Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Mar 1.
Published in final edited form as: J Pharm Sci. 2022 Oct 25;112(3):680–690. doi: 10.1016/j.xphs.2022.10.017

Conformational Changes and Drivers of Monoclonal Antibody Liquid-Liquid Phase Separation

Nicholas R Larson 1,2,a,, Yangjie Wei 1,2,b,, Thayana Araújo Cruz 1,3, Reza Esfandiary 1,2,c, Cavan K Kalonia 1,2,c, M Laird Forrest 1, C Russell Middaugh 1
PMCID: PMC9974558  NIHMSID: NIHMS1844889  PMID: 36306862

Abstract

Liquid-liquid phase separation is a phenomenon within biology whereby proteins can separate into dense and more dilute phases with distinct properties. Three antibodies that undergo liquid-liquid phase separation were characterized in the protein-rich and protein-poor phases. In comparison to the protein-poor phase, the protein-rich phase demonstrates more blue-shift tryptophan emissions and red-shifted amide I absorbances. Large changes involving conformational isomerization around disulfide bonds were observed using Raman spectroscopy. Amide I and protein fluorescence differences between the phases persisted to temperatures above the critical temperature but ceased at the temperature at which aggregation occurred. In addition, large changes occurred in the structural organization of water molecules within the protein-rich phase for all three antibodies. It is hypothesized that as the proteins have the same chemical potential in both phases, the protein viscosity is higher in the protein-rich phase resulting in slowed diffusion dependent protein aggregation in this phase. For all three antibodies we performed accelerated stability studies and found that the protein-rich phase aggregated at the same rate or slower than the protein-poor phase.

Introduction

Protein liquid-liquid phase separation (LLPS) occurs when a solution can achieve a lower free energy state by partitioning itself into two phases 1. An equilibrium is reached when the chemical potential is equal in the two phases. Thus, phase transitions occur about critical points of solution chemical potential. One phase is more dense, viscous, and protein-rich while the other phase is relatively protein-poor 24. Entropic and enthalpic effects can both promote LLPS. It is possible for non-interacting particles to exhibit phase separation solely via entropic effects (i.e. osmotic depletion) 5. Most proteins, however, do not exhibit phase separation even at high protein volume fractions 24. Therefore protein-specific interactions primarily dictate whether a protein will undergo phase separation and the shape of the liquid-liquid coexistence curve. Ionic, hydrophobic, π-π, van der Waals, hydrogen bonding, and other forces can all contribute to favorable protein-self interactions that drive LLPS 2.

It has become apparent, recently, that many more proteins than once thought demonstrate this phenomenon, especially at high concentrations, including many immunoglobulins (e.g. IgG) 610. The highly variable nature of phase separation in IgG molecules has indicated the sensitivity of phase separation to small changes in structure and solution environment. Weak self-associative interactions at high volume fractions cause proteins in solution to form transient clusters11. In the case of phase separation, these clusters form larger branched networks of proteins which make up the protein-rich phase 12. Neutron spin echo and small-angle neutron scattering experiments have identified the formation of dynamic clusters of proteins (including antibodies) in concentrated solutions 1315. Increases in solution viscosity are a result of higher average protein cluster size in solution 11,16. For monoclonal antibody solutions, increases in viscosity are particularly problematic at high concentrations (>100 mg/ml), and many pharmaceutical formulation approaches have been developed to mitigate high solution viscosities17,18.

LLPS also plays a critical role in compartmentalization within cells2,19. Concentrations of proteins in the cytoplasm may reach 350 mg/mL 20. Accurate characterization of protein phase separated components in cells is performed primarily using imaging approaches. The use of many conventional biophysical tools for structurally characterizing proteins (e.g., CD, temperature melts, UV-Vis absorbance, etc.) is severely limited at such high protein concentrations. As LLPS is a thermodynamically reversible phenomenon, dilution of phases to lower protein concentrations results in solutions with equivalent physical properties (a single phase). Therefore, physical characterization of LLPS cannot be performed by diluting the protein-rich and protein-poor phases to concentrations suitable for experimental characterization.

Here, we use a variety of spectroscopic techniques capable of characterizing protein phase separated components without dilution, thereby preserving integrity of their actual physical state and allowing the study of protein conformational in each phase. Protein structural differences between the protein-rich and protein-poor phases were measured using Raman, fluorescence and FTIR spectroscopy. Several spectroscopic observables undergo consistent shifts for all three antibody proteins in the individual phases. FTIR also was used to investigate the dynamics of water in the protein rich and protein poor phases. Finally, we compared the aggregation rates for the three antibodies in the protein-rich and protein-poor phases and observed greater aggregation rates in the protein-poor phase for two of the antibodies and equivalent aggregation rates in two phases for the third antibody.

Materials and Methods

Materials and sample preparation

One IgG4 (mAb Z) and two IgG1 (mAbs H and I) monoclonal antibody stock solutions were provided by MedImmune LLC. These proteins were extensively dialyzed against buffers (Table 1) in which they could undergo LLPS at room temperature (room temperature, RT, ca. 22 °C) using dialysis cassettes (Slide-A-Lyzer™, 10 kDa MW cutoff, ThermoFisher Scientific, Waltham, MA). The individual phases could not be isolated directly from the dialysis cassettes due to high viscosity. The two resultant phases of each mAb were first mixed inside dialysis cassettes, and then the mixture was transferred to a Falcon tube and kept still overnight at RT until two phases again formed with a clear boundary. Samples of each phase were obtained by decanting the protein-poor phase into a separate Falcon tube. The pH and protein concentrations of the two phases (protein-poor and protein-rich) were measured using a pH meter and a Nanodrop spectrometer (Thermo Scientific), respectively (Table 1). The two phases of each mAb were later subjected to biophysical and analytical analyses.

