Osmolyte Effects on Monoclonal Antibody Stability and Concentration

Article pubs.acs.org/JPCB

Osmolyte Effects on Monoclonal Antibody Stability and Concentration-Dependent Protein Interactions with Water and Common Osmolytes Gregory V. Barnett,† Vladimir I. Razinkov,‡ Bruce A. Kerwin,‡ Steven Blake,§ Wei Qi,§ Robin A. Curtis,∥ and Christopher J. Roberts*,† †

Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, Delaware 19716, United States Drug Product Development, Amgen Inc., Seattle, Washington 98119, United States § Malvern Biosciences Inc., Columbia, Maryland 21046, United States ∥ School of Chemical Engineering and Analytical Science, The University of Manchester, Manchester, U.K. ‡

S Supporting Information *

ABSTRACT: Preferential interactions of proteins with water and osmolytes play a major role in controlling the thermodynamics of protein solutions. While changes in protein stability and shifts in phase behavior are often reported with the addition of osmolytes, the underlying protein interactions with water and/or osmolytes are typically inferred rather than measured directly. In this work, Kirkwood−Buff integrals for protein−water interactions (G12) and protein−osmolyte interactions (G23) were determined as a function of osmolyte concentration from density measurements of antistreptavidin immunoglobulin gamma-1 (AS-IgG1) in ternary aqueous solutions for a set of common neutral osmolytes: sucrose, trehalose, sorbitol, and poly(ethylene glycol) (PEG). For sucrose and PEG solutions, both protein−water and protein−osmolyte interactions depend strongly on osmolyte concentrations (c3). Strikingly, both osmolytes change from being preferentially excluded to preferentially accumulated with increasing c3. In contrast, sorbitol and trehalose solutions do not show large enough preferential interactions to be detected by densimetry. G12 and G23 values are used to estimate the transfer free energy for native AS-IgG1 (ΔμN2 ) and compared with existing models. AS-IgG1 unfolding via calorimetry shows a linear increase in midpoint temperatures as a function of trehalose, sucrose, and sorbitol concentrations, but the opposite behavior for PEG. Together, the results highlight limitations of existing models and common assumptions regarding the mechanisms of protein stabilization by osmolytes. Finally, PEG preferential interactions destabilize the Fab regions of AS-IgG1 more so than the CH2 or CH3 domains, illustrating preferential interactions can be specific to different protein domains.

■

INTRODUCTION

Sugars and other polyhydroxy compoundse.g., sucrose or trehaloseare often observed to increase the unfolding free energy or midpoint unfolding temperature for proteins in vitro.16−18 Sugars are often added to protein solutions as cryoprotectants during freezing or lyophilization processes in an effort to reduce protein unfolding and aggregation.19−21 Similarly, the addition of canonical “kosmotropes” such as sulfate anions or poly(ethylene glycol) (PEG) is expected to increase the chemical potential of folded proteins, and this motivates the use of these compounds to promote protein “salting out” or phase separation.11,22,23 Historically, protein solubility data in PEG or sulfate solutions have been used to predict protein solubility in the absence of such solutes.22,24,25 In the remainder of this report, the term osmolyte is used to refer to small solutes such as sugars as well as PEG.

Interactions between proteins, water, and osmolytes mediate changes in protein stability and phase behavior.1−4 Inside cells, the concentration of proteins and other macromolecules can reach volume fractions of 30−40%, and nonideal interactions between proteins, water, and osmolytes are expected to be the norm rather than the exception.5,6 Often, the addition of sugars, polymers, and other osmolytes to protein solutions alters the chemical potential of the protein in its native and unfolded states.3,7,8 Folding/unfolding equilibria are altered when addition of a given osmolyte shifts the chemical potential of the native state to a different extent than for the unfolded state.9,10 Historically, measurements of protein solubility and/ or protein unfolding thermodynamics have often been used to infer differences between protein−water and protein−osmolyte interactions.11−13 Alternatively, it may also be possible to predict the thermodynamics of unfolding or phase separation if the net interactions between protein, water, and a given osmolyte can be determined accurately.14,15 © 2016 American Chemical Society

Received: January 20, 2016 Revised: March 5, 2016 Published: March 23, 2016 3318

DOI: 10.1021/acs.jpcb.6b00621 J. Phys. Chem. B 2016, 120, 3318−3330

Article

The Journal of Physical Chemistry B

binary systems.14 It can also be shown that the preferential interaction parameter (Γ32) is equal to −c3(G12 − G23).35 Casassa and Eisenberg provided one of the first unified discussions on elucidating thermodynamics and connections to molecular interactions for solutions of biological macromolecules using techniques such as densimetry, light scattering, analytical ultracentrifugation, and osmometry.36 Ben-Naim subsequently developed the “inverse KB” theory to evaluate KB integrals from thermodynamic measurements of partial specific volume, isothermal compressibility, and phase behavior or activity coefficients.32,39 Using the inversion process to obtain all of the possible KB integrals for three component systems has been challenging due to the need for large amounts of precise thermodynamic data and the need to solve a number of equations involved for multicomponent systems.40,41 As shown below, it is possible to obtain a subset of the KB integrals if one works in the dilute-protein limit. Preferential interaction parameters have historically been measured with densimetry and osmometry. Timasheff and coworkers deduced preferential interaction parameters from density and dialysis equilibrium measurements of protein solutions with different osmolytes.16 Record and co-workers have extensively measured closely related, but not exactly equivalent, preferential interaction parameters for proteins with osmolytes using vapor pressure osmometry.35,42 As Schurr has discussed, those preferential interaction parameters differ in what thermodynamic parameters were held constant during different types of osmometry measurements. Recently, a KB approach was applied with an additive transfer free energy model (ATFM) for peptides to capture preferential interaction of osmolytes with different amino acid groups by combining solubility measurements with a version of inverse KB theory.43 Assuming the ATFM assumption holds, and that the contributions scale with the solvent accessible surface area for a given peptide group, this allows one to calculate the free energy for transferring a protein from water to a given osmolyte solution for any protein with a known threedimensional structure.9,44,45 In the present work, densimetry and a version of inverse KB theory32 are used to determine the protein−water KB integral (G12), the protein−osmolyte KB intergral (G23), and the net preferential interactions for a monoclonal antibody in solutions of common neutral osmolytes as a function of osmolyte concentration. The osmolytes are sucrose, trehalose, sorbitol, and poly(ethylene glycol) (number-average molecular weight, Mn = 6000). Results are compared to a priori predictions based on preferential interaction models.8,43 The results highlight quantitative and qualitative limitations of existing models and standard expectations for some cases. The results are also combined with calorimetry measurements to assess whether the changes in native state chemical potential and preferential interactions may be predictive of thermal stability. Taken together, the results illustrate that preferential interactions can change qualitatively with increasing osmolyte concentration and also challenge common assumptions that preferential exclusion (accumulation) of osmolytes from (at) the surface of folded proteins will necessarily result in increased (decreased) unfolding free energies for proteins.

