Increasing IgG Concentration Modulates the Conformational

Apr 3, 2009 - The asynchronous spectra reveal variation in β-sheet and turn regions occur before intensity variations in disordered and α-helical re...
0 downloads 0 Views 1MB Size
Download PDF
J. Phys. Chem. B 2009, 113, 6109–6118

6109

Increasing IgG Concentration Modulates the Conformational Heterogeneity and Bonding Network that Influence Solution Properties Tim J. Kamerzell,* Sonoko Kanai, Jun Liu, Steven J. Shire, and Y. John Wang Department of Late Stage Pharmaceutical and Processing DeVelopment, Genentech, Inc., 1 DNA Way, South San Francisco, California 94080 ReceiVed: January 7, 2009; ReVised Manuscript ReceiVed: March 6, 2009

Multiple molecular driving forces mediate protein stability, association, and recognition in concentrated solutions. Here we investigate the interactions that modulate the nonideal solution behavior of two immunoglobulins (IgG1s) in highly concentrated solutions using two-dimensional vibrational correlation spectroscopy (2D-COS) and principal components analysis (PCA). A specific sequence of changes is observed in the concentration-dependent vibrational spectra of the highly viscous IgG solution that deviates from ideality, whereas that sequence is reversed for all other conditions examined. The asynchronous spectra reveal variation in β-sheet and turn regions occur before intensity variations in disordered and R-helical regions as the concentration is increased for the highly viscous regime. This is in contrast to the sequence observed for all other conditions studied and to the idea that β-sheet regions are resistant to concentration-dependent affects. Finally, we show that increased hydrogen bonding and electrostatics primarily modulate the intermolecular association and nonideal behavior. Specifically, 2D-COS and PCA analysis of the amide II region suggests that Glu and Asp residues trigger the change resulting in increased viscosity and association of one IgG. Introduction Investigating protein and solution behavior in “crowded” or highly concentrated conditions is critical to our understanding of physiological and cellular function as well as the stability, safety, and efficacy of biological therapeutics.1-4 In vivo, proteins function in highly crowded environments where the concentration of molecules approximates 400 g/L or up to 40% of the total cellular volume.4,5 Often homogeneous systems are used to build models that describe functionally complex systems. These studies of increasing macromolecular concentration have been used to describe association rates,6-8 association equilibria,8,9 protein conformational changes,8,10 protein activity,2,8 and protein stability.3,8,11 Similarly, the association state of proteins and model solutes in crowded and concentrated environments have been linked to nonideal solution behavior of which models are beginning to emerge.2,12-19 Our understanding of the effects of highly concentrated and volume occupied solutions on protein structure and stability, however, is incomplete. Recently, the effects of increasing protein concentration on the stability and safety of biological therapeutics have gained significant attention from the biotechnology industry and the Food and Drug Administration (FDA).13,20-27 The physicochemical stability of biological therapeutics may be negatively affected simply by increasing protein concentration. Chemical instability typically follows first-order kinetics with regard to concentration; however, physical instability may result in complex higher order processes. It has been shown that increasing immunoglobulin (IgG) concentration increases selfassociation of these molecules resulting in increased nonideal solution properties and significantly affects the viscosity and rheological behavior.13,14,24-26 The purpose of this work is to better understand the molecular level processes that govern protein self-association and how * To whom correspondence should be addressed. Tel.: 650-225-6630. Fax: 650-225-3613. E-mail: [email protected].

these details modulate the solution properties of two previously studied IgGs.12,13 Two-dimensional correlation vibrational spectroscopy and principal components analysis was used to monitor changes in the vibrational characteristics of these IgG molecules as a function of concentration and solution condition. The synchronous and asynchronous correlation spectra reveal important molecular properties that influence self-association and various solution behaviors. Experimental Procedures (Materials and Methods) Two IgG1 full length monoclonal antibodies (MAb1 and MAb2) comprised of κ-light chains constructed from identical human frameworks were used in this study. The difference in amino acid composition is localized to CDR regions. These antibodies were cloned, expressed in Chinese hamster ovary cell lines, and purified at Genentech (South San Francisco, CA). All reagents were ACS grade. The buffer solutions used in this study were 30 mM histidine, pH 6.0, (150 mM NaCl. Fourier Transform Infrared Spectroscopy. Infrared spectra were collected at 25 °C and a resolution of 4 cm-1 using an attenuated total internal reflectance (ATR) accessory mounted in a Nicolet FTIR spectrometer (Thermo Scientific, Waltham, MA). The spectrometer and chamber enclosing the sample trough were continuously purged with dry air, which was controlled using two flow meters. The concentrations of samples were compared before and after spectral collection to ensure all sample concentrations remained constant. Approximately 256 scans were coaveraged per spectrum using a zinc selenide (ZnSe) crystal with a 45° incidence angle. In this study, various pretreatment methods were applied to each individual spectrum. A series of pretreatment steps were performed according to Czarnik-Matusewicz et al.28 First, the spectra were normalized for the minimum of penetration depth and dependence of penetration depth on concentration. Next, the contribution of water was subtracted using a second-order

10.1021/jp9001548 CCC: $40.75  2009 American Chemical Society Published on Web 04/03/2009

6110 J. Phys. Chem. B, Vol. 113, No. 17, 2009

Kamerzell et al.

least-squares fit and baseline correction.29 Finally, all spectra were normalized for concentration using the intensity integral in the range 1720-1220 cm-1. Two-Dimensional Correlation Spectroscopy. In this work, FTIR spectroscopy was used to monitor the concentration dependence of an IgG and the subsequent spectral intensity variations along the concentration and frequency axis. Matrices of m rows of spectral traces and n columns of spectral intensity variations were created from the concentration-dependent FTIR spectra. From each matrix, the synchronous and asynchronous spectra were calculated according to Noda30-33 using routine functions in Matlab R2007a. The discrete data collected from the IR measurements were represented in matrix notation as previously described by Noda:34