Table 1.

The mAb concentrations and buffer conditions used throughout this study. Measured concentrations and pH for each phase.

Protein sample (isotype) Dialysis buffer Phase Concentration (mg/mL) pH

mAb Z (IgG 4) 20 mM histidine, pH 6.8 Protein-poor 35.0 ± 0.1 6.80 ± 0.10
Protein-rich 223.4 ± 1.3 6.81 ± 0.02
mAb H (IgG 1) 20 mM sodium phosphate, pH 7.5 Protein-poor 8.9 ± 0.1 7.51 ± 0.04
Protein-rich 245.7 ± 0.4 7.47 ± 0.02
mAb I (IgG 1) 5 mM sodium phosphate, pH 8.0 Protein-poor 19.9 ± 0.1 8.00 ± 0.01
Protein-rich 207.6 ± 0.6 8.04 ± 0.01

To investigate the temperature effect on the LLPS of these mAbs, freshly dialyzed samples were stored in Eppendorf tubes at 4, 15, 22 and 25 °C. Concentrations of the newly formed protein-poor phases were measured every 24 h until they remained unchanged, indicative of thermodynamic equilibrium. Upon equilibrium, the concentration of each phase was measured using a Nanodrop spectrometer (Thermo Scientific) to construct liquid-liquid coexistence curves.

The liquid-liquid coexistence curves derived from mean field theory were fit as described to: 21

|(CcC)/Cc|=A[(TcT)/Tc]β

Where Cc is the critical protein concentration, Tc is the critical temperature in degrees Kelvin, A parameterizes the width of the coexistence curve, and β is the Ising exponent (0.325). The standard error of the mean on the fitted parameters was calculated as described.22

Quantification of spectroscopic peaks

Spectroscopic peaks from Raman, fluorescence and FTIR were quantified using a variety of methods. The first three central moments of a peak were used to quantify peaks.

The zeroth moment is equivalent to the total peak area or total intensity:

Itot=I(λ)dλ

The first moment is equivalent to the peak mean over a given wavelength range:

λμ=λI(λ)dλI(λ)dλ

The second central moment is the variance of the peak:

[λσ]2=(λλμ)2I(λ)dλI(λ)dλ

The full width at half peak max (FWHM) for a Gaussian peak shape is the peak’s standard deviation λσ multiplied by a factor:

FWHM=2.355λσ

The peak maximum λmax was also used by interpolating between points using a cubic beta spline interpolation to 0.01 cm−1 for determining the wavelength maximum of the amide I band by FTIR. The ratio of peak intensities at two wavelengths (x and y) is designated as Ix/y.

Raman Spectroscopy

Raman spectra of protein samples were collected using a Zetasizer Helix (Malvern Instruments, Columbia, MD) equipped with a 785-nm laser (~280 mW). Protein samples (20 μL) were loaded into a metal micro-cuvette (Malvern) assembled with two quartz windows. For protein concentrations and buffers for each mAb, see Table 1. Samples were measured using 10 acquisitions of 20 sec each. Raw spectra were processed and analyzed using the Zetasizer Helix Analyze software (Malvern Instruments). Raman spectra were first buffer subtracted and normalized according to the phenylalanine peak (1003 cm−1). The amide I band first moment and FWHM were calculated from 1630 – 1690 cm−1.

FTIR Spectroscopy

FTIR analysis was performed using a Tensor-27 FTIR spectrometer (Bruker, Billerica, MA). For protein concentrations and buffers for each mAb, see Table 1. Protein samples were measured from 800 – 4000 cm−1 with a resolution of 4 cm−1. A total of 256 scans were performed for both samples and buffers at 25 °C. For the thermal melting experiments, 64 scans were obtained, and samples were scanned from 25 to 90 °C using an increment of 2.5 °C/step and an equilibration time sufficient to reach equilibrium (2 min) at each step. Buffer subtraction, water vapor and CO2 compensation, baseline correction, and normalization of the amide I band (1700 – 1600 cm−1) were sequentially performed on the raw FTIR spectra using the OPUS V6.5 software (Bruker, Billerica, MA). Second derivative FTIR spectra were generated using a Savitzky–Golay filter with a window size of 9.

The bend + liberation FTIR band of water in these protein samples and their buffers were measured by using air as the blank. Samples were scanned from 800 to 4000 cm−1 using a resolution of 2 cm−1. A total of 256 scans were performed. λμ was calculated between 1900 and 2300 cm−1 for the water band.

