Linking the solution viscosity of an IgG2 monoclonal antibody to its structure as a function of pH and temperature.

RESEARCH ARTICLE – Pharmaceutical Biotechnology

Linking the Solution Viscosity of an IgG2 Monoclonal Antibody to Its Structure as a Function of pH and Temperature WEIQIANG CHENG,1 SANGEETA B. JOSHI,1 NISHANT KUMAR JAIN,1 FENG HE,2 BRUCE A. KERWIN,2 DAVID B. VOLKIN,1 C. RUSSELL MIDDAUGH1 1

Macromolecule and Vaccine Stabilization Center, Department of Pharmaceutical Chemistry, University of Kansas, Lawrence, Kansas 66047 2 Amgen Inc. Process and Product Development, Amgen Inc., Seattle, Washington 98119 Received 14 August 2013; revised 16 September 2013; accepted 19 September 2013 Published online 18 October 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23748

ABSTRACT: Although the viscosity of concentrated antibody solutions has been the focus of many recent studies, less attention has been concentrated on how changes in protein structure impact viscosity. This study examines viscosity profiles of an immunoglobulin G (IgG) 2 monoclonal antibody at 150 mg/mL as a function of temperature and pH. Although the structure of the antibody at pH 4.0–7.0 was comparable at lower temperatures as measured by second derivative UV absorbance and Fourier transform infrared spectroscopy, differences in 8-anilino-1-naphthalene sulfonate (ANS) fluorescence intensity indicated small structural alterations as a function of pH. Below the structural transition onset temperature, the viscosity profiles were pH dependent and linearly correlated with fluorescence intensity, and followed semilogarithmic behavior as a function of temperature. The transitions of the viscosity profiles correlated well with the major structure transitions at a protein concentration of 150 mg/mL. The viscosity correlated particularly well with ANS fluorescence intensity at 0.2 mg/mL below and above the structural transition temperatures. These results suggest: (1) ANS can be an important measure of the overall structure and (2) hydrophobic interactions and charge–charge interactions are the two major physical factors that contribute C 2013 Wiley Periodicals, Inc. and the American Pharmacists Association collectively to the high viscosity of concentrated IgG solutions. J Pharm Sci 102:4291–4304, 2013 Keywords: protein; monoclonal antibody; viscosity; structure; high concentration; FT-IR; partial specific volume; absorption spectroscopy; Fluorescence spectroscopy; zeta potential

INTRODUCTION Monoclonal antibodies (mAbs) have become increasingly important as therapeutic agents, especially as treatments for patients with cancer and autoimmune disorders.1 The desired dosage form for patient convenience often requires these molecules to be self-administrated in prefilled syringes via subcutaneous injection. Because of the limitations in the maximum solution volume used for subcutaneous injection (∼1 mL), high-concentration protein solutions (>100 mg/mL) are often required. Highly concentrated antibody solutions have raised significant stability and formulation delivery challenges, including aggregation and solution viscosity.2 Developing a better understanding of the causes and mechanisms of the elevated viscosity of concentrated antibody solutions is an area of active research.3–6 Solution viscosity appears to have a nonlinear relationship to protein concentration and is often strongly affected by solution pH, ionic strength, presence of inorganic salts, as well as the intrinsic structural nature of the antibody itself.3,5,7 Initial assumptions were that the highsolution viscosity at pH values close to the pI value was primarily because of reversible self-association mediated by elecCorrespondence to: C. Russell Middaugh (Telephone: +785-864-5813; Fax: +785-864-5814; E-mail: [email protected]); Feng He (Telephone: +206-2656917; Fax: +206-217-0346; E-mail: [email protected]) Weiqiang Cheng’s present address is Biologics R&D, Genzyme, a Sanofi Company, Framingham, Massachusetts 01702 This article contains supplementary material available from the authors upon request or via the Internet at http://onlinelibrary.wiley.com/. Journal of Pharmaceutical Sciences, Vol. 102, 4291–4304 (2013)

C 2013 Wiley Periodicals, Inc. and the American Pharmacists Association

trostatic interactions.2 Later studies suggested the higher solution viscosity of certain antibodies was because of the networklike associations of Fab regions rather than Fc associations.4 The viscosity of protein solutions at high concentrations can be related to the second virial coefficient (B22)6 and the interaction parameter (Kd ) as represented by the slope of the mutual self-diffusion coefficient (Dm ) as a function of protein concentration.8 Experimental values for B22 and Kd can be obtained by several different approaches, including static9 and dynamic light scattering,10–12 ultracentrifugation,6 and selfinteraction chromatography.13,14 Such measurements, however, require relatively low-concentration antibody solutions and extrapolation to more viscous solutions.9,10,12 Studies have suggested that increases in antibody solution viscosity may be related to charge–charge interactions, excluded volume effects, short-range van der Waals and apolar interactions, dipole–dipole interactions, as well as a variety of other interactions between mAb molecules.5,7,8,15–17 Even in cases where experimental results have pointed to charge–charge interactions as the major source of elevated viscosity behavior, the nature and source of such interactions often remain to be more clearly defined. In addition, the significance and contribution of each of the possible noncovalent interactions, in combination with charge–charge interactions, to the overall viscosity of a high-concentration protein solution remain unclear. Two recent studies have shown that hydrophobic organic salts can significantly diminish the viscosity of high-concentration antibody solutions, suggesting these bulky organic salt molecules can disrupt the intermolecular transient networks between protein molecules because of their

Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4291

4292


apolarity.18,19 The results of He et al.20 concerning the effect of sugars on the viscosity of antibody solutions also support a related conclusion because these sugars, lacking both significant apolarity and charge, were found to further increase the viscosity of antibody solutions.20 In a more recent study, the viscosity of antibody mixtures was found to be reasonably described as the geometric sum of the viscosity of each solution,21 suggesting that related protein–protein interactions are nonspecific in nature. Although significant progress has been achieved, a further understanding of the factors affecting viscosity without careful consideration of their structural basis would be incomplete. In earlier studies concerning to structure and viscosity relationships, antibodies were assumed to have a similar structure,3,5 although results4,16 using circular dichroism to link antibody secondary structure to solution viscosity were preliminary and inconclusive. Liu et al.22 recently showed that viscosity measurements can be used as a tool to monitor the thermal stability of IgG antibodies. As expected, the unfolding behavior could be inferred by examination of changes in viscosity.22 Kamerzell et al.23 revealed possible protein-specific secondary structure changes as a function of viscosity caused by increasing protein concentrations using two-dimensional vibrational correlation spectroscopy and argued that hydrogen bonding and electrostatics are primarily responsible for the intermolecular association that results in nonideal viscosity behavior.23 However, so far there have not been any additional studies to examine protein solution viscosity profiles as a function of structural changes under different environmental conditions. Therefore, the purpose of this study is to characterize solution viscosity profiles and structural changes of a model IgG2 mAb at 150 mg/mL under different conditions of pH and temperature. A better understanding of the impact of antibody structure on the viscosity of high-concentration antibody solutions can lead to additional insights into how to better develop protein formulations with improved solution properties.