[

y˜(ν1, c1) y˜(ν2, c1) y˜(ν1, c2) y˜(ν2, c2) Y) ... ... y˜(ν1, cm) y˜(ν1, cm)

... y˜(νn, c1) ... y˜(νn, c2) ... ... ... y˜(νn, cm)

]

wherey˜(ν,c) is the set of dynamic spectra, from which the synchronous 2D correlation spectrum is defined as the inner product where each column of the matrix Y is a vector:

Φ(ν1, ν2) )

1 y˜(ν )T y˜(ν2) m-1 1

or

1 YTY m-1 (covariance matrix)

Φ)

It has also been shown that the synchronous spectrum is similar to the covariance matrix if the measurements are recorded at fixed intervals.34 The covariance matrix (C) of n sets of variates (X1),..., (Xn), was calculated and defined as

Cijmn ) 〈(xi - µi)m(xj - µj)n 〉 where µi is the mean. The diagonal elements of the covariance matrix represent the autocorrelation of intensity variations with time at a given wavenumber, and the cross-peaks indicate the simultaneous change of intensity between wavenumbers. The matrix R of statistical correlation coefficients is related to the covariance matrix C(i,j) and was calculated from the following well-known relationship:

R(i, j) )

C(i, j)

√C(i, i)C(j, j)

where C(i,i) and C(j,j) are elements of the covariance matrix. In addition, a matrix of p-values for testing the hypothesis of no correlation was calculated. Each p-value is the probability of getting a correlation as large as the observed value by random chance, when the true correlation is zero. The asynchronous correlation spectrum Ψ(ν1,ν2), which represents the dissimilarity of spectral variations as a function of concentration, can be represented in matrix notation and defined as

Ψ(ν1, ν2) )

1 y˜(ν )TNy˜(ν2) m-1 1

or

Ψ)

F(ν1, ν2) ) Ψ(ν1, ν2) ⁄ Φ(ν1, ν2) This ratio was used to confirm correlation peaks and has been suggested to be a measure of the degree of coherence similar to the global correlation phase angle proposed by Noda.36 All calculations were performed and plotted using routine functions in Matlab R2007a. Central Moments. The central moments of matrix X of various orders (k) were calculated using the following notation,

mk ) E(x - µ)k where E(x) is the expected value of x. The first central moment is zero, and the second moment is the variance, while higher moments describe asymmetry and shape parameters or peakedness of a distribution. Higher moments magnify the importance of those points which deviate from the sample mean or normal distribution. These parameters may be used to further emphasize changes in specific vibrational band frequencies with increasing protein concentration. Principal Component Analysis. Principal component analysis (PCA) is a well-established method used in statistics and chemometrics to reduce the dimensionality of a data set while retaining much of the variation present.37 PCA is used as an independent data analysis method and compliments the 2D correlation approach. For example, it has been shown that the autopower spectrum across the diagonal of a synchronous spectrum is similar to the first PCA loading vector.38,39 All PCA analysis was performed using Matlab. Results The concentration-dependent (20-120 mg/mL) ATR-FTIR spectra are shown in Figure 1 for MAb1 without NaCl. After pretreatments and normalization, the spectra of MAb1 and MAb2 appear similar. Little to no change was observed in percent secondary structure or the number of component bands and positions as a function of protein concentration using standard curve fitting procedures. Covariance Matrices and Synchronous Spectra. The development of perturbation-based 2D correlation spectroscopy was first described by Noda in 1986 30-32 and later extended to generalized 2D correlation spectroscopy.33,34 The use of two dimensions facilitates the deconvolution of complex overlapping spectral bands. Herein, the synchronous correlation spectra describe the simultaneous change in intensity variation between two wavenumbers along the perturbation axis of concentration.

1 YTNY m-1

where N is the Hilbert-Noda transformation matrix.32 The asynchronous spectrum is antisymmetric, and cross-peaks arise only if the spectral intensity variations change out of phase with each other. Thus, the asynchronous spectrum is useful for interpreting the sequential order of spectral intensity variations. The ratio of asynchronous to synchronous functions F(ν1,ν2) has been determined following the method proposed by Buchet et al. 35 using the following relationship:

Figure 1. Concentration-dependent FTIR spectra prior to mean centering and variance scaling. The numbers correspond to protein concentration in units of milligrams per milliliter.

Molecular Forces That Mediate IgG Nonideality

J. Phys. Chem. B, Vol. 113, No. 17, 2009 6111

Figure 2. Wavenumber power spectra (cm-1) along the diagonal line in the amide I synchronous spectra for (A) MAb1 in the absence of 150 mM NaCl, (B) MAb1 + 150 mM NaCl, (C) MAb2 in the absence of 150 mM NaCl, and (D) MAb2 + 150 mM NaCl.

Correlation peaks arise along the diagonal and off-diagonal (cross-peaks) positions of the synchronous spectrum. Mathematically, the diagonal is equivalent to the autocorrelation or variance of spectral intensity. The amide I autopower spectra along the diagonal line of the synchronous matrix with indicated peak positions are shown for MAb1 and MAb2 in the presence and absence of 150 mM NaCl (Figure 2). The power spectrum is composed of wellresolved bands and represents the overall intensity variation of components potentially unidentified in the broadened onedimensional spectrum. Two bands are observed at similar positions (1637 and 1653 cm-1) in the amide I region under all solution conditions tested. The bands at higher wavenumbers, >1655 cm-1 arise at distinctly different positions. Interestingly, peaks arise at 1622 cm-1 in the power spectrum of MAb1 in solution without 150 mM salt, while a slightly higher wavenumber is observed for MAb2 in solution without 150 mM NaCl. This peak is well-resolved for MAb1 and may be attributed to associated molecules, extended structure and increased intermolecular hydrogen bonding.40,41 The synchronous maps (not shown) were used to identify the position (ν1,ν2) and sign (() of potentially relevant auto- and cross-peaks. The autopower spectra of the amide II region for both mAb’s are shown in Figure 3. Many of the same peaks that arise from charged residues are observed with differing intensity and resolution for all conditions studied, suggesting the degree of backbone hydration and/or molecular environments of charged residues are noticeably affected upon increasing protein con-

