Charge-charge interactions influence the denatured state ensemble and contribute to protein stability

Several recent studies have shown that it is possible to increase protein stability by improving electrostatic interactions among charged groups on the surface of the folded protein. However, the stability increases are considerably smaller than predicted by a simple Coulomb's law calculation, and in some cases, a charge reversal on the surface leads to a decrease in stability when an increase was predicted. These results suggest that favorable charge-charge interactions are important in determining the denatured state ensemble, and that the free energy of the denatured state may be decreased more than that of the native state by reversing the charge of a side chain. We suggest that when the hydrophobic and hydrogen bonding interactions that stabilize the folded state are disrupted, the unfolded polypeptide chain rearranges to compact conformations with favorable long-range electrostatic interactions. These charge-charge interactions in the denatured state will reduce the net contribution of electrostatic interactions to protein stability and will help determine the denatured state ensemble. To support this idea, we show that the denatured state ensemble of ribonuclease Sa is considerably more compact at pH 7 where favorable charge-charge interactions are possible than at pH 3, where unfavorable electrostatic repulsion among the positive charges causes an expansion of the denatured state ensemble. Further support is provided by studies of the ionic strength dependence of the stability of charge-reversal mutants of ribonuclease Sa. These results may have important implications for the mechanism of protein folding.     

Amino acid contribution to protein solubility: Asp, Glu, and Ser contribute more favorably than the other hydrophilic amino acids in RNase Sa

Poor protein solubility is a common problem in high-resolution structural studies, formulation of protein pharmaceuticals, and biochemical characterization of proteins. One popular strategy to improve protein solubility is to use site-directed mutagenesis to make hydrophobic to hydrophilic mutations on the protein surface. However, a systematic investigation of the relative contributions of all 20 amino acids to protein solubility has not been done. Here, 20 variants at the completely solvent-exposed position 76 of ribonuclease (RNase) Sa are made to compare the contributions of each amino acid. Stability measurements were also made for these variants, which occur at the i+1 position of a type II beta-turn. Solubility measurements in ammonium sulfate solutions were made at high positive net charge, low net charge, and high negative net charge. Surprisingly, there was a wide range of contributions to protein solubility even among the hydrophilic amino acids. The results suggest that aspartic acid, glutamic acid, and serine contribute significantly more favorably than the other hydrophilic amino acids especially at high net charge. Therefore, to increase protein solubility, asparagine, glutamine, or threonine should be replaced with aspartic acid, glutamic acid or serine.     

The pH dependence of hydrogen exchange in proteins

The static accessibility modified discrete charge model for electrostatic interactions in proteins is extended to the prediction of the pH dependence of hydrogen exchange reactions. The exchange rate profiles of buried amide protons are shown to follow the calculated pH dependence of the electrostatic component of protein stability. Rate profiles are calculated for individual buried amide protons in ribonuclease S and bovine pancreatic trypsin inhibitor. The electrostatic free energy of stabilization of the protein and the energy required to bring the catalytic ion to an exchange site are expressed as an apparent, pH-dependent contribution to the activation energy. Changes in the electrostatic stabilization of the proteins affect the calculated exchange rate for buried amide protons by more than 1000, while local field effects raise or lower the predicted exchange rates by less than 100. The pH dependence of exchangeable protons at the protein surface, such as the C-2 imidazole protons, is shown to follow the estimated energy required to introduce the catalytic ion at the exchange site. These calculations are discussed in terms of current models for proton exchange which incorporate the dynamic nature of the structure to explain exchange data from the interior of a protein.     

A new member of the bacterial ribonuclease inhibitor family from Saccharopolyspora erythraea

We have identified Sti, the gene of a ribonuclease inhibitor from Saccharopolyspora erythraea, by using a T7 phage display system. A specific phage has been isolated from a genome library by a biopanning procedure, using RNase Sa3, a ribonuclease from Streptomyces aureofaciens, as bait. Sti, a protein of 121 amino acid residues, with molecular mass 13059 Da, is a homolog of barstar and other microbial ribonuclease inhibitors. To overexpress its gene in Escherichia coli, we optimized the secondary structure of its mRNA by introducing a series of silent mutations. Soluble protein was isolated and purified to homogeneity. Inhibition constants of complex of Sti and RNase Sa3 or barnase were determined at pH 7 as 5 x 10(-12) or 7 x 10(-7), respectively.     

How optimal are the binding energetics of barnase and barstar?

