Increased folding stability of TEM-1 beta-lactamase by in vitro selection

In vitro selections of stabilized proteins lead to more robust enzymes and, at the same time, yield novel insights into the principles of protein stability. We employed Proside, a method of in vitro selection, to find stabilized variants of TEM-1 β-lactamase from Escherichia coli. Proside links the increased protease resistance of stabilized proteins to the infectivity of a filamentous phage. Several libraries of TEM-1 β-lactamase variants were generated by error-prone PCR, and variants with increased protease resistance were obtained by raising temperature or guanidinium chloride concentration during proteolytic selections. Despite the small size of phage libraries, several strongly stabilizing mutations could be obtained, and a manual combination of the best shifted the profiles for thermal unfolding and temperature-dependent inactivation of β-lactamase by almost 20 °C to a higher temperature. The wild-type protein unfolds in two stages: from the native state via an intermediate of the molten-globule type to the unfolded form. In the course of the selections, the native protein was stabilized by 27 kJ mol^− 1 relative to the intermediate and the cooperativity of unfolding was strongly increased. Three of our stabilizing replacements (M182T, A224V, and R275L) had been identified independently in naturally occurring β-lactamase variants with extended substrate spectrum. In these variants, they acted as global suppressors of destabilizations caused by the mutations in the active site. The comparison between the crystal structure of our best variant and the crystal structure of the wild-type protein indicates that most of the selected mutations optimize helices and their packing. The stabilization by the E147G substitution is remarkable. It removes steric strain that originates from an overly tight packing of two helices in the wild-type protein. Such unfavorable van der Waals repulsions are not easily identified in crystal structures or by computational approaches, but they strongly reduce the conformational stability of a protein.     

Optimization of the gbeta1 domain by computational design and by in vitro evolution: structural and energetic basis of stabilization

Computational design and in vitro evolution are major strategies for stabilizing proteins. For the four critical positions 16, 18, 25, and 29 of the B domain of the streptococcal protein G (Gbeta1), they identified the same optimal residues at positions 16 and 25, but not at 18 and 29. Here we analyzed the energetic contributions of the residues from these two approaches by single and double mutant analyses and determined crystal structures for a variant from the calculation (I16/L18/E25/K29) and from the selection (I16/I18/E25/F29). The structural analysis explains the observed differences in stabilization. Residues 16, 18, and 29 line an invagination, which results from a packing defect between the helix and the beta-sheet of Gbeta1. In all stabilized variants, residues with larger side-chains occur at these positions and packing is improved. In the selected variant, packing is better optimized than in the computed variant. Such differences in side-chain packing strongly affect stability but are difficult to evaluate by computation.     

Evolutionary protein stabilization in comparison with computational design

Two major strategies are currently used for stabilizing proteins: in vitro evolution and computational design. Here, we used gene libraries of the beta1 domain of the streptococcal protein G (Gbeta1) and Proside, an in vitro selection method, to identify stabilized variants of this protein. In the Gbeta1 libraries, the codons for the four boundary positions 16, 18, 25, and 29 were randomized. Many Gbeta1 variants with strongly increased thermal stabilities were found in 11 selections performed with five independent libraries. Previously, Mayo and co-workers used computational design to stabilize Gbeta1 by sequence optimization at the same positions. Their best variant ranked third within the panel of the selected variants. None of the ten computed sequences was found in the Proside selections, because several computed residues for positions 18 and 29 were not optimal for stability.     

Optimized variants of the cold shock protein from in vitro selection: structural basis of their high thermostability

The bacterial cold shock proteins (Csp) are widely used as models for the experimental and computational analysis of protein stability. In a previous study, in vitro evolution was employed to identify strongly stabilizing mutations in Bs-CspB from Bacillus subtilis. The best variant found by this approach contained the mutations M1R, E3K and K65I, which raised the midpoint of thermal unfolding of Bs-CspB from 53.8 degrees C to 83.7 degrees C, and increased the Gibbs free energy of stabilization by 20.9 kJ mol(-1). Another selected variant with the two mutations A46K and S48R was stabilized by 11.1 kJ mol(-1). To elucidate the molecular basis of these stabilizations, we determined the crystal structures of these two Bs-CspB variants. The mutated residues are generally well ordered and provide additional stabilizing interactions, such as charge interactions, additional hydrogen bonds and improved side-chain packing. Several mutations improve the electrostatic interactions, either by the removal of unfavorable charges (E3K) or by compensating their destabilizing interactions (A46K, S48R). The stabilizing mutations are clustered at a contiguous surface area of Bs-CspB, which apparently is critically important for the stability of the beta-barrel structure but not well optimized in the wild-type protein.     