MATERIALS AND METHODS Materials The mAb was purified according to Amgen platform conditions. The antibody solution was dialyzed into a series of 20 mM citrate–phosphate (C–P) buffers at an ionic strength of 0.15 (adjusted with NaCl). The C–P buffers were prepared using a 0.4 M citric acid solution employing citric acid monohydrate (Fisher Scientific, Pittsburgh, Pennsylvania) and a 0.4 M sodium phosphate dibasic solution prepared from sodium phosphate dibasic anhydrous (Sigma, St. Louis, Missouri). All reagents were purchased from Sigma unless otherwise noted. Protein dialyses were performed at 4◦ C using Slide-A-Lyzer dialysis cassettes, 10,000 MWCO (molecular weight cut-off) (Pierce, Rockford, Illinois) against C–P buffers at pH 4.0–7.0. The protein was concentrated using CENTRICON centrifugal filter devices (with a MWCO of 50k or less) centrifuged at 2000–3000 g at lower temperature. The concentration was then adjusted to approximately 150 mg/mL with the corresponding buffers. The protein concentration was measured with an Agilent 8453 UV–Visible spectrophotometer (Palo Alto, California) employing a theoretical extinction coefficient of 1.47 mL/(mg.cm) after dilution to approximately 0.2 mg/mL with the corresponding buffers. R

R


Infrared Spectroscopy Fourier transform infrared (FT-IR) spectra were collected using a Bruker Tensor 27 spectrometer (Bruker Optics GmbH, Ettlington, Germany) equipped with an MCT detector and a Haake DC 30/K20 temperature controller (Thermo Electron, Newington, NH) . The FT-IR instrument employing a BioATR cell was purged with nitrogen throughout all experiments. Single-beam interferograms at 150 mg/mL were acquired using OPUS software (version 6.5) (Bruker Optics Inc., Billerica, MA) from 10◦ C to 90◦ C at 2.5◦ C intervals. The interferograms of a corresponding buffer blank were collected at the same temperatures. FT-IR spectra of the protein were obtained in the range of 4000−900 cm−1 at a resolution of 4 cm−1 . The resultant FT-IR spectra were processed with atmospheric (H2 O and CO2 ) compensation and with a nine-point Savitsky–Golay smoothing function. The spectra in the amide I region (1710– 1590 cm−1 ) were used for further data processing. The resultant spectra were then normalized using a vector normalization function. The vector normalization first calculates the average absorbance value of each spectrum from 1710 to 1590 cm−1 . The average value calculated was then subtracted from the spectrum, which causes the spectrum to be centered around y = 0. This step is followed by calculating the sum of squares of all y values, and the respective spectrum is then divided by the square root of this sum. The spectrum at 10◦ C was then subtracted from the spectra at each temperature to obtain the differential spectra under each pH condition. UV Absorption Spectroscopy Protein samples at 150 mg/mL at pH 4.0–7.0 were loaded into a 10-:m path length cuvette and UV absorbance spectra were collected with the corresponding buffers used as blanks using an Agilent 8453 UV–Visible Spectrophotometer. The UV absorbance spectra of the protein at 0.5 unit intervals were collected at a temperature interval of 2.5◦ C from 10◦ C to 87.5◦ C. The equilibration and integration time at each temperature were 300 and 15 s, respectively. The total time for each run was approximately 4.5 h. The optical density (OD) reading at 350 nm was used to monitor the aggregation behavior of the antibody. Second-derivative spectra were calculated using a nine-point data filter and a third degree polynomial function and then smoothed by 99 interpolated points between each raw data point. The peak positions were obtained using the peakpicking function of Origin 7 software (OriginLab Corporation, Northampton, MA). Extrinsic 8-Anilino-1-Naphthalene Sulfonate Fluorescence and Light Scattering Intensity 8-Anilino-1-naphthalene sulfonate (ANS) fluorescence was used to measure tertiary structural changes upon subjecting the protein to thermal stress. ANS was added to the protein solution at dye to protein molar ratio of 0.5, because higher molar ratios could result in protein precipitation as well as reducing the viscosity of high-concentration protein solutions.19,24 ANS fluorescence spectra of the above samples at pH 4.0–7.0 were collected using a cuvette method in duplicates as reported earlier.25 A spectrofluorometer (Photon Technology International, Lawrenceville, New Jersey) equipped with a turreted four-position Peltier-controlled cell holder and a xenon lamp was used in the study. The fluorescence of each sample (165 :L) in rectangular cuvettes (2 × 10 mm2 ) was excited at 374 nm and DOI 10.1002/jps.23748


emission spectra from 400–600 nm were collected using 2 nm slit widths for both excitation and emission. The fluorescence scan rate was 1 nm/s. The spectra were obtained at 2.5◦ C intervals from 10◦ C to 87.5◦ C and the temperature ramping rate was 1◦ C/min. A spectrum of buffer-containing ANS was subtracted from each spectrum of protein and ANS prior to data analysis. Maximum ANS fluorescence intensities were plotted as a function of temperature to evaluate the binding of ANS to the protein. Light scattering intensities of the antibody at 150 mg/mL were simultaneously acquired along with the fluorescence spectra with a second photomultiplier placed at 180◦ to the fluorescence emission detector. Light scattering intensities at wavelengths ranging from 250 to 450 nm were collected with a slit width of 0.5 nm. The light scattering intensity of the corresponding buffer was also collected along with the sample. The net light scattering intensity of the sample was obtained after subtraction of the light scattering intensity of the corresponding buffer, and the net light scattering intensity at 373–375 nm from duplicate experiments were averaged and used as the scattering intensity of the antibody. Partial Specific Volume The antibody at pH 4.0 and 7.0 at 5.0 mg/mL was prepared at room temperature by dilution of dialyzed high-concentration antibody solutions with the appropriate dialysis buffer. The density of the buffers and the antibody solutions at 5.0 mg/mL were measured two or three times with a DMA-5000 digital density meter (Anton Paar, Graz, Austria) at 2.5◦ C intervals from 7.5◦ C to 50◦ C after the instrument was calibrated with dry air and water. The protein concentration at other temperatures [C(T)] was calculated from the protein concentration (5.0 mg/mL) at room temperature (20◦ C) from the following equation: C(T) =

D(T) × C (20◦ C) D(20◦ C)

where D(T) is the density of the protein solution at temperature T, D(20◦ C) and C(20◦ C) are the density and concentration (5.0 mg/mL) of the protein solution at 20◦ C. The partial specific volume (n) of the protein as a function of temperature was calculated using the following equation26 : n=

D(T) − D0 (T) 1 1− D0 (T) C(T)

where D0 (T) is the density of buffer solutions at temperature T. Viscosity Measurement The viscosity of the antibody solutions at approximately 150 mg/mL was measured using a dynamic light scattering method as reported previously27 with a slight modification. In this study, 0.4 :L of a 150-nm diameter polystyrene particle standard (Thermo Fisher Scientific Inc., Fremont, CA) was mixed with 40 :L of the concentrated antibody solution in a 384 well plate. The resultant antibody solutions containing the particle standard were measured in duplicate in the pH range 4.0–7.0 at 0.5 unit intervals. The apparent diameter of the particle standard was measured from 20◦ C to 62.5◦ C at 2.5◦ C intervals using a Wyatt DynaPro dynamic light scattering DOI 10.1002/jps.23748

4293

instrument (Santa Barbara, California). The auto-attenuation time for each experiment was 30 s, the acquisition time was 30 s, and the temperature ramping rate was 0.5◦ C/min, with the solvent set as water with the viscosity of water at 20◦ C as 1.0 centipoises (cP). The total measurement time for these samples from 20◦ C to 62.5◦ C was approximately 33.5 h and the total time for the protein at each temperature interval was approximately 112 m. The intensity signal of the particle standard can easily be distinguished from that of the protein in the resultant intensity distribution histograms. The apparent diameter (d) of the particle standard was obtained from the autocorrelation function. The apparent diameter (d) versus its actual diameter (150 nm diameter, d0 ) of the particle standard and the viscosity of water (0 0 )at each temperature were used to calculate the viscosity of the antibody solution (0) using the function: 0= 0 0 d/d0 . Values for the dynamic viscosity of water from 10◦ C to 65◦ C at 5◦ C intervals were obtained from literature28 and were used to obtain the viscosity at 2.5◦ C intervals by fitting the literature data at 5◦ C to a third-order polynomial. The viscosity of water (0 0 ) at each temperature at 2.5◦ C intervals was then used to calculate the viscosity of the protein solutions. Initial studies with mVROC and Stabinger viscometer gave similar results but the better reproducibility of the dynamic light scattering method and its superior throughput resulted in its selection for viscosity measurement in this study.

Zeta Potential The zeta potential of the protein solution at 2.0 mg/mL in the C–P buffers was measured using a Malvern Zeta Sizer Nano Series light scattering instrument (Model ZEN3600; Malvern Instruments Ltd., Worcestershire, UK). The zeta potential value for each measurement was obtained from 100 measurements, with each sample measured in triplicate. The measurement temperature was set at 20◦ C, with 3 min of equilibration prior to measurements.