Intrinsic Fluorescence Spectroscopy

Intrinsic fluorescence spectra were obtained using a fluorescence plate reader as previously described 23,24. Samples were loaded into a 384-well plate and silicone oil was layered to avoid sample evaporation during thermal ramps. The plate was centrifuged at 2,200 × g for 1 min to remove air bubbles. For protein concentrations and buffers for each mAb, see Table 1. Samples were excited at 295 nm (>95% tryptophan emission) using an acquisition time of 100 ms, and emission light from 300 to 450 nm was collected. Temperature was ramped from 10 to 100 °C with an increment of 1 °C per step and an equilibration time of 2 min at each step. The mean fluorescence emission wavelength, λμ, was calculated between 300 and 400 nm. The value of λμ provides a better signal-to-noise ratio than the actual peak position (i.e. maximal point), although it is often 5 to 10 nm higher than the latter due to the asymmetry of protein intrinsic fluorescence spectra. The integral of fluorescence spectra between 300 and 400 nm was taken as total fluorescence intensity (Itot). Both parameters (λμ and Itot) were plotted as a function of temperature to construct the melting curves. Their first derivative curves were generated using a Savizky-Golay smoothing function with a polynomial order of 2 and window size of 11 built by employing Origin software (OriginLab Corporation, Northampton, MA). The local maximum was taken as the melting temperature.

Differential Scanning Calorimetry (DSC)

The overall conformational stability of the proteins was investigated using a VP-Capillary micro-calorimeter (Malvern, UK). The buffers for each mAb are listed in Table 1. The three mAb samples (at 1 mg/mL) in their corresponding dialysis buffers were loaded into a 96-well plate kept in the 5 °C sample chamber. Samples was scanned from 10 to 100 °C with a ramp rate of 2 °C per min. Samples were pre-equilibrated for 15 min prior to each scan. Raw DSC thermograms were buffer subtracted, normalized for concentration, and baseline corrected using the DSC Origin software (OriginLab Corporation).

Static light scattering

Static light scattering was obtained using a fluorescence spectrometer (Photon Technology International, Birmingham, NJ) equipped with a xenon arc lamp, excitation and emission monochromators, and a temperature controllable 4-position cuvette holder. For protein concentrations and buffers for each mAb, see Table 1. Light scattering (450 nm) was measured at a 90° detection angle with an acquisition time of 1 sec. Temperature was ramped from 25 to 100 °C using an increment of 1 °C per step and an equilibration time of 120 s.

Protein aggregation in Two Phases by Size Exclusion Chromatography (SEC)

Samples were held at 40 °C in the SEC autosampler for a two-week stability study. For protein concentrations and buffers for each mAb, see Table 1. The protein-poor and protein-rich phases were separated using positive displacement pipettes and placed in HPLC vials with insets (150 μL total volume). To prevent sample evaporation, a small amount (15 μL) of silicone oil was overlaid on the sample. Injections were performed without diluting the sample. For the protein-rich phase, 0.2 μL of sample was injected onto the column. For the protein-poor phase, the amount injected was different for each antibody but determined so that the total protein mass of the injection was equivalent to the protein-rich phase. This was done by multiplying 0.2 μL by the concentration ratio of the protein-rich and protein-poor phases. SEC was performed employing an HPLC instrument (Shimadzu, Kyoto, Japan) equipped with an autosampler, a column oven, and a UV detector. Samples were injected on a TSKgel SWXL guard column (6.0 mm × 40 mm) followed by a TSKgel G3000SWXL SEC column (7.8mm × 300 mm) at a flow rate of 0.7 mL/min for a total run time of 30 min. The mobile phase used was 0.2-M sodium phosphate, pH 6.8, and the column oven was set at 30 °C. Protein was detected at 280 nm. Total peak areas remained approximately constant throughout the stability study and protein recovery was approximately one hundred percent.

Results

LLPS of Three Different mAbs

The three mAb samples had been observed previously to undergo LLPS in low ionic strength solutions at pH values near their pI. We extensively dialyzed each of these three proteins against specific buffers in which LLPS occurred at room temperature as described in Table 1. High concentrations of proteins are known to exhibit the Gibbs-Donnan effect during dialysis due to preferential retention of oppositely charged ions by charged protein molecules resulting in an offset in solution conditions (pH and ion concentrations) between dialysis buffers and dialyzed samples 25. Therefore, we measured the pH of the dialyzed mAb samples to evaluate the magnitude of the Gibbs-Donnan effect during dialysis. Both protein-rich and protein-poor phases of each mAb were found to have pH values close to the pH of their dialysis buffers (±0.04), suggesting complete dialysis and an insignificant Gibbs-Donnan effect. This may be attributed to the net neutral charge of mAbs in their dialysis buffers with pH values close to their pI. It therefore seems reasonable to assume a similar buffer composition in the two phases for each of these three mAbs.

Temperature-Dependent LLPS of Three mAb Proteins

Temperature dictates solution entropy contributions and the magnitude of protein-protein interactions (PPIs), which are important driving forces for the formation of LLPS 26,27. We investigated phase separation behavior at several buffer and temperature conditions for the three mAbs, (Table 1). Liquid-liquid coexistence curves are shown in Figure 1, where solutions are homogeneous above the curve and phase separated below with distinct phase concentrations. LLPS was induced for each of the three mAbs at 4, 15, 22 and 25 °C, with a concave-down profile typical of most proteins 8.

Figure 1.

Figure 1.

Liquid-liquid coexistence curves of mAbs Z, H, and I reflecting their temperature dependent LLPS behavior. Error bars represent the standard deviation of three experiments. Lines represent best fit to mean field theory. See Table 1 for protein concentrations and solution conditions.