The extracellular ribonuclease barnase and its intracellular inhibitor barstar bind fast and with high affinity. Although extensive experimental and theoretical studies have been carried out on this system, it is unclear what the relative importance of different contributions to the high affinity is and whether binding can be improved through point mutations. In this work, we first applied Poisson-Boltzmann electrostatic calculations to 65 barnase-barstar complexes with mutations in both barnase and barstar. The continuum electrostatic calculations with a van der Waals surface dielectric boundary definition result in the electrostatic interaction free energy providing the dominant contribution favoring barnase-barstar binding. The results show that the computed electrostatic binding free energy can be improved through mutations at W44/barstar and E73/barnase. Furthermore, the determinants of binding affinity were quantified by applying COMparative BINding Energy (COMBINE) analysis to derive quantitative structure-activity relationships (QSARs) for the 65 complexes. The COMBINE QSAR model highlights approximately 20 interfacial residue pairs as responsible for most of the differences in binding affinity between the mutant complexes, mainly due to electrostatic interactions. Based on the COMBINE model, together with Brownian dynamics simulations to compute diffusional association rate constants, several mutants were designed to have higher binding affinities than the wild-type proteins.     

Preferential interactions determine protein solubility in three-component solutions: the MgCl2 system

The correlation between protein solubility and the preferential interactions of proteins with solvent components was critically examined with aqueous MgCl2 as the solvent system. Preferential interaction and solubility measurements with three proteins, beta-lactoglobulin, bovine serum albumin, and lysozyme, resulted in similar patterns of interaction. At acid pH (pH 2-3) and lower salt concentrations (less than 2 M), the proteins were preferentially hydrated, while at higher salt concentrations, the interaction was either that of preferential salt binding or low salt exclusion. At pH 4.5-5, all three proteins exhibited either very low preferential hydration or preferential binding of MgCl2. These results were analyzed in terms of the balance between salt binding and salt exclusion attributed to the increase in the surface tension of water by salts, which is invariant with conditions. It was shown that the increase in salt binding at high salt concentration is a reflection of mass action, while its decrease at acid pH is due to the electrostatic repulsion between Mg2+ ions and the high net positive charge on the protein. The preferential interaction pattern was paralleled by the variation of protein solubility with solvent conditions. Calculation of the transfer free energies from water to the salt solutions for proteins in solution and in the precipitate showed dependencies on salt concentration. This indicates that the nature of interactions between proteins and solvent components is the same in solution and in the solid state, which implies no change in protein structure during precipitation. Analysis of the transfer free energies and preferential interaction parameter in terms of the salting-in, salting-out, and weak ion binding contributions has led to the conclusions that, when the weak ion binding contribution is small, the predominant protein-salt interaction must be that of preferential salt exclusion most probably caused by the increase of the surface tension of water by addition of the salt. A necessary consequence of this is salting-out of the protein, if the protein structure is to remain unaltered.     

Temperature dependence of the preferential interactions of ribonuclease A in aqueous co-solvent systems: thermodynamic analysis.

The temperature dependence of preferential solvent interactions with ribonuclease A in aqueous solutions of 30% sorbitol, 0.6 M MgCl2, and 0.6 M MgSO4 at low pH (1.5 and 2.0) and high pH (5.5) has been investigated. This protein was stabilized by all three co-solvents, more so at low pH than high pH (expect 0.6 M MgCl2 at pH 5.5). The preferential hydration of protein in all three co-solvents was high at temperatures below 30 degrees C and decreased with a further increase in temperature (for 0.6 M MgCl2 at pH 5.5, this was not significant), indicating a greater thermodynamic instability at low temperature than at high temperature. The preferential hydration of denatured protein (low pH, high temperature) was always greater than that of native protein (high pH, high temperature). In 30% sorbitol, the interaction passed to preferential binding at 45% for native ribonuclease A and at 55 degrees C for the denatured protein. Availability of the temperature dependence of the variation with sorbitol concentration of the chemical potential of the protein, (delta mu(2)/delta m3)T,p,m2, permitted calculation of the corresponding enthalpy and entropy parameters. Combination with available data on sorbitol concentration dependence of this interaction parameter gave (approximate) values of the transfer enthalpy, delta H2,tr, and transfer entropy delta S2,tr. Transfer of ribonuclease A from water into 30% sorbitol is characterized by positive values of the transfer free energy, transfer enthalpy, transfer entropy, and transfer heat capacity. On denaturation, the transfer enthalpy becomes more positive. This increment, however, is small relative to both the enthalpy of unfolding in water and to the transfer enthalpy of the native protein from water a 30% sorbitol solution.     