The underlying basis for the phenomena summarized above is often couched in terms of the preferential-interactions framework that was a focus of seminal work by Timasheff, Schellman, and others.16,26−31 Briefly, differences between competing protein−water and protein−osmolyte interactions lead to changes in protein chemical potential and solution behavior.32 For a canonical case where the net protein− osmolyte interactions are less favorable (or more unfavorable) than water−protein interactions, the osmolyte is said to be preferentially excluded from the protein, or the protein is said to be preferentially hydrated.27 As a result, the chemical potential of the protein necessarily increases upon increasing osmolyte molality (m3) at fixed temperature, pressure, and protein molality (m2). The opposite occurs if the net protein− osmolyte interactions are more favorable than those with water, and the osmolyte instead is preferentially accumulated or “weakly bound” near the protein surface.29 Kirkwood and Buff, and others such as Hall, developed statistical mechanical theory that relates spatial fluctuations in number density, and molecular radial distribution functions, to the thermodynamics of multicomponent solutions without the need to assume an underlying model about molecular shape or interactions.33,34 As shown previously by Ben-Naim and others, the thermodynamic relation given by eq 1 establishes a means to directly relate preferential interactions and changes in protein chemical potential via the Kirkwood−Buff (KB) integrals for protein−water (G12) or protein−osmolyte (G23) with changes in osmolyte concentration.32,35 ⎛ ∂μ ⎞ ⎜ 2⎟ ⎝ ∂m3 ⎠T , P , m

2 →0

⎛ ∂μ ⎞ = c3(G12 − G23)⎜ 3 ⎟ ⎝ ∂m3 ⎠T , P , m

2 →0

(1)

Using the nomenclature of Scatchard, as well as the arguments by Casassa and Eisenberg, water and (neutral) osmolyte are denoted as components 1 and 3, respectively.36,37 Protein with any counterions needed to exactly balance its net charge are denoted as component 2.34 c3 is the osmolyte concentration in mol/volume or mass/volume units, depending on whether one expresses G23 and G12 in units of volume/mol or volume/mass, respectively. μ2 is the protein chemical potential, μ3 is the osmolyte chemical potential, T and P denote temperature and pressure, respectively, and the other symbols are defined above. The KB integral for interactions between components i and j can be formally defined via eq 2, where gij(r) is the ensembleaveraged radial distribution function for component i with respect to component j in an open ensemble where all of the degrees of freedom of all components in the mixture have been Boltzmann averaged; r typically denotes the center-to-center distance between a molecule of species i and a molecule of species j, and the factor of 4πr2 accounts for integration over a differential shell. Gij = 4π

∫0

∞

(gij(r ) − 1)r 2 dr

(2)

While KB integrals are applicable for any set of solution conditions and protein concentration, eq 1 holds in the limit of infinite dilution of the protein, and the limit m2 → 0 is implicit in the evaluation of the KB integrals in that expression.32 The derivative on the right-hand side of eq 1 is independent of the identity of the protein because m2 → 0. It is necessarily positive for any equilibrium system38 and can be determined from available data in the literature for a number of osmolyte−water

■

MATERIALS AND METHODS Sample Preparation. Antistreptavidin immunoglobulin gamma-1, AS-IgG1 (>98% monomer), was provided by Amgen as a liquid stock solution at a concentration of ∼30

3319


Article


V̂ 2 was determined from the corresponding t-value and standard error of the slope and intercept.51 Protein−water and protein−osmolyte interactions were determined from the behavior of V̂ 2 as a function of osmolyte concentration. Ben-Naim was the first to provide the relevant expressions for a three-component system at infinite dilution for one of the components (protein in the present case) in a convenient form to deduce G12 and G23 from densimetry.32

mg/mL. The protein was dialyzed with Spectra/Por 7 tubing (10 kDa MWCO, Spectrum Laboratories, Santa Clara, CA) against distilled−deionized Milli-Q water (Millipore, Billerica, MA) as previously described.46 Following dialysis, the protein solution was filtered with a 0.22 μm filter (Millipore). The concentration was determined using the UV−vis absorbance at 280 nm (Agilent 8453 UV−vis, Agilent Technologies, Santa Clara, CA) and the extinction coefficient of 1.586 mL/mg cm for AS-IgG1.47 The protein concentration was measured in triplicate, and the 95% confidence value was 0.05 mg/mL or smaller. Osmolyte stock solutions were prepared by dissolving known masses of sucrose (>99.5% (HPLC grade), Sigma-Aldrich), D(+)-trehalose, dihydrate (>98% Fisher Scientific), D-sorbitol (>99.5% (HPLC grade), Sigma-Aldrich), or poly(ethylene glycol) (average molecular weight 6000 Da, Sigma-Aldrich) with distilled−deionized Milli-Q water (EMD-Millipore). Final solutions were titrated with sodium hydroxide stock or hydrochloric acid stock solutions to pH 6 ± 0.1. This pH was chosen because it required minimal titration, as prior work with this protein showed that the net charge at pH 6 was close to zero, and protein−protein interactions were minimal.48 All concentrations based on serial dilutions were determined gravimetrically, making use of the corresponding densities from measurements described in the next subsection. Osmolyte concentrations (mass/volume) were determined from omolyte molality values and the measured values of solution density. Partial Specific Volume via Densimetry. For each osmolyte of interest, a binary osmolyte stock solution (osmolyte and water) and a ternary protein stock solution (osmolyte, water, and protein) were prepared gravimetrically at constant osmolyte molality using a calibrated analytical microbalance (Denver Instruments) with ±0.02 mg precision. All protein concentrations were below 5 mg/mL to minimize the potential effects of protein−protein interactions at higher concentrations. Ten or more protein solutions (∼1−1.25 mL each) ranging from zero to the highest protein concentration were prepared from the two above-mentioned solutions created at constant molality. Samples were gently mixed and centrifuged at 5000g for 5 min after preparation to eliminate any possible bubbles. All solutions were measured within 24 h of preparation. The density of each protein solution was measured using a DMA 4500 density meter (Anton-Paar, Ashland, VA). As all solutions were in the dilute protein concentration range, the density data were observed to be linear as a function of protein concentration for all cases that were tested. The partial specific volume of the protein (V̂ 2) was determined from density as a function of protein weight fraction using eq 3 as previously described.49,50

V2̂ =

(4)

κT is the isothermal compressibility of the solution, R is the gas constant, MW2 is the protein molecular weight, c3 is the osmolyte concentration (units of mass/volume), and V̂ 3 is the partial specific volume of the osmolyte, which is also determined experimentally in the same manner as described above. The other symbols were defined previously. As V̂ 2 has units of volume/mass, the KB integrals in eq 4 implicitly have the same units as V̂ 2. As aqueous solutions far from the critical point have small values for the isothermal compressibility (∼0.1−1 GPa−1), the first term of the right-hand side of eq 4 is negligible compared to the values of the other terms, within typical experimental precision for V̂ 2.52 The change of V̂ 2 as a function of osmolyte concentration then provides G12 and G23 based on slope and intercept or based on the local tangent when V̂ 2 changes nonlinearly with c3. Model Predictions of Preferential Interactions. Predictions for values of (∂μ2/∂m3)T,P,m2→0 and therefore (G12 − G23) for sucrose, trehalose, and sorbitol were calculated using the additive tripeptide preferential interaction model developed by Auton and Bolen.43 The transfer free energy, ΔμN2tr, was calculated using the solvent accessible surface area (ASA) of the three-dimensional model for AS-IgG1 and ATFM model.43 ASA values were determined using GET_AREA,53 Surface Racer,54 or ProtSA55,56 algorithms. Predictions for the difference in Kirkwood−Buff integrals for protein−water and protein−osmolyte, (G12 − G23)pred, were calculated using (∂μ2/∂m3)T,P,m3 values calculated above, combined with eq 1. Concentration dependent values for (∂μ2/∂m3)T,P,m2→0 were determined from analytical expressions for a given activity coefficient as a function of osmolyte concentration (sorbitol and sucrose).14 For trehalose, this quantity was determined from available thermodynamic data (see Supporting Information), based on published water activity of trehalose-water solutions using the isopiestic method.57 Predictions for the AS-IgG1 and PEG preferential interaction were determined using the additive atom transfer preferential interaction model developed by Record and co-workers.8 In this model, values were calculated from the ASA of each type of atom (e.g., aliphatic carbon, amide nitrogen, etc.), which were determined using the available ASA algorithms mentioned above. Predicted values for (G12 − G23)pred were determined using eq 1 and concentration-dependent values of (∂μ3/ ∂m3)T,P, which were determined from available water activity of water−PEG solutions.58,59 To compare the predicted values of (∂μ2/∂m3)T,P,m2 and (G12 − G23)pred to densimetry results gathered in the present work, predicted values for the protein partial specific volume were calculated as a function of osmolyte concentration. Starting at the measured value for V̂ 02 with no osmolyte present, the value for partial specific volume of protein as a function of c3 was

