Portal de Periódicos da CAPES

Structural Determinants of Cold Adaptation and Stability in a Large Protein

2001; Elsevier BV; Volume: 276; Issue: 28 Linguagem: Inglês

10.1074/jbc.m102741200

1083-351X

Salvino D’Amico, Charles Gerday, Georges Feller,

Enzyme Production and Characterization

The heat-labile α-amylase from an antarctic bacterium is the largest known protein that unfolds reversibly according to a two-state transition as shown by differential scanning calorimetry. Mutants of this enzyme were produced, carrying additional weak interactions found in thermostable α-amylases. It is shown that single amino acid side chain substitutions can significantly modify the melting point T m, the calorimetric enthalpyΔH cal, the cooperativity and reversibility of unfolding, the thermal inactivation rate constant, and the kinetic parameters k cat andK m. The correlation between thermal inactivation and unfolding reversibility displayed by the mutants also shows that stabilizing interactions increase the frequency of side reactions during refolding, leading to intramolecular mismatches or aggregations typical of large proteins. Although all mutations were located far from the active site, their overall trend is to decrease bothk cat and K m by rigidifying the molecule and to protect mutants against thermal inactivation. The effects of these mutations indicate that the cold-adapted α-amylase has lost a large number of weak interactions during evolution to reach the required conformational plasticity for catalysis at low temperatures, thereby producing an enzyme close to the lowest stability allowing maintenance of the native conformation. The heat-labile α-amylase from an antarctic bacterium is the largest known protein that unfolds reversibly according to a two-state transition as shown by differential scanning calorimetry. Mutants of this enzyme were produced, carrying additional weak interactions found in thermostable α-amylases. It is shown that single amino acid side chain substitutions can significantly modify the melting point T m, the calorimetric enthalpyΔH cal, the cooperativity and reversibility of unfolding, the thermal inactivation rate constant, and the kinetic parameters k cat andK m. The correlation between thermal inactivation and unfolding reversibility displayed by the mutants also shows that stabilizing interactions increase the frequency of side reactions during refolding, leading to intramolecular mismatches or aggregations typical of large proteins. Although all mutations were located far from the active site, their overall trend is to decrease bothk cat and K m by rigidifying the molecule and to protect mutants against thermal inactivation. The effects of these mutations indicate that the cold-adapted α-amylase has lost a large number of weak interactions during evolution to reach the required conformational plasticity for catalysis at low temperatures, thereby producing an enzyme close to the lowest stability allowing maintenance of the native conformation. pig pancreatic α-amylase P. haloplanktis α-amylase B. amyloliquefaciens α-amylase differential scanning calorimetry The cold-active and heat-labile α-amylase from the antarctic bacterium Pseudoalteromonas haloplanktis is an exceptional case because it is the largest known protein undergoing a reversible unfolding according to a two-state reaction pathway. It has been proposed that this unusual behavior is dictated by the requirement for an improved flexibility or plasticity of the protein molecule to perform catalysis at near-zero temperatures (1Feller G. d'Amico D. Gerday C. Biochemistry. 1999; 38: 4613-4619Crossref PubMed Scopus (148) Google Scholar). Structural investigations have suggested that this is achieved by decreasing the number and strength of weak interactions stabilizing the native conformation (2Feller G. Payan F. Theys F. Qian M. Haser R. Gerday C. Eur. J. Biochem. 1994; 222: 441-447Crossref PubMed Scopus (195) Google Scholar, 3Aghajari N. Feller G. Gerday C. Haser R. Structure. 