centration. Peaks of interest include bands at 1529, 1549, and 1514 cm-1 for MAb1 without NaCl and 1558 and 1533 cm-1 for MAb1 in the presence of 150 mM NaCl. Similar bands near 1558 and 1533 cm-1 appear to dominate the MAb2 autopower spectra in the presence and absence of NaCl. The asynchronous spectra describe the sequence of events leading to these changes (see below). That is, a description of the residues which trigger significant changes when the concentration is increased. The significance of these peaks in the autopower spectra is discussed below. The covariance matrix is a symmetric square matrix mathematically equivalent to the synchronous correlation spectra. The diagonal elements (autopeaks) represent the variance of spectral signal fluctuations as a function of time for the representative spectral variable, while the cross-peaks correspond to the covariance between spectral signal fluctuations at two separate spectral variables. The covariance matrices (not shown) were nearly identical to the synchronous spectra for all conditions studied. Asynchronous Correlation Spectra and Sequence of Spectral Variation. Asynchronous correlation spectroscopy is a powerful method which allows the deconvolution of overlapping spectral bands while providing detailed information regarding the sequence of events that give rise to changes in the spectra. The asynchronous amide I spectra are shown in Figure 4. Using the signs of the asynchronous cross-peaks, the following sequence of events was determined for MAb1 without NaCl according to Noda’s rules: 1639 > 1647, 1657 cm-1. This

6112 J. Phys. Chem. B, Vol. 113, No. 17, 2009

Kamerzell et al.

Figure 3. Power spectra along the diagonal line in the amide II synchronous spectra for (A) MAb1 in the absence of 150 mM NaCl, (B) MAb1 + 150 mM NaCl, (C) MAb2 in the absence of 150 mM NaCl, and (D) MAb2 + 150 mM NaCl.

sequence suggests that changes in concentration affect MAb1 β-sheet/turn structures before disordered and helix regions. Interestingly, the sequence of events is reversed (1670-1680, 1641, 1657 > 1653 > 1637 cm-1) for MAb1 in the presence of 150 mM NaCl indicating changes in turn, disordered, and helix structures before β-sheet structure. The sequence of amide I changes for MAb2 in the presence and absence of 150 mM NaCl is disordered > disordered/helix > β-sheet. In addition, the change observed at 1622-1624 wavenumbers is observed before all other changes for MAb1 without 150 mM NaCl and is not observed with MAb1 + NaCl or MAb2 ( NaCl. The amide II sequence of events follow a similar order compared to amide I vibrational band changes. For MAb1 in the absence of 150 mM NaCl, changes at 1549 and 1560 cm-1 occur prior to molecular variations reflected near 1529 cm-1. Interestingly, the sequence of spectral changes is reversed for MAb1 in the presence of salt and MAb2 ( 150 mM NaCl. A discussion of the significance of these changes is described below. Central Moments. Central moments have been used historically in physics and probability theory to better understand diffusion, polymer chain conformations, statistical thermodynamics, visual pattern recognition, and shape discrimination, and for the characterization of chromatographic peaks.42-48 The central moments describe various peak characteristics including the center of gravity, width, asymmetry, and flattening. In this work, the first through fourth central moments have been calculated to probe the changes in vibrational spectra as a function of solution condition and mAb concentration. Indeed the central moments confirm the results obtained from two-

dimensional analysis regarding the wavenumber position and peak resolution of the most relevant spectroscopic changes (data not shown). Statistical Correlation Coefficients and Coherence. Correlation coefficient mapping was first introduced by Barton et al. to explore the correlation between near-IR and mid-IR regions,49 and numerous variations of this method soon followed.50 The matrix of statistical correlation coefficients is related to the covariance matrix and is a normalized measure of the strength of the relationships between two variables. Thus, all possible scalar products between vectors will capture values between -1 and +1, indicating absolute correlation at these two values. In contrast, 0 represents the absence of correlation. The matrix of correlation coefficients was calculated and the relevant values listed in Table 1. In addition, a matrix of p-values for testing the hypothesis of no correlation was calculated (Table 1). The ratio of asynchronous to synchronous correlation functions, Ψ(ν1,ν2)/Φ(ν1,ν2), was also calculated to identify false peaks according to the method proposed by Buchet et al.35 These values support the results observed from the analysis of synchronous and asynchronous maps. Principal Component Analysis. Principal component analysis is a well-established method used in statistics and chemometrics to reduce the dimensionality of a data set while retaining much of the variation present. Furthermore, PCA has been successfully combined with 2D-COS to better understand many complex systems.34,51-54 Mutually independent events are distinguished using PCA because all the principal components are orthogonal, so there is no redundant information, thus making this method powerful for our purpose.

Molecular Forces That Mediate IgG Nonideality

J. Phys. Chem. B, Vol. 113, No. 17, 2009 6113

Figure 4. Asynchronous correlation spectra for (A) MAb1 in the absence of 150 mM NaCl, (B) MAb1 + 150 mM NaCl, (C) MAb2 in the absence of 150 mM NaCl, and (D) MAb2 + 150 mM NaCl.

