Biochemistry 1987, 26, 2674-2682
2674
Art ides
Effect of Single Amino Acid Replacements on the Folding and Stability of Dihydrofolate Reductase from Escherichia colit Kathleen M. Perry,',$ James J. Onuffer,' Nancy A. Touchette,t.ll Cinda S. Herndon,*J Mitchell S . Gittelman,* C . Robert Matthews,*.t Jin-Tan Chen,' Ruth J. Mayer,*v# Kazunari Taira,' Stephen J. Benkovic,* Elizabeth E. Howell,A and Joseph Krautv Department of Chemistry, The Pennsylvania State University, University Park, Pennsylvania 16802, The Agouron Institute, La Jolla, California 92037, and Department of Chemistry, University of California, San Diego, La Jolla, California 92093 Received September 5, 1986; Revised Manuscript Received December 30, 1986
ABSTRACT: The role of the secondary structure in the folding mechanism of dihydrofolate reductase from
Escherichia coli was probed by studying the effects of amino acid replacements in two a helices and two strands of the central p sheet on the folding and stability. The effects on stability could be qualitatively understood in terms of the X-ray structure for the wild-type protein by invoking electrostatic, hydrophobic, or hydrogen-bonding interactions. Kinetic studies focused on the two slow reactions that are thought to reflect the unfolding/refolding of two stable native conformers to/from their respective folding intermediates [Touchette, N. A., Perry, K. M., & Matthews, C. R. (1986) Biochemistry 25, 5445-54521. Replacements a t three different positions in helix aB selectively alter the relaxation time for unfolding while a single replacement in helix aC selectively alters the relaxation time for refolding. This behavior is characteristic of mutations that change the stability of the protein but do not affect the rate-limiting step. In striking contrast, replacements in strands PF and PG can affect both unfolding and refolding relaxation times. This behavior shows that these mutations alter the rate-limiting step in these native-to-intermediate folding reactions. It is proposed that the intermediates have an incorrectly formed /3 sheet whose maturation to the structure found in the native conformation is one of the slow steps in folding.
A
current working model for the mechanism of protein folding, the framework model, proposes that elementary units of hydrogen-bonded secondary structure form early in the process and direct the rapid and efficient folding to the native conformation (Kim 8i Baldwin, 1982). Studies on the kinetics of formation of secondary structure by circular dichroism (Kuwajima, 1977; Labhardt, 1984) are consistent with this model; however, detailed structural information of the sort provided by nuclear magnetic resonance or X-ray spectroscopy has not been obtained because of the short lifetimes and low populations expected for these species. Recent studies on the a subunit of tryptophan synthase from Escherichia coli (Beasty et al., 1986) suggest that site-directed mutagenesis can be used to test the role of the secondary structure in directing folding. In these studies, it was found This work was supported by National Science Foundation Grant PCM-81-04495 (C.R.M.), U S . Public Health Service Grant GM 24129 (S.J.B.), and Office of Naval Research Contract N00014-85-K-0663 (J.K.). C.R.M. is the recipient of Research Career Development Award KO4 AGO0153 from the U S . Public Health Service, 1981-1986. C.S.H. was the recipient of an American Cancer Society Postdoctoral Fellowship, 1984-1985, and E.E.H. is the recipient of U S . Public Health Service Fellowship F32-GM 09375. *The Pennsylvania State University. 8 Present address: Department of Biochemistry and Biophysics, UCSF Medical Center, San Francisco, CA 94143. 11 Present address: Department of Physiology, The Johns Hopkins University, Baltimore, MD 21205. Present address: Technical Diversification Department, Air Products and Chemicals, Inc., Allentown, PA 18105. +Y Present address: Smith, Kline & Beckman Laboratories, Philadelphia, PA 19101. ATheAgouron Institute. University of California, San Diego.
that mutations can alter the stability of the protein, the rate of a specific folding reaction, or both. Unfortunately, the lack of a high-resolution crystal structure for the a subunit prevents the association of these effects with recognized elements of secondary structure. Dihydrofolate reductase (DHFR,' EC 1.5.1.3) from E . coli is an excellent vehicle for such a study. The refined crystal structure at 1.7 A (Bolin et al., 1982) shows that reductase has an a / @supersecondarystructure: seven parallel /3 strands and one antiparallel strand form a hydrophobic @ sheet which acts as the structural core of the protein and four amphiphatic a helices which bind to the sheet and assist in its solubility. A kinetic folding model for DHFR has recently been proposed (Touchette et al., 1986) that involves a series of native, intermediate, and unfolded forms. Basically, this kinetic model proposes that the unfolded form collapses rapidly to an intermediate which is then slowly processed to the native form. The unfolding reaction is completely controlled by the slow native-to-intermediate step. Using a formalism described elsewhere (Beasty et al., 1986), we can use measurements of the rates of unfolding and refolding for the native-to-intermediate reaction to determine if a particular mutation alters the equilibrium and/or kinetic properties. We report the results of studies on single amino acid replacements at three positions in one a helix, one mutation in a second helix, and two different positions in the 0 sheet of DHFR. These results indicate that the effects observed depend upon the nature of the secondary structure in which the re-
'
Abbreviations: DHFR, dihydrofolatereductase: NaDodSO,, sodium dodecyl sulfate; K,EDTA, ethylenediaminetetraacetic acid dipotassium salt.
0006-2960187 10426-2674%01.50/0 0 1987 American Chemical Society
MUTATIONS AND FOLDING OF DIHYDROFOLATE REDUCTASE
placement is made and suggest that formation of the 0sheet found in the native conformation is a key slow step in folding.