Dimer Formation of a Stabilized Gβ1 Variant: A Structural and Energetic Analysis

In previous work, a strongly stabilized variant of the β1 domain of streptococcal protein G (Gβ1) was obtained by an in vitro selection method. This variant, termed Gβ1-M2, contains the four substitutions E15V, T16L, T18I, and N37L. Here we elucidated the molecular basis of the observed strong stabilizations. The contributions of these four residues were analyzed individually and in various combinations, additional selections with focused Gβ1 gene libraries were performed, and the crystal structure of Gβ1-M2 was determined. All single substitutions (E15V, T16L, T18I, and N37L) stabilize wild-type Gβ1 by contributions of between 1.6 and 6.0 kJ mol^− 1 (at 70 °C). Hydrophobic residues at positions 16 and 37 provide the major contribution to stabilization by enlarging the hydrophobic core of Gβ1. They also increase the tendency to form dimers, as shown by dependence on the concentration of apparent molecular mass in analytical ultracentrifugation, by concentration-dependent stability, and by a strongly increased van't Hoff enthalpy of unfolding. The 0.88-Å crystal structure of Gβ1-M2 and NMR measurements in solution provide the explanation for the observed dimer formation. It involves a head-to-head arrangement of two Gβ1-M2 molecules via six intermolecular hydrogen bonds between the two β strands 2 and 2′ and an adjacent self-complementary hydrophobic surface area, which is created by the T16L and N37L substitutions and a large 120° rotation of the Tyr33 side chain. This removal of hydrophilic groups and the malleability of the created hydrophobic surface provide the basis for the dimer formation of stabilized Gβ1 variants.     

Key role of coulombic interactions for the folding transition state of the cold shock protein

The cold shock protein CspB shows a five-stranded beta-sheet structure, and it folds rapidly via a native-like transition state. A previous Phi value analysis showed that most of the residues with Phi values close to one reside in strand beta1, and two of them, Lys5 and Lys7 are partially exposed charged residues. To elucidate how coulombic interactions of these two residues contribute to the energetic organisation of the folding transition state we performed comparative folding experiments in the presence of an ionic denaturant (guanidinium chloride) and a non-ionic denaturant (urea) and a double-mutant analysis. Lys5 contributes 6.6 kJ mol(-1) to the stability of the transition state, and half of it originates from screenable coulombic interactions. Lys7 contributes 5.3 kJ mol(-1), and 3.4 kJ mol(-1) of it are screened by salt. In the folded protein Lys7 interacts with Asp25, and the screenable coulombic interaction between these two residues is fully formed in the transition state. This suggests that long-range coulombic interactions such as those originating from Lys5 and Lys7 of CspB can be important for organizing and stabilizing native-like structure early in protein folding.     

Stabilization of the cold shock protein CspB from Bacillus subtilis by evolutionary optimization of Coulombic interactions

The bacterial cold shock proteins (Csp) are used by both experimentalists and theoreticians as model systems for analyzing the Coulombic contributions to protein stability. We employ Proside, a method of directed evolution, to identify stabilized variants of Bs-CspB from Bacillus subtilis. Proside links the increased protease resistance of stabilized protein variants to the infectivity of a filamentous phage. Here, three cspB libraries were used for in vitro selections to explore the stabilizing potential of charged amino acids in Bs-CspB. In the first library codons for nine selected surface residues were partially randomized, in the second one random mutations were introduced non-specifically by error-prone PCR, and in the third one the spontaneous mutation rate of the phage in Escherichia coli was used. Stabilizing mutations were found at the surface positions 1, 3, 46, 48, 65, and 66. The contributions of these mutations to stability were characterized by analyzing them individually and in combination. The best combination (M1R, E3K, K65I, and E66L) increased the midpoint of thermal unfolding of Bs-CspB from 53.8 to 85.0 degrees C. The effects of most mutations are strongly context dependent. A good example is provided by the E3R mutation. It is strongly stabilizing (DeltaDeltaGD=11.1kJ mol(-1)) in the wild-type protein, but destabilizing (DeltaDeltaGD=-4.0kJ mol(-1)) in the A46K/S48R/E66L variant. The stabilizations by charge mutations did not correlate well with the corresponding changes in the protein net charge, and they could not be ascribed to the formation of ion pairs. Previous theoretical analyses did not identify the stabilization caused by the mutations at positions 1, 46, and 48. Also, electrostatics calculations based on protein net charge or charge asymmetry did not predict well the stability changes that occur when charged residues in Bs-CspB are mutated. It remains a challenge to model the Coulombic interactions of charged residues in a protein and to determine their contributions to the Gibbs free energy of protein folding.     