Out of these three proteins in their respective buffers, mAbs Z and I had low Tc while mAb H had the highest. Full parameters of the fit to mean-field theory equations are provided in Table 2. The three antibodies studied here have similar critical protein concentrations near 125 mg/ml. The smaller critical concentration of antibodies relative to other protein types (e.g. crystallin, lysozyme) is presumably a consequence of network formation at high concentrations 14,28,29. IgG antibodies are Y-shaped homodimers. If any region of the Fab (antigen binding fragment) is involved in protein self-association there are two binding sites per protein, thus the resulting dimer also has two unoccupied binding sites. This permits formation of long chains of antibody molecules that results in a small critical protein concentration.

Table 2.

Critical concentration, temperature and coexistence width for antibody phase coexistence curves. Cc and Tc are the respective concentration and temperature at the critical point of the best fit coexistence curves. A parameterizes the width of the coexistence curve.

mAb (isotype) Cc (mg/ml) Tc (°C) A

mAb Z (IgG 4) 133.8 ± 3.3 29.8 ± 1.9 2.14 ± 0.14
mAb H (IgG 1) 129.0 ± 1.6 85.9 ± 28.7 1.60 ± 0.16
mAb I (IgG 1) 118.9 ± 2.4 34.4 ± 2.8 2.16 ± 0.13

Structural Comparisons of the Three mAb Proteins in the Two Phases

Raman spectroscopy

The high protein concentrations (>200 mg/ml) in rich phases limited the biophysical techniques able to spectroscopically characterize protein secondary and tertiary structure. Raman spectroscopy was able to detect several dissimilarities between the protein-rich and -poor phases, although the overall normalized protein spectra were highly similar between phases (Supplementary Figure S1). The largest dissimilarities were observed in the spectral region of disulfide bands (Figure 2). IgG1 and IgG4 antibodies both have twelve intrachain and four interchain disulfide bonds 30. The SS stretching vibrations of disulfides in a protein provide specific information about the geometry of the dihedral angles. A disulfide bond has two carbon-sulfur (CS) bonds adjacent to the sulfur-sulfur (SS) bond. The CSSC dihedral angle is typically gauche (G), with angles near 90° or −90°. The CS bonds adjacent to the SS bond usually take a gauche (G) or trans (T) conformation. GGG conformers have a vibrational stretching frequency near 510 cm−1, and the trans conformations in CS bonds adds roughly 15 cm−1 of bond strain. Therefore, GGT conformations have a stretching frequency of approximately 525 cm−1 while TGT have a stretching frequency near 540 cm−1. These three peaks are present in the spectra shown in Figure 2 for all three antibodies.

Figure 2.

Figure 2.

Raman spectra for mab Z (top panels), mab H (middle panels) and mab I (bottom panels). Left column displaying the spectral region associated with SS stretching vibrations, middle column – tyrosine Fermi doublet, and right column the amide I band. See Table 1 for protein concentrations and solution conditions.

We compared the relative intensities of the TGT and GGG bands using the ratio of their intensities (I540/510). For the protein-rich phase of mAbs I and H, the ratio of GGG to TGT conformers is half that of the protein-poor phase (Table 3). For mAb Z, the GGG peak is nearly absent in the protein-poor phase resulting in a I540/510 ratio of 7.4. The changes in disulfide structures may be isotype specific since interchain disulfides in the Fab domain are distinct in IgG1 and IgG4. A peak at 490 cm−1 is present in all three protein-rich phases but missing in the spectra of mAb H and I protein-poor phases. A disulfide stretching vibration of 490 cm−1 can originate from dihedral angles of about 25° 31. The absence of this band in the protein-poor IgG1 phases again suggests a significantly altered state of disulfide bonds in the protein-rich phases of IgG1 molecules. It is expected that the bond energies would be much higher for disulfide bonds with SS angles at 25°. The hinge regions of antibodies have long been recognized for their intrinsic flexibility 32. Disulfide bonds covalently stabilize the Ig domains of antibodies. The additional strain in these regions suggests that domain positional adjustments may be necessary to form protein networks in solutions of antibodies.

Table 3.

Summary of Raman data.

mAb Phase SS ratio I540/510 Tyr ratio I855/830 Amide I mean Amide I FWHM

Z Protein-poor 7.37 ± 0.70 1.42 ± 0.02 1663.3 ± 0.1 34.4 ± 0.1
Z Protein-rich 2.41 ± 0.04 1.39 ± 0.02 1664.3 ± 0.1 33.7 ± 0.1
H Protein-poor 1.15 ± 0.31 1.77 ± 0.03 1663.2 ± 0.5 35.3 ± 0.3
H Protein-rich 2.12 ± 0.12 1.57 ± 0.01 1664.3 ± 0.1 33.7 ± 0.1
I Protein-poor 1.11 ± 0.22 1.28 ± 0.02 1664.3 ± 0.1 33.8 ± 0.1
I Protein-rich 2.03 ± 0.08 1.22 ± 0.02 1665.0 ± 0.1 32.6 ± 0.2

The tyrosine ratio of the Fermi doublet (I855/830) was used to quantify the hydrogen bond donor and acceptor preference of tyrosine in the two phases 33. For all three proteins the I855/830 was lower in the protein rich phase (Table 3) signifying a shift to tyrosine as a stronger hydrogen bond donor. This effect was most pronounced for mAb H with a shift from 1.77 to 1.57. The amide I band λmax at 1670 cm−1 indicates the antibodies are primarily beta sheet structure as expected. The amide I band moment exhibited a blue-shift of ~1 cm−1 for all three antibodies in the protein-rich phase compared to the protein-poor phase (Table 3). Band narrowing was also observed for the protein-rich phase as well; the FWHM of the amide I band decreased by 1 – 2 cm−1. Band narrowing of the amide I band could be the result of restriction of the number of accessible conformational states in the protein-rich phase.

