Coarse-grained strategy for modeling protein stability in concentrated solutions

Anisotropic coarse-grain Monte Carlo simulations of lysozyme, lactoferrin, and NISTmAb by precomputing atomistic models

We develop a multiscale coarse-grain model of the NIST Monoclonal Antibody Reference Material 8671 (NISTmAb) to enable systematic computational investigations of high-concentration physical instabilities such as phase separation, clustering, and aggregation. Our multiscale coarse-graining strategy captures atomic-resolution interactions with a computational approach that is orders of magnitude more efficient than atomistic models, assuming the biomolecule can be decomposed into one or more rigid bodies with known, fixed structures. This method reduces interactions between tens of thousands of atoms to a single anisotropic interaction site. The anisotropic interaction between unique pairs of rigid bodies is precomputed over a discrete set of relative orientations and stored, allowing interactions between arbitrarily oriented rigid bodies to be interpolated from the precomputed table during coarse-grained Monte Carlo simulations. We present this approach for lysozyme and lactoferrin as a single rigid body and for the NISTmAb as three rigid bodies bound by a flexible hinge with an implicit solvent model. This coarse-graining strategy predicts experimentally measured radius of gyration and second osmotic virial coefficient data, enabling routine Monte Carlo simulation of medically relevant concentrations of interacting proteins while retaining atomistic detail. All methodologies used in this work are available in the open-source software Free Energy and Advanced Sampling Simulation Toolkit.

Predicting Experimental B22 Values and the Effects of Histidine Charge States for Monoclonal Antibodies Using Coarse-Grained Molecular Simulations

Static light scattering (SLS) was used to characterize five monoclonal antibodies (MAbs) as a function of total ionic strength (TIS) at pH values between 5.5 and 7.0. Second osmotic virial coefficient (B₂₂) values were determined experimentally for each MAb as a function of TIS using low protein concentration SLS data. Coarse-grained molecular simulations were performed to predict the B₂₂ values for each MAb at a given pH and TIS. To include the effect of charge fluctuations of titratable residues in the B₂₂ calculations, a statistical approach was introduced in the Monte Carlo algorithm based on the protonation probability based on a given pH value and the Henderson-Hasselbalch equation. The charged residues were allowed to fluctuate individually, based on the sampled microstates and the influence of electrostatic interactions on net protein-protein interactions during the simulations. Compared to static charge simulations, the new approach provided improved results compared to experimental B₂₂ values at pH conditions near the pK_a of titratable residues.

Toward a Suite of Coarse-Grained Models for Molecular Simulation of Monoclonal Antibodies and Therapeutic Proteins

A series of coarse-grained models for molecular simulation of proteins are considered, with emphasis on the application of predicting protein-protein self-interactions for monoclonal antibodies (MAbs). As an illustrative example and for quantitative comparison, the models are used to predict osmotic virial coefficients over a broad range of attractive and repulsive self-interactions and solution conditions for a series of MAbs where the second osmotic virial coefficient has been experimentally determined in prior work. The models are compared based on how well they can predict experimental behavior, their computational burdens, and scalability. An intermediate-resolution model is also introduced that can capture specific electrostatic interactions with improved efficiency and similar or improved accuracy when compared to the previously published models. Guidance is included for the selection of coarse-grained models more generally for capturing a balance of electrostatic, steric, and short-ranged nonelectrostatic interactions for proteins from low to high concentrations.

Comparison of Huggins Coefficients and Osmotic Second Virial Coefficients of Buffered Solutions of Monoclonal Antibodies

The Huggins coefficient kH is a well-known metric for quantifying the increase in solution viscosity arising from intermolecular interactions in relatively dilute macromolecular solutions, and there has been much interest in this solution property in connection with developing improved antibody therapeutics. While numerous kH measurements have been reported for select monoclonal antibodies (mAbs) solutions, there has been limited study of kH in terms of the fundamental molecular interactions that determine this property. In this paper, we compare measurements of the osmotic second virial coefficient B22, a common metric of intermolecular and interparticle interaction strength, to measurements of kH for model antibody solutions. This comparison is motivated by the seminal work of Russel for hard sphere particles having a short-range "sticky" interparticle interaction, and we also compare our data with known results for uncharged flexible polymers having variable excluded volume interactions because proteins are polypeptide chains. Our observations indicate that neither the adhesive hard sphere model, a common colloidal model of globular proteins, nor the familiar uncharged flexible polymer model, an excellent model of intrinsically disordered proteins, describes the dependence of kH of these antibodies on B22. Clearly, an improved understanding of protein and ion solvation by water as well as dipole-dipole and charge-dipole effects is required to understand the significance of kH from the standpoint of fundamental protein-protein interactions. Despite shortcomings in our theoretical understanding of kH for antibody solutions, this quantity provides a useful practical measure of the strength of interprotein interactions at elevated protein concentrations that is of direct significance for the development of antibody formulations that minimize the solution viscosity.