() 1

d ρ 1 + V2̂ = ρ0 dw2

RTκT − G12 + c3V3̂ (G12 − G23) MW2

(3)

where ρ is the overall solution density for a given protein concentration and solution composition and ρ0 is the solution density as the weight fraction of protein (w2) approaches zero. All results are implicitly at constant temperature (25.00 °C) and ambient pressure. A linear regression of the reciprocal of solution density as a function of w2 gives an intercept and slope, which are the first and second terms, respectively, on the righthand side of eq 3. The 95% confidence interval for the value of 3320


Article


Figure 1. Binary solutions. (A) Water−osmolyte solutions: density as a function of osmolyte weight percent for sucrose (squares), trehalose (circles), sorbitol (triangles), and poly(ethylene glycol) Mn = 6000 g/mol (diamonds). (B) Water−protein solutions: reciprocal of the density plotted as a function of AS-IgG1 weight fraction. (C) Conversion between osmolyte molarity and c3V̂ 3 values.

calculated by numerically combining (G12 − G23)pred values with eq 3. Finally, the protein chemical potential, relative to its value in pure water, was calculated by numerically integrating (∂μ2/∂m3)T,P,m2 from eq 1 and results from the densimetry measurements, or directly from the models. AS-IgG1 Unfolding via Differential Scanning Calorimetry. Capillary differential scanning calorimetry (MicroCal capDSC, Malvern Instruments, Malvern, UK) was performed on AS-IgG1 solutions at different osmolyte concentrations. A buffer−buffer scan and subsequent protein−buffer scans (1 mg/mL AS-IgG1) were performed at 1 °C/min from 20 to 95 °C. With no osmolyte present, each protein scan (after the first scan) showed no reversibility (data not shown) as expected for AS-IgG1 based on previous work.47 After subtracting the buffer scan, the absolute partial specific heat capacity, Cp, was calculated from the DSC thermograms using standard expressions.60,61 The midpoint unfolding temperature (Tm) and van’t Hoff enthalphy (ΔH) of the each discernible unfolding transition were determined using the Peak Analyzer function in Origin Pro (Origin Lab Corporation, Northampton, MA).

(diamonds). For the binary system of AS-IgG1 and water in Figure 1B, the reciprocal of the solution density is plotted as a function of weight fraction of protein. The maximum protein concentration was limited to ∼5 mg/mL to avoid potential effects of protein−protein interactions (i.e., to ensure the m2 → 0 limit holds). Both panels in Figure 1 show linear increases in solution density as a function of weight fraction of osmolyte or protein, consistent with high quality fits from regression to a linear function. Similarly linear behavior was found for each choice of osmolyte and osmolyte concentration (data not shown). The partial specific volume of the protein in a given osmolyte was determined from linear regression of the reciprocal of density as a function of protein weight fraction (e.g., as shown in Figure 1B) and calculated using eq 3. The partial specific volume for each osmolyte (V̂ 3) in the absence of protein was calculated using the data in Figure 1A in the same manner. Table 1 reports the fitted values and 95% confidence intervals for V̂ 3 of each osmolyte in water. All data with protein present are in the dilute protein limit (m2 → 0); therefore, V̂ 3 is independent of protein concentration. The data in Figure 1A illustrate that V̂ 3 is also independent of c3 for the conditions

■

RESULTS AND DISCUSSION The protein and osmolyte partial specific volumes were determined from a series of density measurements at constant osmolyte molality, at 25 °C and ambient pressure. Figure 1 depicts illustrative results for density measurements for the binary system of water and a given osmolyte. Figure 1A shows the solution density as a function of osmolyte weight fraction for sucrose (squares), trehalose (circles), sorbitol (triangles), and poly(ethylene glycol) with an average Mn = 6000 g/mol

Table 1. Osmolyte Partial Specific Volumea osmolyte sucrose trehalose sorbitol PEG (Mn = 6K) a

3321

V̂ 3 (cm3/g) 0.6206 0.6127 0.6600 0.8362

± ± ± ±

0.0007 0.0014 0.0020 0.0010

Error bars correspond to 95% confidence intervals. DOI: 10.1021/acs.jpcb.6b00621 J. Phys. Chem. B 2016, 120, 3318−3330

Article


Figure 2. Ternary solutions with dilute protein (m2 → 0). AS-IgG1 partial specific volume as a function of c3V̂ 3 for (A) sucrose, (B) PEG, (C) sorbitol, and (D) trehalose. The dashed black curves correspond to a priori preferential interaction predictions based on available models8,43,62 using an AS-IgG1 homology model based on human IgG1 crystal structure (PDB 1HZH). Error bars correspond to 95% confidence intervals. Solid curves are empirical fits to interpolate the experimental values.

Figure 2B shows V̂ 2 as a function of c3V̂ 3 for AS-IgG1 with PEG (Mn = 6000 g/mol). Initially, values of V̂ 2 increase before reaching a maximum at value of c3V̂ 3 of approximately 0.15. Similar to what was observed with sucrose, at higher PEG concentrations the values of V̂ 2 decrease. Therefore, as with sucrose, one observes preferential exclusion of PEG at low PEG concentrations followed by preferential accumulation at higher PEG concentrations. Model predictions (dashed lines in Figure 2B) are qualitatively opposite to the experimental behavior as one increases PEG concentrations to levels that are similar to those used in common practice (i.e., PEG concentrations on the order of 10−20 wt %).27 Figures 2C and 2D illustrate the behavior of the partial specific volume of AS-IgG1 as a function of c3V̂ 3 for sorbitol and trehalose, respectively. There does not appear to be a significant trend for values of V̂ 2 over these osmolyte concentrations, indicating that the magnitude of preferential interactions for this IgG and either of these osmolytes is too small to be directly observed via this experimental approach. Trehalose is often assumed to be preferentially excluded from protein surfaces and to be excluded to a greater degree than sucrose for a given protein17 The results in Figure 2 clearly show this is not the case for AS-IgG1. Existing preferential interaction models (dashed curves in Figure 2C,D) predict significant upward curvature for AS-IgG1−sorbitol solutions and qualitatively similar results for trehalose. However, the predicted increases in protein specific volume are within the error bars of the measurements, which suggest that densimetry may not be sufficiently precise to determine the weak preferential interactions in those cases.

tested here, and therefore the product c3V̂ 3 in eq 1 is a linear function of c3. Figure 1C shows this relationship graphically for each osmolyte listed in Table 1, for use by the reader to more easily convert between osmolyte molarity and c3V̂ 3 in the analysis and discussion below. Partial Specific Volume of AS-IgG1 versus Osmolyte Concentration and Identity. Figure 2 plots the values for the protein partial specific volume as a function of c3V̂ 3 for AS-IgG1 with sucrose (A), PEG (B), sorbitol (C), and trehalose (D). The solid curves in Figure 2 correspond to best fits of V̂ 2 vs c3V̂ 3, and dashed curves are a priori predictions from the preferential interaction models (cf. Methods section and Supporting Information).8,9,43,62 Inspection of Figure 2A shows values of V̂ 2 increase initially at low sucrose concentrations. However, as the concentration of sucrose increases further, the value of V̂ 2 reaches a maximum before decreasing. Based on eq 4 and inspection of the local tangent of V̂ 2 vs c3V̂ 3, (G12 − G23) is initially positive but decreases with increasing c3 until reaching zero at the maximum for V̂ 2 at approximately 0.3 M sucrose. At higher sucrose concentrations, (G12 − G23) is increasingly negative. This indicates that sucrose is preferentially excluded from the protein at lower sucrose concentrations but becomes preferentially accumulated near the protein at higher osmolyte concentrations. The dashed curve represents an a priori prediction of preferential interactions using available models8,43 and shows a shallow upward curvature for V̂ 2 with increasing osmolyte concentration. As such, the model underpredicts the preferential exclusion of sucrose at low osmolyte concentration, and is qualitatively incorrect at higher sucrose concentrations. 3322


Article


Figure 3. Average protein−water interactions (G12, dashed curves) and protein−osmolyte interactions (G23, solid curves) as a function of c3V̂ 3 for (A) sucrose, (B) PEG, and (C) sorbitol or trehalose.