EXPERIMENTAL PROCEDURES Protein Source. Wild-type dihydrofolate reductase (DHFR) was isolated from Escherichia coli strain MC294 containing plasmid pTY 1. This plasmid was constructed by isolation of the two Hue111 fragments from pCV29 (574 and 408 base pairs in length) containing the DHFR gene (Smith & Calvo, 1980). These two fragments were blunt-end ligated and subcloned into the BamHI site of a pBR322 derivative lacking the EcoRI site (a gift from D. Shortle). The construction of the Asp-27 Asn mutant gene and the purification of the protein have been described (Villafranca et al., 1983). The Leu-28 Arg protein was isolated from plasmid pG93. This plasmid appeared spontaneously after exposure of cells containing this plasmid to high levels of trimethoprim (T. Yaegashi and J.-T. Chen, unpublished results). DNA sequencing showed a T G replacement at the second base in the codon corresponding to amino acid position 28 that accounted for the amino acid change. The Phe-3 1 Val mutant was isolated from plasmid pCK3 14 whose construction is described elsewhere (Taira et al., 1986). The Arg-44 Leu mutant was isolated from plasmid pCHl which was constructed by oligonucleotide mutagenesis (Dalbadie-McFarland et al., 1982). Plasmid pTYl was partially digested with EcoRI endonuclease in the presence of ethidium bromide. Singlestrand nicking at the unique EcoRI site in this plasmid was followed by exonuclease I11 digestion to expose single-stranded DNA in the region of interest. The oligonucleotide, a 23-mer containing a G to T change at the first base in the codon corresponding to amino acid position 44 to obtain the desired mutation and a C to G change at the third base in the codon corresponding to amino acid position 43 to remove a Hue111 restriction enzyme site and thus facilitate the isolation of the mutant, was synthesized on an Applied Biosystems DNA synthesizer, Model 380A. The oligonucleotide, annealed to the single-stranded DNA, served as a primer for DNA polymerase. Ligation was done with DNA ligase. Following transformation into E. coli strain MC294, the mutant was isolated by colony hybridization. The Maxam-Gilbert method (Maxam & Gilbert, 1977) was used to determine the DNA sequence and confirm the mutation. The construction of the Thr-113 Val mutation is described elsewhere (Chen et al., 1985). The Glu-139 Lys mutant was isolated from plasmid pKP1. This plasmid was constructed by nicking pTYl with EcoRI endonuclease as described above. Because the mutation was made at the EcoRI site, this region was exposed by digesting with both exonuclease I11 and DNA polymerase I. The oligonucleotide, an 18-mer with a G to A change at the first base in the codon corresponding to residue 139, was annealed to the single-stranded DNA; the gaps were filled with the Klenow fragment of DNA polymerase I and closed with DNA ligase. This hybrid plasmid was used to transform strain MC294, purified, digested with EcoRI endonuclease, and used in a second transformation. Because the desired base change removes the EcoRI site, only mutant plasmids survived the EcoRI digestion as supercoiled circles. Individual colonies were selected from the second transformation and examined for the loss of the EcoRI site; approximately 60% of the colonies satisfied this criterion. DNA sequencing confirmed the desired base change. Protein Purification. With the exception of the Asp-27 Asn mutant, all proteins were purified by the procedure of Baccanari et al. (1975, 1977). The purification of the Asp-27 Asn protein has been described previously (Villafranca et
-
-
-
-
-
-
-
-
-
VOL. 26, NO. 10, 1987
2675
al., 1983). Purity was ascertained by the observation of single bands on Coomassie blue stained native and NaDodSOc polyacrylamide gels. Protein concentration was measured by the absorbance at 280 nm = 3.11 X lo4 M-' cm-I; D. Baccanari, personal communication). Enzymatic activity was determined as described previously (Hillcoat et al., 1967). The specific activities of the wild-type and mutant proteins at pH 7.0 were found to be the following: wild type, 70 units mgl; Leu-28 Arg, 3 units mg-I; Phe-3 1 Val, 26 units mg-l; Arg-44 Leu, 19 units mg-I; Thr-113 Val, 35 units mg-'; Glu-139 Lys, 51 units mg-I. The k,,, and K, (dihydrofolate) for the Asp-27 Asn protein at pH 7.0 were 0.10 s-l and 44 pM, respectively;these parameters for the wild-type protein are 30 s-l and 1.2 pM (Howell et al., 1986). The retention of enzyme activity in these mutant proteins demonstrates that the tertiary structure must be very similar. Howell et al. (1986) have reported the high-resolution structure of the Asp-27 Asn protein. The only changes in structure introduced by this mutation involve the placement of two water molecules in the active-site pocket. Experimental Conditions. The buffer used throughout the folding experiments was 10 mM potassium phosphate, pH 7.8, 0.2 mM K,EDTA, and 1 mM 2-mercaptoethanol, The temperature was maintained at 15 "C by a Haake Model A80 constant-temperature bath. The reversibility of the folding for all mutant proteins under these conditions was greater than 95%, as judged by optical spectroscopy (unpublished results). Spectroscopic Methods. The equilibrium unfolding transition was monitored by difference ultraviolet spectroscopy at 293 nm on a Cary 118CX spectrophotometer. Difference spectra were obtained by using the tandem cell technique (Herskovits, 1967). All samples were allowed to fully equilibrate before spectra were collected. Kinetic data were obtained in most cases from difference spectroscopy measurements using manual mixing methods that have been described (Beasty et al., 1986). For relaxation times less than 10 s, data were collected on a Durrum 110 stopped-flow spectrophotometer in the fluorescence mode. Fluorescence measurements were made by exciting at 290 nm with a slit width of 5 mm. Emission intensity was monitored at wavelengths greater than 340 nm with a Corning C.S. 0-52 ground-glass filter. It has been shown previously that difference ultraviolet and fluorescence spectroscopies monitor the same folding reactions (Touchette et al., 1986). Data were collected on a PDP 1 1/23 computer that has been interfaced to both spectrophotometers. Data Fitting. The spectroscopic data were converted to plots of the apparent fraction of protein unfolded, Fa,,, by Fa,, = (eobsd - tN)/(tU - tN) where eobsd is the observed extinction coefficient and tN and tu are the extinction coefficients of the native and unfolded forms, respectively. Values for tN and tu in the transition region were obtained by linear extrapolation. The data for the dependence of Fa,, on the urea concentration were fit to the equation (Tanford, 1968):
---
--
-
Fapp
= exP(-AGapp/R73/ t1
+ ex~(-AGapp/RT)l
(1)
where AG,,, is the apparent free energy difference between the native and unfolded forms, R is the gas constant, and T is the absolute temperature. AG,,, was assumed to depend linearly on the urea concentration (Schellman, 1978): AG,,, = AGY$
+ A[urea]
(2)
where AG,,, is the apparent free energy difference between native and unfolded species at a given urea concentration,
2676
BIOCHEMISTRY
PERRY ET A L .
-Iv. I: fast
1,
Ul
v. slow
slow
fast
I2
1
v. slow
fast
13v. slow
1
1 4 -
fast u4
FIGURE 1: Proposed kinetic folding model for DHFR (Touchette et al., 1986). The reactions monitored in this study are the slower NI II step whose relaxation time is designated by T~ and the faster N2 I2 step whose relaxation time is T ~ .