Herein, the principal components (PC) and the percent total variability explained by each principal component have been calculated as a function of varying protein concentration for MAb1-2 ( 150 mM NaCl. The first three PCs describe greater than 95% of the amide I and II spectral variation for both MAb’s with and without salt (Figures 5 and 6). The position of the peaks from the first loadings plot is similar to the autopower spectra of the synchronous spectra for MAb1-2 ( salt, although some new peaks are observed in the synchronous spectra (Table 2). The shape and symmetry of the peaks from the loadings plot, however, are distinctive compared to the autopower spectra in all cases. New peaks are identified in the second loadings plot for MAb1-2 only in the presence of 150 mM NaCl (Figures 5 and 6 and Table 2). Discussion The importance of macromolecular crowding and confinement affects on protein structure, stability, and function have

been exhaustively and unquestionably demonstrated.4,5,8,21,55 Indeed, our understanding of these processes has increased dramatically over the past 2 decades.2,8,17,56,57 For example, general qualitative effects of crowding and volume occupation on the rates and equilibria of interactions involving macromolecules can now be reasonably well predicted using statistical thermodynamic models.8,55 Similarly, the association state of proteins and model solutes in crowded and concentrated environments have been linked to nonideal solution behavior of which models are beginning to emerge.2,12-19 Our understanding of the affects of highly concentrated and volume occupied solutions on protein structure and stability, however, is incomplete. This work aims to understand the molecular level interactions at high concentrations that modulate the self-association of two MAbs with similar sequences but widely differing solution behavior. Herein, the affects of increasing protein concentration and confinement on the molecular properties of two well-

6114 J. Phys. Chem. B, Vol. 113, No. 17, 2009

Kamerzell et al.

TABLE 1: Cross-Peaks from Amide I Asynchronous Spectra IgG1((150 mM NaCl) MAb1

MAb1 + 150 mM NaCl

MAb2

MAb2 + 150 mM NaCl

c

ν1, ν2

ra

p-valueb

1622, 1637 1637, 1645 1637, 1653 1603, 1637 1645, 1655 1645, 1622 1622, 1655 1637, 1641 1641, 1653 1637, 1653 1637, 1657 1653, 1657 1653, 1672 1653, 1626 1672, 1637 1672, 1626 1635, 1645 1635, 1624 1624, 1645 1624, 1653 1662, 1653 1662, 1645 1662, 1635 1662, 1624 1635, 1653 1637, 1653 1637, 1684 1637, 1626 1653, 1684

0.98 0.97 0.97 0.95 0.96 0.95 0.91 0.98 0.70 0.68 0.89 0.88 0.78 0.58 0.85 0.89 0.81 0.90 0.68 0.73 0.91 0.92 0.63 0.39 0.89 0.92 -0.40 0.97 -0.72

0.0004 0.001 0.001 0.03 0.003 0.005 0.005 0.0008 0.12 0.14 0.01 0.02 0.07 0.22 0.03 0.02 0.10 0.04 0.21 0.16 0.03 0.03 0.25 0.52 0.04 0.03 0.51 0.006 0.17

|Ψ(ν1,ν2)/Φ(ν1,ν2)|

c

0.51 0.42 0.32 0.48 0.0016 0.036 0.0022 0.32 0.24 0.09 0.22 0.24 0.19 2.59 0.22 0.33 0.11 0.11 0.23 0.06 0.11 0.05 0.06 0.14 0.06 0.06 0.14 0.35 0.49

a r, correlation coefficient. b Each p-value is the probability of obtaining a correlation as large as the observed value by random chance. F(ν1,ν2) ) Ψ(ν1,ν2)/Φ(ν1,ν2), ratio of asynchronous to synchronous.

Figure 5. Percent variation described by each principal component for (A) MAb1 in the absence of 150 mM NaCl, (B) MAb1 + 150 mM NaCl, (C) MAb2 in the absence of 150 mM NaCl, and (D) MAb2 + 150 mM NaCl.

studied IgG’s have been explored. The most important changes in MAb1 association appear to be correlated with hydrogen bonding and electrostatic effects.

It has been shown that MAb1 self-associates at high concentrations, primarily through Fab-Fab interactions, resulting in nonideal solution behavior and increased viscosity, while

Molecular Forces That Mediate IgG Nonideality

J. Phys. Chem. B, Vol. 113, No. 17, 2009 6115

Figure 6. Concentration-dependent principal components for (A) MAb1 in the absence of 150 mM NaCl, (B) MAb1 + 150 mM NaCl, (C) MAb2 in the absence of 150 mM NaCl, and (D) MAb2 + 150 mM NaCl. PC1 (red), PC2 (blue), and PC3 (black).

the addition of 150 mM NaCl decreases MAb1 association and viscosity.12,13 In contrast, MAb2 does not appreciably associate in the presence or absence of NaCl and was used for comparison in this work. We have shown that significant protein conformational changes do not preclude or result from self-association, but rather conformational heterogeneity and bonding networks modulate association and increased viscosity with increasing mAb concentration. Multiple vibrational bands change as a function of protein concentration. The vibrational frequency and magnitude and sequence of those changes is distinctive to each condition and follows the solution behavior from previous studies.12,13 The asynchronous spectra reveal variation in β-sheet and turn regions occur before intensity variations in disordered and R-helical regions as the concentration is increased for MAb1 in the absence of salt (SAS, self-associated state; HV, high-viscosity regime) (Figure 4). Interestingly, this sequence was reversed for MAb1 in the presence of salt and MAb2 ( salt (no association) and is comparable to that observed for β-lactoglobulin.28 Also interesting is that, in all instances including

β-lactoglobulin, except the highly viscous self-associated state of MAb1, β-sheet regions are more resistant to concentrationdependent changes. In addition, the presence of a strong component at 1622 cm-1 in the power spectra and loadings from PCA for MAb11 may suggest the formation of intermolecular extended chains or networks and increased intermolecular hydrogen bonding strength and/or number as a function of concentration (Figure 2, Table 2). This band has been often observed and associated with intermolecular β-sheet formation and often irreversible protein association.40,58-63 It should be noted, however, that irreversible protein association was not observed in any of our studies. Increasing hydrogen bond strength and number typically decreases the frequency of stretching vibrations and increases the frequency of bending vibrations.64 A similar vibrational band was observed with increasing concentration for MAb2 in the absence of 150 mM NaCl; however, the peak is broadened and positioned at higher frequency indicating highly dynamic and weaker H-bonding, respectively. This data also suggest that MAb2 may selfassociate in the absence of 150 mM NaCl. Previously, 2D-IR

6116 J. Phys. Chem. B, Vol. 113, No. 17, 2009

Kamerzell et al.