1998; 6: 1503-1516Abstract Full Text Full Text PDF PubMed Scopus (188) Google Scholar, 4Aghajari N. Feller G. Gerday C. Haser R. Protein Sci. 1998; 7: 564-572Crossref PubMed Scopus (161) Google Scholar). Accordingly, one can expect that any mutation designed to improve the stability of this fragile edifice will induce significant perturbations of the denaturation pattern, thereby giving evidence for the structural features linked to the unfolding parameters. Small globular proteins that unfold according to a pure two-state process, i.e. reversibly and without a stable intermediate between the native and the unfolded states, have allowed researchers to establish the thermodynamic stability function accounting for the Gibbs energy change associated with their denaturation over physiological and nonphysiological temperature ranges. The current availability of this thermodynamic function, mainly gained by differential scanning calorimetry, represents much more than the simple physicochemical characterization of a polymer because it offers a powerful tool for investigation of the still unexplained "second genetic code,"i.e. the driving force allowing a polypeptide to fold into a predetermined native and biologically active conformation (5Rumbley J. Hoang L. Mayne L. Englander S.W. Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 105-112Crossref PubMed Scopus (132) Google Scholar). The validity of the bell-shaped stability curve, which predicts protein unfolding at both high and subzero temperatures, has been established by the experimental demonstration of cold unfolding (6Privalov P.L. Crit. Rev. Biochem. Mol. Biol. 1990; 25: 281-305Crossref PubMed Scopus (809) Google Scholar). Proteins are said to be marginally stable because the positive Gibbs free energy change is only a small difference between the large enthalpic and entropic contributions in the temperature range restrained between the cold and hot melting temperatures. Basically, denaturation at high temperature is entropy-driven, because the large and positive thermal dissipative force exceeds the corresponding enthalpy change. By contrast, cold unfolding is enthalpy-driven and arises from the negative contribution of the hydration Gibbs energy of protein groups, mainly polar, at low temperatures (7Makhatadze G.I. Privalov P.L. Adv. Protein. Chem. 1995; 47: 307-425Crossref PubMed Scopus (1000) Google Scholar). Four parameters are usually assigned to the unfolding event recorded by differential scanning calorimetry, (i) the melting point of the unfolding transition; (ii) ΔCp, the difference in heat capacity between the native and unfolded states arising from the transfer of buried groups in the protein to water, in the random coil; (iii) ΔH cal, the calorimetric enthalpy corresponding to the total amount of heat absorbed during unfolding and recorded by the calorimeter; and (iv)ΔH eff, the effective or van't Hoff enthalpy calculated from the slope of the transition and describing the cooperativity of the unfolding reaction. In the case of a pure two-state model, ΔH cal andΔH eff are nearly identical. However, large proteins generally unfold irreversibly as a result of strong heat-induced aggregation, with large deviations from the two-state model because their multidomain structure induces a stepwise or overlapping unfolding of domains, behaving as individual calorimetric units (Fig. 1). We report here the calorimetric characterization of mutants from the cold-active α-amylase defined on the basis of its strong homology with heat-stable mesophilic α-amylases. These mutants reveal the unsuspected contribution of single amino acid substitutions to the thermodynamic parameters characteristic of large heat-stable proteins and provide new insight into the molecular adaptations of enzymes from psychrophilic organisms. Mutations in theP. haloplanktis α-amylase gene were introduced by polymerase chain reaction as described (8Feller G. le Bussy O. Houssier C. Gerday C. J. Biol. Chem. 1996; 271: 23836-23841Abstract Full Text Full Text PDF PubMed Scopus (85) Google Scholar). The nucleotide sequence of the constructions was checked on an Amersham Pharmacia Biotech ALF DNA sequencer. The recombinant wild-type α-amylase and the mutant enzymes were expressed in Escherichia coli at 18 °C and purified by DEAE-agarose, Sephadex G-100, and Ultrogel AcA54 column chromatography, as described previously (9Feller G. le Bussy O. Gerday C. Appl. Environ. Microbiol. 1998; 64: 1163-1165Crossref PubMed Google Scholar). Pig pancreatic α-amylase (PPA)1 was from Roche. Measurements were performed using a MicroCal MCS-DSC instrument as detailed (1Feller G. d'Amico D. Gerday C. Biochemistry. 