Reaching new levels of realism in modeling biological macromolecules in cellular environments

An increasing number of studies are aimed at modeling cellular environments in a comprehensive and realistic fashion. A major challenge in these efforts is how to bridge spatial and temporal scales over many orders of magnitude. Furthermore, there are additional challenges in integrating different aspects ranging from questions about biomolecular stability in crowded environments to the description of reactive processes on cellular scales. In this review, recent studies with models of biomolecules in cellular environments at different levels of detail are discussed in terms of their strengths and weaknesses. In particular, atomistic models, implicit representations of cellular environments, coarse-grained and spheroidal models of biomolecules, as well as the inclusion of reactive processes via reaction-diffusion models are described. Furthermore, strategies for integrating the different models into a comprehensive description of cellular environments are discussed.

Concentrated dispersions of equilibrium protein nanoclusters that reversibly dissociate into active monomers

Stabilizing proteins at high concentration is of broad interest in drug delivery, for treatment of cancer and many other diseases. Herein, we create highly concentrated antibody dispersions (up to 260 mg/mL) comprising dense equilibrium nanoclusters of protein (monoclonal antibody 1B7, polyclonal sheep immunoglobulin G, and bovine serum albumin) molecules which, upon dilution in vitro or administration in vivo, remain conformationally stable and biologically active. The extremely concentrated environment within the nanoclusters (∼700 mg/mL) provides conformational stability to the protein through a novel self-crowding mechanism, as shown by computer simulation, while the primarily repulsive nanocluster interactions result in colloidally stable, transparent dispersions. The nanoclusters are formed by adding trehalose as a cosolute which strengthens the short-ranged attraction between protein molecules. The protein cluster diameter was reversibly tuned from 50 to 300 nm by balancing short-ranged attraction against long-ranged electrostatic repulsion of weakly charged protein at a pH near the isoelectric point. This behavior is described semiquantitatively with a free energy model which includes the fractal dimension of the clusters. Upon dilution of the dispersion in vitro, the clusters rapidly dissociated into fully active protein monomers as shown with biophysical analysis (SEC, DLS, CD, and SDS-PAGE) and sensitive biological assays. Since the concept of forming nanoclusters by tuning colloid interactions is shown to be general, it is likely applicable to a variety of biological therapeutics, mitigating the need to engineer protein stability through amino acid modification. In vivo subcutaneous injection into mice results in indistinguishable pharmacokinetics versus a standard antibody solution. Stable protein dispersions with low viscosities may potentially enable patient self-administration by subcutaneous injection of antibody therapeutics being discovered and developed.

Computational methods to predict therapeutic protein aggregation

Protein based biotherapeutics have emerged as a successful class of pharmaceuticals. However, these macromolecules endure a variety of physicochemical degradations during manufacturing, shipping, and storage, which may adversely impact the drug product quality. Of these degradations, the irreversible self-association of therapeutic proteins to form aggregates is a major challenge in the formulation of these molecules. Tools to predict and mitigate protein aggregation are, therefore, of great interest to biopharmaceutical research and development. In this chapter, a number of such computational tools developed to understand and predict the various steps involved in protein aggregation are described. These tools can be grouped into three general classes: unfolding kinetics and native state thermal stability, colloidal stability, and sequence/structure based aggregation liabilities. Chapter sections introduce each class by discussing how these predictive tools provide insight into the molecular events leading to protein aggregation. The computational methods are then explained in detail along with their advantages and limitations.

