will certainly be more than complex equation (2). Nevertheless, equation (2) fits the experimental da.ta (Fig. 3) and was used to extrapolate D,, to zero ionic strength
Table 8 Gradient
(h) of Hill plots for pK, -h
Substrate WA-4.P-Fp-NA
bz-V-GKp-NA
Ionic strength (M) 0.005 0.01 04)25 0.10 0.50 1.00 0.10 1.00
(1)
Wild-type 140+0~01 1~00~041 1.00 & 0.0% 0.99 * 0.02 1~00~0~01 1.03 + 0.02 140~0~02 0.99 f 042
Asp --t SW99
Glu - Ser156
0.99 * 092 1.01*0@2 0.99 f 0.03 140,042 0.99 + 0.01 1.02 f 042 l~oo+o~Ol 0.99fOa2
1.03 * 0.03 0.99f04)l 0.98 & 0.03 0.99 + 043 1.02 + 0.02 140 * 0.02
The Hill plot is log [(k,,,/KM),,,(k,,,/K~),,,]/[(k,,,/K,),,,1 wrsus pH, where (k,,,/KM),,, limiting value at, high pH and (kca,/KM)obr is the value at the particular pH.
is the
A. J. Russell et al.
810
'6
7
'6
8
7 PH
PH
1=0.025~
'6
I=
o-5
I=
M
-I
'6
7
7 PH
PH
8
PH
6
7
l.OOM
0
PH
Figure 2. Hill plots (log [(k,,,/K,),,,(k,,,/h’,),,,]/[(k,,,/K,),,,] versus pH, where (k,,,/K,),,, is the limiting value at is the value at the particular pH) for the hydrolysis of su-A-A-P-Fp-NA by wild-type and high PH and (kJ&)Obs mutant Asp -+ 8er99 subtilisin. For ionic strengths 0405, 04lO1, 0.025, 0.1 and 1.0 M, the average values of the data for each pH value are plotted; for ionic strength 0.5 M, the individual values (2 for each pH value) are shown to illustrate t,he spread. (Fig. 3). The extrapolation is, in any case, small were made down to measurements since 1 = O-005 M. Figure 3 also nicely illustrates the effect of ionic atmosphere on the effective dielectric constant. The effective dielectric constant between
Asp99 and His64 extrapolated to I = 0 is 42( +4). The lowest observed values are 47 t,o 55 and are found in the range I = O-005 t,o 0.025 M. The observed values between Glu156 and His64 are between 40 and 50 in this range.
The Active
811
Site Cleft of Subtilisin
Table 9 Effective dielectric constant (D,,,) at varying ionic strength Ionic strength (M)
PK Wild-type
Asp + Ser99
04)05t 0.01t
6.99 + 0.01
6.61 kO.01
0.38+0+2
0425t
7.08k0.02 7.04 k 0.01
6.66kO.02
0.42f0.03
0.10
7.17f0.021
6.68 + 0.02 690+0.02$
0.36 f 0.03 0.27kO.03
0.50
7.1OkO.02 7.13kO.02:
7GO~O.02 7.08+0.02$
O.lOf0.03 0.05+0.03
I .oo
LVild-type 0~005t 0.01t
o-025t
0.10
I7eff
APK,
52+3 47+3 55f5 73k8 200+60 400&-240
Glu + Ser156
6.99 k 0.01
7.08 k 0.02 7.04+0.01 7.17 + 0+2$
t DMSO conm. O.O3”i,. data $ Averaged for bz-V-G-Rp-NA.
3’
6.67 k 0.03
0.32 kO.04
6.66 + 0.02 6.63 k 0.01
0.42 f 0.03
52k7 40*3
6.92 + 0.03
0.41 f0.02 0.25 + 0.04
4152 67k9
both
su-A-A-P-Fp-NA
I 0.1
I 0.2
I 0.3
1 0.4
I 0.5
I 0.6
I 0.7
I 0.0
rh
and
Figure 3. The effect of ionic strength on effective dielectric constant between Asp99 and His64. The experimental results are plotted according to eqn (2), although it is expected that a more complex equation will be required to fit’ data of higher accuracy (see t,he text for details).