1999; 38: 4613-4619Crossref PubMed Scopus (148) Google Scholar). Samples (∼3 mg/ml) were dialyzed overnight against 30 mm MOPS, 50 mm NaCl, and 1 mm CaCl2, pH 7.2. Thermograms were analyzed according to a non-two-state model in whichT m, ΔH cal, andΔH eff of individual transitions are fitted independently using the MicroCal Origin software (version 2.9). The magnitude and source of the errors in the T m and enthalpy values have been discussed elsewhere (10Matouschek A. Matthews J.M. Johnson C.M. Fersht A.R. Protein Eng. 1994; 7: 1089-1095Crossref PubMed Scopus (95) Google Scholar). Fitting standard errors on a series of three DSC measurements made under the same conditions in the present study were found to be ±0.05 K onT m and ±1% on both enthalpies. Thermograms generated by starch hydrolysis were recorded on a MCS isothermal titration calorimeter. The technical background for such an experiment has been described elsewhere (11Lonhienne T. Baise E. Feller G. Bouriotis V. Gerday C. Biochim. Biophys. Acta. 2001; 1545: 349-356Crossref PubMed Scopus (59) Google Scholar). Starch concentration was 2% in 50 mmHepes, 50 mm NaCl, 1 mm CaCl2, pH 7.2, and was saturating during the time course of the experiment. Reactions were initiated by injecting 10–50 µl of a 30 µg/ml enzyme solution into the same buffer. The activity decay at 45 °C was recorded continuously for 30 min, and data were fitted on a monoexponential function to determine the first order constant rate of enzyme inactivation at 45 °C, S.E. <1%. The k catvalues were determined at 25 °C using 3.5 mm4-nitrophenyl-α-d-maltoheptaoside-4,6-O-ethylidene as substrate (8Feller G. le Bussy O. Houssier C. Gerday C. J. Biol. Chem. 1996; 271: 23836-23841Abstract Full Text Full Text PDF PubMed Scopus (85) Google Scholar). The k cat /K mvalues for substrate hydrolysis were determined from initial rates using the following relation.k cat /K m = v 0/S0 E 0, which is valid at S0 ≪ K m for systems that obey Michaelis-Menten kinetics. The validity of this equation was ascertained by performing determinations at three substrate concentrations. The psychrophilic α-amylase from P. haloplanktis (AHA) belongs to chloride-dependent α-amylases. This group includes all known animal α-amylases and the enzymes from some Gram-negative bacteria, covering a large range of living temperatures (12D'Amico S. Gerday C. Feller G. Gene ( Amst. ). 2000; 253: 95-105Crossref PubMed Scopus (92) Google Scholar). The high degree of amino acid sequence identity between chloride-dependent α-amylases allowed us to construct a multiple sequence alignment of the available primary structures (12D'Amico S. Gerday C. Feller G. Gene ( Amst. ). 2000; 253: 95-105Crossref PubMed Scopus (92) Google Scholar) and to detect about 20% of 453 residues that are specific to the cold-adapted enzyme. The involvement of these residues in the weak stability of the psychrophilic α-amylase was assessed by comparison of its crystal structure (3Aghajari N. Feller G. Gerday C. Haser R. Structure. 1998; 6: 1503-1516Abstract Full Text Full Text PDF PubMed Scopus (188) Google Scholar, 4Aghajari N. Feller G. Gerday C. Haser R. Protein Sci. 1998; 7: 564-572Crossref PubMed Scopus (161) Google Scholar) with that of PPA, the closest structural homolog, as well as with the structure of α-amylases from human salivary and pancreatic glands and from the insect Tenebrio molitor. Molecular modeling was then used to check the consistency of mutations aimed to introduce weak interactions found in mesophilic α-amylases in the heat-labile AHA. The selected mutations are listed in Table I along with the restored interactions. Following production in E. coli and purification, the stability and kinetic parameters of 14 mutant enzymes were determined (Table II). Mutants are listed according to their main properties, as discussed below. Fig. 1 depicts the microcalorimetric behavior of the recombinant wild-type AHA and of heat-stable α-amylases, including PPA taken as reference for mutation selection.Table ISelected mutations in this studyMutantsRestored interactions1-a(A) and (C) refer to the central (β/α)8 barrel and to the globular C-terminal domain, respectively.K300RAdditional H-bond in the Cl− binding site (A)N150DSalt bridge with Lys-190 (A)V196FDouble aromatic interaction with Tyr-82 and Phe-198 (A)N150D/V196FDouble mutant (A)Q164I/V196FDouble mutant (A)Q164IHydrophobic interaction in a core cluster (A)T232VHydrophobic interaction in a core cluster (A)N288VHydrophobic interaction in a core cluster (A)M379VHydrophobic interaction in a core cluster (C)R64EHelix dipole stabilization (A)L219RAmino-aromatic interaction with Tyr-298 (A)Del6Deletion of 6 nonconstrained residues at the C terminus possibly involved in unzipping (C)N12RSalt bridge with Asp-15 (A)E279WAromatic interaction with Phe-307 (A)1-a (A) and (C) refer to the central (β/α)8 barrel and to the globular C-terminal domain, respectively. Open table in a new tab Table IIStability and kinetic parameters of wild-type and mutant α-amylasesTmΔH calΔH effΔH cal/ΔH effReversibilityk inactk catK mk cat/K m°Ckcal mol−1kcal mol−1%10−3s−1s−1µms−1µm−1Heat-labile α-amylaseAHA44.02142031.051002.27697 ± 33234 ± 182.98Enthalpic stabilizationK300R45.22232171.03870.59326 ± 17141 ± 22.31N150D44.82712401.13971.53682 ± 31241 ± 102.83V196F45.42342960.79600.95752 ± 5223 ± 123.37N150D/V196F46.42722711910.54642 ± 25174 ± 113.69Calorimetric domainsQ164I/V196F(1)2-a(1) and (2) refer to calorimetric domains identified by deconvolution of DSC thermograms.44.7831210.69510.87501 ± 12153 ± 73.27(2)46.21253080.41Q164I(1)42.7871170.74641.77514 ± 18118 ± 114.36(2)45.21172400.49T232V(1)42.71021140.89852.51735 ± 16164 ± 224.48(2)45.71042020.51N288V(1)38.1881330.6630350 ± 14139 ± 72.52(2)45.9591140.52M379V(1)42.6127961.32952.25633 ± 24159 ± 43.98(2)47.7841370.61Unfolding cooperativityR64E42.62361122.11962.22719 ± 6220 ± 243.27L219R42.52171871.16811.17469 ± 42144 ± 33.26Del642.42112850.74892.07726 ± 21264 ± 72.75N12R43.31932310.84252.48670 ± 27178 ± 63.76E279W41.41862340.79762.40744 ± 22201 ± 143.70Heat-stable α-amylasePPA(1)61.71471271.1600291 ± 865 ± 44.48(2)65.61482460.62-a (1Feller G. d'Amico D. Gerday C. Biochemistry. 1999; 38: 4613-4619Crossref PubMed Scopus (148) Google Scholar) and (2Feller G. Payan F. Theys F. Qian M. Haser R. Gerday C. Eur. J. Biochem. 1994; 222: 441-447Crossref PubMed Scopus (195) Google Scholar) refer to calorimetric domains identified by deconvolution of DSC thermograms. Open table in a new tab Enthalpic stabilization involves an increase of the denaturation enthalpy and of the melting point. Such stabilization of the heat-labile α-amylase is obtained when a salt bridge (N150D), polar aromatic interactions (V196F), or an additional H-bond (K300R) is introduced in the protein. Thermal unfolding of some of these mutants is illustrated in Fig. 2. These additional electrostatic interactions induce a slightly higherT m value (Table II) associated, however, with a large increase of the calorimetric enthalpy (ΔH cal). Characterization of the double mutant N150D/V196F revealed several interesting features. The effect of the mutations on the melting point is additive (Table II, Fig. 2). Moreover, both mutations affectΔH cal andΔH eff differently, whereas their cumulative effect in the double mutant leads to aΔH cal/ΔH effratio at unity, indicative of a perfect two-state unfolding pattern. Aliphatic side chains were introduced in AHA in order to reconstruct hydrophobic core clusters found in mesophilic α-amylases (Q164I, T232V, N288V, and M379V). Interestingly, each substitution results in a non-two-state unfolding of the mutant enzymes and in the appearance of two calorimetric units or domains (Fig. 3). As also shown in Fig. 3, the separation between both calorimetric units in the mutants can be improved by adjusting the experimental conditions, providing evidence for the coexistence of two protein domains of distinct stability. Deconvolution of the heat capacity function revealed that one domain always has an increasedT m compared with that of AHA (Table II), underlining the stabilization induced by the reconstituted hydrophobic cluster. By contrast, the lower T m of the first domain seems to be the consequence of strain imposed in one domain that is not compensated in other protein regions. This is supported by the combination of mutations Q164I and V196F resulting in a mutant enzyme displaying two stabilized domains, as far as T m is concerned. The mutant T232V deserves special comments. Indeed, the shape of its heat capacity function is similar to that of PPA (Fig. 3) as well as the relative contribution of each domain toΔH cal andΔH eff (Table II). Therefore, this mutant consistently reproduces the microcalorimetric unfolding pattern of the heat-stable PPA but at lower temperatures. The comparison ofΔH cal andΔH eff values allows us to analyze the unfolding cooperativity by microcalorimetry. Several mutations listed in Table II affect the unfolding cooperativity without noticeable change of ΔH cal. Thermal