Monte Carlo study of the molecular mechanisms of surface-layer protein self-assembly

The molecular mechanisms guiding the self-assembly of proteins into functional or pathogenic large-scale structures can be only understood by studying the correlation between the structural details of the monomer and the eventual mesoscopic morphologies. Among the myriad structural details of protein monomers and their manifestations in the self-assembled morphologies, we seek to identify the most crucial set of structural features necessary for the spontaneous selection of desired morphologies. Using a combination of the structural information and a Monte Carlo method with a coarse-grained model, we have studied the functional protein self-assembly into S(surface)-layers, which constitute the crystallized outer most cell envelope of a great variety of bacterial cells. We discover that only few and mainly hydrophobic amino acids, located on the surface of the monomer, are responsible for the formation of a highly ordered anisotropic protein lattice. The coarse-grained model presented here reproduces accurately many experimentally observed features including the pore formation, chemical description of the pore structure, location of specific amino acid residues at the protein-protein interfaces, and surface accessibility of specific amino acid residues. In addition to elucidating the molecular mechanisms and explaining experimental findings in the S-layer assembly, the present work offers a tool, which is chemical enough to capture details of primary sequences and coarse-grained enough to explore morphological structures with thousands of protein monomers, to promulgate design rules for spontaneous formation of specific protein assemblies.

Controlling protein crystal growth rate by means of temperature

We have proposed a model to analyze the growth kinetics of lysozyme crystals/aggregates under non-isothermal conditions. The model was formulated through an analysis of the entropy production of the growth process which was obtained by taking into account the explicit dependence of the free energy on the temperature. We found that the growth process is coupled with temperature variations, resulting in a novel Soret-type effect. We identified the surface entropy of the crystal/aggregate as a decisive ingredient controlling the behavior of the average growth rate as a function of temperature. The behavior of the Gibbs free energy as a function of temperature is also analyzed. The agreement between theory and experiments is very good in the range of temperatures considered.

Molecular dynamics of an embedded-charge model of lysozyme aqueous solutions

The onset of liquid-vapor separation in an interaction site model of a lysozyme aqueous solution is investigated by means of molecular dynamics (MD). Calculations are performed for a soft-core version of a potential early introduced by Carlsson et al. (J. Phys. Chem. B 2001, 105, 9040; 2001, 105, 12189.) whose liquid-vapor coexistence was studied by Rosch and Errington (J. Phys. Chem. B 2007, 111, 12591.); our modified model preserves the tailoring onto the experimental lysozyme solution properties embodied by those descriptions. We first show that the structural results obtained by Carlsson et al. at ambient conditions are quite well reproduced by our approach. Then, we cool the system along an isochoric path by monitoring the structural and dynamical properties at various temperatures. We thus show that a fluid-fluid separation takes place at a temperature 15% below the presumed binodal; in particular, we observe that density inhomogeneities develop rather early in the MD run and evolve over tens of nanoseconds into two dense aggregates that eventually merge, after approximately 100 ns more, into a single liquid phase separated from a vapor region by a well-defined planar interface. The densities of the two coexisting fluids are compatible with previous determinations of the binodal line. The connections of this work to the overall scenario of phase stability investigations in protein solutions, as well as possible developments based on the use of more refined models, are discussed.

A didactic model of macromolecular crowding effects on protein folding

A didactic model is presented to illustrate how the effect of macromolecular crowding on protein folding and association is modeled using current analytical theory and discrete molecular dynamics. While analytical treatments of crowding may consider the effect as a potential of average force acting to compress a polypeptide chain into a compact state, the use of simulations enables the presence of crowding reagents to be treated explicitly. Using an analytically solvable toy model for protein folding, an approximate statistical thermodynamic method is directly compared to simulation in order to gauge the effectiveness of current analytical crowding descriptions. Both methodologies are in quantitative agreement under most conditions, indication that both current theory and simulation methods are capable of recapitulating aspects of protein folding even by utilizing a simplistic protein model.

Competition between folding and aggregation in a model for protein solutions

We study the thermodynamic and kinetic consequences of the competition between single-protein folding and protein-protein aggregation using a phenomenological model, in which the proteins can be in the unfolded (U), misfolded (M) or folded (F) states. The phase diagram shows the coexistence between a phase with aggregates of misfolded proteins and a phase of isolated proteins (U or F) in solution. The spinodal at low protein concentrations shows non-monotonic behavior with temperature, with implications for the stability of solutions of folded proteins at low temperatures. We follow the dynamics upon "quenching" from the U-phase (cooling) or the F-phase (heating) to the metastable or unstable part of the phase diagram that results in aggregation. We describe how interesting consequences to the distribution of aggregate size, and growth kinetics arise from the competition between folding and aggregation.