RESULTS The secondary structure of the IgG2 mAb in solution at 150 mg/mL was monitored by FT-IR spectroscopy. Figure 1a shows the amide I region of the absorbance and secondderivative absorbance spectra of the mAb at 10◦ C and 85◦ C at pH 4.0 and 7.0. At 10◦ C, the protein shows a major band with a peak at 1635 cm−1 , arising primarily from the expected intramolecular $-sheet.29,30 There was no significant difference among the spectra at 10◦ C between pH 4.0 and 7.0. Increasing the temperature resulted in a significant change in the spectra (Fig. S1). For example, at 85◦ C, the absorbance at 1635 cm−1 diminished, and a new band appeared between 1605 and 1635 cm−1 with a peak at approximately 1623 cm−1 appeared because of the formation of intermolecular $-sheet.31 In addition, a new, broad band in the absorbance spectra and several minor peaks in the second-derivative spectra (Fig. 1a) appeared between 1640 and 1700 cm−1 , corresponding to turns, intramolecular $-sheet structure, and disordered structures.29,32,33 There appeared to be a slight difference in the relative intensity of these two new bands at 87.5◦ C under different pH conditions (Fig. 1a), possibly because of the difference in the nature of the resulting material caused by structural Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4294


Figure 1. Fourier transform infrared analysis of IgG2 mAb as a function of pH and temperature. Representative FT-IR absorbance and secondderivative spectra of the IgG2 mAb at indicated solution pH before and after the heat treatment (a). FT-IR differential absorbance of IgG2 mAb as function of pH and temperature as monitored at 1637(b), 1620(c), and 1668 cm−1 (d). FT-IR spectra were baseline corrected and vector normalized between 1590 and 1710 cm−1 . Differential absorbance data at 1668 cm−1 was parallely shifted by one or multiples of 0.002 units for clearer presentation of the data. Error bars represent 1 SD with n = 2.

differences and the light penetration depth as a function of wavelength. To track changes in the protein’s overall secondary structure as a function of the temperature and pH, differential spectra of IgG2 at each pH condition versus the respective initial Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

IR spectrum at 10◦ C were obtained. The differential spectra showed two positive peaks at 1620 and 1668 cm−1 and a negative peak at 1637 cm−1 as the temperature increased (Fig. S2). Figures 1 b–1d show the changes of differential absorbance at 1620 (Fig. 1b), 1637 (Fig. 1c), and 1668 cm−1 (Fig. 1d). There was DOI 10.1002/jps.23748


a monotonic linear change of differential absorbance at each of these three wavelengths at lower temperatures before more significant changes occurred at elevated temperatures. The reason for this could be related to the change in partial specific volume as a function of temperature as discussed below and subsequent vector normalization (see Materials and Methods section). At pH 4.0, the peak at 1637 (Fig. 1b) and at 1620 cm−1 (Fig. 1c) appeared to deviate from this linear behavior at approximately 40◦ C, and was followed by more significant changes starting at 60◦ C. Similar changes were seen at pH 5.0 but at slightly higher temperatures than that at pH 4.0. The initiating temperature for the first transition at 1637 cm−1 (Fig. 1b) appeared to follow the sequence: pH 4.0 < pH 4.5 < pH 5.0–5.5 < pH 6.0–7.0. A similar sequence for the first transition onset temperature at 1620 cm−1 (Fig. 1c) was observed. At 1668 cm−1 , the onset temperatures of the early transitions of the protein were much less obvious than the corresponding transitions seen at 1637 (Fig. 1b) and 1620 cm−1 (Fig. 1c). For example, at pH 4.0, a deviation of the differential absorbance from the linear increase was found at approximately 45◦ C, whereas a similar deviation was found at 57.5◦ C and 65◦ C for pH 4.5 and 5.0, respectively. A more drastic increase in the differential absorbance at 1668 cm−1 was seen at approximately 70◦ C, and the transition temperature for this change followed the sequence: pH 4.0 < pH 4.5 < pH 5.0 < pH 5.5–7.0. The transition temperatures measured by absorbance changes at 1637, 1620, and 1668 cm−1 by FT-IR at pH 4.0 and 5.0 with 150 mg/mL protein solutions were significantly lower (∼5.0◦ C–7.5◦ C) than that observed previously with this mAb with CD spectroscopy measured in 0.2 mg/mL protein solutions.25 Under other pH conditions, the thermal transitions measured by FT-IR in this study and CD in a previous study25 were comparable. Figure 2a shows OD values at 350 nm of the 150 mg/mL protein solution as a function of pH and temperature as a measure of protein aggregation. The initial OD values below 50◦ C were comparable across all pH values examined. The OD 350 nm values at higher pH (6.5 and 7.0) were slightly greater than at lower pH (4.0–6.0), presumably because of the slightly higher initial aggregation levels. As the temperature was increased, the slight shifts of the baseline OD values of the protein seen at 40◦ C–47.5◦ C and 50◦ C–52.5◦ C for pH 7.0 and pH 5.5, respectively, were probably because of air bubbles or a dust in the light path. At pH 4.0, the OD350 values did not change below 55◦ C. These values increased slightly when the temperature increased to approximately 55◦ C–60◦ C or higher. More obvious changes in OD 350 values occurred in the pH range of 5.0–7.0. The overall magnitude of the changes at temperatures above 60◦ C followed the pH sequence 7.0 > 6.5 > 6.0 > 5.5 > 5.0 > 4.5 > 4.0. These results were consistent with the observation of the appearance of gelled material at high temperature. The transparency of the gels differs as a function of pH, being a white opaque gel at pH 7.0 and a transparent gel at pH 4.0. Figure 2b shows Raleigh scattering of the antibody at 150 mg/mL as a function of pH and temperature at 374 nm (in the presence of ANS while ANS fluorescence spectra were collected simultaneously; see below). The major transition temperatures observed by Raleigh scattering are consistent with the OD measurement in that the light scattering intensity did not increase appreciably until the temperature reached 55◦ C–60◦ C for all pH conditions examined. These results indicate that the major aggregation event started at approxiDOI 10.1002/jps.23748

4295

mately 55◦ C. In addition, a less obvious pH-dependent transition at 40◦ C, 47.5◦ C, and 52.5◦ C for pH 4.0, 4.5, and 5.0, respectively, was also observed before the initiation of major transition (Fig. 2b, insert), indicating there was no detectable aggregation events below these temperatures under the corresponding pH conditions. Second-derivative UV spectra of the IgG2 mAb at 150 mg/mL were used to monitor the overall tertiary structure of the protein as sensed by changes in the environment of the aromatic residues. The combination of short pathlength and derivative analysis minimized any light scattering artifacts. As shown in Figure 3a, the second-derivative UV absorption spectra show four minor negative peaks at approximately 253, 259, 269, and 277 nm as well as two major negative peaks at 284 and 293 nm. These minor peaks arise primarily from the absorbance of phenylalanine (253, 259, and 269 nm) and tyrosine (277 nm) residues. The major peak at 284 nm is from the collective effect of the absorbance peaks of tyrosine and tryptophan residues, whereas the peak at 292 nm originates solely from indole side chains.34 The peak positions at the lower temperatures are all comparable across the 4.0–7.0 pH range. As the temperature was increased, the peak positions at pH 4.0–5.5 could be obtained because of the transparent nature of the sample throughout the temperature range studied. In contrast, the peak positions at higher pH (6.0–7.0) could not be accurately obtained at higher temperatures (e.g., above 65◦ C) because of the significant reduction in the light transparency of the sample. The peak positions for the Phe peaks at 253, 259, and 269 nm show either a linear upward shift (at 253 and 259 nm) without any clear transition or no obvious change (at 269 nm) throughout the measured temperature range (data not shown). The peak positions at 284 (Fig. 3b) and 292 nm (Fig. 3c) drifted slightly at lower temperatures and then showed distinct transitions. The linear shift in peak positions was also seen in model phenylalanine, tyrosine, and tryptophan compounds,34 and the slope of these peaks has been shown to reflect changes in the dynamic properties of proteins.35 At pH 4.0, there were two notable structural transitions as a function of temperature, the first transition starting at approximately 30◦ C, followed by a second that initiates at approximately 50◦ C and ends at 75◦ C. A structural transition was also observed for the protein at pH 4.5 starting at approximately 45.0◦ C, or at a slightly higher temperature at pH 5.0 or above with the transition onset temperature difficult to define. Figure 4a shows representative ANS fluorescence spectra at 10◦ C and 60◦ C in the presence of the IgG2 mAb at 150 mg/mL, pH 5.5. Similar fluorescence spectra were obtained at other pH conditions (data not shown). The ANS fluorescence spectra showed a broad peak at approximately 480–490 nm. The shape of the spectra suggests the presence of a second peak. The major peak was at 480 nm, comparable to that obtained at 0.2 mg/mL protein.25 The second peak or the peak shoulder has a position near 490 nm, a much lower value than the 500 nm that was seen previously at 0.2 mg/mL protein.25 The major peak that was independent of protein concentration is probably from high-affinity apolar binding of ANS to a protein region that is less exposed to the surrounding environment. The second peak or the peak shoulder, which varies with protein concentration at 0.2 and 150 mg/mL, may reflect the binding of ANS molecules to more exposed apolar surfaces. The intensity profiles at different temperatures were generally parallel between 400 and 600 nm except for those acquired during a structural Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4296