unfolding of mutants Del6, L219R and R64E (Fig.4) clearly illustrates the relation between ΔH eff and the steepness of the transition. Furthermore, the thermogram slope of mutants L219R and R64E is considerably decreased and approaches the profile of large and stable proteins such as Bacillus amyloliquefaciensα-amylase (BAA) (Fig. 1). In these experiments, the mutant enzymes were 100% unfolded, then allowed to refold during a cooling period of 15 min at 15 °C. Reversibility was calculated by the recovery of ΔH calduring a second scan (Table II). Substitution of a single amino acid side chain is sufficient to drastically alter this critical parameter of protein folding, as demonstrated by mutant N12R that displays only 25% renaturation. Mutants Q164I, V196F, and Q164I/V196F that are stabilized by hydrophobic groups also decrease the unfolding reversibility significantly. In the same conditions, unfolding of the heat-stable PPA and BAA is completely irreversible. Determination of inactivation rate constants usually requires the recording of the residual enzyme activity after incubation at high temperature. In the case of AHA and of its mutants, the model of Lumry and Eyring (13Lumry R. Eyring H. J. Phys. Chem. 1954; 58: 110-120Crossref Scopus (729) Google Scholar) is valid: N⇌k−1k1U→k2I(Eq. 1) where k 2 is the first order rate constant of thermal inactivation. However, the various degrees of reversibility noted for the mutants (kinetically characterized byk −1) strongly impair the validity of the results obtained by conventional methods. We have therefore devised a new method using isothermal titration calorimetry and recording activity at 45 °C as the heat released by the hydrolysis of starch glycosidic bonds. This method provides a direct and continuous monitoring of activity at the denaturing temperature (Fig.5). As a prerequisite, the same experiment was followed by fluorescence (without substrate) showing that the structural modifications accompanying a temperature shift from 15 to 45 °C are fast with a k 1 value of about 0.04 s−1. Therefore, the transition N → U is one order of magnitude faster than the transition U → I, and the disappearance of the active state N can be taken as a measure ofk 2, as reported in Table II. It is significant that 10 of 14 mutants are protected against thermal inactivation and that mutants carrying newly introduced electrostatic interactions display the lowest inactivation rate constants (Fig. 5). A plot of unfolding reversibility versus thermal inactivation (Fig.6 A) shows that these parameters for the mutant enzymes are clustered between the respective values for AHA and PPA.Figure 6Correlation of stability (A) and kinetic (B) parameters for wild-type (○) and mutant (●) α-amylases.View Large Image Figure ViewerDownload Hi-res image Download (PPT) The kinetic parametersk cat and K m were recorded using a chromogenic maltoheptaose oligosaccharide (Table II). From these results, four main aspects should be pointed out. (i) As reported previously, AHA possesses a higher activity (k cat) than PPA in order to compensate for the slow chemical reaction rates at low temperature. (ii) The general trend of the mutations is to decrease both k cat andK m. Fig. 6 B shows that both kinetic parameters for mutants all fall within a narrow region of the plot, between the values of AHA and PPA. (iii) As a result, the mutations tend to raise the specificity constant or catalytic efficiencyk cat /K m to values close to that of PPA (Table II). (iv) Most mutations protecting the enzymes against thermal inactivation also concomitantly decrease thek cat value. The magnitude of the contribution, as recorded by microcalorimetry, of a single amino acid side chain to the stability parameters of a large protein is one of the most significant features revealed by engineering the heat-labile α-amylase. Such individual contributions are generally masked in stable proteins by the complex network of interactions involved in the native state conformation and by the fact that frequently, heat-induced unfolding of large proteins is kinetically driven as a result of pronounced irreversibility. Our results show that the selected electrostatic interactions provide the largest contribution to stability per se, by increasing both T mand ΔH cal and by protecting mutants against thermal inactivation (Table II). Thermal unfolding of the double mutant N150D/V196F follows the two-state