--
AC:$ is the free energy difference in the absence of urea, and A is a parameter which describes the cooperativity of the unfolding transition. The fits were done by using a nonlinear least-squares fitting program, NLIN (SAS Institute Inc., Cary, NC) . The kinetic data were fit to a sum of exponentials as described previously, again using NLIN. Reagents. Ultrapure urea was purchased from Schwarz/ Mann and used without further purification; fresh solutions were prepared daily. The methotrexate affinity resin used in protein purification was obtained from Pierce. All other chemicals were reagent grade.
tUreo1.M
1
c I
B
1000
RESULTS A kinetic folding model has been proposed for DHFR (Touchette et al., 1986) and is shown in Figure 1. The model proposes that two slowly interconverting native conformers, N 1and N2, are populated in the absence of denaturant. At pH 7.8, 15 OC, Nl and N 2 comprise 15% and 85% of the population, respectively. When denaturant is added, these species unfold to intermediates I1 and I2 which are then subsequently converted to unfolded forms U1and U2. If DHFR is allowed to fully equilibrate at high denaturant concentrations, two additional unfolded forms, U3 and U4,become populated as well. A reasonable hypothesis for this collection of unfolded forms is isomerization at X-Pro peptide bonds (Brandts et al., 1975). In refolding, these unfolded forms rapidly (7 = 233 ms at 0.54 M urea, 15 "C) collapse to four intermediates, II-14. These intermediates then slowly fold to either the native forms, N1 and N2, or the nativelike intermediates, IN, and INI, depending upon their particular arrangement of isomers at the essential X-Pro peptide bonds. The two nativelike intermediates then convert to the two stable native forms at equilibrium. The model implies that proline isomerization also differentiates both the intermediates and the native and nativelike forms. Only the N1 I1 and N2 I2 reactions whose relaxation times are designated 71 and 72 were examined in this study, because both unfolding and refolding relaxation times can be measured. The method of analysis for the effect of mutations on folding requires information on both the unfolding and refolding relaxation times (Beasty et al., 1986). To test the roles of various elements of secondary structure on these two folding reactions, amino acid replacements were made at positions 27, 28, and 31 in helix aB, position 44 in helix aC, position 1 1 3 in strand PF, and position 139 in strand PG (Bolin et al., 1982). The equilibrium and kinetic properties of the folding reaction for DHFR were primarily monitored by difference ultraviolet spectroscopy at 293 nm. Changes in absorbance at this wavelength correspond to changes in exposure to solvent of one or more of the five tryptophan residues. Stopped-flow
-
-
t*
Y
" 0 1 2 3 4 5 6 7 8 9
t Urea 3 , M (A) Dependence of the apparent fraction of protein unfolded, Fa ,on the urea concentration for the Asp27 Asn mutant protein The unfolding transition curve for the wild-type protein under the same conditions is shown as a solid line for reference (Touchette et al., 1986). The lines represent fits of the data to a two-state unfolding model. The buffer contained 10 mM potassium phosphate, pH 7.8,0.2 mM K,EDTA, and 1 mM 2-mercaptoethanol; the temperature was 15 O C . (B) Dependence of the relaxation times for the T~ (0, W) and T~ (0,0 ) folding reactions on the final urea concentration for the Asp-27 Asn mutant protein at pH 7.8, 15 O C . Unfolding reactions (closed symbols) were initiated at 0 M urea and ended at the indicated urea concentration. Refolding reactions (open symbols) were initiated at 5.4 M urea and ended at the indicated final urea concentration. The dashed lines are drawn to aid the eye; the solid lines represent the results for the wild-type protein under the same conditions (Touchette et al., 1986).
-
FIGURE 2:
(07.
-
kinetic studies were performed by using fluorescence spectroscopy which also monitors the tryptophan residues at the excitation wavelength used, 290 nm. The effect of the Asp27 Asn replacement on the stability of DHFR is shown in Figure 2A. The dependence of the apparent fraction unfolded, Fapp, on the urea concentration for this mutant protein, like the wild-type DHFR, is that expected for a two-state unfolding reaction. The midpoint of the transition occurs at 3.6 M urea, 0.5 M urea higher than wild type, and the slope of the curve is slightly greater. Assuming a linear dependence of the free energy of folding on the urea concentration (Schellman, 1978) and a two-state folding model, the stability in the absence of denaturant for the mutant protein is predicted to be 7.3 kcal mol-'. The
-
V O L . 2 6 , N O . 10, 1987
MUTATIONS A N D FOLDING OF DIHYDROFOLATE REDUCTASE
2677
Table I: Stabilities and Transition Midpoints for the Urea-Induced Unfolding of Wild-Type DHFR and Six Mutant Proteins
A' [kcal mol-' (M protein mol-') Cmb(M, urea) urea)-'] -1.9 f 0.1 5.9 f 0.3 3.1 f 0.1 wild-type DHFR -2.0 0.4 7.3 i 1.3 3.6 i 0.1 Asp-27 Asn -2.3 0.2 7.6 i 0.7 3.3 0.1 Leu-28 Arg -1.8 f 0.3 4.4 i 0.5 2.4 i 0.1 Phe-31 Val -2.4 f 0.5 7.4 f 1.7 3.1 f 0.1 Arg-44 Leu 2.8 i 0.1 -1.6 f 0.2 Thr-113 Val 4.7 f 0.7 2.1 f 0.1 -2.3 f 0.5 Glu- 139 LYS 4.9 f 1.0 "The apparent free energy of unfolding in the absence of urea at pH 7.8, 15 "C. Errors are 95% confidence intervals from the nonlinear least-squares fits. Midpoint of the urea-induced unfolding transition. The value of C, is calculated from C, = -AGF$/A. The errors were calculated by propagation of errors analysis. CTheparameter A is defined in eq 2. Errors are 95% confidence intervals from the nonlinear least-squares fits. AGHfOa
(kS3
----
*
* *
stability of the wild-type protein is 5.9 kcal mol-' (Table I). The urea concentration dependence of the 7' and r 2 relaxation times for the Asp-27 Asn mutant protein is shown in Figure 2B. The relaxation times for unfolding for both phases increase by approximately a factor of 2. The effect on the refolding relaxation times is less and approaches the estimated experimental error. The slower, T ' , phase decreases by 30% while the small differences for the faster, r 2 , phase reflect a change in slope. The replacement of Leu by Arg at position 28, the adjacent position in the helix, has an effect on the equilibrium and kinetic properties of folding that is rather similar to that for Asp-27 Asn. The equilibrium curve (Figure 3A) is displaced to higher urea concentrations and has a slightly greater slope than wild-type protein; the stability in the absence of denaturant is calculated to be 7.6 kcal mol-' (Table I). The kinetic data (Figure 3B) again show that for unfolding both T , and r2 increase, in this case by a factor of 3. For refolding, the 7 , phase is coincident with that of wild type while the r 2 phase