Kirkwood Buff Integrals for Protein−Water and Protein−Osmolyte Interactions. Values of the protein− water KB integral (G12) and the protein−osmolyte KB integral (G23) were determined from regression of the data in Figure 2 for the cases where values of V̂ 2 changed with osmolyte concentration (i.e., sucrose and PEG). Data for V̂ 2 vs c3V̂ 3 were fit to a quadratic equation (see Supporting Information for details), and the fitted coefficients were used to calculate the KB integrals. Briefly, the tangent of V̂ 2 as a function of c3V̂ 3 is equal to (G12 − G23), and the y-intercept of the tangent line is equal to −G12 (cf. eq 4).32 Therefore, if G12 and G23 are independent of osmolyte concentration, then one expects V̂ 2 to be linear with respect to c3V̂ 3. The slope is positive (G12 > G23) for preferential exclusion of osmolytes, or vice versa for preferential accumulation of the osmolyte near the protein. The cases where V̂ 2 is independent of osmolyte concentration correspond to sufficiently weak preferential interactions that G12 and G23 are not discernably different (Figures 2C and 2D). The units for Gij are volume (or volume per molecule) when one uses concentration units of molecules/volume. Gij are reported here in terms of more experimentally conventional units (volume/mass), which can be converted to those in eq 4 by use of the protein molecular weight and Avogadro’s number. Positive (negative) values for KB integrals imply a net attractive (repulsive) interaction for the two components relative to an ideal (noninteracting) mixture, when averaged over all centerto-center distances between a molecule of component i and that of component j. As noted earlier, significant preferential interactions exist for the protein when the difference in KB integrals (i.e., G12 − G23) is non-negligible. However, the interactions between protein and water (G12) and those

between protein and osmolyte (G23) may still be large when the net preferential interactions are small. Figure 3 shows G12 (dashed curves) and G23 (solid curves) for sucrose (panel A), PEG (panel B), and sorbitol or trehalose (panel C). Inspection of Figure 3 shows all values for Kirkwood−Buff integrals are negative for both protein−water and protein−osmolyte. This follows from the large steric contribution to the KB integrals, as the osmolyte or water molecules cannot overlap with the protein. Purely steric interactions will result in negative values of any KB integrals, although they will not necessarily equal those from simple gasphase or implicit solvent estimates.63 In Figures 3A and 3B, one observes that G12 or G23 values depend strongly on osmolyte concentration. In the absence of a given osmolyte (c3 = 0), G12 is equal to −V̂ 2 in the limit of low protein concentration,32 and the initial value for G23 follows from the initial slope for V̂ 2 in Figure 2. For sucrose, G23 starts at lower values than G12, as the osmolyte is initially preferentially excluded. In Figure 3A, one observes that the value of G12 decreases with increasing sucrose concentration, suggesting a decrease in the effective concentration (or density) of water molecules near the protein surface. As G12 decreases, the value of G23 increases, suggesting an increase in the effective concentration of sucrose near the protein surface. The net result is that G12 > G23 (i.e., preferential exclusion of sucrose) at low osmolyte concentration, and this switches to the opposite behavior above a sucrose concentration of approximately 0.3 M. A qualitatively similar behavior is observed for PEG solutions (Figure 3B), but the switchover from preferentially excluded PEG to preferentially accumulated PEG occurs at a higher PEG concentration (∼10 wt %). In contrast, Figure 3B shows G12 (dashed line) and G23 (solid line) for trehalose and sorbitol 3323


Article


Figure 4. AS-IgG1 native state transfer free energy as a function of c3V̂ 3 for (A) sucrose, (B) PEG, and (C) sorbitol or trehalose. Dashed curves correspond to predictions using the preferential interaction models (see main text).

osmolytes. In contrast, there were not significant changes in V̂ 2 vs c3 in Figure 2C,D. Therefore, G12 is approximately equal to G23, implying preferential interactions were not large enough to be detected within the sensitivity of the densimetry measurements for trehalose and sorbitol. AS-IgG1 Native State Chemical Potential. As noted above and shown by eq 1, thermodynamics establishes a direct relation between preferential interactions, −c3(G12 − G23) or Γ32, and changes in chemical potential of the protein with respect to osmolyte concentration at fixed temperature and pressure for dilute protein solutions. The change in osmolyte chemical potential with respect to osmolyte concentration, (∂μ3/∂m3)T,P, is needed for this conversion between KB integrals and the derivative of the protein chemical potential with respect to c3. As mentioned above, (∂μ3/∂m3)T,P,m2→0 does not depend on the identity of the protein because it is evaluated in the limit of m2 → 0. Using values of (G12 − G23) determined from densimetry, and literature values of(∂μ3/∂m3)T,P,m2→0,14,57−59 the quantity N /∂m3)T,P,m2 was determined via eq 1 for a given (∂Δμ2,tr osmolyte concentration, where the superscript N denotes that the experimental densimetry results are all for folded or native AS-IgG1. The analysis assumes the unfolded state of AS-IgG1 is not significantly populated, as the densimetry measurements were performed at 25 °C, far below the temperature range where measurable unfolding occurs for any of the domains for this antibody (see also DSC results below). Values of ΔμN2,tr were then determined as a function of osmolyte concentration by numerically integrating the derivative with respect to osmolyte molality. Figure 4 shows the results for ΔμN2,tr as a

function of c3V̂ 3 for sucrose (A), PEG (B), and sorbitol or trehalose (C), which are based on the results in Figure 3. The solid curves in Figure 4 are based upon best fits of V̂ 2 as a function of c3V̂ 3 (shown above), and dashed curves are a priori predictions from the preferential interaction model(s) that were used to produce the dashed curves in Figure 2 (cf. Methods section and Supporting Information). Inspection of Figure 4A shows the value of ΔμN2,tr increases at low sucrose concentration, reaches a maximum value of about 20 kJ/mol at a value of c3V̂ 3 value near 0.1 and then decreases at higher osmolyte concentrations. From Figure 4B, one observes similarly quadratic behavior of ΔμN2,tr with increasing PEG concentration for AS-IgG1, and the value of μN2,tr reaches a maximum value at c3V̂ 3 slightly greater than 0.1. Notably, preferential interaction models predict the opposite behavior, with monotonically increasing ΔμN2,tr with increasing PEG concentration. In Figure 4C, values of ΔμN2,tr from densimetry for sorbitol and trehalose are not plotted, as these osmolytes did not appear to show any preferential interaction within the sensitivity of the densimetry measurement. The preferential interaction model developed by Bolen and co-workers predicts a linear increase in chemical potential for sorbitol and trehalose. The results here cannot conclusively indicate why the a priori model predictions deviate so greatly from the experimental observations for AS-IgG1. Preliminary data with a globular protein (not shown) display similar behavior to AS-IgG1 for this set of osmolytes, indicating that this is not simply an anomalous case for the IgG1. One possible simple reason for the disagreement between the model predictions and the results here could be that preferential interactions are generally dependent on osmolyte concentration(s) even for neutral 3324


Article


Figure 5. AS-IgG1 differential scanning calorimetry. DSC thermograms with increasing (A) sucrose and (B) PEG concentration. (C) CH2-Fab and (D) CH3 Tm values as a function of osmolyte concentration. In panels C and D, symbols correspond to trehalose (squares), sucrose (diamonds), sorbitol (circles), and PEG (triangles).