Origins of the high stability of an in vitro-selected cold-shock protein

In previous work, we had identified stabilized forms of the cold-shock protein Bs-CspB from Bacillus subtilis in a combinatorial library by an in vitro selection procedure. In this library, the sequence positions 2, 3, 46, 64, 66, and 67 had been randomized, because Bs-CspB differs from the naturally thermostable homolog Bc-Csp from Bacillus caldolyticus, among others, at these six positions. For the most stable selected variant, the midpoint of thermal unfolding (tM) increased by 28.2 deg. C and the Gibbs free energy of unfolding (deltaG(D)) by 19 kJ/mol. Here, we analyzed by site-directed mutagenesis how the selected residues contribute individually to this strong stabilization. Val3 and Val66, which replace Glu3 and Glu66 of wild-type Bs-CspB, each contribute about 7 kJ/mol to stability, the Thr64Arg substitution contributes 4.5 kJ/mol, and 3.2 kJ/mol originate from the Ala46Leu replacement. Gly67 at the carboxy terminus is unimportant for stability, the Arg selected at position 2 is overall slightly destabilizing but improves the coulombic interactions. The best variant differs from Bc-Csp at all six positions; nevertheless, natural and in vitro selection followed similar principles. In both cases, negatively charged residues at the adjacent positions 3 and 66 are avoided, and a positively charged residue is introduced into this area of the protein surface. Its exact location is unimportant. It can be at position 3, as in the thermophilic Bc-Csp, or at positions 2 or 64, as in the most stable selected variant. These positively charged residues contribute to stability not by engaging in pairwise coulombic interactions with a specific carboxyl group, but by generally improving the charge distribution in this particular region of the protein surface. These coulombic effects contribute significantly to the thermostability of the cold-shock proteins. They are only weakly interdependent and best explained by the presence of a flexible ion network at the protein surface. Our results emphasize that surface positions are very good candidates for optimizing protein stability.     

In-vitro selection of highly stabilized protein variants with optimized surface

Thermostable proteins are of prime importance in protein science, but it has remained difficult to develop general strategies for stabilizing a protein. Site-directed mutagenesis based on comparisons with thermophilic homologs is rarely successful because the sequence differences are too numerous and dominated by neutral mutations. Here we used a method of directed evolution to increase the stability of a mesophilic protein, the cold shock protein Bs-CspB from Bacillus subtilis. It differs from its thermophilic counterpart Bc-Csp from Bacillus caldolyticus at 12 surface-exposed positions. To elucidate the stabilizing potential of exposed amino acid residues, six of these variant positions were randomized by saturation mutagenesis, the corresponding library of sequences was inserted into the gene-3-protein of the filamentous phage fd, and stabilized variants were selected by the Proside technique. Proside links the increased protease resistance of stabilized protein variants with the infectivity of the phage. Many strongly stabilized variants of Bs-CspB were identified in two selections, one in the presence of a denaturant and the other at elevated temperature. Several of them are significantly more stable than the naturally thermostable homolog Bc-Csp, and the best variant reaches Tm-Csp (the homolog from the hyperthermophile Thermotoga maritima) in stability. Remarkably, this variant differs from Tm-Csp at five and from Bc-Csp at all six randomized positions. This indicates that proteins can be strongly stabilized by many different sets of surface mutations, and Proside selects them efficiently from large libraries. The course of the selection could be directed by the conditions. In an ionic denaturant non-polar surface interactions were optimized, whereas at elevated temperature variants with improved electrostatics were selected, pointing to two different strategies for stabilization at protein surfaces.     