again shows a greater slope than wild type. To determine if the selective effect on the relaxation time for unfolding is specific to positions 27 and 28 or perhaps reflects a more general property of helix aB, DHFR containing the Phe-3 1 -,Val mutation was examined. This replacement causes a large decrease in the stability of DHFR (Figure 4A). The midpoint of the transition decreases to 2.4 M urea and the stability to 4.4 kcal mol-' (Table I). The effects on the relaxation times are equally as dramatic (Figure 4B). The unfolding relaxation times for the r 1 and T~ phases decrease in magnitude by a factor of 8 and 13, respectively, at 4 M urea; a part of this decrease is due to an increase in slope for both phases. The refolding relaxation times are within experimental error of those for the wild-type protein. Therefore, like the replacements at positions 27 and 28, the replacement at position 3 1 selectively alters the rate of unfolding. These results can be compared with those from a mutation in helix aC,Arg-44 -,Leu. The midpoint of the equilibrium curve occurs at 3.1 M urea, the same as that for wild type; however, the slope appears to be slightly greater (Figure 5A). The stability is calculated to be 7.4 kcal mol-' in the absence of denaturant (Table I). The refolding kinetic data (Figure 5B) show a striking difference from wild-type protein: below 3 M urea, the refolding relaxation times for both phases become urea independent. Above 3 M urea, the logarithm of the unfolding and refolding relaxation times follows the inverted V dependence observed for protein folding reactions (Tanford et al., 1973; Crisanti & Matthews, 1981; Kelley et al., 1986). The effect of the Arg-44 Leu replacement is
0
1
2
3
4 5 [Ureal, M
6
7
8
-
-
-
I : : : : ; : 3,A b CUrea
M
-
of protein unfolded, Fap,on the urea concentration for the Leu-28 Arg mutant protein (05 at pH 7.8, 15 OC. The transition curve for the wild-type protein under the same conditions is shown as a reference (Touchette et al., 1986). The lines represent fits of the data to a two-state folding model. (B) Semilog plot of the dependence of the relaxation times for the T , (0, m) and T~ (0, 0 ) folding reactions on the final urea Arg mutant protein at pH 7.8, 15 concentration for the Leu-28 OC. Open and closed symbols correspond to refolding and unfolding reactions, respectively, as described in Figure 2B. The dashed lines are drawn to aid the eye; the solid line indicates the results for the wild-type protein under the same conditions. Experimental conditions are described in Figure 2A. FIGURE 3: (A) Dependence of the apparent fraction
-
-
to decrease the relaxation time between 3 and 5 M urea where the I N refolding reaction dominates and to leave the relaxation times unaffected above 5 M urea where the N -, I unfolding reaction dominates.2 The important observation is that the N I refolding reaction is selectively accelerated for this mutation in helix aC. This is a markedly different result than those for all three replacements in helix a B where only the relaxation time for unfolding was altered. To determine if selective effects on unfolding or refolding are unique to helices, mutations at two positions in the central p sheet were examined. Replacement of Thr at position 113 with Val decreases the stability of DHFR (Figure 6A); the midpoint decreases to 2.8 M urea and the apparent free energy of unfolding to 4.7 kcal mol-' (Table I). The effect on kinetics
-
* This type of change in the urea dependence of both r , and 72 relaxation times at 3 M urea has been observed previously for the a subunit of tryptophan synthase (Matthews et al., 1983) and has been attributed to the coupling of folding and proline isomerization phases.
2678
BIOCHEMISTRY
PERRY ET AL.
4 II
!/
9"
0.0
0
1
2
3
4 5 [Ureal, M
6
7
8
I
B
I
I
I
I
I
I
I
I
I
L B
1000
t\
8
1
2
3
4 5 6 CUrea1,M
FIGURE 4:
7
8
1 o 0 1 2 L3Urea 4 3,5M 6 7 8 9
9
-
(A) Dependence of the apparent fraction of protein unfolded, Fa ,on the urea concentration for the Phe-31 Val mutant protein at pH 7.8, 15 "C. The transition curve for the wild-type protein under the same conditions is shown as a reference (Touchette et al., 1986). The lines represent fits of the data to a two-state folding model. (B) Semilog plot of the dependence of the relaxation times for the 71 ( 0 ,u) and Q (0, 0 ) folding reactions on the final urea concentration for the Phe-31 Val mutant protein at pH 7.8, 15 "C. Open and closed symbols correspond to refolding and unfolding reactions, respectively, monitored by difference spectroscopy as described in Figure 2B. Half-filled symbols indicate refolding relaxation times monitored by stopped-flow fluorescence spectroscopy. The dashed lines are drawn to aid the eye; the solid line indicates the results for the wild-type protein under the same conditions. Experimental conditions are described in Figure 2A.
(07
-
is complex and quite different from mutations in the two helices. The two slow phases observed previously are present; however, unfolding and refolding relaxation times have decreased for both the T~ and 7 2 phases (Figure 6B). For unfolding, the phase decreases by a factor of 4 and the 7 2 phase by a factor of 16 at 6 M urea, and for refolding, T~ decreases by a factor of 4 and T~ by a factor of 10 at 1 M urea. The slope of the T~ phase in refolding is also somewhat less than that for wild-type protein while the slope of the 72 phase in refolding is somewhat greater. The effects of an amino acid replacement in a second /3 strand, Glu-139 Lys, on the equilibrium and kinetic properties of folding are shown in Figure 7A,B. This mutation shifts the midpoint of the transition to 2.1 M urea and decreases the apparent free energy of unfolding to 4.9 kcal mol-' (Table I). Both the and T~ relaxation times for unfolding are decreased by a factor of 4-5 at 5 M urea; the slope of the 7,phase is greater than that of wild-type DHFR. In contrast to the other mutants, Glu-139 Lys has a differential effect
-
-
-
FIGURE 5: (A) Dependence of the apparent fraction of protein unfolded, Fa , on rhe urea concentration for the Arg-44 Leu mutant protein at pH 7.8, 15 "C. The transition curve for the wild-type protein under the same conditions is shown as a reference (Touchette et al., 1986). The lines represent fits of the data to a two-state folding model. (B) Semilog plot of the dependence of the relaxation times for the T, (0, u) and T* (0,0 ) folding reactions on the final urea concentration for the Arg-44 Leu mutant protein at pH 7.8, 15 "C. Open and closed symbols correspond to refolding and unfolding reactions, respectively, as described in Figure 2B. The dashed lines are drawn to aid the eye; the solid line indicates the results for the wild-type protein under the same conditions. Experimentalconditions are described in Figure 2A.