osmolytes, but historically it was impractical to experimentally elucidate that for many systems, or systems of interest were restricted to a relatively small range of osmolyte concentrations (e.g., the ranges encountered in particular living organisms). That would suggest that existing models could be adapted to provide improved predictions; however, it would require a highly nontrivial amount of new data to characterize quantities such as the osmolyte-concentration dependence for each type of amino acid and a given osmolyte. Independent of any model limitations or assumptions, thermodynamic arguments show that preferential exclusion (accumulation) of osmolytes leads to increases (decreases) in chemical potential of the protein. Mechanistically, the preferential exclusion of sugars and hydrophilic polymers is thought to originate from the difference in size of the osmolytes compared to water, along with the generally hydrophilic nature of those solutes. Together, these effects are often argued to promote favorable water−protein interactions at the surface of the protein, relative to protein−sugar or protein−polymer interactions.17,19 Alternatively, preferentially accumulated osmolytes such as urea are thought to have favorable interactions such as hydrogen bonding with the protein surface.44,64 In both cases, it is common to assume that the magnitude of preferential interactions scale with the solvent-exposed protein surface area. Building on that reasoning, it is common to then assume that an osmolyte that is preferentially excluded from (accumulated near) the native protein will be preferentially excluded (accumulated) to a greater extent by the unfolded protein. That assumption is based on the idea that the degree of exclusion (accumulation) scales with the amount of solvent-

exposed surface area for a given protein. The next subsection considers whether those assumptions hold in the present case. AS-IgG1 Unfolding via DSC. With the above arguments in mind, and the relatively large changes in chemical potential observed in Figure 4, one would expect biophysical properties of the IgG, such as the free energy of unfolding to be significantly altered in the presence of osmolytes such as sucrose and PEG. Similarly, on the basis of the small changes in protein partial specific volume in Figures 2C and 2D, one would anticipate that trehalose and sorbitol would be much less effective in that regard. To help test that hypothesis, differential scanning calorimetry was performed on AS-IgG1 solutions over similar concentration ranges of sucrose, trehalose, sorbitol, and PEG as those used in the densimetry measurements. Figures 5A and 5B illustrate DSC profiles as a function of sucrose and PEG concentration, respectively. DSC profiles are vertically offset for easier viewing, and the arrow denotes the direction of increasing osmolyte concentration. Prior work with AS-IgG1 and other mAbs showed that the number of unfolding transition endotherms (peaks or shoulders) depends on pH.46,47,65,66 At low pH (e.g., pH 4) three transitions were observed, and prior work showed that the unfolding transitions correspond, respectively, from the lowest to highest transition temperatures: the constant heavy chains of the second domain (CH2), the fragment antigen binding region (Fab), and the constant heavy chains of the third domain (CH3). In contrast, at pH 6, the unfolding transitions for the CH2 and Fab overlap, and therefore only two transitions were observed by visible inspection.46,47,66 From inspection of Figure 5A one observes two unfolding transitions, consistent with prior work at pH 6. The larger peak 3325


Article


Figure 6. Schematic of (A) classic theory and (B) concentration-dependent preferential interactions manifesting as changes in unfolding thermodynamics. Panels illustrate qualitative behavior of the protein chemical potential for the native (N) and unfolded (U) states as a function of increasing osmolyte concentration.

sorbitol.43 However, doing so requires one to assume a magnitude for the change in the ASA between the native and denatured states of the protein, which is highly nontrival55 (see Supporting Information). In addition, the DSC thermograms are not at equilibrium because unfolding is convoluted with aggregation, and it is not clear whether one should consider the lower Tm transition to be simply an overlap of CH2 and Fab transitions in the DSC signal or a single transition that requires both domains to unfold as a group. Those quantitative limitations notwithstanding, it is possible to at least derive qualitative conclusions by comparing the results in Figures 2−4 with those in Figures 5 and 6. From the densimetry data for sucrose and for PEG solutions, the native state chemical potential for AS-IgG1 increases within increasing c3 at low c3, reaches a maximum, then decreases with further increases in c3. For trehalose and sorbitol, it is possible that the native state chemical potential of AS-IgG1 increases with increasing c3, but this was not discernible experimentally. In what follows below, we assume that it was simply too weak of an effect to be detected with these experimental methods. The DSC results qualitatively show that the chemical potential dif ference upon unfolding of the CH3 domain of ASIgG1 increases with increasing c3 for sucrose, trehalose, and sorbitol, while it decreases for PEG as the osmolyte. A similar argument holds for the overlapping transition of the Fab and CH2 domains. Even though the DSC thermograms are not reversible, irreversible processes such as aggregation do not cause increases in Tm values and only cause decreases.67 These results for Tm vs c3 were unexpected based on the densimetry results and canonical models for protein unfolding in the presence of osmolytes.26 That is, canonical expectations are that osmolytes such as sucrose, trehalose, and PEG are always preferentially excluded from the native state and excluded even more so from the unfolded state.16,68 As a result, the chemical potential of both the folded and unfolded states are expected to increase by adding such osmolytes, but the increase is greater for the unfolded state. This has the net effect of increasing the unfolding free energy, and Tm values, upon addition of such osmolytes. If one were to consider the DSC results for sucrose or trehalose in isolation, they would normally be interpreted mechanistically in that way.16−18 That would also commonly be done for sorbitol, although it has been found in some cases that sorbitol can change from being preferentially excluded to

at lower temperatures corresponds to overlapping unfolding transitions of the CH2 and Fab, while the smaller peak at higher temperatures corresponds to that of the CH3. The apparent midpoint unfolding temperature (Tm) increases with increasing sucrose concentration (see Figure 5C). However, for the conditions tested, the unfolding transitions were not fully reversible and may be convoluted with aggregation (data not shown). As such, the transition temperatures are referred to only as apparent Tm values. For trehalose, sucrose, and sorbitol, the CH2/Fab unfolding transitions occurred together even with the addition of more osmolyte (see Supporting Information for trehalose and sorbitol thermograms). However, as PEG concentration increases, the CH2 and Fab peaks become distinguishable (cf. Figure 5B). Additionally, the Fab unfolding transition temperature decreases more so than for that of the CH2. This suggests PEG preferentially interacts with (accumulates near) the Fab to a greater degree that it does with the CH2. If one assumes the change in solvent exposed surface area of a given domain scales with the size of the folded domain, then one would expect preferential interactions to be largest for the Fab domains, smaller for the CH3 domain, and smallest for the CH2 domain of a typical antibody. That may help to rationalize the above results, although the data here cannot rule out the possibility of specific interactions of PEG with the Fab domains. Tm values for CH2/Fab peaks and CH3 peaks were interpolated numerically (see Methods section). Figure 5C (5D) plots Tm values as a function of osmolyte concentration for CH2/Fab (CH3) unfolding transition. For PEG, Fab Tm values are reported since the Fab transition was distinguishable in that case. In both figures, symbols correspond to sucrose (diamonds), trehalose (squares), sorbitol (circles), and PEG (triangles). Solid lines in Figures 5C,D show linear fits determined from regression for each data set. Interestingly, one observes from Figure 5C,D that sucrose, trehalose, and sorbitol increase AS-IgG1 unfolding transitions, while PEG decreases the transitions. Preferential Interactions and Their Effect on Protein Unfolding and Stability. The free energy of unfolding is determined by the difference in the protein chemical potential of the native and denatured state. The preferential interaction model of Bolen and co-workers can be used to predict unfolding free energy changes for sucrose, trehalose, and 3326