Figure 2. Thermal aggregation of the IgG2 mAb as a function of solution pH. (a) OD at 350 nm, and (b) static light scattering (at 374 nm) measurements as a function of pH and temperature at 150 mg/mL (n = 2–4). The insert in (b) shows more detailed changes in light scattering intensity at the temperature below 60◦ C. The error bars for some data sets were omitted to better present the data. Error bars represent 1 SD.

transition. Figures 4b and 4c show, respectively, ANS peak intensity and peak wavelength as a function of pH and temperature in the presence of 150 mg/mL mAb. The initial ANS peak intensity at a temperature before the first transition started follows the order: pH 4.0 > pH 4.5 > pH 5.0 > pH 6.0–7.0. This decreasing trend of ANS fluorescence intensity as a function of pH in the presence of 150 mg/mL mAb was further confirmed in a separate experiment (data not shown). As the temperature increased, ANS intensity decreased monotonically prior to a transition (Fig. 4b), which was also accompanied by a blue shift in peak wavelength from approximately 480 to 467 nm (Fig. 4c). The transition onset temperature (the temperature at the minimum intensity) was approximately 27.5◦ C, 37.5◦ C, 45◦ C, 52.5◦ C, 52.5◦ C, 55◦ C, and 55◦ C for pH 4.0, 4.5, 5.0, 5.5, 6.0 6.5, and 7.0, respectively. Unlike the intensity changes in ANS fluorescence intensity at 0.2 mg/mL protein at an ANS/protein molar ratio being 28,25 the alteration in ANS fluorescence intensity for the peak at 480 nm in the presence of 150 mg/mL protein during the transition was probably not caused by binding of adCheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

ditional unbound ANS molecules to the antibody molecules as the ANS/protein molar ratio was 0.5 at this high concentration. Therefore, the changes in ANS fluorescence intensity was more likely because of the collective effects of the movement of bound ANS molecules from relatively less apolar to more apolar regions36 during the structure transition and the resulting increase in viscosity37 (see below). In addition, fewer significant transitions in ANS intensity were observed at 150 mg/mL protein concentration at pH 4.0 than those obtained at 0.2 mg/mL at temperatures above the first transition.25 This could be because of the collective effects of the saturation of fluorescence signals as well as the solidification (gelation) of the solution at 150 mg/mL. The partial specific volume was also measured to characterize the overall volume of the protein molecules under different pH and temperature conditions. At pH 7.0, the partial specific volume (N) of the protein as a function of temperature (T, ◦ C) followed the equation N = 3.73 × 10−4 T + 0.709 (R2 = 0.999) and increased linearly from 0.719 mL/g at 10◦ C to 0.733 mL/g DOI 10.1002/jps.23748


4297

Figure 3. Second-derivative UV absorbance analysis of the IgG2 mAb (150 mg/mL). Representative spectra at pH 4.0 at 10.0◦ C and 80◦ C (a). UV absorbance peak positions for Trp/Tyr (b) and Trp peaks (c) as a function of pH and temperature. Error bars represent 1 SD with n = 2–4.

at 50◦ C. A similar relationship, N = 3.40 × 10−4 T + 0.716 (R2 = 0.996), between the partial specific volume and temperature was also observed for the protein at pH 4.0. The differences in partial specific volume at the two different pH conditions are within the measurement error of the actual protein concentration using the current dilution procedure. At both pH conditions, there was no clear transition observed over the entire range of measured temperatures, indicating that the partial DOI 10.1002/jps.23748

specific volume is not sensitive to the limited alterations in the structure of this protein seen here. The viscosity of the IgG2 mAb solution at 150 mg/mL under different pH conditions in C–P buffer at 20◦ C was measured as shown in Figure 5a. The viscosity at 0.2 mg/mL was between 1.09 and 1.10 cP, which is comparable to that of the buffer alone (data not shown). There was no clear correlation between the viscosity and pH of the antibody solution at this low protein Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4298


Figure 4. Representative ANS fluorescence emission spectra at pH 5.5 (a), ANS fluorescence peak intensity (b), and ANS peak wavelength (c) as a function of pH and temperature in the presence of 150 mg/mL IgG2 mAb. ANS to protein molar ratio was 0.5. Error bars represent 1 SD with n = 2 or 4.

concentration. In contrast, the solution viscosity of the antibody at 150 mg/mL was highly pH dependent. It increased slightly from 5.9 to 7.9 cP at pH 4.0–5.5 but more significantly from 7.9 to 23.9 cP from pH 5.5 to 7.0. The viscosity values were independent of the measurement method employed. Figure 5b shows the viscosity of the antibody solution at 150 mg/mL as a function of pH and temperature. At pH 4.0, the viscosity was 5.5–5.9 cP at temperatures from 20◦ C to 27.5◦ C. When the temperature reached at 30◦ C–37.5◦ C, the viscosity increased significantly from 6.2 to 12.6 cP, and another phase of viscosity increase from 12.7 to 22.9 cP was observed from 40◦ C to 47.5◦ C. At pH 4.5, there was a slight decreasing trend in the viscosity from 6.8 to 3.8 cP until the temperature reached 42.5◦ C, where the viscosity increased from 3.8 to 12.1 cP at Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

57.5◦ C in apparently two phases. At pH 5.0 and 5.5, a similar trend of decrease in viscosity was observed until the temperature reached 50.0◦ C and 52.5◦ C, respectively, at which point the viscosity increased from 3.2 and 3.1 to 7.8 and 4.7 at 60◦ C, respectively. At pH 6.0, 6.5, and 7.0, solution viscosity decreased at a variable rate depending on the initial viscosity, until the temperature reached 60◦ C and above. At higher temperature, the viscosity either became so high that an accurate measurement was not possible, (e.g., at pH 4.0) because of the loss of Brownian motion of the polystyrene beads when gelation of the protein solution took place, or accurate measurement was impractical because data from the polystyrene beads was difficult to identify because of the interference from protein aggregation (e.g., at pH 7.0) (Fig. 2). Measurement of the viscosity by other DOI 10.1002/jps.23748


4299

Figure 5. The dynamic viscosity of IgG2 mAb solution at 150 mg/mL protein as a function of pH at 20◦ C (a) and as a function of pH and temperature (b). The error bars are 1 SD and are hidden within the symbols at some time points (n ≥ 2).

methods (Stabinger and mVROC viscometers) were equally difficult. To better understand the relationship among viscosity, protein structural change, and the net charge of the protein, zeta potential values were measured. Zeta potential decreased nonlinearly from 5.3 (at pH 4.0) to 0.6 mV at pH 5.0 and then further decreased to −1.0 and −1.3 mV at pH 6.0 and 7.0, respectively. The trend in zeta potential as a function of solution pH observed for this mAb was overall consistent with the theoretical pI value of the antibody (8.7) without considering posttranslational modifications. These values were consistent with those reported for other antibodies,16 although of lower magnitude.