(07
-
on the T~ and 72 phases in refolding. The relaxation time for the 71 phase, which can only be resolved in unfolding jumps, is a factor of 3 smaller than wild type at 2.5 M urea. Although the relaxation time at this urea concentration is a composite of both the unfolding and refolding rate constants, the refolding rate constant dominates below the maximum in the In 7 vs. [urea] plot (Matthews, 1986). Therefore, the 71 relaxation time is presumed to also be faster than that for wild-type DHFR at lower urea concentrations, since it cannot be resolved from the 7 2 phase by the fitting procedure. The relaxation time for the r2 phase in refolding appears to be nearly unaffected by the mutation. DISCUSSION Structural Basis for the Effects of Mutations on the Folding Kinetics. Amino acid replacements in a helices have markedly different effects on the folding kinetics of DHFR than replacements in j3 strands. The mutations in helix aB, at positions 27, 28, and 31, all selectively alter the relaxation times for unfolding; the refolding relaxation times are nearly unaffected. This behavior classifies these as equilibrium
MUTATIONS A N D FOLDING OF DIHYDROFOLATE REDUCTASE
t
2679
0
t
‘ 0 1 2 3 4 5 6 7 6 CUreol, M
-
(A) Dependence of the apparent fraction of protein unfolded, Fa ,on the urea concentrations for the Thr- 113 Val mutant FIGURE 6:
VOL. 26, N O . 10, 1987
(3
protein at pH 7.8, 15 O C . The transition curve for the wild-type protein under the same conditions is shown as a reference (Touchette et al., 1986). The lines represent fits of the data to a two-state folding model. (B) Semilog plot of the dependence of the relaxation times for the 7, (0, m) and r2 (0,0 ) folding reactions on the final urea concentration for the Thr-113 Val mutant protein at pH 7.8, 15 “C. Difference ultraviolet spectroscopy was used to collect the r1phase data (0,m) and stopped-flow fluorescence spectroscopyto collect the r2phase data (0, 0 ) . Open and closed symbols correspond to refolding and unfolding reactions, respectively,as described in Figure 2B. The dashed lines are drawn to aid the eye; the solid line indicates the results for the wild-type protein under the same conditions. Experimental conditions are described in Figure 2A.
-
mutants (Beasty et al., 1986) and reflects a selective effect of the mutation on the free energy of the native conformations relative to the intermediate forms and the transition states linking these two species. Similarly, the Arg-44 Leu mutation in helix aC is an equilibrium mutant. However, in this case, the selective effect on the relaxation time for refolding implies that the mutation differentially alters the free energy of the intermediate conformations with respect to the transition states and the native conformations. Therefore, none of these replacements has altered the rate-limiting step linking the native and intermediate forms. In cantrast, the two mutations in /3 strands do have an effect on the rate-limiting step. For the Thr-113 Val mutant protein, unfolding and refolding relaxation times decrease by nearly equal amounts for both the T~ and r2 phases. This behavior is characteristic of kinetic mutants (Beasty et al., 1986) and identifies this position as playing a key role in the
-
-
1
0
1
2
3
4
5
6
[Urea 1, M
7
8
-
FIGURE 7: (A) Dependence of the apparent fraction of protein unfolded, Fa ,on the urea concentration for the Glu-139 Lys mutant protein at pH 7.8, 15 OC. The transition curve for the wild-type protein under the same conditions is shown as a reference (Touchette et al., 1986). The lines represent fits of the data to a two-state folding model. (B) Semilog plot of the dependence of the relaxation times for the r1 (0, W) and r2 (0, 0 ) folding reactions on the final urea concentration for the Glu-139 Lys mutant protein at pH 7.8, 15 O C . Open and closed symbols correspond to refolding and unfolding reactions, respectively, as described in Figure 2B. The half-filled symbol represents the results for unfolding and refolding jumps to the same final conditions. The dashed lines are drawn to aid the eye; the solid line indicates the results for the wild-type protein under the same conditions. Experimental conditions are described in Figure
(6
-
2A.
-
interconversion of native and intermediate forms. The behavior of the Glu-139 Lys mutant is more complex. The effect on the 71 phase is that expected for a kinetic mutation, with both unfolding and refolding accelerated. The effect on the 72 phase is characteristic of an equilibrium mutation with only the unfolding reaction altered. A hypothesis can be drawn from these results concerning the nature of the rate-limiting step in this slow folding reaction in DHFR. Considering a single folding channel in Figure 1 slow
N-I-U
fast
the unfolded protein collapses to one or more rapidly interconverting intermediates that have substantial secondary and tertiary structure. The presumption of substantial structure is based upon the observation that the I N reaction is very slow (7 > 10 s). According to the Eyring formalism, the activation free energy must be greater than 18 kcal mol-I. Therefore, a number of noncovalent interactions must be disrupted. The observation of a selective effect of replacements in the 0 strands on the rate-limiting step can be integrated into
-
2680
BIOCHEMISTRY
PERRY ET AL.