Decrease in protein solubility and cataract formation caused by the Pro23 to Thr mutation in human gamma D-crystallin

The P23T mutation in the human gammaD-crystallin gene has in recent years been associated with a number of well known cataract phenotypes. To understand the molecular mechanism of lens opacity caused by this mutation, we expressed human gammaD-crystallin (HGD), the P23T mutant, and other related mutant proteins in Escherichia coli and compared the structures and thermodynamic properties of these proteins in vitro. The results show that the cataract-causing mutation P23T does not exhibit any significant structural change relative to the native protein. However, in marked contrast to the native protein, the mutant shows a dramatically lowered solubility. The reduced solubility results from the association of the P23T mutant to form a new condensed phase that contains clusters of the mutant protein. The monomer-cluster equilibrium is represented by a solubility curve in the phase diagram. When the solubility limit is exceeded, the mutant protein forms the condensed phase after a nucleation time of 10-20 min. We found that the solubility of the P23T mutant exhibits an inverse dependence on temperature, i.e., the protein clusters are increasingly soluble as the temperature of the solution decreases. The solubility of P23T can be substantially altered by the introduction of specific mutations at or in the immediate vicinity of residue 23. We examined the mutants P23S, P23V, P23TInsP24, and P23TN24K and found that the latter two mutations can restore the solubility of the P23T mutant. These findings may help develop a strategy for the rational design of small molecule inhibitors of this type of condensed phase.     

pK values of histidine residues in ribonuclease Sa: effect of salt and net charge

The primary goal of this study was to gain a better understanding of the effect of environment and ionic strength on the pK values of histidine residues in proteins. The salt-dependence of pK values for two histidine residues in ribonuclease Sa (RNase Sa) (pI=3.5) and a variant in which five acidic amino acids have been changed to lysine (5K) (pI=10.2) was measured and compared to pK values of model histidine-containing peptides. The pK of His53 is elevated by two pH units (pK=8.61) in RNase Sa and by nearly one pH unit (pK=7.39) in 5K at low salt relative to the pK of histidine in the model peptides (pK=6.6). The pK for His53 remains elevated in 1.5M NaCl (pK=7.89). The elevated pK for His53 is a result of screenable electrostatic interactions, particularly with Glu74, and a non-screenable hydrogen bond interaction with water. The pK of His85 in RNase Sa and 5K is slightly below the model pK at low salt and merges with this value at 1.5M NaCl. The pK of His85 reflects mainly effects of long-range Coulombic interactions that are screenable by salt. The tautomeric states of the neutral histidine residues are changed by charge reversal. The histidine pK values in RNase Sa are always higher than the pK values in the 5K variant. These results emphasize that the net charge of the protein influences the pK values of the histidine residues. Structure-based pK calculations capture the salt-dependence relatively well but are unable to predict absolute histidine pK values.     

Nonnative Electrostatic Interactions Can Modulate Protein Folding: Molecular Dynamics with a Grain of Salt

In recent years, a growing number of protein folding studies have focused on the unfolded state, which is now recognized as playing a major role in the folding process. Some of these studies show that interactions occurring in the unfolded state can significantly affect the stability and kinetics of the protein folding reaction. In this study, we modeled the effect of electrostatic interactions, both native and nonnative, on the folding of three protein systems that underwent selective charge neutralization or reversal or complete charge suppression. In the case of the N-terminal L9 protein domain, our results directly attribute the increase in thermodynamic stability to destabilization of the unfolded ensemble, reaffirming the experimental observations. These results provide a deeper structural insight into the ensemble of the unfolded state and predict a new mutation site for increased protein stability. In the second case, charge reversal mutations of RNase Sa affected protein stability, with the destabilizing mutations being less destabilizing at higher salt concentrations, indicating the formation of charge-charge interactions in the unfolded state. In the N-terminal L9 and RNase Sa systems, changes in electrostatic interactions in the unfolded state that cause an increase in free energy had an overall compaction effect that suggests a decrease in entropy. In the third case, in which we compared the β-lactalbumin and hen egg-white lysozyme protein homologues, we successfully eliminated differences between the folding kinetics of the two systems by suppressing electrostatic interactions, supporting previously reported findings. Our coarse-grained molecular dynamics study not only reproduces experimentally reported findings but also provides a detailed molecular understanding of the elusive unfolded-state ensemble and how charge-charge interactions can modulate the biophysical characteristics of folding.     