FTIR Water Band

Antibodies in the protein-rich phase are thought to form extensive interaction networks in an aqueous environment. The volume fraction of mAbs at 250 mg/mL is 17.8% when calculated with a specific volume of 0.71 mL/g for proteins3. We hypothesized such a large change in volume fraction could affect the structure and dynamics of solvent water. The bend + liberation (BL) FTIR band of water red shifts when water’s hydrogen bonding network is weakened by heating or by a chaotrope 34, whereas a kosmotrope (a solute water structure maker) causes a blue shift. Protein-poor phases have equivalent band positions to their corresponding buffers (Figure 3). This suggests that water in low-ionic strength buffers and in the protein-poor phase has similar dynamic properties. A blue shift ( ~2 cm−1) of the BL band, however, was observed for the protein-rich phases compared to the poor phases, suggesting water in the protein-rich phase is more structured. These measurements suggest that while water is only weakly involved in protein phase separation, it may be important in the development of more sophisticated theories for LLPS. Water molecules in the two phases have different dynamic properties but reach thermodynamic equilibrium. It may be interesting to study water dynamics or simulate water diffusion across the phase boundaries to better understand the LLPS of protein solutions.

Figure 3.

Figure 3.

(A) FTIR bend-liberation combination bands of water in the two liquid phases of mAb samples and (B) their mean frequency between 1900 and 2300 wavenumbers cm−1. See Table 1 for protein concentrations and solution conditions.

FTIR Amide I region

The amide I FTIR band (between 1700 and 1600 cm−1) primarily derives from the stretching of the amide carbonyl group and therefore reflects the backbone pattern of hydrogen bonding of a polypeptide. Various secondary structure types show distinct amide I bands (e.g. alpha helix, intramolecular beta sheet, intermolecular beta sheet, beta turn, and random coil). The amide I band of the protein-rich phase red shifted by ~2 cm−1 relative to that of the more depleted phase for all three mAbs, indicative of subtle structural differences (Supplementary Figures S2 and S5). However, the amide I spectra of rich phases blue shifted so that they were identical to the deleted phase, once the rich phase was diluted to an equivalent concentration, indicating these changes were reversible.

Protein-rich phases showed significantly more intermolecular β sheet (1614 cm−1) than the protein-poor phases (Supplementary Figure S2). A red shift of the stronger main peak is accompanied by an increase in intermolecular beta sheet. The main peak maximum corresponds to primarily beta sheet structure. The beta sheet rich structures of antibodies participate in extensive intramolecular hydrogen bonds. This red shift in the amide I band is observed for all three proteins. The decrease in energy of the amide I band could arise from alterations in the beta sheet structure such as the sheet twist angle. The red shift could also result from small specific changes in sidechain positions. It is also possible there are some stabilizing non-bonded interactions such as n → π* interactions of carbonyl lone pairs and carbonyl π* orbitals at the protein-protein interface, which lower the energy of the amide I band 35. While we are uncertain about the exact nature of the red shift, it is consistent across the three antibodies studied as also seen with the blue shift observed in tryptophan fluorescence (see below). The amide I red-shift in FTIR and blue-shift in Raman spectra may be opposite in direction because of the different selection rules for the two methods (dipole moment and polarizability changes for FTIR and Raman, respectively).

The FTIR peak position of spectra (Supplementary Figure S5) was followed as a function of temperature (Figure S6). All three mAbs for both phases exhibited large red shifts upon aggregating to a peak at 1624 cm−1 (Figure S6). This peak has previously been assigned to amorphous aggregates of IgG1 antibodies following thermal unfolding and aggregation 36. Spectra prior to the melting temperature do not exhibit significant changes, indicating that the overall secondary structure differences between the proteins in the two phases are maintained until structural alterations and aggregation initiates.

Intrinsic Tryptophan Fluorescence

Tryptophan residues can be used as intrinsic fluorescent probes for protein structural changes since their fluorescence is sensitive to the polarity of their immediate environment. The protein-rich phases exhibited a blue-shifted λμ by 1 – 2 nm compared to their corresponding depleted phases at 25 °C, suggesting Trp residues are on average more buried in the protein-rich phase, which could be due to tertiary structural changes or their involvement in protein-protein interactions at interfaces. The observed blue-shifts are not artifacts due to inner filter effects since such an artifact usually results in a red-shifted emission 37. The twelve indole sidechains in antibodies which participate in nonbonded interactions with intrachain disulfides are designated Trp triads 38. Conformational changes within the Trp triad could be responsible for the observed blue shifts in Trp fluorescence that are accompanied by the changes in the disulfide dihedral angles seen with Raman spectroscopy.