TABLE 2: Amide I and II Band Positions (cm-1) amide I sample

a

b

PC1

PC2

Syn

PC1

1622 1637 1645 1655

1622 1637 1645

1603 1622 1637 1645 1655

1514 1529 1549

MAb1

MAb1 + NaCl

MAb2

amide II c

1610 1626 1637 1645 1653 1672 1624 1635 1643 1662

1657 1678 1626 1639 1653 1662

1626 1637 1645 1653 1672 1624 1635 1645 1653 1662

1558 1572 1508 1524 1543 1558

PC2

Syn

1506

1502 1514 1529 1549 1560 1572

1522 1549 1560 1576 1506 1533 1549 1558 1493 1512 1543 1566 1578

MAb2 + NaCl 1637 1651 1684

1626 1633 1641 1653

1637

1502 1506

1653 1684

1541 1558

1533 1549

1506 1533 1549 1558 1574 1493 1508 1524 1543 1558 1566 1502 1512 1533 1549

a PC1 ) first principal component. b PC2 ) second principal component. c Syn ) synchronous component.

and near-IR COS was used to study the temperature-dependent spectral variations of self-associating N-methylacetamide.65 This work attributes oligomer and chain size to frequency, suggesting here that MAb1 in the absence of 150 mM NaCl forms slightly larger reversibly associated complexes consistent with previous observations.13 Inherently, it is difficult to distinguish between increased intermolecular hydrogen bonding resulting from protein-protein interactions with that from increased protein-H2O hydrogen bonding. However, it is clear that this band is an important component that increases in this system. This band may be the result of increased intermolecular bonding, changes in hydrogen bonding, or some combination. The potential change in hydrogen bonding as a function of concentration for MAb11 may not be entirely surprising considering the well-documented relationship between viscosity and hydrogen bonding. This specific feature was not evident in the spectra of MAb1 in the presence of 150 mM NaCl or MAb2 ( 150 mM NaCl, all of which exhibit little or no self-association and low viscosity.13 Interestingly, the only differences in the amino acid sequence between the two MAbs studied occur in CDR regions which are highly solvent exposed loops with varying proportions of hydrogen bonding donors and acceptors. Perhaps not surprisingly the CDR of MAb1 is composed of significantly more strong H-bonding partners including His, Ser, Thr, Tyr, Asp, and Glu residues. Aggelli et al. designed peptides that self-assemble to form β-sheet tapes with nonideal solution behavior. These peptides form highly cooperative intermolecular hydrogen bonds and β-sheet structure in polar hydrogen bonding solvents with characteristic IR bands near 1620 cm-1.66 Furthermore, in the solvent 1,1,1,3,3,3-haxafluoroisopropanol (HFIP), the solution viscosity was minimal and the peptides structure was composed primarily of helices and random coils. In addition, Miller et al. investigated the self-assembly of helical β-peptides and show that the mechanism of self-association was primarily the result

of electrostatic interactions.67 It is reasonable to consider our system using similar logic. At high concentrations, MAb1 forms highly cooperative intermolecular hydrogen bonds through “peptide-like” exposed CDR regions resulting in extended structures or networks with an important contribution arising from electrostatics. The increasing high wavenumber component (1672, 1684 cm-1) observed in the autopower spectra and loadings plot for MAb1 + NaCl and MAb2 ( NaCl may be attributed to changes in β-sheet or turn (Figure 2, Table 2). It is plausible that this component is the consequence of increased band splitting or exciton splitting resulting from transition dipole coupling (TDC) in all solutions except the associated, highly viscous MAb1 solution. TDC is a resonant interaction between oscillating dipoles near one another similar to energy transfer in fluorescence experiments and has been used to explain the splitting of the amide I band of proteins with high β-sheet structure.68-73 Both through space and through bond vibrational mode couplings arise from electrostatics and/or covalent contributions involving the amide backbone and depend on the correlation of electron density and nuclear position. In addition, coupling depends on the relative orientations of and distance between oscillating dipoles which intuitively would be greater for MAb1 without 150 mM NaCl, unless the dipoles were oscillating with different frequencies or localized on one oscillator, the oscillating dipoles were out of phase, or the molecular geometry was not conducive to coupling. In fact, Kobayashi et al. have shown that for certain linear polymers the energy arising from dipole coupling is absent for extended chain polymer crystals but maximum for folded conformers.74 It is plausible that subtle conformational heterogeneity modulates the TDC effect for MAb11. Finally, increased solvation has also been shown to decrease the vibrational frequency of R-helices through formation of hydrogen bonding with the solvent.75-77 Our results suggest solvation of MAb2 ( 150 mM NaCl is greater in volume occupied solutions compared to MAb1 in the absence of 150 mM NaCl since solvated helices absorb at lower frequencies compared to nonsolvated helices (Figures 2 and 6 and Table 2). The amide II region was used to investigate changes in amino acid side chain absorption and backbone hydrogen bonding.40,78-81 The side chain absorptions of specific residues may overlap with other residues or the protein backbone; however, only the amino acids with particularly strong absorption coefficients due to vibrations of polar groups have been assigned. Specifically, deprotonated Asp and Glu and protonated Lys residues strongly absorb near 1572-1579, 1550-1560, and 1526, respectively.40 The asynchronous and autopower spectra of MAb1 in the absence of 150 mM NaCl suggest perturbation of the molecular environment near deprotonated Asp and Glu residues occur before those near protonated Lys or Tyr (asynchronous not shown, see Figure 2 and Table 2). The sequence of events was reversed for MAb1 NaCl and MAb2 ( 150 mM NaCl. Herein, we have further described the molecular level processes responsible for the increased self-association of a particular monoclonal antibody and the resulting changes in solution properties. Interestingly, increased intermolecular hydrogen bonding and electrostatics appear to modulate the selfassociation and subsequent solution viscosity. Furthermore, the order of the concentration-dependent changes appears to be an important component of the driving force for the nonideal solution behavior.