Exploration micromechanism of VP35 IID interaction and recognition dsRNA: A molecular dynamics simulation

Multifunctional viral protein (VP35) encoded by the highly pathogenic Ebola viruses (EBOVs) can antagonize host double‐stranded RNA (dsRNA) sensors and immune response because of the simultaneous recognition of dsRNA backbone and blunt ends. Mutation of select hydrophobic conserved basic residues within the VP35 inhibitory domain (IID) abrogates its dsRNA‐binding activity, and impairs VP35‐mediated interferon (IFN) antagonism. Herein the detailed binding mechanism between dsRNA and WT, single mutant, and double mutant were investigated by all‐atom molecular dynamics (MD) simulation and binding energy calculation. R312A/R322A double mutations results in a completely different binding site and orientation upon the structure analyses. The calculated binding free energy results reveal that R312A, R322A, and K339A single mutations decrease the binding free energies by 17.82, 13.18, and 13.68 kcal mol^?1, respectively. The binding energy decomposition indicates that the strong binding affinity of the key residues is mainly due to the contributions of electrostatic interactions in the gas phase, where come from the positively charged side chain and the negatively charged dsRNA backbone. R312A, R322A, and K339A single mutations have no significant effect on VP35 IID conformation, but the mutations influence the contributions of electrostatic interactions in the gas phase. The calculated results reveal that end‐cap residues which mainly contribute VDW interactions can recognize and capture dsRNA blunt ends, and the central basic residues (R312, R322, and K339) which mainly contribute favorable electrostatic interactions with dsRNA backbone can fix dsRNA binding site and orientation. Proteins 2017; 85:1008–1023. © 2017 Wiley Periodicals, Inc.     

Electrostatic calculations of amino acid titration and electron transfer, Q-AQB-->QAQ-B, in the reaction center.

The titration of amino acids and the energetics of electron transfer from the primary electron acceptor (QA) to the secondary electron acceptor (QB) in the photosynthetic reaction center of Rhodobacter sphaeroides are calculated using a continuum electrostatic model. Strong electrostatic interactions between titrating sites give rise to complex titration curves. Glu L212 is calculated to have an anomalously broad titration curve, which explains the seemingly contradictory experimental results concerning its pKa. The electrostatic field following electron transfer shifts the average protonation of amino acids near the quinones. The pH dependence of the free energy between Q-AQB and QAQ-B calculated from these shifts is in good agreement with experiment. However, the calculated absolute free energy difference is in severe disagreement (by approximately 230 meV) with the observed experimental value, i.e., electron transfer from Q-A to QB is calculated to be unfavorable. The large stabilization energy of the Q-A state arises from the predominantly positively charged residues in the vicinity of QA in contrast to the predominantly negatively charged residues near QB. The discrepancy between calculated and experimental values for delta G(Q-AQB-->QAQ-B) points to limitations of the continuum electrostatic model. Inclusion of other contributions to the energetics (e.g., protein motion following quinone reduction) that may improve the agreement between theory and experiment are discussed.     

Ion Specificity and Nonmonotonic Protein Solubility from Salt Entropy

The addition of salt to protein solutions can either increase or decrease the protein solubility, and the magnitude of this effect depends on the salt used. We show that these effects can be captured using a theory that includes attractive and repulsive electrostatic interactions, nonelectrostatic protein-ion interactions, and ion-solvent interactions via an effective solvated ion radius. We find that the ion radius has significant effects on the translational entropy of the salt, which leads to salt specificity in the protein solubility. At low salt, the dominant effect comes from the entropic cost of confining ions within the aggregate, whereas at high concentrations, the salt drives a depletion attraction that favors aggregation. Our theory explains the reversal in the Hofmeister series observed in lysozyme cloud point measurements and semi-quantitatively describes the solubility of lysozyme and chymosin crystals. We present a comparison of the contributions to the free energy and give guidelines for when salting in or salting out should be expected.     