4. Discussion (a) Precision
of kinetic determination pK, values
of
Kinetics provides a very sensitive and convenient, means of determining the pK, values of active sites. Nuclear magnetic resonance spectroscopy is, of course, a far more general procedure for determining the pK, values of groups in proteins, especially histidine residues. But, for subtilisin, kinetics has two significant advantages. First, the activity of the enzyme tends to zero at low pH values and so the rate data fit an ionization equation of just two unknowns; the pK, and limiting value of rate constant at high pH. Nuclear magnetic resonance spectroscopy requires estimating three unknowns; the pK, and the limiting values of the chemical shift at both high and low pH values. Second, the experiments are performed at low concentration of enzyme and so autolysis is minimized. We find that the pH dependence of k,,,lK, between pH 6 and 8 fits with high precision the expected equation for the ionization of a single base, both to direct titration curves and derivative plots such as the Hill plot (Figs 1 and 2). Provided there is no change of ratedetermining step with pH, the pH dependence of kcaJKM gives the pK, of the unligated enzyme or the free substrate. The substrates employed here do not ionize in the region of the measurements and so t,he observed pK, values should be those of the enzyme. The reproducibility of the data is indicated from three different measurements at ionic strength enzyme: 7.15( -&0.02), 0.1 M of the pK, of wild-type ‘i.l’i(f0.02) and 7.18( kO.01). These values were determined for the positively charged substrate bz-V-G-Ap-NA and the negatively charged substrate su-A-A-P-Fp-NA in the absence and
presence of EGTA, respectively. (A wide range of ester substrates has given a value of 7.15 for the pK, of wild-type enzyme at ionic strength 0.1 M (Philipp et al., 1979).) The corresponding values for mutant Asp -+ Ser99 are: 6.92( + 0.02)) the 6.88( kO.02) and 6.91( kO.02). The changes in pK, respond to changes in ionic strength in the predicted manner: the differences are suppressed at high ionic strength and gradually increase with decreasing ionic strength. The perturbation of pK, on the mutation Asp + Ser99 is not a unique event. The mutation Glu --+ Ser156 also causes a similar change in pK,. We are therefore confident that we are measuring real and statistically significant effects on the pK, of the active site of the enzyme. (b) Magnitude
of pK,
shift
The results of previous experiments using extensive chemical modification predicted that we would find negligible changes in pK, on the modification of single charges remote from the active site histidine. Valenzuela & Bender (1971) reported a change of 1-O unit in the pK, of chymotrypsin (His57) on altering the overall surface charge by 28 units (i.e. changing 14 positively charged lysine residues to 14 negatively charged succinate half-amides by succinylation with succinic anhydride). A slightly lower effect was noted for the conversion of 13 surface carboxylates to positively charged residues by coupling with ethylenediamine. Acetylation of all surface lysine residues of trypsin lowers the pK, of its His57 by only 0.2 unit (Spomer & Wootton, 1971). Such drastic changes in surface charge could well alter the properties of the enzyme and/or surrounding solvent. Reorientation of solvent and surrounding
812
A. J. Russell et al.
counterions or an expansion of the protein caused by electrostatic repulsion of surface residues could affect’ the pK, of the active site by structural changes in addition to direct electrostatic interactions. (c) Dielectric constant across the active site cleft It is debatable whether the term dielectric constant,, which represents a macroscopic phenomenon, should be applied to proteins which are very heterogeneous in their dielectric properties. Nevertheless, it is important to know or calculate the transmission of electrostatic effects in proteins and hence know values of effective dielectric constants. The measured dielectric constant’ between Asp99 and His64 tends to about, 42 at zero ionic strength. Values of 47 to 55 are observed in the range of ionic strengths of 0.005 to 0.025 M. The measured dielectric constant between Glu156 and His64 is in the range 40 to 50 in this range of ionic strength. Rogers et aE. (1985) report a value of 27 between the iron of cytochrome cssl and a propionate 8.2 L%distant, measured from the change in redox potential on ionization of the propionate. But, both ions are buried in the protein by at least 5 A in this example. A change of ionic strength from 0.007 to 0.1 M also does not effect the change in redox potential on ionization of the propionate. The crystal structure of subtilisin shows predominantly protein between Asp99 and His64. There is no protein whatsoever in a direct line between Glu156 and His64. The observed values of the dielectric constant. are about half the value of 78.5 expected for water. There could be several reasons for this. First, there is evidence that ions may lower the dielectric constant of solvent water by a combination of solvent orientation and inductive polarization (e.g. see Mehler & Eichele. 