Molecular Forces That Mediate IgG Nonideality Acknowledgment. The authors thank Sherry Martin-Moe and Susan Hershenson for support and guidance. In addition, the authors thank C. Russell Middaugh for helpful discussions and insight. References and Notes (1) Zimmerman, S. B.; Harrison, B. Macromolecular crowding increases binding of DNA polymerase to DNA: An adaptive effect. Proc. Natl. Acad. Sci. U.S.A. 1987, 84 (7), 1871–5. (2) Minton, A. P.; Wilf, J. Effect of macromolecular crowding upon the structure and function of an enzyme: Glyceraldehyde-3-phosphate dehydrogenase. Biochemistry 1981, 20, 4821–6. (3) Minton, A. P. Influence of macromolecular crowding upon the stability and state of association of proteins: Predictions and observations. J. Pharm. Sci. 2005, 94 (8), 1668–75. (4) Ellis, R. J.; Minton, A. P. Cell biology: Join the crowd. Nature (London) 2003, 425 (6953), 27–8. (5) Alberts, B. The cell as a collection of protein machines: Preparing the next generation of molecular biologists. Cell 1998, 92 (3), 291–4. (6) Minton, A. P., Holobiochemistry: An integrated approach to the understanding of biochemical mechanism that emerges from the study of proteins and protein associations in Volume-occupied solutions; WileyLiss:New York, 1990. (7) Zhou, H. X.; Szabo, A. Theory and simulation of the time-dependent rate coefficients of diffusion-influenced reactions. Biophys. J. 1996, 71 (5), 2440–57. (8) Zhou, H. X.; Rivas, G.; Minton, A. P. Macromolecular crowding and confinement: Biochemical, biophysical, and potential physiological consequences. Annu. ReV. Biophys. 2008, 37, 375–97. (9) Minton, A. P. Confinement as a determinant of macromolecular structure and reactivity. Biophys. J. 1992, 63 (4), 1090–100. (10) Cheung, M. S.; Thirumalai, D. Nanopore-protein interactions dramatically alter stability and yield of the native state in restricted spaces. J. Mol. Biol. 2006, 357 (2), 632–43. (11) Ellis, R. J.; Minton, A. P. Protein aggregation in crowded environments. Biol. Chem. 2006, 387 (5), 485–97. (12) Kanai, S.; Liu, J.; Patapoff, T. W.; Shire, S. J. Reversible selfassociation of a concentrated monoclonal antibody solution mediated by Fab-Fab interaction that impacts solution viscosity. J. Pharm. Sci. 2008, 97 (10), 4219–27. (13) Liu, J.; Nguyen, M. D.; Andya, J. D.; Shire, S. J. Reversible selfassociation increases the viscosity of a concentrated monoclonal antibody in aqueous solution. J. Pharm. Sci. 2005, 94 (9), 1928–40. (14) Jimenez, M.; Rivas, G.; Minton, A. P. Quantitative characterization of weak self-association in concentrated solutions of immunoglobulin G via the measurement of sedimentation equilibrium and osmotic pressure. Biochemistry 2007, 46, 8373–8. (15) Abbott, A. P.; Hope, E. G.; Palmer, D. J. Effect of solutes on the viscosity of supercritical solutions. J. Phys. Chem. B 2007, 111, 8114–8. (16) Wilf, J.; Minton, A. P. Evidence for protein self-association induced by excluded volume. Myoglobin in the presence of globular proteins. Biochim. Biophys. Acta 1981, 670 (3), 316–22. (17) Minton, A. P.; Lewis, M. S. Self-association in highly concentrated solutions of myoglobin: A novel analysis of sedimentation equilibrium of highly nonideal solutions. Biophys. Chem. 1981, 14 (4), 317–24. (18) Hall, C. G.; Abraham, G. N. Reversible self-association of a human myeloma protein. Thermodynamics and relevance to viscosity effects and solubility. Biochemistry 1984, 23, 5123–9. (19) Hall, C. G.; Abraham, G. N. Size, shape, and hydration of a selfassociating human IgG myeloma protein: Axial asymmetry as a contributing factor in serum hyperviscosity. Arch. Biochem. Biophys. 1984, 233 (2), 330– 7. (20) Rosenberg, A. S. Historical overview of the relationship of protein aggregates to immunogenicity and formation of consortium to address key issues. Workshop on protein aggregation, Breckenridge, CO, 2006. (21) Shire, S. J.; Shahrokh, Z.; Liu, J. Challenges in the development of high protein concentration formulations. J. Pharm. Sci. 2004, 93 (6), 1390–402. (22) Harn, N.; Allan, C.; Oliver, C.; Middaugh, C. R. Highly concentrated monoclonal antibody solutions: Direct analysis of physical structure and thermal stability. J. Pharm. Sci. 2007, 96 (3), 532–46. (23) Harn, N. R.; Jeng, Y. N.; Kostelc, J. G.; Middaugh, C. R. Spectroscopic analysis of highly concentrated suspensions of bovine somatotropin in sesame oil. J. Pharm. Sci. 2005, 94 (11), 2487–95. (24) Saluja, A.; Badkar, A. V.; Zeng, D. L.; Kalonia, D. S. Ultrasonic rheology of a monoclonal antibody (IgG2) solution: Implications for physical stability of proteins in high concentration formulations. J. Pharm. Sci. 2007, 96 (12), 3181–95. (25) Saluja, A.; Badkar, A. V.; Zeng, D. L.; Nema, S.; Kalonia, D. S. Ultrasonic storage modulus as a novel parameter for analyzing protein-