Contribution of active site residues to the activity and thermal stability of ribonuclease Sa

We have used site-specific mutagenesis to study the contribution of Glu 74 and the active site residues Gln 38, Glu 41, Glu 54, Arg 65, and His 85 to the catalytic activity and thermal stability of ribonuclease Sa. The activity of Gln38Ala is lowered by one order of magnitude, which confirms the involvement of this residue in substrate binding. In contrast, Glu41Lys had no effect on the ribonuclease Sa activity. This is surprising, because the hydrogen bond between the guanosine N1 atom and the side chain of Glu 41 is thought to be important for the guanine specificity in related ribonucleases. The activities of Glu54Gln and Arg65Ala are both lowered about 1000-fold, and His85Gln is totally inactive, confirming the importance of these residues to the catalytic function of ribonuclease Sa. In Glu74Lys, k(cat) is reduced sixfold despite the fact that Glu 74 is over 15 A from the active site. The pH dependence of k(cat)/K(M) is very similar for Glu74Lys and wild-type RNase Sa, suggesting that this is not due to a change in the pK values of the groups involved in catalysis. Compared to wild-type RNase Sa, the stabilities of Gln38Ala and Glu74Lys are increased, the stabilities of Glu41Lys, Glu54Gln, and Arg65Ala are decreased and the stability of His85Gln is unchanged. Thus, the active site residues in the ribonuclease Sa make different contributions to the stability.     

Fluctuations between stabilizing and destabilizing electrostatic contributions of ion pairs in conformers of the c-Myc-Max leucine zipper

In solution proteins often exhibit backbone and side-chain flexibility. Yet electrostatic interactions in proteins are sensitive to motions. Hence, here we study the contribution of ion pairs toward protein stability in a range of conformers which sample the conformational space in solution. Specifically, we focus on the electrostatic contributions of ion pairs to the stability of each of the conformers in the NMR ensemble of the c-Myc-Max leucine zipper and to their average energy minimized structure. We compute the electrostatic contributions of inter- and intra-helical ion pairs and of an ion pair network. We find that the electrostatic contributions vary considerably among the 40 NMR conformers. Each ion pair, and the network, fluctuates between being stabilizing and being destabilizing. This fluctation reflects the variability in the location of the ion pairing residues and in the geometric orientation of these residues, both with respect to each other and with respect to other charged groups in the rest of the protein. Ion pair interactions in the c-Myc-Max leucine zipper in solution depend on the protein conformer which is analyzed. Hence, the overall stabilizing (or destabilizing) contribution of an ion pair is conformer population-dependent. This study indicates that free energy calculations performed using the continuum electrostatics methodology are sensitive to protein conformational details.     

Analysis of amino acid contributions to protein solubility using short peptide tags fused to a simplified BPTI variant

Protein solubility is usually characterized in terms of a hydrophobicity scale, which refers to the free energy of transfer of a molecule from an aqueous to a nonpolar solution and is not a "solubility propensity scale" per se. Using a "host-guest" approach, we measured the effects of short poly-amino-acid tags (guests) on the solubility of a host protein, a simplified bovine pancreatic trypsin inhibitor (BPTI), to which they were fused at the C-terminus. We analyzed 10 amino acid types, representing the full range of biophysical properties (acidic, basic, polar, and hydrophobic). As anticipated, positively charged residues significantly increased the solubility of the model protein, at both pH 4.7 and 7.7, whereas very hydrophobic poly-Ile markedly reduced the solubility of BPTI. Poly-Asp and poly-Glu barely affected BPTI solubility at pH 4.7, but induced an eight to ten-fold increase at pH 7.7, attributable to the ionization of their side chains. Although Pro is the most soluble amino acid, poly-Pro did not affect the protein's solubility. The effects of the other tags on BPTI solubility ranged from none to an eight-fold increase. To ensure that the measured solubility values were context independent and could provide a "solubility propensity scale", we confirmed that the tags remained independent of the structure, thermal stability, and biochemical activity of the host protein. These observations suggest that this approach is valuable for measuring the solubility propensities of amino acids, which could eventually allow the calculation of a polypeptide's relative solubility from its amino acid sequence.     

Free energy simulations: The meaning of the individual contributions from a component analysis

A theoretical analysis is made of the decomposition into contributions from individual interactions of the free energy calculated by thermodynamic integration. It is demonstrated that such a decomposition, often referred to as “component analysis,” is meaningful, even though it is a function of the integration path. Moreover, it is shown that the path dependence can be used to determine the relation of the contribution of a given interaction to the state of the system. To illustrate these conclusions, a simple transformation(Cl^? to Br^? in aqueous solution) is analyzed by use of the Reference Interaction Site Model-Hypernetted Chain Closure integral equation approach; it avoids the calculational difficulties of macromolecular simulation while retaining their conceptual complexity. The difference in the solvation free energy between chloride and bromide is calculated, and the contributions of the Lennard-Jones and electrostatic terms in the potential function are analyzed by the use of suitably chosen integration paths. The model is also used to examine the path dependence of individual contributions to the double free energy differences (ΔΔG or ΔΔA) that are often employed in free energy simulations of biological systems. The alchemical path, as contrasted with the experimental path, is shown to be appropriate for interpreting the effects of mutations on ligand binding and protein stability. The formulation is used to obtain a better understanding of the success of the Poisson-Boltzmann continuum approach for determining the solvation properties of polar and ionic systems. © 1994 Wiley-Liss, Inc.     