Thermodynamic properties of BPTI variants with highly simplified amino acid sequences

We report the first detailed thermodynamic analysis of simplified proteins by differential scanning calorimetry (DSC). The experiments were carried out with five simplified BPTI variants, whose structures and activities have been reported, in which several residues not essential for specifying the tertiary structure were replaced by alanine. In most aspects, the thermodynamics of simplified proteins were very similar to, if not essentially identical with, those of natural proteins. In particular, they undergo a highly cooperative two-state thermal unfolding process with a large enthalpy change, which is a thermodynamic hallmark of the native state of natural globular proteins. Furthermore, the specific enthalpy and entropy changes upon unfolding at 110 degrees C were close to values invariably observed for small natural globular proteins (55 J g(-1) and ~16 J K(-1) g(-1), respectively). On the other hand, two simplified BPTI variants, BPTI-21 and BPTI-22 (containing 21 and 22 alanine residues), were enthalpically stabilized while entropically destabilized with respect to the reference BPTI-[5,55] molecule. This peculiar type of entropy-enthalpy compensation is in sharp contrast to the usual enthalpy destabilization/entropy stabilization observed in mutational studies of natural proteins. Overall, we conclude that a thermodynamic native state can be achieved by proteins encoded with extensively simplified sequences.     

Changing the determinants of protein stability from covalent to non-covalent interactions by in vitro evolution: a structural and energetic analysis

The three disulfide bonds of the gene-3-protein of the phage fd are essential for the conformational stability of this protein, and it unfolds when they are removed by reduction or mutation. Previously, we used an iterative in vitro selection strategy to generate a stable and functional form of the gene-3-protein without these disulfides. It yielded optimal replacements for the disulfide bonds as well as several stabilizing second-site mutations. The best selected variant showed a higher thermal stability compared with the disulfide-bonded wild-type protein. Here, we investigated the molecular basis of this strong stabilization by solving the crystal structure of this variant and by analyzing the contributions to the conformational stability of the selected mutations individually. They could mostly be explained by improved side-chain packing. The R29W substitution alone increased the midpoint of the thermal unfolding transition by 14 deg and the conformational stability by about 25 kJ mol^− 1. This key mutation (i) removed a charged side chain that forms a buried salt bridge in the disulfide-containing wild-type protein, (ii) optimized the local packing with the residues that replace the C46-C53 disulfide and (iii) improved the domain interactions. Apparently, certain residues in proteins indeed play key roles for stability.     

Consequences of Domain Insertion on the Stability and Folding Mechanism of a Protein

SlyD, the sensitive-to-lysis protein from Escherichia coli, consists of two domains. They are not arranged successively along the protein chain, but one domain, the “insert-in-flap” (IF) domain, is inserted internally as a guest into a surface loop of the host domain, which is a prolyl isomerase of the FK506 binding protein (FKBP) type. We used SlyD as a model to elucidate how such a domain insertion affects the stability and folding mechanism of the host and the guest domain. For these studies, the two-domain protein was compared with a single-domain variant SlyDΔIF, SlyD* without the chaperone domain (residues 1-69 and 130-165) in which the IF domain was removed and replaced by a short loop, as present in human FKBP12. Equilibrium unfolding and folding kinetics followed an apparent two-state mechanism in the absence and in the presence of the IF domain. The inserted domain decreased, however, the stability of the host domain in the transition region and decelerated its refolding reaction by about 10-fold. This originates from the interruption of the chain connectivity by the IF domain and its inherent instability. To monitor folding processes in this domain selectively, a Trp residue was introduced as fluorescent probe. Kinetic double-mixing experiments revealed that, in intact SlyD, the IF domain folds and unfolds about 1000-fold more rapidly than the FKBP domain, and that it is strongly stabilized when linked with the folded FKBP domain. The unfolding limbs of the kinetic chevrons of SlyD show a strong downward curvature. This deviation from linearity is not caused by a transition-state movement, as often assumed, but by the accumulation of a silent unfolding intermediate at high denaturant concentrations. In this kinetic intermediate, the FKBP domain is still folded, whereas the IF domain is already unfolded.     

Stability enhancement of cytochrome c through heme deprotonation and mutations

The chemical denaturation of Pseudomonas aeruginosa cytochrome c(551) variants was examined at pH 5.0 and 3.6. All variants were stabilized at both pHs compared with the wild-type. Remarkably, the variants carrying the F34Y and/or E43Y mutations were more stabilized than those having the F7A/V13M or V78I ones at pH 5.0 compared with at pH 3.6 by ~3.0-4.6 kJ/mol. Structural analyses predicted that the side chains of introduced Tyr-34 and Tyr-43 become hydrogen donors for the hydrogen bond formation with heme 17-propionate at pH 5.0, but less efficiently at pH 3.6, because the propionate is deprotonated at the higher pH. Our results provide an insight into a stabilization strategy for heme proteins involving variation of the heme electronic state and introduction of appropriate mutations.     