DISCUSSION Several recent studies4,6,7,17–20 have considered factors that may contribute to the viscosity of high-concentration solutions of various antibodies. The high viscosity of antibody solutions results from the self-association of proteins because of noncovalent interactions, usually reversible in nature. The nature of such self-association has not been well defined. For example, self-interaction was initially assigned primarily to Fab–Fab interactions, and in later studies to various noncovalent forces, such as charge–charge and charge–dipole interactions, dipole– dipole contacts, hydrophobic effects, and other short-range and long-range interactions.4,5,8,15,38 DOI 10.1002/jps.23748

In the few CD studies that have examined the relationship between the viscosity of antibody solutions to protein structure, it was concluded that the overall secondary structure of the different antibodies were largely unaltered.4,16,17 Even though all IgG antibodies are homodimers that contain two binding regions (Fab) and two constant region fragments (Fc), together forming a typical “Y” structure,39 different antibodies or the same antibody under different conditions may display small variations in their overall secondary and tertiary structure. Measuring CD profiles in the far (190–240 nm) as well as the near-UV region (250–300 nm) may not be sufficient to conclude either an overall similarity or differences in higher-order structure among different antibodies. Because of the low-resolution nature of many characterization methods, it is generally more informative to use a combination of methods to probe antibody structure in solution. To this end, the secondary structure of the IgG2 mAb at high protein concentration was examined in this work by FTIR spectroscopy. There are numerous approaches to process and analyze FT-IR data of proteins, such as area normalization followed by a curve fitting,32 as well as second-derivative analyses.29 These data processing methods can introduce subjectivity and/or amplify the noise during assignment of each structural component.40 Therefore, vector normalization treatment of FT-IR spectra at each temperature, followed by subtraction of the initial FT-IR spectra in the amide I region at 10◦ C, was used here to minimize the distortions that could otherwise be introduced.40,41 FT-IR spectra showed overall Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4300


comparable secondary structure at all pH values ranging from 4.0 to 7.0 at temperatures below 40◦ C (Fig. 1). Characterization of the tertiary structure of the IgG2 at 150 mg/mL was examined by a combination of secondderivative UV absorbance spectroscopy (Fig. 3) and ANS fluorescence (Fig. 4). Absorbance spectroscopy provides tertiary structure information as represented by the positions of absorbance peak wavelengths for each collection of aromatic amino acids. These data suggest that below 30◦ C, the IgG2 protein assumes the same structure from pH 4.0 to 7.0. The difference in ANS fluorescence peak intensity (Fig. 4) between different pH conditions, however, suggested that the antibody at 150 mg/mL at lower pH formed a structure different from that at higher pH (e.g., pH 5.0–7.0) even at 10◦ C. This result is consistent with the ANS intensity data reported previously for this mAb at 0.2 mg/mL.25 This lack of consistency in the data concerning the nature of the mAb’s overall tertiary structure at low pH (i.e., the contrast between ANS fluorescence intensity and second-derivative UV absorbance peak wavelength changes) is not necessarily surprising as ANS measurements primarily reflect binding of ANS to accessible surface or charged regions of the protein, whereas UV spectrophotometric monitoring of the aromatic side chains generally involves amino acid residues more buried inside the apolar interior of a protein. The transition temperatures of the protein at 150 mg/mL obtained by UV absorbance peak wavelength in this work were approximately 5◦ C–10◦ C lower than that previously observed25 at 0.2 mg/mL by tryptophan fluorescence measurement under the same pH conditions. In addition, the transition onset temperature for ANS intensity data at the 150 mg/mL protein concentration (using the cuvette method with the same experiment settings) was slightly higher than that reported earlier (27.5◦ C vs. 25◦ C at pH 4.0 and 45◦ C vs. 40◦ C at pH 5.0) at 0.2 mg/mL protein concentration.25 The differences observed between 0.2 and 150 mg/mL are consistent with the results of earlier studies examining the effect of protein concentration on structure as well as conformational stability of IgG molecules.42,43 Results obtained with CD and tryptophan fluorescence at 0.2 mg/mL in an earlier study,25 as well as the UV absorbance and FT-IR results of this study at 150 mg/mL, from pH 4.0 to 7.0, could lead to the conclusion that the protein at temperatures ranging from 10◦ C to 30◦ C have similar structure. ANS fluorescence peak intensity data from the current work at 150 mg/mL and previous work at 0.2 mg/mL protein concentration,25 however, suggest a subtle difference in tertiary structures of this IgG2 molecule between low and high pH conditions at lower temperature. ANS fluorescence has been commonly used to probe the accessibility of hydrophobic patches.36 Interestingly, ANS intensity data at pH 4.0–7.0 at both 150 and 0.2 mg/mL was reasonably well correlated with zeta potential measurements (Fig. 6). The positive zeta potential value at low pH (e.g., pH 4.0) supports the argument that binding of ANS could be because of the interactions between the negative-charged sulfonate group and positive-charged amino acid side chains.44 This argument, however, is not supported by the ANS fluorescence peak wavelength (480–490 nm), as ionic binding between negatively charged sulfonate group of ANS and positive-charged protein side-chain residues would be expected to result in a peak near 516 nm at low protein concentration had only ionic interaction been involved.45 Although ANS intensity increased significantly as the temperature increased at each pH examined, this effect was not accompanied Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

by significant aggregation at lower temperatures as measured by OD (350 nm) and static light scattering (e.g., pH 4.0 below 42.5◦ C, Fig. 2b). It is also likely that such ionic interactions would facilitate apolar interactions, as binding of the sulfonate group to positively charged side chains is thermally favorable.44 Instead, the repulsive force from additional positive charges of the protein at low pH could outweigh the attractive force of apolar interactions and therefore result in a conformation with largely intact native-like secondary structure and tertiary structure as measured by second-derivative UV absorbance but a less tightly packed tertiary conformation as seen by ANS fluorescence intensity. Such states fit the definition of molten globules.46–48 Optical density results at 350 nm (Fig. 2a) and static light scattering data at 374 nm (Fig. 2b) reflect self-association of the protein.49,50 Increasing the temperature did not significantly change the OD values and light scattering intensity until the temperature reached approximately 55◦ C–60◦ C (across the pH range of 5.0–7.0). These results, when compared with the changes in solution viscosity (Fig. 5b), for the protein across all measured temperatures and pH conditions suggest no obvious correlation between OD (350 nm) and solution viscosity for this particular IgG2 mAb. These results, together with FTIR (Fig. 1), second-derivative UV absorbance (Fig. 3), and ANS fluorescence data (Fig. 4) at 150 mg/mL, suggest that the initial structural transitions of this antibody were not necessarily accompanied by aggregation. This result is consistent with the findings on the structure–aggregation relationship at low pH conditions seen by others.51 Another observation from the present study was that gels of different opacities were formed upon heating, depending on pH. A transparent gel appeared to be formed from an array of unfolded protein molecules because of the strong electrostatic repulsive forces at low pH conditions, whereas a white opaque gel seemed to be formed from aggregation of the protein in native-like states because of the low electrostatic repulsive forces at high pH.52,53 The trend in the change of the partial specific volume of the protein as a function of temperature was generally consistent with previous observations except that the linear relationship seen here differs from some other reports.54 This difference may be because of the change in protein concentration caused by temperature-dependent volume expansion, which was taken into account in this study. Even though the change of partial specific volume as a function of temperature was very small in magnitude, the partial specific volume data do indicate that the protein volume significantly increased as a function of temperature, and that this increase was independent of the observed thermal transitions at pH 4.0 as measured by all of the structural characterization methods used in this and our earlier study.25 Partial specific volume has been found to have a reasonably good correlation with the compressibility, that is, fluctuations and relaxations of a protein molecule in solution.55,56 Although we have found no previous attempt to use partial specific volume values to directly probe protein dynamic properties, the linear change of partial specific volume does correlate well with the temperature and the initial shift of peak positions of aromatic residues in the second-derivative UV absorption spectra, in which the slope of the shift of the peak position was considered to be a measure of dynamic aspects of protein behavior.35 There is a monotonic nonlinear decrease of solution viscosity as the temperature increases at values below the structural DOI 10.1002/jps.23748


4301

Figure 6. The correlation of zeta potential values with ANS fluorescence intensity (for IgG2 mAb solutions at different pH values). Solutions contain antibody at either 150 or 0.2 mg/mL prepared in a 20-mM C–P buffer at a total ionic strength of 0.15 adjusted with NaCl in the pH range of 4.0–7.0 at 25◦ C. Error bars represent 1 SD with n = 2.