J. Phys. Chem. B, Vol. 113, No. 17, 2009 6117 protein interactions in high protein concentration solutions: Correlation with static and dynamic light scattering measurements. Biophys. J. 2007, 92 (1), 234–44. (26) Saluja, A.; Kalonia, D. S. Nature and consequences of proteinprotein interactions in high protein concentration solutions. Int. J. Pharm. 2008, 358 (1-2), 1–15. (27) Shire, S. J.; Cromwell, M.; Liu, J. Concluding summary: Proceedings of the AAPS Biotec open forum on “aggregation of protein therapeutics”. Aaps J 2006, 8 (4), E729–30. (28) Czarnik-Matusewicz, B.; Murayama, K.; Wu, Y.; Ozaki, Y. Twodimensional attenuated total reflection/infrared correlation spectroscopy of adsorption-induced and concentration-dependent spectral variations of β-lactoglobulin in aqueous solutions. J. Phys. Chem. B 2000, 104, 7803– 11. (29) Dousseau, F.; Therrien, M.; Pezolet, M. On the spectral subtraction of water from the FT-IR spectra of aqueous solutions of proteins. Appl. Spectrosc. 1989, 43 (3), 538–542. (30) Noda, I. Two-dimensional infrared (2-D IR) spectroscopy of synthetic and biopolymers. Bull. Am. Phys. Soc. 1986, 31, 520. (31) Noda, I. Two-dimensional infrared spectroscopy. J. Am. Chem. Soc. 1989, 111, 8116–8. (32) Noda, I. Two-dimensional infrared (2D IR) spectroscopy: Theory and applications. Appl. Spectrosc. 1990, 44 (4), 550–561. (33) Noda, I. Generalized two-dimensional correlation method applicable to infrared, Raman, and other types of spectroscopy. Appl. Spectrosc. 1993, 47 (9), 1329–36. (34) Noda, I.; Ozaki, Y. , Two-dimensional correlation spectroscopy: Applications in vibrational and optical spectroscopy Wiley: Hoboken, NJ, 2004; p xii, 295 pp. (35) Buchet, R.; Wu, Y.; Lachenal, G.; Raimbault, C.; Ozaki, Y. Selecting two-dimensional cross-correlation functions to enhance interpretation of near-infrared spectra of proteins. Appl. Spectrosc. 2001, 55 (2), 155– 162. (36) Noda, I. Recent mathematical developments in 2D correlation spectroscopy. In AIP Conference Procedures; Ozaki, Y. , Noda, I. , Eds.; American Institute of Physics: New York, 2000; pp 2001-204. (37) Jolliffe, I. T., Principal component analysis, 2nd ed.; Springer: New York, 2004. (38) Wu, Y.; Murayama, K.; Ozaki, J. Two-dimensional infrared spectroscopy and principle component analysis studies of the secondary structure and kinetics of hydrogen-deuterium exchange of human serum albumin. J. Phys. Chem. B 2001, 105, 6251–6259. (39) Murayama, K.; Czarnik-Matusewics, B.; Wu, Y.; Tsenkova, R.; Ozaki, Y. Comparison between conventional spectral analysis methods, chemometrics, and two-dimensional correlation spectroscopy in the analysis of near-infrared spectra of protein. Appl. Spectrosc. 2000, 54 (7), 978–985. (40) Barth, A.; Zscherp, C. What vibrations tell us about proteins. Q. ReV. Biophys. 2002, 35 (4), 369–430. (41) Krimm, S.; Bandekary, J. Vibrational spectroscopy and conformation of peptides, polypeptides, and proteins. AdV. Protein Chem. 1986, 38, 181–367. (42) Giddings, J. C.; Eyring, H. A. A molecular dynamic theory of chromatography. J. Phys. Chem. 1955, 59, 416–421. (43) Grushka, E.; Myers, M. N.; Schettler, P. D.; Gidding, J. C. Computer characterization of chromatographic peaks by plate height and higher central moments. Anal. Chem. 1969, 41, 889–92. (44) Ivanov, G.; Ponomarchuk, Y. V.; Chashkin, Y. R. Statistical procedures for using approximate variance equality for unknown fourth central moments. Meas. Tech. (Engl. Transl.) 2005, 48 (4), 322–326. (45) Dityatev, A. E.; Altinbaev, R.; Astrelin, A. V.; Voronin, L. L. Combining principal component and spectral analyses with the method of moments in studies of quantal transmission. J. Neurosci. Methods 2003, 130 (2), 173–99. (46) Haefner, D.; Sukhov, S.; Dogariu, A. Stochastic scattering polarimetry. Phys. ReV. Lett. 2008, 100 (4), 043901. (47) Dill, K. a. B. S. Molecular driVing forces: Statistical thermodynamics in chemistry and biology; Garland Science: New York, 2003. (48) Wu, K. L.; Wei, J. H.; Lai, S. K.; Okabe, Y. Energy and enthalpy distribution functions for a few physical systems. J. Phys. Chem. B 2007, 111, 8946–58. (49) Barton, F. E.; Himmelsbach, D. S.; Duckworth, J. H.; Smith, M. J. Two-dimensional vibration spectroscopy: Correlation of mid- and nearinfrared regions. Appl. Spectrosc. 1992, 46 (3), 420–429. (50) Sasic, S.; Ozaki, Y. Statistical two-dimensional correlation spectroscopy: Its theory and applications to sets of vibrational spectra. Anal. Chem. 2001, 73, 2294–301. (51) Wu, Y.; Murayama, K.; Ozaki, Y. Two-dimensional infrared spectroscopy and principle component analysis studies of the secondary structure and kinetics of hydrogen-deuterium exchange of human serum albumin. J. Phys. Chem. B 2001, 105, 6251–6259. (52) Raussens, V.; Ruysschaert, J. M.; Goormaghtigh, E. Analysis of 1H/2H exchange kinetics using model infrared spectra. Appl. Spectrosc. 2004, 58 (1), 68–82.