1984). A tabulation of values by Conway et al. (1951) gives a value of dielectric constant of 65 at 12.4 A from a singly charged ion and 72 at 14.6 8. Second, some of the electrostatic field of the ion will be transmitted via the low dielectric constant of the protein. Third, the water molecules in the active site may differ from bulk water because of the orienting effects of the cavity and its amino acid side-chains. The observed increase in effective dielectric constant with increasing ionic strength is in accord with the classical Debye-Hiickel theory (Hill, 1956). It has been observed only rarely that ionic solutions have greater dielectric constants than do the pure solvents (Pethig, 1979). (d) Apparent dielectric constant The term effective dielectric constant is generally used to denote the observed values of dielectric constant when the dielectric is inhomogeneous. We suggest the use of the term apparent dielectric constant to describe the observed values of dielectric constant that are measured by site-
directed mutagenesis since, as discussed next. the observed value of the dielectric constant may have additional components from changes in wat,er structure. Mut’ation of residues that int’eract directly with a substrate may have complications caused by local changes in protein conformation that affect other interactions. Small local changes in conformabion are less likely to cause problems in experiments such as those described here, where a remote surface residue is mutated and long-range electrostatic interactions are monitored. But there could be effects transmit.ted by solvent’. The static dielectric constant of a polar liquid has contributions from two components: polarization bv induction and orientation of dipoles. For water. orientation of dipoles is by far the overwhelming contribution, accounting for about, 98Y; of the observed dielectric constant. (The cont~ribution from polariza,tion is calculated to he 1.79 from t)hfl refractive index, n, using Maxwell’s equation. U = n’.) hlutation of a charged side-chain of an amino acid may result in changes in the structure of water around the protein and. hence, the orientation of wat,er. Any change in dielet:t,ric consta,nt brought) about by reorientat)ion of solvent by t,he presence of an ion is, of course, an inhrrent component of t’he effect’ive dielectric constant of the medium surrounding that charge. In the abxencc of other charges affecting the pK, value of the a#ctivc site, the measured value of the dielectric constant would be identical with t,he classical effective dielectric constant. But if there are ot,her cha.rges present. t,hey could be affected indirectIT 1)~ changes in solvent structure. Mutat,ion of one residue could alter the effective dielectric constant between other charged residues and the ionizing group at the active site and, consequently. influence the magnitudes of other electrostatic interact’ions. These changes will be reflected in the measured effective dielectric constant. Because of this, we suggest the use of the’ term apparent dielectric constant, to acknowledge that bhe measured value of t,he dielectric constant contains additional components. Other interactions measured by sit~edirected tnutagenesis have component,s from changes in water stru&ure (e.g. hydrogen honds. Fersht et al., 1985) in addition to changes in protein conformation, and the term apparent is generally useful, since it describes experimentally observable results (Wells &, Fersht, 1986). We art’ condu&ng an extensive survey of the effects of mutation of surface charges on the ph’, of histidine residues in subtilisin t,o t,est, for effects of water strut:t.ure anti also to map the microhet,erogeneit’y of dielectric constant within other regions. Wr thank Dr M. N. (2. tJames for unpublished CC)ordinates of the complex between subtilisin Novo and t,hr barley CT-2 inhibitor and Dr R. Bott for measurements of interatomic distances in subtilisin Xovo. We thank also Drs D. Estell, ,J. Wells a.nd T. Graycar for exchange of’ unpublished information.
The Active Site Cleft of Subtilisin
References Bender, M. I,.. Begoue-Canton, M. L., Blakeley, R. L., Brubacher, I,. ,J., Feder, J., Gunter, C. R., Kezdy, F. J., Killheffer, J. V., Marshall, T. H., Miller, C. J.. Roeske, R. W. & Stoops, J. K. (1966). J. Amer. Chem,. Sot. 88. 5890-5914. Rode. W., Papamokos, E., Musil. D.. Seemueller, C. 8r Fritz, H. (1986). EMBO J. 5, 813-818. (lonway. B. E.. Bockris, J. O’M. & Ammar, I. (1951). Trans.
Faraday
Sot. 47. 756-766.