To further probe the spectral shifts observed in the protein-rich phase, fluorescence was observed during thermal ramps from 25 to 90 °C (Figure 4AC). The blue-shift of λμ for the protein-rich phase persisted prior to the thermal transitions for all three mAbs. Typical red shifts of λμ were observed upon melting, although the protein-rich phases (mAb Z, mAb H, and the second transition of mAb I) exhibited stronger thermal transitions than their corresponding protein-poor phases. The post-transition baselines of the two phases nearly overlapped for each mAb. This indicates a disruption of the underlying difference between the protein-rich and protein-poor phases upon thermal unfolding.

Figure 4.

Figure 4.

Thermal stability of the proteins in the two liquid phases of the mAb samples measured by intrinsic tryptophan fluorescence spectroscopy. The mean fluorescence emission wavelength (λμ) of (A) mAb Z, (B) mAb H and (C) mAb I. See Table 1 for protein concentrations and solution conditions. (D) DSC thermograms for mAb Z, H and I (at 1 mg/ml).

Compared to λμ, total fluorescence intensity (Itot) correlates linearly with the degree of unfolding of a protein and therefore was used to generate Tm values (Supplementary Table S1)24. Similar to results in Figure 4, mAb Z showed only one transition, while the other two mAbs manifested two (Supplementary Figure S3AC). Their first derivative plots also confirm the number of transitions (Supplemental Figure S3D – F). First derivative plots of the protein-poor phase samples show thermal profiles consistent with DSC thermograms (Figure 4). Thermal stability of the proteins in the two liquid phases of the mAb samples was measured by intrinsic tryptophan fluorescence spectroscopy. The mean fluorescence emission wavelength (λμ) and DSC thermograms for mAb Z, H and I (at 1 mg/ml), illustrate the ability of fluorescence melts to resolve multiple thermal transitions. Hence, first derivative plots of thermal transitions were used to delineate domain specific transitions of both protein-rich and -poor phases, which could not be obtained directly using DSC due to a clogging issue at high temperature by gelled highly concentrated mAbs. The protein-rich phase of mAb Z had a lower Tm than its protein-poor phase by ~1.6 °C. The same trend was observed for mAb I, whose protein-rich phase is thermally less stable by ~ 3.9 °C (Tm1) and 1.2 °C (Tm2) than the protein-poor phase. Conversely, mAb H’s enriched phase showed higher Tm values than its protein-poor phase. Tm1 of its protein-rich and protein-poor phases are 64.4 ± 0.1 and 61.1 ± 0.1 °C, respectively, although their Tm2 showed only a small difference. The increase of Tm1 for mAb H may correspond the fact that the Tm1 of mAb H is below its critical temperature since mAb H has the strongest stabilizing PPIs of the three mAbs.

Differences seen in their thermal unfolding profiles can result from the relative thermal stability of each domain and/or the cooperativity of their thermal unfolding events. For instance, mAb I showed a small early transition and a large second one. The first transition is often attributed to the unfolding of the CH2 domain because of its relatively low stability 39, and the second one probably reflects a combination of the Fab and the CH3 domain. In contrast, the magnitude of mAb H’s first transition is comparable to the second one. This may suggest that mAb H has a relatively less-stable Fab, which contributes to the first transition alone or in combination with the CH2 domain. Since mAb H’s Tm1 increases in the protein-rich phase, this indicates that PPIs in protein-rich phase probably stabilize mAb H’s Fab. This was also supported by hydrogen-deuterium exchange studies on these proteins, which we have previously reported.40 MAb Z showed one broad transition consisting of several overlapping peaks, indicating that the domains of mAb Z thermally unfold in a more cooperative manner. The first thermal transition of mAb I’s two phases shared a similar magnitude (Figure 4C). As mentioned above, such a transition may represent the thermal unfolding of the CH2 domain. This indicates that thermal unfolding of the CH2 domain in mAb I did not significantly disrupt PPIs in its protein-rich phase. Therefore, the CH2 domain is probably not part of the PPI interface for mAb I. 40

Static Light Scattering

Static light scattering (SLS) was employed to monitor the aggregation of the protein in the two phases upon heating. Dynamic light scattering could not be used because of multiple scattering by the proteins in the high concentration solutions. SLS was measured with 450 nm light to avoid any absorption of lamp light by highly concentrated chromophores (Phe, Tyr, and Trp residues). As expected, protein-rich phases had higher SLS signals than corresponding protein-poor phases at 25 °C (Figure 5). Increasing temperature resulted in different trends in the pre-transition baselines of these three mAbs. For instance, both phases of mAb Z showed declining linear-like baselines, with the protein-rich phase decaying slightly faster. The protein-rich phase of mAb I also showed a decreasing baseline prior to its thermal transition. Its protein-poor phase, however, had a nearly flat baseline. In contrast, both mAb H’s two phases manifested flat pre-transition baselines, suggesting the interaction networks formed by mAb H are not as strongly affected by temperature. The differences seen in pre-transition baselines of the three mAbs can be attributable to the temperature dependence of their PPIs. MAb Z seems to show the highest temperature dependence and mAb H the lowest. This trend agrees with the temperature dependent phase behaviors illustrated in their coexistence curves (Figure 1).

Figure 5.

Figure 5.

Static light scattering vs temperature of the protein-rich and protein-poor phases of (A) mAb Z, (B) mAb H and (C) and mAb I. See Table 1 for protein concentrations and solution conditions.