Energetic coupling between native-state prolyl isomerization and conformational protein folding

In folded proteins, prolyl peptide bonds are usually thought to be either trans or cis because only one of the isomers can be accommodated in the native folded protein. For the N-terminal domain of the gene-3 protein of the filamentous phage fd (N2 domain), Pro161 resides at the tip of a beta hairpin and was found to be cis in the crystal structure of this protein. Here we show that Pro161 exists in both the cis and the trans conformations in the folded form of the N2 domain. We investigated how conformational folding and prolyl isomerization are coupled in the unfolding and refolding of N2 domain. A combination of single-mixing and double-mixing unfolding and refolding experiments showed that, in unfolded N2 domain, 7% of the molecules contain a cis-Pro161 and 93% of the molecules contain a trans-Pro161. During refolding, the fraction of molecules with a cis-Pro161 increases to 85%. This implies that 10.3 kJ mol(-1) of the folding free energy was used to drive this 75-fold change in the Pro161 cis/trans equilibrium constant during folding. The stabilities of the forms with the cis and the trans isomers of Pro161 and their folding kinetics could be determined separately because their conformational folding is much faster than the prolyl isomerization reactions in the native and the unfolded proteins. The energetic coupling between conformational folding and Pro161 isomerization is already fully established in the transition state of folding, and the two isomeric forms are thus truly native forms. The folding kinetics are well described by a four-species box model, in which the N2 molecules with either isomer of Pro161 can fold to the native state and in which cis/trans isomerization occurs in both the unfolded and the folded proteins.     

Evolutionary stabilization of the gene-3-protein of phage fd reveals the principles that govern the thermodynamic stability of two-domain proteins

The gene-3-protein (G3P) of filamentous phage is essential for their propagation. It consists of three domains. The CT domain anchors G3P in the phage coat, the N2 domain binds to the F pilus of Escherichia coli and thus initiates infection, and the N1 domain continues by interacting with the TolA receptor. Phage are thus only infective when the three domains of G3P are tightly linked, and this requirement is exploited by Proside, an in vitro selection method for proteins with increased stability. In Proside, a repertoire of variants of the protein to be stabilized is inserted between the N2 and the CT domains of G3P. Stabilized variants can be selected because they resist cleavage by a protease and thus maintain the essential linkage between the domains. The method is limited by the proteolytic stability of G3P itself. We improved the stability of G3P by subjecting the phage without a guest protein to rounds of random in vivo mutagenesis and proteolytic Proside selections. Variants of G3P with one to four mutations were selected, and the temperature at which the corresponding phage became accessible for a protease increased in a stepwise manner from 40 degrees C to almost 60 degrees C. The N1-N2 fragments of wild-type gene-3-protein and of the four selected variants were purified and their stabilities towards thermal and denaturant-induced unfolding were determined. In the biphasic transitions of these proteins domain dissociation and unfolding of N2 occur in a concerted reaction in the first step, followed by the independent unfolding of domain N1 in the second step. N2 is thus less stable than N1, and it unfolds when the interactions with N1 are broken. The strongest stabilizations were caused by mutations in domain N2, in particular in its hinge subdomain, which provides many stabilizing interactions between the N1 and N2 domains. These results reveal how the individual domains and their assembly contribute to the overall stability of two-domain proteins and how mutations are optimally placed to improve the stability of such proteins.     

Electrostatic interactions in leucine zippers: thermodynamic analysis of the contributions of Glu and His residues and the effect of mutating salt bridges