this hypothesis by presuming that the supersecondary structure of the intermediate resembles but is not identical with that found in the native conformation. In particular, the /3 sheet is not correctly formed. This intermediate folds to the native conformation in a slow process whose rate-limiting step selectively involves the /3 sheet. Possible errors in the /3 sheet might include an incorrect twist (Richardson, 1981), strand exchange, strand inversion (parallel to antiparallel or vice versa), strand register, and the absence or incorrect placement of a P bulge in strand PG (Bolin et al., 1982). Further support for the hypothesis that the secondary structure and not the individual residues play a key role in this folding reaction comes from the Asp-27 Asn and Thr- 113 Val mutants. The crystal structure (Bolin et al., 1982) shows that Asp27 and Thr-113 are hydrogen bonded to each other and therefore occupy contiguous positions in DHFR. The observation that Asp27 Asn is an equilibrium mutant while Thr-113 Val is a kinetic mutant demonstrates that substitutions in the same region of the protein can have drastically different effects. Although this result is consistent with the role of secondary structure in the N * I reaction, this hypothesis must be directly tested by constructing multiple replacements of the same position and testing the effects on folding. Such studies are now in progress. The amphipathic helices may also form in the intermediate and bind to the hydrophobic surface of the B sheet. Certainly, the selective effect of the Arg-44 Leu mutation on the free energy of the intermediate demonstrates its participation in the structure and suggests that the remainder of the helix exists as well. Further mutations in this helix are required to prove this point. The selective effect of mutations in helix a B at positions 27, 28, and 31 on the free energy of the native form does not permit an assessment of its existence in the intermediate. If helix a B is present, its position in the intermediate and the transition state must remain unchanged so as not to affect the relaxation time for refolding. It is possible that this helix only forms and binds to the /3 sheet after the rate-limiting step is complete. This latter possibility is supported by the results of an inhibitor binding study which found that methotrexate does not bind to the intermediates (Touchette et al., 1986). Because helix a B forms a part of the binding site for methotrexate, this helix may not exist in the intermediate. It should be mentioned that the observation of a selective effect of amino acid replacement on either the unfolding or the refolding relaxation time for the N I reaction cannot be uniquely interpreted in terms of an effect only on the energy of either the native form or the intermediate form. It is also possible that the replacement alters equally the energy of both the transition state and the complementary form. For example, for the Phe-31 Val protein (Figure 4B), either the selectively accelerated unfolding could reflect an increase in the energy of the native conformation while the energies of the transition state and intermediate forms are unchanged or the energies of both the transition state and intermediate conformations decrease by the same amounts and the energy of the native conformation is unchanged. The latter explanation seems less likely; however, it is certainly a possibility because the rate constants only reflect differences in free energies. For this reason, structural interpretations of the unfolding/refolding kinetic data must be made with caution. Structural Basis for the Effects of Mutations on Stability. The goodness of fit of the equilibrium data for all the proteins in this study to a two-state model may appear to be rather surprising in view of the complexity of the folding model for DHFR (Figure 1). However, previous studies have shown that
-
-
-
-
-
-
-
the only species expected to be highly populated at equilibrium are N1,N2,U1, U2, U3, and U4 (Touchette et al., 1986). If N, and N2 are spectroscopically indistinguishable and are of similar stability and if U1, U2, U3, and U4 are also spectroscopically similar, then a two-state model is expected to describe the equilibrium data. The goodness of the fit supports these assumptions. In this system, the calculated stability must be regarded as an apparent free energy difference since each unfolding reaction, N, U,and N2 * U2, will contribute to the apparent stability; the 15:85 ratio for Nl:N2corresponds to a free energy difference between N1 and N2 of 1 kcal mol-'. The errors shown for AC:$ in Table I for the Asp-27 Asn, Arg-44 Leu, and Glu-139 Lys mutant proteins are comparable to the differences from the value of A G:$ for the wild-type protein. Note that these errors are 95% confidence limits obtained from the nonlinear least-squares fits and are roughly comparable to two standard deviations. We believe Asn protein that the increase in stability for the Asp-21 and decrease for the Glu- 139 Lys protein are real because (1) the midpoints of the transitions are measurably different from wild type (Figures 2A and 7A and Table I) and (2) the urea dependence of Fapp, which is reflected in the parameter A, is the same or nearly the same for both mutants. Therefore, extrapolation to 0 M urea should not change the rank order of the apparent stabilities and should preserve the differences in stabilities observed at higher urea concentrations. Using the same reasoning, one would conclude that the difference in Act;: between the Arg-44 Leu mutant protein and the wild-type protein is not significant. Given this assessment of the error analysis, the effects of these amino acid replacements on the stability can be qualitatively explained in terms of the X-ray structure of DHFR (Bolin et al., 1982). Asp-27 is located at the bottom of a hydrophobic crevice and has a small exposure to solvent. The apparent effect of sequestering the carboxylic acid moiety in a hydrophobic environment is to raise the pK value to -6.5 (C. Fierke, S.J. Benkovic, and K. A. Johnson, personal communication) so that the protonated form can participate in the catalytic reaction (Stone & Morrison, 1984; Howell et al., 1986). This increase in the pKvalue must come at the expense of the protein stability; an increase in the pK for an aspartic acid residue from 4.5 to 6.5 is expected to destabilize the protein by 2.8 kcal mol-' at pH 7.8 (Cantor & Schimmel, 1980). Replacement of Asp-27 by Asn may increase the stability of DHFR by removing this partially buried negative charge. In addition, mutation of Leu-28 to Arg may increase the stability by partially neutralizing the neighboring Asp. The increase in stability for both of these mutants is in accord with this prediction. Phe-31 forms a part of the hydrophobic core of DHFR. Its replacement by the smaller Val could destabilize the protein by disrupting the close packing found in the interior of proteins (Richards, 1974). The 1.5 kcal mol-' decrease in stability could also reflect the difference in the free energy of transfer for these two hydrophobic residues from water to the protein interior. According to the measurements of Nozaki and Tanford (1 97 1) on the free energy of transfer from ethanol to water, the Phe-3 1 protein would be predicted to be 1.O kcal mol-' more stable than the Val-3 1 variant. The uncertainty in the value for AGF$ for the Arg-44 Leu mutant protein precludes any discussion of the effect of the mutation on the stability. Thr-113 participates in a hydrogen-bonding network with Asp-27 and a water molecule. Its replacement by Val would disrupt several hydrogen bonds which could certainly account