Electrostatic contributions to the stability of a thermophilic cold shock protein

The thermophilic Bacillus caldolyticus cold shock protein (Bc-Csp) differs from the mesophilic Bacillus subtilis cold shock protein B (Bs-CspB) in 11 of the 66 residues. Stability measurements of Schmid and co-workers have implicated contributions of electrostatic interactions to the thermostability. To further elucidate the physical basis of the difference in stability, previously developed theoretical methods that treat electrostatic effects in both the folded and the unfolded states were used in this paper to study the effects of mutations, ionic strength, and temperature. For 27 mutations that narrow the difference in sequence between Bc-Csp and Bs-CspB, calculated changes in unfolding free energy (Delta G) and experimental results have a correlation coefficient of 0.98. Bc-Csp appears to use destabilization of the unfolded state by unfavorable charge-charge interactions as a mechanism for increasing stability. Accounting for the effects of ionic strength and temperature on the electrostatic free energies in both the folded and the unfolded states, explanations for two important experimental observations are presented. The disparate ionic strength dependences of Delta G for Bc-Csp and Bs-CspB were attributed to the difference in the total charges (-2e and -6e, respectively). A main contribution to the much higher unfolding entropy of Bs-CspB was found to come from the less favorable electrostatic interactions in the folded state. These results should provide insight for understanding the thermostability of other thermophilic proteins.     

The catalytic effect of dihydrofolate reductase and its mutants is determined by reorganization energies

The effect of distant mutations on the catalytic reaction of dihydrofolate reductase (DHFR) is reexamined by empirical valence bond simulations. The simulations reproduce for the first time the changes in the observed rate constants (without the use of adjustable parameters for this purpose) and show that the changes in activation barriers are strongly correlated with the corresponding changes in the reorganization energy. The preorganization of the polar groups of enzymes is the key catalytic factor, and anticatalytic mutations destroy this preorganization. Some anticatalytic mutations in DHFR also increase the distance between the donor and acceptor, but this effect is not directly related to catalysis since the native enzyme and the uncatalyzed reaction in water have similar average donor-acceptor distances. Insight into the effect of a mutation is provided by constructing the relevant free energy surfaces in terms of the generalized solute-solvent coordinates. It is shown how the mutations change the reaction coordinate and the activation barrier, and it is clarified that the corresponding changes do not reflect dynamical effects. It is also pointed out that all reactions in a condensed phase involve correlated motions (both in enzymes and in solution) and that the change of such motions upon mutations is a result of the change in the shape of the multidimensional reaction path on the solute-solvent surface, rather than the reason for the change in rate constant. Thus, as far as catalysis is concerned, the change in the activation barrier is due to the change in the electrostatic preorganization energy.     

Effects of pH on protein-protein interactions and implications for protein phase behavior

The effects of pH on protein interactions and protein phase behavior were investigated by measuring the reduced second osmotic virial coefficient (b2) for ovalbumin and catalase, and the aggregate and crystal solubilities for ovalbumin, beta-lactoglobulin A and B, ribonuclease A and lysozyme. The b2 trends observed for ovalbumin and catalase show that protein interactions become increasingly attractive with decreasing pH. This trend is in good agreement with ovalbumin phase behavior, which was observed to evolve progressively with decreasing pH, leading to formation of amorphous aggregates instead of gel bead-like aggregates, and spherulites instead of needle-like crystals. For both acidic and basic proteins, the aggregate solubility during protein salting-out decreased with decreasing pH, and contrary to what is commonly believed, neither aggregate nor crystal solubility had a minimum at the isoelectric point. beta-Lactoglobulin B was the only protein investigated to show salting-in behavior, and crystals were obtained at low salt concentrations in the vicinity of its isoelectric point. The physical origin of the different trends observed during protein salting-in and salting-out is discussed, and the implications for protein crystallization are emphasized.