DrlMar, E. G.. Largman, C., Brodrick, J. W. & Goekas, M. C. (1979). Anal. Biochem. 99, 316-320. Drenth, ,J. & Hol. W. G. ,J. (1967). J. Mol. Biol. 28, 543551. Chemistry. Edsall, J. T. & Wyman, ?J. (1958). Biophysical p. 510. Academic Press. Xew York. Estell. I). A., Graycar. T. P. & Wells, J. A. (1985). J. Biol. (‘hem. 260, 6518-6521. i-\. R. ( 1985). In Enzyme Structure and Fersht. Mechanism, 2nd edit., chapt. 5, W. H. Freeman. Xew York and Oxford. Fersht. A. R. & Renard, M. (1974). Biochemistry, 13. 1416-1426. Fersht. A. R,.. Shi, J. I’.. Knill-Jones, J., Lowe, D. M.. Wilkinson. A. J.. Blow. D. M., Brick, I’., Carter. P.. Waye. M, M. Y. & Winter. G. (1985). Nature (London), 3 14, 235-238. Hill, T. L. (1956). ,J. Phys. Ch,em. 60, 253-255. *Jacobs. M.. Eliasson. M.. Uhlen, M. & Flock. J. (1985). LVucl. Acids ties. 13. 8913-8927. Kurihara. M.. Markland, F. & Smith. E. (1972). .I. BioZ. (‘hem,. 247, 5619-5631. Maryuardt. 1). W. (1963). J. Sot. Jndust. Appl. Math. 11, 431-441. Matsubara, H.. Kasper. C. B.. Brown, D. M. & Smith. E. L. (1965). J. Biol. Chem. 240, 1125-1130. Matthew, J. H. (1985). Annu. Rev. Biophys. Chem. 14, 387-417. McPhalen. (‘. ,4.. Schnebli. H. P. & James. M. I$, G. (19850). FEBS Letters, 188, 55-58.
813
McPhalen, C. A., Svendsen, I., Jonassen, I. & James, M. N. G. (1985b). Proc. Nat. Acad. Ski.. V.S. A. 82, 7242-7246. 23. Mehler, E. L. & Eichele, G. (1984). Biochemistry, 3887-3891. Xedov, P., Oberthur, W. & Braunitzer, G. (1983). HoppeSeyler’s 2. Physiol. Chem. 364, 1537-1540. Perutz, M. (1978). Science, 201, 1187-1191. Pethig, R. (1979). In Dielectric and Electronic Properties qf Biological Materials, p. 137, Wiley, Chichester. Philipp, M. & Bender, M. L. (1983). Mol. (‘ell. Biochem. 51. 5-32. Philipp. 11,. Tsai. I. H. & Bender. M. L. (1979). Biochemistry, 18, 3769-3773. I’oulos, T. I,., Alden, R. A., Freer. S. T., Birktoff, J. J. $ Kraut, J. (1976). J. Biol. Chem. 251. 1097-1103. Rees, D. C. (1980). J. Mol. Biol. 173. 323.-326. Rogers. h‘. K. (1984). D.Phil. thesis, Oxford University. Rogers. h’. K., Moore, G. R. & Sternberg, M. ,J. E. (1985). J. Mol. Biol. 182, 613-616. Spomer, W. E. & Wootton, ,J. F. (1971). Biochim. Biophys. Acta, 235, 164-171. Stahl. M. L. & Ferrari. E. (1984). .I. Bacterial. 158, 411L 418. Thomas, P. G., Russell. A. J. CGFersht. A. R. (1985). Xatuw (London), 318, 375-376. Valenzuela. P. & Bender, M. T,. (1971). Biochim. Hiophys. Acta, 250, 538-548.
Warshel, A., Russell, S. T. & Churg; A. K. (1984). I’roc. Nat. Acad. Ski., U.S.A. 81. 4785-4789. Wells. *J. A., Ferrari, E., Henner. D. J.. Estell. D. A. & Chen. E. Y. (1983). Nucl. Acids Res. 11. 7911-7925. Wells. J. A., Cunningham, B. C.. Graycar, T. I’. & Estell. I). A. (I 986). Phil. Tmnx Roy. Sot. ser. A. 317. 415.. 423. Wells. T. ?rj. C. 8: Fersht. A. R. (1986). Biochemistry, 25, 1881-1886. Wright, C. S.. Alden, R. A. & Kraut. ,J. (1969). Nature (Lon,don) , 221, 235-242.
Edited by G. A. Gilbert