6118 J. Phys. Chem. B, Vol. 113, No. 17, 2009 (53) Wu, Y.; Hao, Y. Q.; Li, M.; Guo, C.; Ozaki, Y. Two-dimensional infrared spectroscopy and principal component analysis studies on a new azobenzene derivative supramolecular system based on hydrogen bonds. Appl. Spectrosc. 2003, 57 (8), 933–42. (54) Jung, Y. M.; Sato, H.; Noda, I. Characterization of the thermal behavior of poly(hydroxybutylate) by principal component analysis-based two-dimensional correlation spectroscopy. Anal. Sci. 2007, 23 (7), 881–9. (55) Rosgen, J.; Pettitt, B. M.; Bolen, D. W. Uncovering the basis for nonideal behavior of biological molecules. Biochemistry 2004, 43, 14472– 84. (56) Zhou, H. X. Effect of mixed macromolecular crowding agents on protein folding. Proteins 2008, 72 (4), 1109–13. (57) Jarvis, T. C.; Ring, D. M.; Daube, S. S.; von Hippel, P. H. “Macromolecular crowding”: Thermodynamic consequences for proteinprotein interactions within the T4 DNA replication complex. J. Biol. Chem. 1990, 265 (25), 15160–7. (58) Krishnan, S.; Chi, E. Y.; Webb, J. N.; Chang, B. S.; Shan, D.; Goldenberg, M.; Manning, M. C.; Randolph, T. W.; Carpenter, J. F. Aggregation of granulocyte colony stimulating factor under physiological conditions: Characterization and thermodynamic inhibition. Biochemistry 2002, 41, 6422–31. (59) Haris, P. I.; Chapman, D. The conformational analysis of peptides using Fourier transform IR spectroscopy. Biopolymers 1995, 37 (4), 251– 63. (60) Saluja, A.; Crampton, S.; Kras, E.; Fesinmeyer, R. M.; Remmele, R. L., Jr.; Narhi, L. O.; Brems, D. N.; Gokarn, Y. R. Anion binding mediated precipitation of a peptibody. Pharm. Res. 2009, 26 (1), 152–60. (61) van de Weert, M.; Haris, P. I.; Hennink, W. E.; Crommelin, D. J. Fourier transform infrared spectrometric analysis of protein conformation: Effect of sampling method and stress factors. Anal. Biochem. 2001, 297 (2), 160–9. (62) Paquet, M. J.; Laviolette, M.; Pezolet, M.; Auger, M. Twodimensional infrared correlation spectroscopy study of the aggregation of cytochrome c in the presence of dimyristoylphosphatidylglycerol. Biophys. J. 2001, 81 (1), 305–12. (63) Kendrick, B. S.; Cleland, J. L.; Lam, X.; Nguyen, T.; Randolph, T. W.; Manning, M. C.; Carpenter, J. F. Aggregation of recombinant human interferon gamma: Kinetics and structural transitions. J. Pharm. Sci. 1998, 87 (9), 1069–76. (64) Straughan, B. P. Spectroscopy; Chapman and Hall: London, 1976; Vol. 2. (65) Ozaki, Y.; Liu, Y.; Noda, I. Two-dimensional infrared and nearinfrared correlation spectroscopy: Applications to studies of temperaturedependent spectral variations of self-associated molecules. Appl. Spectrosc. 1997, 51 (4), 526–535.

Kamerzell et al. (66) Aggeli, A.; Bell, M.; Boden, N.; Keen, J. N.; Knowles, P. F.; McLeish, T. C.; Pitkeathly, M.; Radford, S. E. Responsive gels formed by the spontaneous self-assembly of peptides into polymeric beta-sheet tapes. Nature (London) 1997, 386 (6622), 259–62. (67) Miller, C. A.; Hernandez-Ortiz, J. P.; Abbott, N. L.; Gellman, S. H.; de Pablo, J. J. Dipole-induced self-assembly of helical beta-peptides. J. Chem. Phys. 2008, 129 (1), 015102. (68) Miyazawa, T. Perturbation treatment of the characteristic vibrations of polypeptide chains in various configurations. J. Chem. Phys. 1960, 32, 1647–1652. (69) Abe, Y.; Krimm, S. Normal vibrations of polyglycine II. Biopolymers 1972, 11 (9), 1841–53. (70) Abe, Y.; Krimm, S.; Normal vibrations of crystalline polyglycine, I. Biopolymers 1972, 11 (9), 1817–39. (71) Krimm, S.; Abe, Y. Intermolecular interaction effects in the amide I vibrations of polypeptides. Proc. Natl. Acad. Sci. U.S.A. 1972, 69 (10), 2788–92. (72) Chirgadze, Y. N.; Nevskaya, N. A. Infrared spectra and resonance interaction of amide-I vibration of the antiparallel-chain pleated sheet. Biopolymers 1976, 15 (4), 607–25. (73) Chirgadze, Y. N.; Nevskaya, N. A. Infrared spectra and resonance interaction of amide-I vibration of the paraellel-chain pleated sheets. Biopolymers 1976, 15 (4), 627–36. (74) Kobayashi, M.; Sakashita, M. Morphology dependent anomalous frequency shifts of infrared absorption bands of polymer crystals: Interpretation in terms of transition diple-dipole coupling theory. J. Chem. Phys. 1992, 96 (1), 748. (75) Parrish, J. R., Jr.; Blout, E. R. The conformation of poly-L-alanine in hexafluoroisopropanol. Biopolymers 1972, 11 (5), 1001–20. (76) Reisdorf, W. C., Jr.; Krimm, S. Infrared amide I′ band of the coiled coil. Biochemistry 1996, 35, 1383–6. (77) Gilmanshin, R.; Dyer, R. B.; Callender, R. H. Structural heterogeneity of the various forms of apomyoglobin: Implications for protein folding. Protein Sci. 1997, 6 (10), 2134–42. (78) MyshakinaN. S. AhmedZ. AsherS. A. Dependence of amide vibrations on hydrogen bonding J. Phys. Chem. B, in press. (79) Myshakina, N. S.; Asher, S. A. Peptide bond vibrational coupling. J. Phys. Chem. B 2007, 111, 4271–9. (80) Barth, A. The infrared absorption of amino acid side chains. Prog. Biophys. Mol. Biol. 2000, 74 (3-5), 141–73. (81) Chirgadze, Y. N.; Fedorov, O. V.; Trushina, N. P. Estimation of amino acid residue side-chain absorption in the infrared spectra of protein solutions in heavy water. Biopolymers 1975, 14 (4), 679–94.

JP9001548