-
-
-
-
-
-
-
-
MUTATIONS A N D FOLDING OF DIHYDROFOLATE REDUCTASE
for the observed 1.2 kcal mol-' decrease in stability. Glu- 139 forms a hydrogen-bonded salt bridge with His- 141. Replacement by Lys would disrupt this noncovalent interaction and decrease the stability. The contribution to stability of the Glu-139-His-141 salt bridge can be estimated from the magnitude of the increase in the pK value of His-141. The pK of His-141 has been reported as 7.4 (Poe et al., 1979), 0.4-0.9 pH unit higher than that expected for free His (Cantor & Schimmel, 1980). For a completely protonated His residue, this shift in pK value corresponds to an interaction energy of 0.6-1.2 kcal mol-'. At pH 7.8, where the folding studies were done, the His residue is only -30% protonated. Therefore, the electrostatic contribution to the salt bridge would be reduced accordingly. At a minimum, the loss in stability caused by the Glu to Lys change should be 0.2-0.4 kcal mol-' at pH 7.8. The observed decrease in stability, 1.0 kcal mol-', is larger and suggests that other factors including perhaps the loss of a hydrogen bond between the neutral His-141 and Glu-139 or repulsive effects between Lys-139 and His-141 may also contribute. Other Comments. The effects of five of the six mutations on the folding kinetics were quite similar for both the 71 and 72 phases. The lone exception was the Glu-139 Lys mutant where a kinetic effect was observed for the 71 phase and an equilibrium effect for the 7 2 phase. Because these two phases are postulated to result from two native conformers which are in slow equilibrium (Touchette et al., 1986), differential effects on these two folding reactions highlight those regions of the protein where N 1 and N 2 differ. The results show that the region near residue 139 in the N, conformer is still coupled to the rate-limiting step; however, the same region in the N z conformer is not. Although further mutagenic and spectroscopic studies are required to determine the difference in these conformers, this result demonstrates that mutagenesis can be useful in focusing the efforts of high-resolution structural studies such as NMR. In several instances, it was observed that amino acid replacements appear to alter the slopes of In 7 vs. the concentration of denaturant plots (Figures 2B, 3B, 4B, 6B, and 7B). Although it is not clear exactly what these changes mean, one way of considering the slopes of such plots has been presented by Tanford (1970). In this scheme, the transition state linking the native and intermediate conformations is proposed to be at some position a along a normalized reaction coordinate between N and I, where the value of a varies from 0 for a nativelike transition state to 1 for an intermediate-like transition state. The unfolding and refolding slopes of a plot of In 7 vs. denaturant concentration define the value of a. Therefore, changes in slope correspond to changes in the value of a and may be related to differential changes in the structures of N, the transition state, and the intermediate state. Differential effects on the slopes for the 71 and 7 2 phases serve to emphasize differences between the N1and N2conformers. The variety of effects seen for this limited collection of mutants precludes the possibility of a structural interpretation. The hypothetical nature of this analysis also suggests that caution be used in interpreting changes in these slopes. The scheme used to evaluate the results of these kinetic studies of folding is based upon transition-state theory (Amdur & Hammes, 1966). This theory was developed for extremely simple chemical transformations, and one may question the validity of its application to a protein folding reaction. Certainly one expects that the reaction coordinate for a folding reaction would not be associated with any single bond length. Rather, it would represent the simultaneous change of a large
-
VOL. 2 6 , NO. 10, 1 9 8 7
2681
number of individual coordinates, much as happens in a normal mode analysis of a macromolecule. Currently, the application of transition-state theory to protein folding is justified by the experimental data, which show that the I N interconversion behaves kinetically as a simple two-state reaction. The relaxation time across the transition varies in a way expected for such a simple reaction, and the temperature dependence of the rate constants (k = 7-l) for unfolding and refolding obeys the Arrhenius law. Therefore, until data are obtained which refute this simple analysis, it appears to be a useful approach for understanding the effects of mutations on folding kinetics. The hypothesis concerning the relationship between a rate-limiting step in the folding of DHFR and its secondary structure must be considered preliminary at this time. However, it is sufficiently specific that tests using site-directed mutagenesis are readily constructed. For example, do multiple replacements at the same site in a helii or strand have the same selective effects on the folding kinetics? If so, then the effect is related to the placement of the amino acid in a particular element of secondary structure and not to the specific amino acid. Further, do replacements at any position in the same element of secondary structure have the same selective effect on kinetics? If so, then that element of secondary structure may be acting as a structural unit in folding. What are the effects of replacements in other elements of structure such as /3 turns and loops? Mutagenesis appears to provide a means of answering these questions and greatly improving our understanding of the structural aspects of protein folding.
-
ACKNOWLEDGMENTS We thank Mark Hurle and Dr. Neil Tweedy for helpful comments and assistance with data analysis and also Gail Feldman for typing the manuscript. REFERENCES Amdur, I., & Hammes, G. G. (1966) Chemical Kinetics, McGraw-Hill, New York. Baccanari, D., Phillips, A,, Smith, S.,Sinski, D., & Burchall, J. (1975) Biochemistry 14, 5267. Baccanari, D., Averett, D., Briggs, C., & Burchall, J. (1977) Biochemistry 16, 3566. Beasty, A. M., Hurle, M. R., Manz, J. T., Stackhouse, T., Onuffer, J. J., & Matthews, C. R. (1986) Biochemistry 25, 2965. Bolin, J. T., Filman, D. J., Matthews, D. A,, Hamlin, R. C., & Kraut, J. (1982) J . Biol. Chem. 257, 13650. Brandts, J. F., Halvorsen, H. R., & Brennan, M. (1975) Biochemistry 14, 4953. Cantor, C. R., & Schimmel, P. R. (1980) in Biophysical Chemistry, pp 42-50, W. H. Freeman, San Francisco, CA. Chen, J.-T., Mayer, R. J., Fierke, C. A., & Benkovic, S . J. (1985) J . Cell. Biochem. 29, 73. Crisanti, M. M., & Matthews, C. R. (1981) Biochemistry 20, 2700. Dalbadie-McFarland, G., Cohen, L. W., Riggs, A. D., Morin, C., Itakura, K., & Richards, J. H. (1982) Proc. Natl. Acad. Sci. U.S.A. 79, 6409. Herskovits, T. T. (1967) Methods Enzymol. 11, 748. Hillcoat, B. L., Nixon, P. F., & Blakley, R. L. (1967) Anal. Biochem. 21, 178. Howell, E. E., Villafranca, J. E., Warren, M. S., Oatley, S . J., & Kraut, J. (1986) Science (Washington, D.C.) 231, 1123. Kelley, R. F., Wilson, J., Bryant, C., & Stellwagen, E. (1986) Biochemistry 25, 728.