Article


While it is common in current practice to assume that the osmolytes obey one of the two scenarios in Figure 6A, Timasheff69 and others70 provided cautionary examples to the contrary. Prior work has shown that transfer free energy models were able to predict ΔμN−D with the addition of osmolytes for 2 some proteins.71 Those models rely on accurate solvent accessible surface area values, which is difficult to determine for unfolded proteins.72 If the only issue was accurate surface area calculations, then the predictions should only differ quantitatively (not qualitatively) from the experimental behavior for this IgG. The discussion above highlights potential reasons for these disagreements but generally supports the idea that preferential interactions change with osmolyte concentration and can even switch qualitatively. Recent data also highlight that PEG interactions depend on the size of the PEG molecules in question and that PEG preferential interactions are not simply the steric repulsions expected for an idealized hydrophilic polymer.8,62 Despite those previous findings, sucrose and high-molecular-weight PEG are historically canonical examples of osmolytes that not only are preferentially excluded but also do not show significant changes in their behavior as a function of osmolyte concentration.8,27,62,68 However, preferential interaction coefficients have been shown to be concentration dependent for some neutral osmolytes73,74 and recent work has suggested that concentrations dependence of preferential interaction may be due to a balance of enthalpic and entropic interactions.75,76 The results here confirm that caution is needed in assuming such simplified models hold for a given protein, and they also illustrate experimental means to test such assumptions. Finally, the results for sucrose may be of interest from another context. Sucrose and trehalose have long been used as lyoprotectants and cryoprotectants for proteins during freezing and drying processes.19,21 While sucrose and trehalose have similar chemical structure, many studies have found that these molecules interact differently with proteins.19,21 In the current report, trehalose increased thermal unfolding of AS-IgG1 more so than sucrose. On the other hand, sucrose showed larger preferential exclusion from AS-IgG1 at low osmolyte concentrations, followed by preferential accumulation at higher osmolyte concentrations, while preferential interactions with trehalose were not significant. It is not clear from the present results why, at a molecular level, trehalose and sucrose have such different net interactions with this IgG, as G12 and G23 are ensemble-averaged quantities. One qualitative mechanism by which sucrose and trehalose putatively operate to stabilize proteins in that context is by serving as “water replacers”, in that they are able to hydrogen bond to the protein as the protein “loses” water due to water removal by drying or freezing of ice.17,19 Such a “waterreplacement” process would be expected to manifest as preferential accumulation of the sugar molecules at the protein surface, at least for sufficiently high sugar concentrations. The present results extend only to sugar concentrations on the order of 30 w/w %, which is a factor of 2 or more lower than sugar concentrations in freeze-concentrated or freeze-dried states.77 However, from a qualitative perspective the results for sucrose clearly indicate that a switchover from preferential exclusion to preferential accumulation is possible and even indicate that it occurs more readily for sucrose than for trehalose in the case of AS-IgG1. While it is beyond the scope of this report, preliminary results with a globular protein show qualitatively similar behavior and suggest this observation is

preferentially accumulated (e.g., as a function of temperature).69 The DSC results for PEG are unexpected for multiple reasons. PEG is not expected to be preferentially accumulated at the protein surface unless one deals with much smaller PEG molecules,8 and addition of PEG is not expected to decrease Tm values.27 Furthermore, the results here show that PEG interacts more strongly (greater preferential accumulation) with the Fab domains of AS-IgG1 than it does with the CH2 domain. This is evidenced by the observation that addition of PEG causes the peak for the Fab domain to become discernible from the CH2 domain in Figure 5B (top). While the DSC results for sorbitol, sucrose, and trehalose would suggest the canonical models or interpretations hold for AS-IgG1, the densimetry results show this is not the case. Sucrose is preferentially excluded from the native state at low c3 but accumulated at high c3. If the same were to hold for the unfolded state, then the Tm values should increase with increasing c3 at low c3 but decrease at higher c3. From Figure 5, this is clearly not the case. For PEG, it is similarly complex. By similar reasoning to that for sucrose, one would expect the Tm values to first increase and then decrease with increasing PEG concentration. Instead, PEG is a destabilizer (in terms of unfolding) at all PEG concentrations tested here. Finally for sorbitol and trehalose, the canonical model might be correct, but the data here cannot discriminate whether that is the case. To help summarize the key results and their interpretation graphically, Figure 6A illustrates the qualitative behavior expected from canonical interpretations preferential interactions by plotting the protein chemical potential as a function of osmolyte concentration, for native (N) and unfolded (U) states. A simple two-state unfolding example is used for ease of depiction. For a multistate protein such as an antibody, one could generalize to have unfolding intermediates in addition to the N and U states. Figure 6A (top) shows the canonical example described above for preferential exclusion. The opposite behavior shown in the bottom panel is expected for classic chemical denaturants such as urea.27,68 The black bar denotes the native state (N) and the gray bar denotes the unfolded state (U). Units on the y-axis are arbitrary for the chemical potential of a given state. Double arrows are included to more easily see how the ) increases or decreases with a unfolding free energy (ΔμN−D 2 change in osmolyte concentration. The results in Figure 5 for sorbitol and trehalose are consistent with Figure 6A, although with relatively small changes to the native state chemical potential with increasing osmolyte concentration. Figure 6B shows a schematic that is more appropriate for the cases of AS-IgG1 with sucrose (top) and PEG (bottom). The chemical potential of the native state changes nonmonotonically with sucrose concentration, but the value of Δμ2N−D increases monotonically with the addition of sucrose. As such, preferential interactions of sucrose with the unfolded state must also depend on sucrose concentration in a nontrivial way. For PEG, it is a similar but reversed scenario to sucrose, in that ΔμN−D decreases monotonically with the addition of PEG. 2 Overall, these results indicate that preferential interactions of osmolytes with the native state should not be generally assumed to be predictive of interactions with the unfolded state(s). Osmolytes that are preferentially excluded in the native state (e.g., PEG or sucrose) may increase or decrease protein unfolding free energies. 3327



■

more general than what is shown here for AS-IgG1; that study will be the topic of a future report.

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.6b00621. Additional details regarding the density data and analysis, DSC thermograms for trehalose and sorbitol, and details about predicted preferential interaction (PDF)

■

REFERENCES

(1) Cheung, J. K. Insights Into Crowding Effects on Protein Stability From a Coarse-Grained Model. J. Biomech. Eng. 2009, 131, 071002. (2) Shen, V. K.; Cheung, J. K.; Errington, J. R.; Truskett, T. M. Coarse-Grained Strategy for Modeling Protein Stability in Concentrated Solutions. II: Phase Behavior. Biophys. J. 2006, 90, 1949−1960. (3) Annunziata, O.; Asherie, N.; Lomakin, A.; Pande, J.; Ogun, O.; Benedek, G. B. Effect of Polyethylene Glycol on the Liquid−liquid Phase Transition in Aqueous Protein Solutions. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 14165−14170. (4) D’Amico, M.; Raccosta, S.; Cannas, M.; Martorana, V.; Manno, M. Existence of Metastable Intermediate Lysozyme Conformation Highlights the Role of Alcohols in Altering Protein Stability. J. Phys. Chem. B 2011, 115, 4078−4087. (5) Zimmerman, S. B.; Trach, S. O. Estimation of Macromolecule Concentrations and Excluded Volume Effects for the Cytoplasm of Escherichia Coli. J. Mol. Biol. 1991, 222, 599−620. (6) Ellis, R. J. Macromolecular Crowding: An Important but Neglected Aspect of the Intracellular Environment. Curr. Opin. Struct. Biol. 2001, 11, 114−119. (7) Scherer, T. M. Cosolute Effects on the Chemical Potential and Interactions of an IgG1Monoclonal Antibody at High Concentrations. J. Phys. Chem. B 2013, 117, 2254−2266. (8) Knowles, D. B.; Shkel, I. A.; Phan, N. M.; Sternke, M.; Lingeman, E.; Cheng, X.; Cheng, L.; O’Connor, K.; Record, M. T. Chemical Interactions of Polyethylene Glycols (PEGs) and Glycerol with Protein Functional Groups: Applications to Effects of PEG and Glycerol on Protein Processes. Biochemistry 2015, 54, 3528−3542. (9) Auton, M.; Bolen, D. W. Predicting the Energetics of OsmolyteInduced Protein Folding/unfolding. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 15065−15068. (10) Schneck, E.; Horinek, D.; Netz, R. R. Insight into the Molecular Mechanisms of Protein Stabilizing Osmolytes from Global Force-Field Variations. J. Phys. Chem. B 2013, 117, 8310−8321. (11) Banks, D. D.; Latypov, R. F.; Ketchem, R. R.; Woodard, J.; Scavezze, J. L.; Siska, C. C.; Razinkov, V. I. Native-State Solubility and Transfer Free Energy as Predictive Tools for Selecting Excipients to Include in Protein Formulation Development Studies. J. Pharm. Sci. 2012, 101, 2720−2732. (12) Trevino, S. R.; Scholtz, J. M.; Pace, C. N. Measuring and Increasing Protein Solubility. J. Pharm. Sci. 2008, 97, 4155−4166. (13) Nick Pace, C.; Trevino, S.; Prabhakaran, E.; Martin Scholtz, J. Protein Structure, Stability and Solubility in Water and Other Solvents. Philos. Trans. R. Soc., B 2004, 359, 1225−1235. (14) Rösgen, J.; Pettitt, B. M.; Bolen, D. W. Protein Folding, Stability, and Solvation Structure in Osmolyte Solutions. Biophys. J. 2005, 89, 2988−2997. (15) Auton, M.; Bolen, D. W. Additive Transfer Free Energies of the Peptide Backbone Unit That Are Independent of the Model Compound and the Choice of Concentration Scale †. Biochemistry 2004, 43, 1329−1342. (16) Lee, J. C.; Timasheff, S. N. The Stabilization of Proteins by Sucrose. J. Biol. Chem. 1981, 256, 7193−7201. (17) Jain, N. K.; Roy, I. Effect of Trehalose on Protein Structure. Protein Sci. 2008, 18, 24−36. (18) Arakawa, T.; Timasheff, S. N. Stabilization of Protein Structure by Sugars. Biochemistry 1982, 21, 6536−6544. (19) Chang, L. L.; Shepherd, D.; Sun, J.; Ouellette, D.; Grant, K. L.; Tang, X. C.; Pikal, M. J. Mechanism of Protein Stabilization by Sugars during Freeze-Drying and Storage: Native Structure Preservation, Specific Interaction, And/or Immobilization in a Glassy Matrix? J. Pharm. Sci. 2005, 94, 1427−1444. (20) Stärtzel, P.; Gieseler, H.; Gieseler, M.; Abdul-Fattah, A. M.; Adler, M.; Mahler, H.-C.; Goldbach, P. Freeze Drying of L-Arginine/ Sucrose-Based Protein Formulations, Part I: Influence of Formulation and Arginine Counter Ion on the Critical Formulation Temperature, Product Performance and Protein Stability. J. Pharm. Sci. 2015, 104, 2345−2358.