Protein-poor phases showed greater increases in their SLS signals during thermal transitions than the protein-rich phase for the three mAbs. Gelation upon thermal unfolding in the protein-rich phase could be responsible for reduced transmittance and the smaller changes in SLS signals during thermal transitions. We also noticed that the protein-rich phase of mAb I had an early transition starting at ~60 °C, which was not seen in its depleted phase. We attribute this transition to the unfolding induced contacts of the CH2 domain, based on the DSC and fluorescence data discussed above. This suggests that the thermal unfolding of the CH2 domain of mAb I initiates aggregation in the protein-rich phase but not in the protein-poor phase, probably because of the much higher protein concentration in the protein-rich phase. This early transition was, however, not observed for mAbs Z and H.

Size Exclusion Chromatography

Size exclusion chromatograms (SEC) of the protein phases displayed a preference for more dimer in the protein-rich phase and more fragments in the protein-poor phase. There was 2.2, 2.1 and 2.0 % more dimer in the protein-rich phase than the protein-poor one for mAbs Z, H and I, respectively (Table 4). Fragments eluting at ~11.5 mL were observed for all three mAbs with the greatest difference seen in mAb H, where 1 % more fragments were observed in the protein-poor phase. The chromatograms for mAb H and mAb Z also exhibited small shoulders on either side of the main peak (Figure 6). We hypothesize that the protein-rich phase contains more aggregates because dimers can participate in the dynamic networks that drive phase separation while fragments only have one binding site and, therefore, could only exist at the edges of a network.

Table 4.

Initial percent of species observed at 25 °C after phase separation by SEC. The standard deviation on these measurements is ± 0.1%.

mAb Phase % Monomer % Dimer % Fragments

Z Protein-poor 99.4 0.5 0.1
Z Protein-rich 97.6 2.3 0.1
H Protein-poor 97.5 1.2 1.3
H Protein-rich 96.3 3.3 0.4
I Protein-poor 99.0 0.7 0.3
I Protein-rich 97.1 2.7 0.2

Figure 6.

Figure 6.

SEC chromatograms of the three mAbs in the poor and rich phases. Samples were diluted to 1 mg/ml before injection on the column. Absorbance was measured at 280 nm. The y-axis has been scaled to show the difference observed in the dimer and fragment peaks for each antibody. The inset of each panel shows the full scale of the chromatogram.

Aggregation Rates

The two phases were separated and injected onto the column without any dilution to measure the aggregation rates for each phase at 40 °C. Since the protein-rich and protein-poor phases are in equilibrium, the two phases have identical chemical potentials. Aggregation kinetics are a function of the chemical potential of protein in solution, therefore we hypothesized that protein aggregation rates would be equivalent in both phases. One caveat is the difference observed in HMWS when directly injecting the protein-rich phase (4.5 % HMWS for mAb I at time zero) in comparison to diluting the protein-rich phase prior to injection (2.7 % HMWS for mAb I at time zero). This difference (4.5% without dilution to 2.7% following dilution) reflects that some fraction of the aggregates observed by SEC without dilution of the protein-rich phase are reversible. This was observed for all three immunoglobulins and most drastically for mAb H. Therefore, the fraction of reversible and non-reversible aggregates at each time point could not be determined by SEC. Our analysis is simplified by interpreting HMWS as the sum of reversible and non-reversible aggregates.

All three antibodies exhibited higher or equivalent aggregation rates in the protein-poor compared to the protein-rich phase (Figure 7). MAb Z’s protein-rich phase remained approximately constant in the amount of high molecular weight species (HMWS) for the two-week period (Table 5). However, mAb Z’s protein-poor phase had an aggregation rate of 0.84 %/wk. MAb H had two small decreases in amount of HMWS. A slower equilibration process in the protein-rich network could alter the fraction of HMWS at these time scales. The protein-poor phase of mAb H aggregated faster than the rich-phase at a rate of 0.58 %/wk. MAb I aggregated at equivalent rates in the protein-rich (0.82 %/wk) and protein-poor phases (0.78 %/wk). Typically, aggregation rates increase with protein concentration since thermodynamic activity increases with concentration; however, all three mAbs in this study showed slower or equivalent rates in the protein-rich phase compared to the protein-poor phase.

Figure 7.

Figure 7.

Size fractions of mAbs for the protein-poor and protein-rich phases during a two-week accelerated stability study at 40 °C as measured by SEC.

Table 5.

Rates of changes in % per week of species observed with SEC chromatography at 40 °C.

mAb Phase HMWS (%/ wk) Monomer (%/wk) Clips (%/wk)

Z Protein-poor 0.84 ± 0.01 −1.16 ± 0.06 0.31 ± 0.06
Z Protein-rich −0.01 ± 0.04 −5.08 ± 0.28 5.09 ± 0.03
H Protein-poor 0.58 ± 0.02 −1.06 ± 0.03 0.48 ± 0.01
H Protein-rich −0.55 ± 0.06 −0.28 ± 0.06 0.83 ± 0.03
I Protein-poor 0.78 ± 0.02 −1.04 ± 0.03 0.27 ± 0.01
I Protein-rich 0.82 ± 0.03 −1.25 ± 0.05 0.33 ± 0.02