2682
Biochemistry 1987, 26, 2682-2689
Kim, P. S., & Baldwin, R. L. (1982) Annu. Rev. Biochem. 51, 459. Kuwajima, K. (1977) J . Mol. Biol. 114, 241. Labhardt, A. M. (1984) Proc. Natl. Acad. Sci. U.S.A. 81, 7674. Matthews, C. R. (1986) Methods Enzymol. (in press). Matthews, C. R., Crisanti, M. M., Manz, J. T., & Gepner, G. L. (1983) Biochemistry 22, 1445. Maxam, A., & Gilbert, W. (1977) Proc. Natl. Acad. Sci. U.S.A. 74, 560. Nozaki, Y., & Tanford, C. (1 971) J. Biol. Chem. 246, 221 1. Poe, M., Hoogsteen, K., & Matthews, D. A. (1979) J. Biol. Chem. 254, 8143. Richards, F. M. (1974) J . Mol. Biol. 82, 1. Richardson, J. S. (1981) Adu. Protein Chem. 34, 168. SAS Institute Inc. (1985) S A S User’s Guide: Statistics,
Version 5 Edition, p 575, SAS Institute Inc., Cary, NC. Schellman, J. A. (1978) Biopolymers 17, 1305. Smith, D. R., & Calvo, J. M. (1980) Nucleic Acids Res. 8, 2255. Stone, S. R., & Morrison, J. F. (1984) Biochemistry 23, 2753. Taira, K., Chen, J.-T., Mayer, R. J., & Benkovic, S. J. (1986) Bull. Chem. SOC.Jpn. (in press). Tanford, C. (1968) Adu. Protein Chem. 23, 218. Tanford, C. (1970) A h . Protein Chem. 24, 1. Tanford, C., Aune, K. C., & Ikai, A. (1973) J. Mol. Biol. 73, 185. Touchette, N. A., Perry, K. M., & Matthews, C. R. (1986) Biochemistry 25, 5445-5452. Villafranca, J. E., Howell, E. E., Voet, D. H., Strobel, M. S., Ogden, R. C., Abelson, J. N., & Kraut, J. (1983) Science (Washington, D.C.) 222, 782.
Mechanism of Slow-Binding Inhibition of Human Leukocyte Elastase by Trifluoromethyl Ketones Ross L. Stein,*,*Anne M . Strimpler,* Phillip D. Edwards,§ Joseph J. Lewis,$ Russel C. Mauger,l Jack A. Schwartz,g Mark M. Stein,$ D. Amy Trainor,! Richard A. Wildonger,s and Mark A. Zottolas Departments of Pharmacology and Medicinal Chemistry, Stuart Pharmaceuticals. A Division of ICI Americas Inc., Wilmington, Delaware 19897 Received October 8, 1986; Revised Manuscript Received December 24, 1986
ABSTRACT: Kinetics of inhibition have been determined for the interaction of human leukocyte elastase (HLE) with two series of peptide trifluoromethyl ketones (TFMKs): X-Val-CF,, X-Pro-Val-CF,, X-Val-Pro-Val-CF3, and X-Lys(Z)-Val-Pro-Val-CF,, where X is MeOSuc or Z. These compounds are “slow-binding” inhibitors of H L E and, thus, allow the determination of Ki, the dissociation constant for the stable complex of inhibitor and enzyme, as well as k,, and kOff,the rate constants for formation and decomposition of this complex. Maximal potency is reached with Z-Lys(Z)-Val-Pro-Val-CF3,which displays a Ki < 0.1 nM. Upon binding to HLE, these compounds undergo addition by the hydroxyl of the active site serine to form a hemiketal. The evidence supporting a hemiketal intermediate includes (i) Ki values of 1.6 and 80000 nM for ZVal-Pro-Val-CF, and its alcohol analogue, (ii) linear free energy correlations between inhibitory potency and catalytic efficiency for structurally related TFMKs and substrates, and (iii) the pH dependence of k,, for the inhibition of H L E by Z-Val-Pro-Val-CF,, which is sigmoidal and displays a pK, of 6.9. Hemiketal formation is probably not rate limiting, however. Kinetic solvent isotope effects of unity suggest that k,, cannot be rate limited by a reaction step, like hemiketal formation, that is subject to protolytic catalysis. A general mechanism that is consistent with these results is one in which formation of the hemiketal is rapid and is followed or preceded by a slow step that rate limits k,,. This step must be insensitive to both the isotopic composition of the solvent and the p H and may be a conformational change of one of the enzyme-inhibitor complexes preceding the final complex.
Fluorine-substituted ketones are known inhibitors of a variety of hydrolytic enzymes, including acetylcholine esterase (Brodeck et al., 1979; Gelb et al., 1985), juvenile hormone esterase (Prestwich et al., 1984), carboxypeptidase A (Gelb et al., 1985), angiotensin converting enzyme (Gelb et al., 1985), pepsin (Gelb et al., 1985), phospholipase A2 (Gelb, 1986), porcine pancreatic elastase (Imperiali & Abeles, 1986), and chymotrypsin (Imperiali & Abeles, 1986). For serine hydrolases, such as AChE’ and PPE, trifluoromethyl ketones resembling specific substrates have been prepared and shown
* Address correspondence to this author at his present address: Department of Enzymology, Merck Institute for Therapeutic Research, Rahway, N J 07065. $Department of Pharmacology. 8 Department of Medicinal Chemistry. 0006-2960/87/0426-2682$01.50/0
to be potent inhibitors. TFMKs are thought to inhibit these enzymes by combining with the active site serine to form a hemiketal that resembles the tetrahedral intermediate or transition states that occur during serine acylation by peptide substrates. In this paper we explore the mechanism of inhibition of the serine protease human leukocyte elastase (Stein et al., 1985) by TFMKs. Our studies suggest a mechanism in which for-
’
Abbreviations: HLE, human leukocyte elastase; PPE, porcine pancreatic elastase; AChE, acetylcholine esterase; TFMK, trifluoromethyl ketone; MeOSuc, A’-(methoxysuccinyl); pNA, p-nitroanilide; Z , N(carbobenzoxy); ONP, p-nitrophenyl ester; Tris, tris(hydroxymethy1)aminomethane; CAPS, 3-(cyclohexylamino)propanesulfonic acid; CHES, 2-(cyclohexylamino)ethanesulfonic acid: HEPES, 4-(2-hydroxyethyl)- 1 piperazineethanesulfonic acid.
0 1987 American Chemical Society