transition onset temperatures (Fig. 5b). The viscosity values as a function of temperature below the onset transition temperature at pH 4.5–7.0 were well fit by the equation Ln(0) = aT + b (Table 1), where 0 is the viscosity and T is the temperature, with a and b the slope and intercept of the function, respectively. These fitting results were consistent with earlier studies of the viscosity of water,28 antibodies, and other protein solutions, which can be quantitatively described by the Arrhenius equation.57,58 Because of the limited number of data points below the transition temperature in this study, the viscosity data at pH 4.0 were not fit to this equation. There appears to be an overall good correlation between the solution viscosity as a function of temperature with the IgG2 mAb’s higher-order structure (in terms of the behavior of the major transition) as represented by the various structural characterization results, including second-derivative UV absorbance, FT-IR spectroscopy, and ANS fluorescence, as well as tryptophan fluorescence, CD, and differential scanning calorimetry results from an earlier study.25 The monotonic change of solution viscosity with temperature (below the mAb’s transition temperatures) at each pH correlates reasonably well with changes in both tryptophan (correlation data not shown) and ANS fluorescence intensity at 0.2 mg/mL as seen previously,25 as well as ANS fluorescence intensity at protein concentrations of 150 mg/mL in this study. Figures 7a and 7b show the correlation of the viscosity of the IgG2 mAb solutions at 150 mg/mL with ANS fluorescence intensity as a function of temperature in the presence of the antibody at 150 and 0.2 mg/mL, respectively. The good correlation between fluorescence intensity and viscosity follows the relation-

ship known as the Forster–Hoffmann equation.59 This relationship has been verified with other fluorescence molecules, such as 9-(dicyanovinyl)-julolidine and 9-(2-carboxy-2-cyano)vinyljulolidine,60 in that increases in viscosity can lead to higher fluorescence intensities. This equation does not, however, explain the correlation of fluorescence intensity with viscosity as a function of pH observed in this study, as ANS fluorescence intensity showed a reverse relationship with viscosity as a function of pH. The correlation of viscosity of the mAb at 150 mg/mL with ANS fluorescence intensity at protein concentration of 0.2 mg/mL under the corresponding pH conditions (Fig. 7b) is particularly noteworthy. The protein solution viscosity at pH 4.0 had good correlation with ANS intensity values during its structural transition (25◦ C–37.5◦ C) as well. This transition was not, however, accompanied by aggregation of the antibody at 150 mg/mL (Fig. 2). The correlation coefficient value of 0.97 or better suggests that the binding of ANS to the protein (at 0.2 mg/mL) could potentially be used to monitor structural factors that impact the viscosity of high-concentration protein solutions. The good correlation between the viscosity and the binding as represented by ANS fluorescence intensity in the presence of the antibody at 0.2 mg/mL as a function of temperature might suggest that intermolecular hydrophobic interactions are the predominant force that accounts for the observed high viscosity of the protein solution at pH 4.0 at elevated temperatures. One should, however, be cautioned that the relevant charge– charge interactions associated with structural changes might also differently contribute as the relative position of charged side chains alter. The correlation of the viscosity data with ANS fluorescence intensity in the presence of the antibody at

Table 1. Viscosity (0) of IgG2 mAb Solutions at 150 mg/mL as a Function of Temperature (T), at Temperatures Below the Onset Transition Temperature, Were Fitted to the Equation Ln (0) = aT + b pH condition Upper fitting limit (◦ C) a b R2

4.5 40.0 − 0.0311 2.54 − 0.995

5.0 47.5 − 0.0300 2.54 − 0.995

5.5 55.0 − 0.0289 2.62 − 0.995

6.0 60.0 − 0.0282 2.86 − 0.997

6.5 60.0 − 0.0282 3.40 − 0.994

7.0 60.0 − 0.0386 4.03 − 0.993

The antibody solution contained 20 mM C–P buffer at indicated pH at a total ionic strength of 0.15 by addition of NaCl. DOI 10.1002/jps.23748


4302


Figure 7. The correlation of the solution viscosity of the IgG2 at 150 mg/mL with ANS fluorescence peak intensity in 20 mM C–P buffer. (a) The solution viscosity was correlated with ANS fluorescence peak intensity in the presence the IgG2 at 150 mg/mL at pH 4.5 (20◦ C–37.5◦ C), pH 5.0 (20◦ C–45◦ C), pH 5.5 (20◦ C–52.5◦ C), and pH 6.0–7.0 (20◦ C–55 ◦ C). (b) The solution viscosity was correlated with ANS fluorescence peak intensity in the presence of the IgG2 mAb at 0.2 mg/mL at pH 4.0 (20◦ C–37.5◦ C), pH 5.0 (20◦ C–47.5◦ C), pH 6.0 (20◦ C–60◦ C), and pH 7.0 (20◦ C–60◦ C).

150 mg/mL at pH 4.0 at temperatures ranging from 20◦ C to 37.5◦ C was not as good (I = 20530 0 + 377908; R2 = 0.77) as that at 0.2 mg/mL protein concentration. This discrepancy of the viscosity–ANS fluorescence intensity relationship between antibody solutions at 0.2 and 150 mg/mL could be explained by the underlying reasons for the change in fluorescence intensity. ANS fluorescence intensity in the presence of 0.2 mg/mL of antibody was not appreciably affected by the change of viscosity as a result of changes in the structure of the antibody, whereas in the presence of 150 mg/mL protein, ANS fluorescence intensity was not only affected by the changes in structure, but also simultaneously affected by resulting change of viscosity, which has been known to affect fluorescence intensity.37 The correlation of ANS fluorescence intensity measurements at 0.2 mg/mL and solution viscosity at 150 mg/mL was not satisfactory at temperatures far above the IgG2 mAb’s transition onset temperature. For example, at pH 5.0 and temperatures ranging from 50◦ C to 60◦ C, the viscosity showed a small transition, although the extent of viscosity change and the number of data points collected were not sufficient to establish a rigorous relationship. At pH 4.0 and in the temperature range 37.5◦ C–


47.5◦ C, there was a clear trend of ANS fluorescence intensity increase with increases in viscosity. The viscosity was, however, measured over a prolonged period of time and the protein could undergo significant time-dependent changes in structure during the measurement period.25,61,62 Therefore, the change of solution viscosity was observed at a slightly lower temperature than that observed for ANS peak intensity measurements. More viscosity data points with more tightly controlled conditions, however, will be needed to establish a correlation over a wider temperature range and corresponding protein conformational changes events. Even though ANS fluorescence peak intensity measured at 0.2 mg/mL was approximately the same at pH 5.0–7.0 below the onset folding temperature,25 the overall trend in ANS intensity at 150 mg/mL protein concentration appeared to be slightly different from that obtained at 0.2 mg/mL as far as pH-dependent intensity is concerned. The difference in the relationship of ANS intensity with pH may reveal a subtle difference in the structure of the protein between low and high protein concentrations at these pH values. The discrepancy in the relationship of ANS intensity with pH measured at low and high protein

DOI 10.1002/jps.23748


concentration remains to be further confirmed. Whatever the case, the total fluorescence intensity is still very significant in magnitude despite the difference in ANS intensity between pH 5.0 and 7.0 at both low and high protein concentrations. The apolar binding sites for ANS molecules could also be binding sites for apolar interactions between protein molecules. Apolar effects, however, did not result in irreversible aggregation, as the protein at 150 mg/mL, when dialyzed and diluted to 0.2 mg/mL, showed comparable light scattering intensity.25 This indicates that strong repulsive forces must exist to counteract any apolar force. In addition, the FT-IR results, collectively with the viscosity results across different pH conditions, appear to be somewhat contradictory to the findings of Kamerzell et al.23 in that the change of viscosity is not necessarily associated with secondary structural changes. It should be noted, however that in the current study the same protein concentrations as well as the same ionic strength were used across all conditions employed. Therefore, changes in viscosity may not necessarily be limited to tertiary structure changes as detected by ANS fluorescence, but could further involve aspects of the secondary structure of the protein. The interrelationship between protein structure and solution viscosity has only begun to be explored in this work. For example, differences in electrostatic interactions at low pH compared with higher pH could result in structural changes that could be detected by ANS fluorescence intensity, whereas structural transitions at a single pH could result in a different viscosity without changes in the aggregation status seen when the structure was altered by temperature. The viscosity could also be affected by other protein thermodynamic properties, as reflected in the partial specific volume of the protein. Correlations of solution viscosity to zeta potential did not reveal a direct relationship. Differences in the structure of the antibody as a function of pH could further complicate any relationships as the relative location of ionic groups as a function of pH, along with changes in apolarity, could contribute collectively to the viscosity of high-concentration antibody solutions. Even though this study does not offer a method to predict the viscosity of a concentrated solution based on a protein’s structure, it does demonstrate that the structure, which is the collective result of apolar and ionic factors as well as protein thermodynamic behavior, plays an important role in the viscosity of highly concentrated protein solutions. Thus, a close relationship between viscosity and the structure of a protein as a function of temperature and pH clearly exists.