Structure and thermodynamics of colloidal protein cluster formation: comparison of square-well and simple dipolar models

Reversible formation of weakly associated protein oligomers or clusters is a key early step in processes such as protein aggregation and colloidal phase separation. A previously developed cell-based, quasichemical model for lattice fluids [T. M. Young and C. J. Roberts, J. Chem. Phys. 127, 165101 (2007)] is extended here to treat continuous-space systems. It is illustrated using two simplified limiting cases for globular proteins at the isoelectric point: spherical square-well (SW) particles with an isotropic short-ranged attraction and screened dipolar particles with SW attractions and square-shoulder repulsions. Cluster free energies (DeltaA(i)) and structures are analyzed as a function of the reduced second virial coefficient b(2)(*). DeltaA(i) values and the average structures of clusters up to pentamers have distinct differences due to the anisotropic nature of the dipolar interactions. However, DeltaA(i) values can be mapped semiquantitatively between the two cases if compared at common values of b(2)(*). Free energy landscapes of oligomerization are constructed, illustrating significant differences in landscape ruggedness for small clusters of dipolar versus SW fluids, and suggesting a possible molecular interpretation for empirical models of nucleation-dependent aggregation of proteins.

Model for amorphous aggregation processes

The amorphous aggregation of proteins is associated with many phenomena, ranging from the formation of protein wine haze to the development of cataract in the eye lens and the precipitation of recombinant proteins during their expression and purification. While much literature exists describing models for linear protein aggregation, such as amyloid fibril formation, there are few reports of models which address amorphous aggregation. Here, we propose a model to describe the amorphous aggregation of proteins which is also more widely applicable to other situations where a similar process occurs, such as in the formation of colloids and nanoclusters. As first applications of the model, we have tested it against experimental turbidimetry data of three proteins relevant to the wine industry and biochemistry, namely, thaumatin, a thaumatinlike protein, and alpha -lactalbumin. The model is very robust and describes amorphous experimental data to a high degree of accuracy. Details about the aggregation process, such as shape parameters of the aggregates and rate constants, can also be extracted.

Insights Into Crowding Effects on Protein Stability From a Coarse-Grained Model

Proteins aggregate and precipitate from high concentration solutions in a wide variety of problems of natural and technological interest. Consequently, there is a broad interest in developing new ways to model the thermodynamic and kinetic aspects of protein stability in these crowded cellular or solution environments. We use a coarse-grained modeling approach to study the effects of different crowding agents on the conformational equilibria of proteins and the thermodynamic phase behavior of their solutions. At low to moderate protein concentrations, we find that crowding species can either stabilize or destabilize the native state, depending on the strength of their attractive interaction with the proteins. At high protein concentrations, crowders tend to stabilize the native state due to excluded volume effects, irrespective of the strength of the crowder-protein attraction. Crowding agents reduce the tendency of protein solutions to undergo a liquid-liquid phase separation driven by strong protein-protein attractions. The aforementioned equilibrium trends represent, to our knowledge, the first simulation predictions for how the properties of crowding species impact the global thermodynamic stability of proteins and their solutions.

Molecular crowding and solvation: direct and indirect impact on protein reactions

The typical environment for biomolecules in vivo is highly crowded. Under such conditions chemical activities, rather than simply concentrations, govern the behavior of the molecules. In this chapter we discuss the underlying solvation principles that give rise to the chemical activities. We focus on simple experimentally accessible examples, macromolecular crowding, protein folding, and ligand binding under crowded conditions. We discuss effects of high concentrations of both macromolecules and small molecules in terms of the Kirkwood-Buff theory, which couples solution structure to thermodynamics.

Stabilization of beta-galactosidase (from peas) by immobilization onto amberlite MB-150 beads and its application in lactose hydrolysis