CONCLUSIONS Interactions of AS-IgG1 with water and common neutral osmolytes were quantified in terms of Kirkwood−Buff integrals, determined from protein partial specific volume measurements. Sucrose and PEG were preferentially excluded from the protein at low osmolyte concentrations but, surprisingly, became preferentially accumulated at higher concentrations. In contrast, sorbitol and trehalose showed no significant preferential interactions with AS-IgG1 (within the sensitivity of the measurements). Determining Kirkwood−Buff integrals (G12 and G23) offers a direct link to changes in the protein chemical potential with osmolytes, beyond what can be obtained from preferential-interaction measurements that can only probe the difference between G12 and G23. The resulting transfer free energy (ΔμN2,tr) for transfer of protein from water to aqueous osmolyte solutions may therefore depend strongly on osmolyte concentration and can be large compared to typical unfolding free energies. It was also considered whether AS-IgG1 native chemical potential or a priori preferential interaction models based on ASA are indicative of changes in unfolding free energy as a function of osmolyte type and concentration. Thermal denaturation via DSC illustrated net stabilization with trehalose, sorbitol, and sucrose but destabilization with PEG. Particularly, PEG destabilized the Fab domain(s) more so than the CH2 or CH3 domains, suggesting preferential interactions may be specific to different regions of AS-IgG1. The concentrationdependent densimetry results suggest competing protein− osmolyte and protein−water interactions may change with osmolyte concentration. The results are in contrast to some commonly assumed preferential-interaction models that assume the magnitude of preferential interactions scales with the solvent-exposed surface area and highlight potential pitfalls when adopting approaches that assume that experimental or theoretical transfer free energies will be predictive of protein stability (unfolding free energy) as a function of osmolyte type and concentration.

■

Article

AUTHOR INFORMATION

Corresponding Author

*Tel 302-831-0838; fax 302-831-1048; e-mail [email protected] (C.J.R.). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS G.V.B. and C.J.R. gratefully acknowledge support from Amgen, the National Institute of Standards and Technology (NIST 70NANB12H239), and the National Science Foundation (CBET 0931173). 3328


Article

The Journal of Physical Chemistry B (21) Hedoux, A.; Paccou, L.; Achir, S.; Guinet, Y. Mechanism of Protein Stabilization by Trehalose during Freeze-Drying Analyzed by in Situ Micro-Raman Spectroscopy. J. Pharm. Sci. 2013, 102, 2484− 2494. (22) Atha, D. H.; Ingham, K. C. Mechanism of Precipitation of Proteins by Polyethylene Glycols. Analysis in Terms of Excluded Volume. J. Biol. Chem. 1981, 256, 12108−12117. (23) Mukherjee, S. K.; Gautam, S.; Biswas, S.; Kundu, J.; Chowdhury, P. K. Do Macromolecular Crowding Agents Exert Only an Excluded Volume Effect? A Protein Solvation Study. J. Phys. Chem. B 2015, 119, 14145−14156. (24) Arakawa, T.; Timasheff, S. N. Preferential Interactions of Proteins with Salts in Concentrated Solutions. Biochemistry 1982, 21, 6545−6552. (25) Arakawa, T.; Timasheff, S. N. Mechanism of Poly(ethylene Glycol) Interaction with Proteins. Biochemistry 1985, 24, 6756−6762. (26) Timasheff, S. N. The Control of Protein Stability and Association by Weak Interactions with Water: How Do Solvents Affect These Processes? Annu. Rev. Biophys. Biomol. Struct. 1993, 22, 67−97. (27) Bhat, R.; Timasheff, S. N. Steric Exclusion Is the Principal Source of the Preferential Hydration of Proteins in the Presence of Polyethylene Glycols. Protein Sci. 1992, 1, 1133−1143. (28) Schellman, J. A. Protein Stability in Mixed Solvents: A Balance of Contact Interaction and Excluded Volume. Biophys. J. 2003, 85, 108−125. (29) Schellman, J. A. Selective Binding and Solvent Denaturation. Biopolymers 1987, 26, 549−559. (30) Schellman, J. A. Temperature, Stability, and the Hydrophobic Interaction. Biophys. J. 1997, 73, 2960−2964. (31) Schellman, J. A. The Thermodynamics of Solvent Exchange. Biopolymers 1994, 34, 1015−1026. (32) Ben-Naim, A. Statistical Thermodynamics for Chemists and Biochemists, softcover reprint of the original 1st ed. 1992 edition; Springer: S.l., 2013. (33) Kirkwood, J. G.; Buff, F. P. The Statistical Mechanical Theory of Solutions. I. J. Chem. Phys. 1951, 19, 774−777. (34) Hall, D. G. Kirkwood-Buff Theory of Solutions. An Alternative Derivation of Part of It and Some Applications. Trans. Faraday Soc. 1971, 67, 2516−2524. (35) Schurr, J. M.; Rangel, D. P.; Aragon, S. R. A Contribution to the Theory of Preferential Interaction Coefficients. Biophys. J. 2005, 89, 2258−2276. (36) Casassa, E. F.; Eisenberg, H. Thermodynamic Analysis of Multicomponent Solutions. Adv. Protein Chem. 1964, 19, 287−395. (37) Scatchard, G. Physical Chemistry of Protein Solutions. I. Derivation of the Equations for the Osmotic Pressure1. J. Am. Chem. Soc. 1946, 68, 2315−2319. (38) Debenedetti, P. G. Metastable Liquids: Concepts and Principles; Princeton University Press: 1996. (39) Nain, A. K. Inversion of the Kirkwood−Buff Theory of Solutions: Application to Tetrahydrofuran + Aromatic Hydrocarbon Binary Liquid Mixtures. J. Solution Chem. 2008, 37, 1541−1559. (40) Ben-Naim, A. Molecular Theory of Solutions; Oxford University Press: 2006. (41) Smith, P. E. On the Kirkwood−Buff Inversion Procedure. J. Chem. Phys. 2008, 129, 124509. (42) Courtenay, E. S.; Capp, M. W.; Anderson, C. F.; Record, M. T. Vapor Pressure Osmometry Studies of Osmolyte−Protein Interactions: Implications for the Action of Osmoprotectants in Vivo and for the Interpretation of “Osmotic Stress” Experiments in Vitro †. Biochemistry 2000, 39, 4455−4471. (43) Auton, M.; Bolen, D. W.; Rösgen, J. Structural Thermodynamics of Protein Preferential Solvation: Osmolyte Solvation of Proteins, Aminoacids, and Peptides. Proteins: Struct., Funct., Genet. 2008, 73, 802−813. (44) Auton, M.; Holthauzen, L. M. F.; Bolen, D. W. Anatomy of Energetic Changes Accompanying Urea-Induced Protein Denaturation. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 15317−15322.