Electrostatic interactions play a complex role in stabilizing proteins. Here, we present a rigorous thermodynamic analysis of the contribution of individual Glu and His residues to the relative pH-dependent stability of the designed disulfide-linked leucine zipper AB(SS). The contribution of an ionized side-chain to the pH-dependent stability is related to the shift of the pK(a) induced by folding of the coiled coil structure. pK(a)(F) values of ten Glu and two His side-chains in folded AB(SS) and the corresponding pK(a)(U) values in unfolded peptides with partial sequences of AB(SS) were determined by 1H NMR spectroscopy: of four Glu residues not involved in ion pairing, two are destabilizing (-5.6 kJ mol(-1)) and two are interacting with the positive alpha-helix dipoles and are thus stabilizing (+3.8 kJ mol(-1)) in charged form. The two His residues positioned in the C-terminal moiety of AB(SS) interact with the negative alpha-helix dipoles resulting in net stabilization of the coiled coil conformation carrying charged His (-2.6 kJ mol(-1)). Of the six Glu residues involved in inter-helical salt bridges, three are destabilizing and three are stabilizing in charged form, the net contribution of salt-bridged Glu side-chains being destabilizing (-1.1 kJ mol(-1)). The sum of the individual contributions of protonated Glu and His to the higher stability of AB(SS) at acidic pH (-5.4 kJ mol(-1)) agrees with the difference in stability determined by thermal unfolding at pH 8 and pH 2 (-5.3 kJ mol(-1)). To confirm salt bridge formation, the positive charge of the basic partner residue of one stabilizing and one destabilizing Glu was removed by isosteric mutations (Lys-->norleucine, Arg-->norvaline). Both mutations destabilize the coiled coil conformation at neutral pH and increase the pK(a) of the formerly ion-paired Glu side-chain, verifying the formation of a salt bridge even in the case where a charged side-chain is destabilizing. Because removing charges by a double mutation cycle mainly discloses the immediate charge-charge effect, mutational analysis tends to overestimate the overall energetic contribution of salt bridges to protein stability.     

Downhill versus Barrier-Limited Folding of BBL 1: Energetic and Structural Perturbation Effects upon Protonation of a Histidine of Unusually Low pKa

A dispersion of melting temperatures at pH 5.3 for individual residues of the BBL protein domain has been adduced as evidence for barrier-free downhill folding. Other members of the peripheral subunit domain family fold cooperatively at pH 7. To search for possible causes of anomalies in BBL's denaturation behavior, we measured the pH titration of individual residues by heteronuclear NMR. At 298 K, the pK_a of His142 was close to that of free histidine at 6.47 ± 0.04, while that of the more buried His166 was highly perturbed at 5.39 ± 0.02. Protonation of His166 is thus energetically unfavorable and destabilizes the protein by ∼ 1.5 kcal/mol. Changes in C^α secondary shifts at pH 5.3 showed a decrease in helicity of the C-terminus of helix 2, where His166 is located, which was accompanied by a measured decrease of 1.1 ± 0.2 kcal/mol in stability from pH 7 to 5.3. Protonation of His166 perturbs, therefore, the structure of BBL. Only ∼ 1% of the structurally perturbed state will be present at the biologically relevant pH 7.6. Experiments at pH 5.3 report on a near-equal mixture of the two different native states. Further, at this pH, small changes of pH and pK_a induced by changes in temperature will have near-maximal effects on pH-dependent conformational equilibria and on propagation of experimental error. Accordingly, conventional barrier-limited folding predicts some dispersion of measured thermal unfolding curves of individual residues at pH 5.3.     

Thermodynamically stable aggregation-resistant antibody domains through directed evolution

Protein aggregates are usually formed by interactions between unfolded or partially unfolded species, and often occur when a protein is denatured by, for example, heat or low pH. In earlier work, we used a Darwinian selection strategy to create human antibody variable domains that resisted heat aggregation. The repertoires of domains were displayed on filamentous phage and denatured (at 80 °C in pH 7.4), and folded domains were selected by binding to a generic ligand after cooling. This process appeared to select for domains with denatured states that resisted aggregation, but the domains only had low free energies of folding (ΔG_N-D^o = 15-20 kJ/mol at 25 °C in pH 7.4). Here, using the same phage repertoire, we have extended the method to the selection of domains resistant to acid aggregation. In this case, however, the thermodynamic stabilities of selected domains were higher than those selected by thermal denaturation (under both neutral and acidic conditions; ΔG_N-D^o = 26-47 kJ/mol at 25 °C in pH 7.4, or ΔG_N-D^o = 27-34 kJ/mol in pH 3.2). Furthermore, we identified a key determinant (Arg28) that increased the aggregation resistance of the denatured states of the domains at low pH without compromising their thermodynamic stabilities. Thus, the selection process yielded domains that combined thermodynamic stability and aggregation-resistant unfolded states. We suggest that changes to these properties are controlled by the extent to which the folding equilibrium is displaced during the process of selection.