The soluble PsBGAL (from Pisum sativum ) is extremely unstable with loss of over 80% in enzyme activity within 24 h at 4 degrees C when the protein concentration was lower than 0.1 mg/mL. Enzyme immobilization onto Amberlite MB-150 beads (diameter = 5 microm) greatly stabilized the enzyme preparation, with almost no loss for 12 months at room temperature (27 degrees C). Enzyme (21.9 microg) was immobilized by 62.56% onto activated 100 mg of Amberlite MB-150 beads using 4% glutaraldehyde, at pH 6.0 (50 mM, sodium phosphate buffer). Statistical analysis carried out by ANOVA revealed that all parameters used during immobilization were equally important at P < 0.05 (level of significance). An approach toward commercial exploitation of Amberlite-PsBGAL especially in lactose hydrolysis was anticipated due to improved physicochemical properties including broad optimum pH and temperature, with a K(m) of 4.11 +/- 0.21 mM for lactose. Amberlite-PsBGAL hydrolyzed 64.57 and 69.18% of lactose present in milk and milk whey, respectively, within 10 h (at room temperature). Immobilized enzyme has reusability of over 10 batchwise uses, with almost no loss in activity. The easy accessibility of enzyme source, ease of its immobilization on Amberlite, lower cost of Amberlite, enhanced stability of Amberlite-PsBGAL, and comparable lactose hydrolysis in milk and milk whey described here make it a suitable product for future applications at laboratory and industrial scale.

Thermodynamics and kinetics of protein folding under confinement

Understanding the effects of confinement on protein stability and folding kinetics is important for describing protein folding in the cellular environment. We have investigated the effects of confinement on two structurally distinct proteins as a function of the dimension d(c) and characteristic size R of the confining boundary. We find that the stabilization of the folded state relative to bulk conditions is quantitatively described by R(-gamma(c)), where the exponent gamma(c) is approximately 5/3 independent of the dimension of confinement d(c) (cylindrical, planar, or spherical). Moreover, we find that the logarithm of the folding rates also scale as R(-gamma(c)), with deviations only being seen for very small confining geometries, where folding is downhill; for both stability and kinetics, the dominant effect is the change in the free energy of the unfolded state. A secondary effect on the kinetics is a slight destabilization of the transition state by confinement, although the contacts present in the confined transition state are essentially identical to the bulk case. We investigate the effect of confinement on the position-dependent diffusion coefficients D(Q) for dynamics along the reaction coordinate Q (fraction of native contacts). The diffusion coefficients only change in the unfolded state basin, where they are increased because of compaction.

Influence of macromolecular crowding on protein-protein association rates--a Brownian dynamics study

The high total concentration of macromolecules, often referred to as macromolecular crowding, is one of the characteristic features of living cells. Macromolecular crowding influences interactions between many types of macromolecules, with consequent effects on, among others, the rates of reactions occurring in the cell. Simulations to study the influence of crowding on macromolecular association rate were performed using a modified Brownian dynamics protocol. The calculated values of the time-dependent self-diffusion coefficients in different crowding conditions are in a very good agreement with those obtained by other authors. Simulations of the complex formation between the monoclonal antibody HyHEL-5 and its antigen hen egg lysozyme, both represented at atomic level detail, show that the calculated association rates strongly depend on the volume excluded by crowding. The rate obtained for the highest concentration of crowding particles is greater than twofold higher than the rate for proteins without crowding.

Investigation of the phase behavior of an embedded charge protein model through molecular simulation

The phase behavior of an embedded-charge model for lysozyme developed by Carlsson and co-workers (J. Phys. Chem. B 2001, 105, 9040) is investigated using grand canonical transition matrix Monte Carlo simulation. Within this model, protein-protein interactions are approximated through a combination of hard-sphere repulsion, isotropic hydrophobic attraction, and screened electrostatic interactions through a series of embedded point charges located at the positions of charged amino acid groups within lysozyme. Liquid-liquid phase diagrams are constructed for a wide range of solution conditions and compared with experimental data. Our results indicate that the model is generally capable of describing qualitative trends in the evolution of protein phase behavior with variation of pH and ionic strength. From a quantitative perspective, model estimates for both the change in critical temperature with variation of the solution conditions and the critical concentration do not agree with experimental results. We find the width of model coexistence curves to be independent of solution conditions and narrow relative to experimentally obtained phase envelopes. Connections between the value of the second virial coefficient evaluated at the critical temperature and the location of the liquid-liquid phase envelope are also examined.

Macromolecular crowding increases structural content of folded proteins

Here we show that increased amount of secondary structure is acquired in the folded states of two structurally-different proteins (alpha-helical VlsE and alpha/beta flavodoxin) in the presence of macromolecular crowding agents. The structural content of flavodoxin and VlsE is enhanced by 33% and 70%, respectively, in 400 mg/ml Ficoll 70 (pH 7, 20 degrees C) and correlates with higher protein-thermal stability. In the same Ficoll range, there are only small effects on the unfolded-state structures of the proteins. This is the first in vitro assessment of crowding effects on the native-state structures at physiological conditions. Our findings imply that for proteins with low intrinsic stability, the functional structures in vivo may differ from those observed in dilute buffers.