transition model and is essentially reversible. Accordingly, its thermodynamic stability can be analyzed using the modified Gibbs-Helmholtz equation: ΔG(T)=ΔHm(1−T/Tm)+ΔCp(T−Tm)−TΔCp ln(T/Tm)(Eq. 2) This function corresponds to the energy required to disrupt the native state (14Privalov P. Creighton T. Protein Folding. W. H. Freeman and Company, New York1992: 83-126Google Scholar) and is illustrated in Fig.7. One can note that despite a modest increase of the melting point, the conformational energy of the double mutant is twice that of the wild-type AHA around 10 °C. Similarly, the contribution at 20 °C of the ion pair in N150D can be estimated to 4.4 kcal/mol, of the H-bond in K300R to 0.7 kcal/mol, and of a polar aromatic interaction in V196F (assuming two interplanar interactions) to 0.8 kcal/mol. These values are in good agreement with data obtained by ab initio calculations (15Burley S.K. Petsko G.A. Adv. Protein. Chem. 1988; 39: 125-189Crossref PubMed Scopus (801) Google Scholar) or by experimental studies (16Anderson D.E. Becktel W.J. Dahlquist F.W. Biochemistry. 1990; 29: 2403-2408Crossref PubMed Scopus (481) Google Scholar, 17Jaenicke R. Prog. Biophys. Mol. Biol. 1999; 71: 155-241Crossref PubMed Scopus (171) Google Scholar). These free energy gains originate, however, from very large enthalpic-entropic compensation effects. The denaturation enthalpy and entropy can be calculated according to the relations (18Privalov P.L. Adv. Protein. Chem. 1979; 33: 167-241Crossref PubMed Scopus (2203) Google Scholar):ΔH(T)=ΔHm+ΔCp(T−Tm)(Eq. 3) ΔS(T)=ΔHm/Tm+ΔCp ln(T/Tm)(Eq. 4) For instance, the denaturation enthalpy at 20 °C for AHA is 10.5 kcal/mol, and the entropic term (TΔS) amounts to 2.2 kcal/mol. The corresponding values for mutant N150D are 60.9 and 48.2 kcal/mol, respectively. Therefore, the gain of 4.4 kcal/mol in conformational energy provided by the ion pair arises from a 6-fold increase of the denaturation enthalpy balanced by a 20-fold increase of the entropic cost. It should be added that mutant N150D restores a surface salt bridge. Although the net contribution of solvent-exposed ion pairs has been frequently questioned (19Strop P. Mayo S.L. Biochemistry. 2000; 39: 1251-1255Crossref PubMed Scopus (151) Google Scholar), our results demonstrate that such interactions can provide a substantial increment of conformational stability. This is especially relevant for thermophilic and hyperthermophilic proteins that are characterized by an abundance of surface ion pairs (20Cambillau C. Claverie J.M. J. Biol. Chem. 2000; 275: 32383-32386Abstract Full Text Full Text PDF PubMed Scopus (249) Google Scholar), in some instances organized in interconnected networks (21Pappenberger G. Schurig H. Jaenicke R. J. Mol. Biol. 1997; 274: 676-683Crossref PubMed Scopus (91) Google Scholar, 22Vetriani C. Maeder D.L. Tolliday N. Yip K.S. Stillman T.J. Britton K.L. Rice D.W. Klump H.H. Robb F.T. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 12300-12305Crossref PubMed Scopus (243) Google Scholar). On the other hand, about 8 ion pairs, 15 arginine residues and 10 aromatic interactions are lacking in the crystal structure of AHA when compared with the heat-stable PPA (3Aghajari N. Feller G. Gerday C. Haser R. Structure. 1998; 6: 1503-1516Abstract Full Text Full Text PDF PubMed Scopus (188) Google Scholar). The disappearance of these electrostatic based features is obviously a main determinant for both the low T m and melting enthalpy of the psychrophilic α-amylase. It is generally agreed that the hydrophobic effect is the main factor in stabilizing folded conformations (23Matthews B.W. Annu. Rev. Biochem. 1993; 62: 139-160Crossref PubMed Scopus (462) Google Scholar). As shown in Fig. 3 and TableII, the insertion of a nonpolar group into a hydrophobic core cluster creates a stabilization center in the protein and in the peculiar case of AHA promotes the appearance of calorimetric domains. Unfortunately, large deviations from the two-state model preclude any reliable calculation for these mutants. It should be noted, however, that the increases of T m for the stabilized transition are of the same magnitude or even higher (M379V) than that provided by electrostatic interactions. Analysis of the hydrophobic clusters in AHA shows that 20 substitutions decrease the hydrophobicity index of the involved residues relative to PPA. As indicated by the calorimetric behavior of the mutations selected