REFERENCES 1. Nelson AL, Dhimolea E, Reichert JM. 2010. Development trends for human monoclonal antibody therapeutics. Nat Rev Drug Discov 9(10):767–774. 2. Shire SJ, Shahrokh Z, Liu J. 2004. Challenges in the development of high protein concentration formulations. J Pharm Sci 93(6):1390–1402. 3. Liu J, Nguyen MD, Andya JD, Shire SJ. 2005. Reversible selfassociation increases the viscosity of a concentrated monoclonal antibody in aqueous solution. J Pharm Sci 94(9):1928–1940. 4. Kanai S, Liu J, Patapoff TW, Shire SJ. 2008. Reversible selfassociation of a concentrated monoclonal antibody solution mediated by Fab-Fab interaction that impacts solution viscosity. J Pharm Sci 97(10):4219–4227. 5. Yadav S, Laue TM, Kalonia DS, Singh SN, Shire SJ. 2012. The influence of charge distribution on self-association and viscosity behavior of monoclonal antibody solutions. Mol Pharm 9(4):791–802. DOI 10.1002/jps.23748

4303

6. Saito S, Hasegawa J, Kobayashi N, Kishi N, Uchiyama S, Fukui K. 2012. Behavior of monoclonal antibodies: Relation between the second virial coefficient (B22) at low concentrations and aggregation propensity and viscosity at high concentrations. Pharm Res 29(2):397–410. 7. Yadav S, Shire SJ, Kalonia DS. 2010. Factors affecting the viscosity in high concentration solutions of different monoclonal antibodies. J Pharm Sci 99(12):4812–4829. 8. Yadav S, Liu J, Shire SJ, Kalonia DS. 2010. Specific interactions in high concentration antibody solutions resulting in high viscosity. J Pharm Sci 99(3):1152–1168. 9. George A, Wilson WW. 1994. Predicting protein crystallization from a dilute solution property. Acta Crystallogr D Biol Crystallogr 50(Pt 4):361–365. 10. Jing Z, Liu XY. 2003. Effect of protein-protein interactions on protein aggregation kinetics. J Chem Phys 119(20):10972–10976. 11. Jia Y, Narayanan J, Liu XY, Liu Y. 2005. Investigation on the mechanism of crystallization of soluble protein in the presence of nonionic surfactant. Biophys J 89(6):4245–4251. 12. Yadav S, Scherer TM, Shire SJ, Kalonia DS. 2011. Use of dynamic light scattering to determine second virial coefficient in a semidilute concentration regime. Anal Biochem 411(2):292–296. 13. Tessier PM, Lenhoff AM, Sandler SI. 2002. Rapid measurement of protein osmotic second virial coefficients by self-interaction chromatography. Biophys J 82(3):1620–1631. 14. Tessier PM, Sandler SI, Lenhoff AM. 2004. Direct measurement of protein osmotic second virial cross coefficients by cross-interaction chromatography. Protein Sci 13(5):1379–1390. 15. Yadav S, Shire SJ, Kalonia DS. 2011. Viscosity analysis of high concentration bovine serum albumin aqueous solutions. Pharm Res 28(8):1973–1983. 16. Yadav S, Shire SJ, Kalonia DS. 2012. Viscosity behavior of highconcentration monoclonal antibody solutions: Correlation with interaction parameter and electroviscous effects. J Pharm Sci 101(3):998– 1011. 17. Yadav S, Sreedhara A, Kanai S, Liu J, Lien S, Lowman H, Kalonia DS, Shire SJ. 2011. Establishing a link between amino acid sequences and self-associating and viscoelastic behavior of two closely related monoclonal antibodies. Pharm Res 28(7):1750–1764. 18. Guo Z, Chen A, Nassar RA, Helk B, Mueller C, Tang Y, Gupta K, Klibanov AM. 2012. Structure-activity relationship for hydrophobic salts as viscosity-lowering excipients for concentrated solutions of monoclonal antibodies. Pharm Res 29(11):3102–3109. 19. Du W, Klibanov AM. 2011. Hydrophobic salts markedly diminish viscosity of concentrated protein solutions. Biotechnol Bioeng 108(3):632–636. 20. He F, Woods CE, Litowski JR, Roschen LA, Gadgil HS, Razinkov VI, Kerwin BA. 2011. Effect of sugar molecules on the viscosity of high concentration monoclonal antibody solutions. Pharm Res 28(7):1552– 1560. 21. Galush WJ, Le LN, Moore JM. 2012. Viscosity behavior of highconcentration protein mixtures. J Pharm Sci 101(3):1012–1020. 22. Liu C, Vailhe C, Samuel N, Min Q. 2009. A simple and reproducible approach to characterize protein stability using rheology. Int J Pharm 374(1–2):1–4. 23. Kamerzell TJ, Kanai S, Liu J, Shire SJ, Wang YJ. 2009. Increasing IgG concentration modulates the conformational heterogeneity and bonding network that influence solution properties. J Phys Chem B 113(17):6109–6118. 24. Daumantas M, Christoph GB, Victor AB, Rex EL. 1999. 1-Anilino8-naphthalene sulfonate as a protein conformational tightening agent. Biopolymers 49(6):451–458. 25. Cheng W, Joshi SB, He F, Brems DN, He B, Kerwin BA, Volkin DB, Middaugh CR. 2012. Comparison of high-throughput biophysical methods to identify stabilizing excipients for a model IgG2 monoclonal antibody: Conformational stability and kinetic aggregation measurements. J Pharm Sci 101(5):1701–1720. 26. Stothart PH. 1984. Determination of partial specific volume and absolute concentration by densimetry. Biochem J 219(3):1049–1052. Cheng et al., JOURNAL OF PHARMACEUTICAL SCIENCES 102:4291–4304, 2013