(45) Wu, P.; Bolen, D. W. Osmolyte-Induced Protein Folding Free Energy Changes. Proteins: Struct., Funct., Genet. 2006, 63, 290−296. (46) Brummitt, R. K.; Nesta, D. P.; Chang, L.; Chase, S. F.; Laue, T. M.; Roberts, C. J. Nonnative Aggregation of an IgG1 Antibody in Acidic Conditions: Part 1. Unfolding, Colloidal Interactions, and Formation of High-Molecular-Weight Aggregates. J. Pharm. Sci. 2011, 100, 2087−2103. (47) Kim, N.; Remmele, R. L.; Liu, D.; Razinkov, V. I.; Fernandez, E. J.; Roberts, C. J. Aggregation of Anti-Streptavidin Immunoglobulin gamma-1 Involves Fab Unfolding and Competing Growth Pathways Mediated by pH and Salt Concentration. Biophys. Chem. 2013, 172, 26−36. (48) Barnett, G. V.; Razinkov, V. I.; Kerwin, B. A.; Laue, T. M.; Woodka, A. H.; Butler, P. D.; Perevozchikova, T.; Roberts, C. J. Specific-Ion Effects on the Aggregation Mechanisms and Protein− Protein Interactions for Anti-Streptavidin Immunoglobulin Gamma-1. J. Phys. Chem. B 2015, 119, 5793−5804. (49) Pjura, P. E.; Paulaitis, M. E.; Lenhoff, A. M. Molecular Thermodynamic Properties of Protein Solutions from Partial Specific Volumes. AIChE J. 1995, 41 (4), 1005−1009. (50) Lewis, G. N.; Randall, M. Thermodynamics; McGraw Hill: 1961. (51) Ogunnaike, B. A. Random Phenomena: Fundamentals of Probability and Statistics for Engineers; CRC Press: 2011. (52) Green, D.; Perry, R. Perry’s Chemical Engineers’ Handbook, 8th ed.; McGraw Hill Professional: 2007. (53) Fraczkiewicz, R.; Braun, W. Exact and Efficient Analytical Calculation of the Accessible Surface Areas and Their Gradients for Macromolecules. J. Comput. Chem. 1998, 19, 319−333. (54) Tsodikov, O. V.; Record, M. T.; Sergeev, Y. V. Novel Computer Program for Fast Exact Calculation of Accessible and Molecular Surface Areas and Average Surface Curvature. J. Comput. Chem. 2002, 23, 600−609. (55) Estrada, J.; Bernadó, P.; Blackledge, M.; Sancho, J. ProtSA: A Web Application for Calculating Sequence Specific Protein Solvent Accessibilities in the Unfolded Ensemble. BMC Bioinf. 2009, 10, 104. (56) Bernadó, P.; Blackledge, M.; Sancho, J. Sequence-Specific Solvent Accessibilities of Protein Residues in Unfolded Protein Ensembles. Biophys. J. 2006, 91, 4536−4543. (57) Galmarini, M. V.; Chirife, J.; Zamora, M. C.; Pérez, A. Determination and Correlation of the Water Activity of Unsaturated, Supersaturated and Saturated Trehalose Solutions. LWT - Food Sci. Technol. 2008, 41, 628−631. (58) Ninni, L.; Camargo, M. S.; Meirelles, A. J. A. Water Activity in Poly(ethylene Glycol) Aqueous Solutions. Thermochim. Acta 1999, 328, 169−176. (59) Salabat, A.; Nasirzadeh, K. Measurement and Prediction of Water Activity in PEG + (NH4)2SO4 + H2O Systems Using Polymer Scaling Laws. J. Mol. Liq. 2003, 103−104, 349−358. (60) Andrews, J. M.; Roberts, C. J. Non-Native Aggregation of αChymotrypsinogen Occurs through Nucleation and Growth with Competing Nucleus Sizes and Negative Activation Energies †. Biochemistry 2007, 46, 7558−7571. (61) Privalov, P. L.; Potekhin, S. A. [2] Scanning Microcalorimetry in Studying Temperature-Induced Changes in Proteins. Methods Enzymol. 1986, 131, 4−51. (62) Shkel, I. A.; Knowles, D. B.; Record, M. T. Separating Chemical and Excluded Volume Interactions of Polyethylene Glycols with Native Proteins: Comparison with PEG Effects on DNA Helix Formation. Biopolymers 2015, 103, 517−527. (63) Siderius, D. W.; Krekelberg, W. P.; Roberts, C. J.; Shen, V. K. Osmotic Virial Coefficients for Model Protein and Colloidal Solutions: Importance of Ensemble Constraints in the Analysis of Light Scattering Data. J. Chem. Phys. 2012, 136, 175102. (64) Moeser, B.; Horinek, D. Unified Description of Urea Denaturation: Backbone and Side Chains Contribute Equally in the Transfer Model. J. Phys. Chem. B 2014, 118, 107−114. (65) Ionescu, R. M.; Vlasak, J.; Price, C.; Kirchmeier, M. Contribution of Variable Domains to the Stability of Humanized IgG1Monoclonal Antibodies. J. Pharm. Sci. 2008, 97, 1414−1426. 3329


Article

The Journal of Physical Chemistry B (66) Wu, H.; Kroe-Barrett, R.; Singh, S.; Robinson, A. S.; Roberts, C. J. Competing Aggregation Pathways for Monoclonal Antibodies. FEBS Lett. 2014, 588, 936−941. (67) Sanchez-Ruiz, J. M. Theoretical Analysis of Lumry-Eyring Models in Differential Scanning Calorimetry. Biophys. J. 1992, 61, 921. (68) Timasheff, S. N. Protein Hydration, Thermodynamic Binding, and Preferential Hydration. Biochemistry 2002, 41, 13473−13482. (69) Xie, G.; Timasheff, S. N. Temperature Dependence of the Preferential Interactions of Ribonuclease A in Aqueous Co-Solvent Systems: Thermodynamic Analysis. Protein Sci. 1997, 6, 222−232. (70) Schneider, C. P.; Trout, B. L. Investigation of Cosolute−Protein Preferential Interaction Coefficients: New Insight into the Mechanism by Which Arginine Inhibits Aggregation. J. Phys. Chem. B 2009, 113, 2050−2058. (71) Auton, M.; Rösgen, J.; Sinev, M.; Holthauzen, L. M. F.; Bolen, D. W. Osmolyte Effects on Protein Stability and Solubility: A Balancing Act between Backbone and Side-Chains. Biophys. Chem. 2011, 159, 90−99. (72) Gong, H.; Rose, G. D. Assessing the Solvent-Dependent Surface Area of Unfolded Proteins Using an Ensemble Model. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 3321−3326. (73) Harries, D.; Rösgen, J. A Practical Guide on How Osmolytes Modulate Macromolecular Properties; Biology, B.-M. in C., Ed.; Biophysical Tools for Biologists, Vol. One: In Vitro Techniques; Academic Press: 2008; Vol. 84, pp 679−735. (74) Rosgen, J.; Pettitt, B. M.; Bolen, D. W. An Analysis of the Molecular Origin of Osmolyte-Dependent Protein Stability. Protein Sci. 2007, 16, 733−743. (75) Sapir, L.; Harries, D. Is the Depletion Force Entropic? Molecular Crowding beyond Steric Interactions. Curr. Opin. Colloid Interface Sci. 2015, 20, 3−10. (76) Sarkar, M.; Smith, A. E.; Pielak, G. J. Impact of Reconstituted Cytosol on Protein Stability. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 19342−19347. (77) Starzak, M.; Mathlouthi, M. Temperature Dependence of Water Activity in Aqueous Solutions of Sucrose. Food Chem. 2006, 96, 346− 370.

3330


Osmolyte Effects on Monoclonal Antibody Stability and Concentration

Recommend Documents