for this work (Fig. 3), the overall low core hydrophobicity of the heat-labile AHA accounts to a large extent for the lack of discrete unfolding intermediates. It is usually accepted that extreme cooperativity is only achieved in small molecules with tight packing of groups (14Privalov P. Creighton T. Protein Folding. W. H. Freeman and Company, New York1992: 83-126Google Scholar). This obviously is not the case for the large multidomain heat-labile α-amylase. As suggested by the mutation effects, its unfolding cooperativity arises from the very low number of interactions maintaining the native state, which therefore results in the simultaneous disruption of all structural elements and in the unusual two-state unfolding of this large protein. Considering protein adaptation to temperature, a practical implication of the flattening of the unfolding transition in thermostable α-amylases such as BAA (Fig. 1) is to maintain a higher level of native conformation at elevated temperatures. Fig. 4 shows that the same holds true between 45 and 55 °C for mutants of the heat-labile α-amylase with altered cooperativity. The relation between unfolding reversibility and thermal inactivation (Fig. 6 A) is another unsuspected result gained from the α-amylase mutant analysis. Referring to Equation 1, mutants protected against thermal inactivation are also those that refold less efficiently, mimicking the behavior of thermostable α-amylases. This could be explained by the large number of all types of weak interactions required for proper folding and stability in mesophilic and thermophilic enzymes. As a consequence, there will be more possibilities of side reactions during refolding, either by intramolecular mismatch or by aggregation due to replacement of the regular intramolecular pairing by intermolecular interactions (17Jaenicke R. Prog. Biophys. Mol. Biol. 1999; 71: 155-241Crossref PubMed Scopus (171) Google Scholar). It was shown that all 24 side chains in the active site performing H-bonding to a large pseudosaccharide inhibitor in the transition state conformation are strictly conserved in both AHA and PPA crystal structures (4Aghajari N. Feller G. Gerday C. Haser R. Protein Sci. 1998; 7: 564-572Crossref PubMed Scopus (161) Google Scholar, 24Qian M. Haser R. Buisson G. Duee E. Payan F. Biochemistry. 1994; 33: 6284-6294Crossref PubMed Scopus (285) Google Scholar). As a consequence, changes in primary structure that occur far from the active site should be responsible for the high specific activity of AHA. As a matter of fact, all mutations in this study designed to restore a mesophilic character are located far from the catalytic center and tend to decrease bothk cat and K m (Fig.6 B). It has been convincingly argued that increases in flexibility of cold-adapted enzymes lead to a broader distribution of conformational states, translated into higher K mvalues for enzymes devoid of adaptive changes within the active site (25Holland L.Z. McFall-Ngai M. Somero G.N. Biochemistry. 1997; 36: 3207-3215Crossref PubMed Scopus (101) Google Scholar, 26Fields P.A. Somero G.N. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 11476-11481Crossref PubMed Scopus (349) Google Scholar). In this context, rigidifying the molecule or part of the molecule by the newly engineered weak interactions in the psychrophilic α-amylase should contribute to decreased K m values by reducing the conformational entropy of the binding-competent species. The kinetic parameters of the mutants (Fig. 6 B) give experimental support to the previous hypothesis that temperature-adaptive increases in k cat occur concomitantly with increases in K m (26Fields P.A. Somero G.N. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 11476-11481Crossref PubMed Scopus (349) Google Scholar). It has been proposed that stability and activity are not physically linked (27Petrounia I.P. Arnold F.H. Curr. Opin. Biotechnol. 2000; 11: 325-330Crossref PubMed Scopus (149) Google Scholar). Laboratory evolution can indeed yield nonnatural catalysts that are thermostable and active either at high (28Cherry J.R. Lamsa M.H. Schneider P. Vind J. Svendsen A. Jones A. Pedersen A.H. Nat. Biotechnol. 1999; 17: 379-384Crossref PubMed Scopus (259) Google Scholar) or low (29Giver L. Gershenson A. Freskgard P.O. Arnold F.H. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 12809-12813Crossref PubMed Scopus (423) Google Scholar, 30Miyazaki K. Wintrode P.L. Grayling R.A. Rubingh D.N. Arnold F.H. J. Mol. Biol. 2000; 297: 1015-1026Crossref PubMed Scopus (214) Google Scholar) temperatures. Nevertheless, when the selection is only applied on low temperature activity (the relevant parameter for psychrophilic enzymes), especially by in vivo functional selection systems (allowing cell growth, for instance), the evolved enzymes generally display a concomitant reduced thermostability (31Lebbink J.H. Kaper T. Bron P. van der Oost J. de Vos W.M. Biochemistry. 2000; 39: 3656-3665Crossref PubMed Scopus (78) Google Scholar, 32Merz A. Yee M.C. Szadkowski H. Pappenberger G. Crameri A. Stemmer W.P. Yanofsky C. Kirschner K. Biochemistry. 2000; 39: 880-889Crossref PubMed Scopus (108) Google Scholar, 33Wintrode P.L. Miyazaki K. Arnold F.H. J. Biol. Chem. 2000; 275: 31635-31640Abstract Full Text Full Text PDF PubMed Scopus (124) Google Scholar, 34Roovers M. Sanchez R. Legrain C. Glansdorff N. J. Bacteriol. 2001; 183: 1101-1105Crossref PubMed Scopus (24) Google Scholar). It was assumed that the selective advantage for these optimized catalysts depends on a high k cat and not on maintaining stability (27Petrounia I.P. Arnold F.H. Curr. Opin. Biotechnol. 2000; 11: 325-330Crossref PubMed Scopus (149) Google Scholar). However, our structure-based approach demonstrates that natural mutations occurring in the cold-active α-amylase associate high activity and low stability. The convergence of the results from both random and rational studies stresses that the lack of selective pressure for stable proteins at low temperatures does not entirely account for the instability of psychrophilic proteins. On the contrary, heat lability or conformational plasticity seems to be intimately interlaced with sustained activity in cold environments. This is possibly the easiest route taken by natural selection on the evolutionary time scale or because large fluctuations around the native state are required by the energetics of substrate binding (35Ma B. Kumar S. Tsai C.J. Nussinov R. Protein Eng. 1999; 12: 713-720Crossref PubMed Scopus (486) Google Scholar) and catalysis at low temperatures, an aspect that is not documented for such enzymes to date. The mutant N288V represents an extreme case. This mutant possesses a heat-labile domain with the lowestT m (Table II) but is marginally stable even at 4 °C, with a half-life of 2 h at 25 °C and requiring fast purification procedures. This behavior is consistent with the proposal that AHA has reached a state close to the lowest accessible stability of its native state and cannot be further destabilized (1Feller G. d'Amico D. Gerday C. Biochemistry. 1999; 38: 4613-4619Crossref PubMed Scopus (148) Google Scholar). If indeed stability is linked to activity in cold-adapted enzymes, this limitation explains the imperfect adaptation of psychrophilic enzymes,i.e. their activity at low temperatures remains lower than that of mesophiles at 37 °C. This mutational analysis of a large protein demonstrates that the psychrophilic α-amylase has lost numerous weak interactions during evolution to reach the proper conformational flexibility at low temperatures. These adaptive adjustments contribute to improve k cat without alteration of the catalytic mechanism, because the active site architecture is not modified, but at the expense of a weaker substrate binding affinity. On the other hand, thermophilic enzymes strengthen the same type of interaction to gain in structural stability at high temperatures but at the expense of poor activity at room temperature (36Varley P.G. Pain R.H. J. Mol. Biol. 1991; 220: 531-538Crossref PubMed Scopus (104) Google Scholar, 37Zavodszky P. Kardos J. Svingor A. Petsko G.A. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 7406-7411Crossref PubMed Scopus (494) Google Scholar, 38Kohen A. Cannio R. Bartolucci S. Klinman J.P. Nature. 1999; 399: 496-499Crossref PubMed Scopus (504) Google Scholar). These aspects underline the continuum in the strategy of temperature adaptation in proteins and reinforce the concepts of "compromise" (39Somero G.N. Annu. Rev. Physiol. 1995; 57: 43-68Crossref PubMed Scopus (657) Google Scholar) between activity and stability leading to "corresponding states" (40Jaenicke R. Bohm G. Curr. Opin. Struct. Biol. 1998; 8: 738-748Crossref PubMed Scopus (688) Google Scholar) of enzymes adapted to different thermal environments. We thank N. Gérardin and R. Marchand for their skillful technical assistance.