4304


27. He F, Becker GW, Litowski JR, Narhi LO, Brems DN, Razinkov VI. 2010. High-throughput dynamic light scattering method for measuring viscosity of concentrated protein solutions. Anal Biochem 399(1):141– 143. 28. Kestin J, Sokolov M, Wakeham WA. 1978. Viscosity of liquid water in the range −8◦ C to 150◦ C. J Phys Chem Ref Data 7:941. 29. Susi H, Byler DM. 1983. Protein structure by Fourier transform infrared spectroscopy: Second derivative spectra. Biochem Biophys Res Commun 115(1):391–397. 30. Murphy BM, Zhang N, Payne RW, Davis JM, Abdul-Fattah AM, Matsuura JE, Herman AC, Manning MC. 2012. Structure, stability, and mobility of a lyophilized IgG1 monoclonal antibody as determined using second-derivative infrared spectroscopy. J Pharm Sci 101(1):81– 91. 31. Clark AH, Saunderson DH, Suggett A. 1981. Infrared and laserRaman spectroscopic studies of thermally-induced globular protein gels. Int J Pept Protein Res 17(3):353–364. 32. Byler DM, Susi H. 1986. Examination of the secondary structure of proteins by deconvolved FTIR spectra. Biopolymers 25(3):469–487. 33. Ismail AA, Mantsch HH, Wong PT. 1992. Aggregation of chymotrypsinogen: Portrait by infrared spectroscopy. Biochim Biophys Acta 1121(1–2):183–188. 34. Kueltzo LA, Ersoy B, Ralston JP, Middaugh CR. 2003. Derivative absorbance spectroscopy and protein phase diagrams as tools for comprehensive protein characterization: A bGCSF case study. J Pharm Sci 92(9):1805–1820. 35. Esfandiary R, Hunjan JS, Lushington GH, Joshi SB, Middaugh CR. 2009. Temperature dependent 2nd derivative absorbance spectroscopy of aromatic amino acids as a probe of protein dynamics. Protein Sci 18(12):2603–2614. 36. Stryer L. 1965. The interaction of a naphthalene dye with apomyoglobin and apohemoglobin. A fluorescent probe of non-polar binding sites. J Mol Biol 13(2):482–495. 37. Sharafy S, Muszkat KA. 1971. Viscosity dependence of fluorescence quantum yields. J Am Chem Soc 93(17):4119–4125. 38. Chari R, Jerath K, Badkar AV, Kalonia DS. 2009. Long- and shortrange electrostatic interactions affect the rheology of highly concentrated antibody solutions. Pharm Res 26(12):2607–2618. 39. Wang W, Singh S, Zeng DL, King K, Nema S. 2007. Antibody structure, instability, and formulation. J Pharm Sci 96(1):1–26. 40. Surewicz WK, Mantsch HH, Chapman D. 1993. Determination of protein secondary structure by Fourier transform infrared spectroscopy: A critical assessment. Biochemistry 32(2):389–394. 41. Goldberg ME, Chaffotte AF. 2005. Undistorted structural analysis of soluble proteins by attenuated total reflectance infrared spectroscopy. Protein Sci 14(11):2781–2792. 42. Harn N, Allan C, Oliver C, Middaugh CR. 2007. Highly concentrated monoclonal antibody solutions: Direct analysis of physical structure and thermal stability. J Pharm Sci 96(3):532–546. 43. Bhambhani A, Kissmann JM, Joshi SB, Volkin DB, Kashi RS, Middaugh CR. 2012. Formulation design and high-throughput excipient selection based on structural integrity and conformational stability of dilute and highly concentrated IgG1 monoclonal antibody solutions. J Pharm Sci 101(3):1120–1135. 44. Matulis D, Lovrien R. 1998. 1-Anilino-8-naphthalene sulfonate anion-protein binding depends primarily on ion pair formation. Biophys J 74(1):422–429.


45. Gasymov OK, Glasgow BJ. 2007. ANS fluorescence: Potential to augment the identification of the external binding sites of proteins. Biochim Biophys Acta 1774(3):403–411. 46. Chakraborty S, Peng Z. 2000. Hierarchical unfolding of the alphalactalbumin molten globule: Presence of a compact intermediate without a unique tertiary fold. J Mol Biol 298(1):1–6. 47. Buchner J, Renner M, Lilie H, Hinz HJ, Jaenicke R, Kiefhabel T, Rudolph R. 1991. Alternatively folded states of an immunoglobulin. Biochemistry 30(28):6922–6929. 48. Semisotnov GV, Rodionova NA, Razgulyaev OI, Uversky VN, Gripas AF, Gilmanshin RI. 1991. Study of the “molten globule” intermediate state in protein folding by a hydrophobic fluorescent probe. Biopolymers 31(1):119–128. 49. Bateman JB, Weneck EJ, Eshler DC. 1959. Determination of particle size and concentration from spectrophotometric transmission. J Colloid Sci 14(3):308–329. 50. Sukumar M, Doyle BL, Combs JL, Pekar AH. 2004. Opalescent appearance of an IgG1 antibody at high concentrations and its relationship to noncovalent association. Pharm Res 21(7):1087– 1093. 51. Sahin E, Weiss WF 4th, Kroetsch AM, King KR, Kessler RK, Das TK, Roberts CJ. 2012. Aggregation and pH-temperature phase behavior for aggregates of an IgG2 antibody. J Pharm Sci 101(5):1678– 1687. 52. Barbut S. 1995. Effects of calcium level on the structure of preheated whey protein isolate gels. LWT-Food Sci Technol 28(6):598– 603. 53. Doi E. 1993. Gels and gelling of globular proteins. Trends Food Sci Technol 4(1):1–5. 54. Lee S, Tikhomirova A, Shalvardjian N, Chalikian TV. 2008. Partial molar volumes and adiabatic compressibilities of unfolded protein states. Biophys Chem 134(3):185–199. 55. Gekko K, Hasegawa Y. 1986. Compressibility-structure relationship of globular proteins. Biochemistry 25(21):6563–6571. 56. Kamerzell TJ, Ramsey JD, Middaugh CR. 2008. Immunoglobulin dynamics, conformational fluctuations, and nonlinear elasticity and their effects on stability. J Phys Chem B 112(10):3240–3250. 57. Monkos K. 1996. Viscosity of bovine serum albumin aqueous solutions as a function of temperature and concentration. Int J Biol Macromol 18(1–2):61–68. 58. Monkos K, Turczynski B. 1999. A comparative study on viscosity of human, bovine and pig IgG immunoglobulins in aqueous solutions. Int J Biol Macromol 26(2–3):155–159. 59. Haidekker MA, Theodorakis EA. 2010. Environment-sensitive behavior of fluorescent molecular rotors. J Biol Eng 4:11. 60. Sutharsan J, Lichlyter D, Wright NE, Dakanali M, Haidekker MA, Theodorakis EA. 2010. Molecular rotors: Synthesis and evaluation as viscosity sensors. Tetrahedron 66(14):2582–2588. 61. Santiago PS, Carvalho JW, Domingues MM, Santos NC, Tabak M. 2010. Thermal stability of extracellular hemoglobin of Glossoscolex paulistus: Determination of activation parameters by optical spectroscopic and differential scanning calorimetric studies. Biophys Chem 152(1–3):128–138. 62. Kurganov BI, Lyubarev AE, Sanchez-Ruiz JM, Shnyrov VL. 1997. Analysis of differential scanning calorimetry data for proteins. Criteria of validity of one-step mechanism of irreversible protein denaturation. Biophys Chem 69(2–3):125–135.

DOI 10.1002/jps.23748

Observation of small cluster formation in concentrated monoclonal antibody solutions and its implications to solution viscosity.

Mitigation of reversible self-association and viscosity in a human IgG1 monoclonal antibody by rational, structure-guided Fv engineering.

Cysteinylation of a monoclonal antibody leads to its inactivation.

The peritrophic membrane as a barrier: its penetration by Plasmodium gallinaceum and the effect of a monoclonal antibody to ookinetes.

Development of technology for linking photosensitizers to a model monoclonal antibody.

Properties of acid whey as a function of pH and temperature.

Intermolecular Interactions and the Viscosity of Highly Concentrated Monoclonal Antibody Solutions.

Effective protocol for the investigation of physicochemical and conformational stability and aggregation kinetics measurements of therapeutic IgG2 monoclonal antibody.

Freezing-induced perturbation of tertiary structure of a monoclonal antibody.

function relationships between capsid proteins and genetics of susceptibility to antibody action.

The heterogeneity of bovine IgG2--V. Differences in the primary structure of bovine IgG2 allotypes.

The preparation of an ovine monoclonal antibody to progesterone.

Erratum to: Thermodynamic Prediction of Glycine Polymerization as a Function of Temperature and pH Consistent with Experimentally Obtained Results.

Cross-linking of a monoclonal antibody-antigen complex enables detection of parasite antigen in immunoblots and in an expression library.

Crystal structure of Plasmodium knowlesi apical membrane antigen 1 and its complex with an invasion-inhibitory monoclonal antibody.

Density functional theory simulation of hydrogen-bonding structure and vibrational densities of states at the quartz (101)-water interface and its relation to dissolution as a function of solution pH and ionic strength.

Murine monoclonal anti-idiotype antibody (alpha) as a probe to detect human monoclonal antibody bound to human tumor tissues.

An equation to calculate the actual methylene middle parameter as a function of temperature.

Placental transfer of a fully human IgG2 monoclonal antibody in the cynomolgus monkey, rat, and rabbit: a comparative assessment from during organogenesis to late gestation.

Temperature-ramped studies on the aggregation, unfolding, and interaction of a therapeutic monoclonal antibody.

The molecular composition of dissolved organic matter in forest soils as a function of pH and temperature.

Formation of large, membrane skeleton-free erythrocyte vesicles as a function of the intracellular pH and temperature.

Entanglement model of antibody viscosity.

Characterization of a monoclonal antibody that specifically binds to choline phospholipids and its use in immunocytochemistry.