Coarse-grained strategy for modeling protein stability in concentrated solutions. III: directional protein interactions

We extend our coarse-grained modeling strategy described in parts I and II of this investigation to account for nonuniform spatial distributions of hydrophobic residues on the solvent-exposed surfaces of native proteins. Within this framework, we explore how patchy surfaces can influence the solvent-mediated protein-protein interactions, and the unfolding and self-assembly behaviors of proteins in solution. In particular, we compare the equilibrium unfolding and self-assembly trends for three model proteins that share the same overall sequence hydrophobicity, but exhibit folded configurations with different solvent-exposed native-state surface morphologies. Our model provides new insights into how directional interactions can affect native-state protein stability in solution. We find that strongly-directional attractions between native molecules with patchy surfaces can help stabilize the folded conformation through the formation of self-assembled clusters. In contrast, native proteins with more uniform surfaces are destabilized by protein-protein attractions involving the denatured state. Finally, we discuss how the simulation results provide insights into the experimental solution behaviors of several proteins that display directional interactions in their native states.

Analytical model for studying how environmental factors influence protein conformational stability in solution

We introduce an analytical modeling strategy for probing the conformational stability of globular proteins in aqueous solution. In this approach, the intrinsic (i.e., infinite dilution) thermodynamic stability and coarse structural properties of the proteins, as well as the effective protein-protein interactions, derive from a heteropolymer collapse theory that incorporates predicted temperature- and pressure-dependent hydrophobic interactions. Protein concentration effects are estimated by integrating this information into a molecular thermodynamic model, which is an ad hoc generalization of the exact equilibrium theory of a one-dimensional binary mixture of square-well particles that interconvert through an isomerization (i.e., folding) reaction. The end result is an analytical multiscale modeling approach which, although still schematic, can predict that folded proteins exhibit a closed-loop region of stability in the pressure-temperature plane and that protein concentration has a nonmonotonic effect on protein stability, results consistent with qualitative trends observed in both experiments of protein solutions and simulations of coarse-grained protein models.

Heteropolymer collapse theory for protein folding in the pressure-temperature plane

We revisit a heteropolymer collapse theory originally introduced to explore how the balance between hydrophobic interactions and configurational entropy determines the thermal stability of globular proteins at ambient pressure. We generalize the theory by introducing a basic statistical mechanical treatment for how pressure impacts the solvent-mediated interactions between hydrophobic amino-acid residues. In particular, we estimate the strength of the hydrophobic interactions using a molecular thermodynamic model for the interfacial free energy between liquid water and a curved hydrophobic solute. The model, which also reproduces many of the distinctive thermodynamic properties of aqueous solutions in bulk and interfacial environments, predicts that the water-solute interfacial free energy is significantly reduced by the application of high hydrostatic pressures. This allows water to penetrate into folded heteropolymers at high pressure and break apart their hydrophobic cores, a scenario suggested earlier by information theory calculations. As a result, folded heteropolymers are predicted to display the kind of closed region of stability in the pressure-temperature plane exhibited by native proteins. We compare predictions of the collapse theory with experimental data for several proteins.

Coarse-grained strategy for modeling protein stability in concentrated solutions. II: phase behavior

We use highly efficient transition-matrix Monte Carlo simulations to determine equilibrium unfolding curves and fluid phase boundaries for solutions of coarse-grained globular proteins. The model we analyze derives the intrinsic stability of the native state and protein-protein interactions from basic information about protein sequence using heteropolymer collapse theory. It predicts that solutions of low hydrophobicity proteins generally exhibit a single liquid phase near their midpoint temperatures for unfolding, while solutions of proteins with high sequence hydrophobicity display the type of temperature-inverted, liquid-liquid transition associated with aggregation processes of proteins and other amphiphilic molecules. The phase transition occurring in solutions of the most hydrophobic protein we study extends below the unfolding curve, creating an immiscibility gap between a dilute, mostly native phase and a concentrated, mostly denatured phase. The results are qualitatively consistent with the solution behavior of hemoglobin (HbA) and its sickle variant (HbS), and they suggest that a liquid-liquid transition resulting in significant protein denaturation should generally be expected on the phase diagram of high-hydrophobicity protein solutions. The concentration fluctuations associated with this transition could be a driving force for the nonnative aggregation that can occur below the midpoint temperature.