The clipping of antibody molecules happened more rapidly in the protein-rich phase than the protein-poor phase; especially for mAb Z, whose protein-rich phase had a fragmentation rate of 5 %/wk and protein-poor phase a rate of 0.3 %/wk. (Supplementary Figure S4). A major degradation pathway of antibodies at pH > 6 is fragmentation caused by base catalyzed beta elimination of Xaa-Cys residues in the hinge region 41. The hydrogen abstraction required for this mechanism may be more energetically favorable for the GGG conformation only present in the protein-rich phase of mAb Z, which may explain the major discrepancy in the observed rates. For mAb H, the clipping rate was only slightly faster, while for mAb I they were nearly equivalent. The differences in disulfide structures in IgG4 (mAb Z) and IgG1 (mAb H and I) may be important for the differences in the observed clipping rate constants. This explanation is only speculative, however, and differences could also be the result of partitioning of some protease impurities into the protein-rich phase of mAb Z.

Discussion

Phase separation of protein solutions has become a well-recognized phenomena. Studies of proteins such as lens crystallins 3,21,42, monoclonal immunoglobulins 4,43, and a number of other proteins have described the event in some detail in terms of temperature, pH and salt effects 12. The structural state of proteins in each phase, however, has seen few studies. It might at first be assumed the state of the protein in each phase would be the same since they are present at thermodynamic equilibrium. But in a protein/water/buffer or salt system, simple application of the phase rule is not possible. For example, proteins themselves exist in an equilibrium of many microstates which reflect the highly dynamic nature of protein structure. Thus, it is not a surprise that characterization of these three immunoglobulins suggest differences in the structure and behavior of the proteins in each phase. Measurements of the water bend-liberation band suggests an increase in order of the hydrogen-bonding network of a portion of the water in the more concentrated protein phase. This is also expected since the increase in protein concentration will be accompanied by more water of hydration, although self-association of the protein in the more concentrated phase would be expected to reduce hydrating water due to reduction in immunoglobulin surface area. Entrapment of water molecules in aggregates/transient protein clusters offers another possibility. The magnitude of this effect, however, seems to be too small to account for at low protein concentrations since no changes in band position were observed for water in the protein-lean phase..

Is it possible that the apparent conformational changes seen are an artifact of any of the measurement processes? While we cannot entirely exclude this possibility, we think this is unlikely. Spectral artifacts such as inner filter effects in fluorescence, surface adsorption (minimal contribution at the high concentrations employed) and electrostatic effects due to unequal distribution of buffering ions all appear to be negligible. We think two other explanations are more likely. Shifts in the distribution of protein conformational states due to water activity and protein concentration could produce the effects seen and are consistent with conformational changes observed in phase separation.44 More likely, however, is that the micro-aggregation states that appear to exist in concentrated immunoglobulin solutions could lead to structural alterations.13 The change seen in vibrational spectra and intrinsic fluorescence are consistent with this idea. The change seen in disulfide properties (dihedral angles) are consistent with the idea that changes in the dynamic nature of immunoglobulin may be involved in the spectral changes seen. In separate isotope exchange studies of two of the three proteins, we identified regions that form contact points for self-association (data not shown, published separately). Theoretical models of protein aggregation predict increases in aggregation rates with increased protein concentration (activity) since the reaction is at least bimolecular.45,46

Aggregation rates were slowed or equivalent for more highly concentrated solutions of all three phase separated antibodies herein. This atypical observation results from the equivalent chemical potential of the protein-rich and protein-poor phases 47. However, if chemical potential solely dictated aggregation rates one would anticipate equivalent aggregation rates in both phases for all three antibodies. However, only for mAb I were aggregation rates equivalent in both phases, while for mAb H and mAb Z the aggregation rate was slower in the protein-rich phase. If aggregation is diffusion rate-limited, the higher viscosity protein-rich phase could slow aggregation. An alternate explanation for why the aggregation rates are slower in the protein-rich phase is that to form an irreversible aggregate two monomers must diffuse close together to form an encounter complex which needs to rearrange to form a stable aggregate nucleus. If the protein’s native state is stabilized by self-interaction or the proteins are constrained by crowding, the kinetics of forming a stable aggregate nucleus would be expected to be slower.17

Phase separation provides a unique opportunity to study aggregation rates at different protein concentrations but equivalent chemical potentials. Phase separation is known to promote aggregation in cells for some proteins 19. We suspect, however, that in some cases the opposite may be true; the protein-rich droplets may serve to slow protein aggregation processes by slowing diffusion and/or stabilizing a protein’s native state. Conformational changes of proteins in the protein-rich phase may alter other cellular reaction kinetics similar to the increased rate of clipping for mAb Z.

Conclusion

Highly concentrated protein formulations are becoming commonplace in biotechnology products as an increasing number are developed for subcutaneous injection, where volumes are much smaller, compared to conventional intravenous formulations. Whereas intravenous formulations of biotech drugs are often <10mg/mL, the small injection volume of 1–1.5 mL available for subcutaneous administration necessitates much higher protein concentrations. In the case of mAbs, concentrations >150 mg/mL are common 48. Highly concentrated protein solutions conventionally are produced by lengthy ultrafiltration or tangential flow-filtration. Alternatively, proteins solutions are lyophilized and rehydrated in reduced volumes, which introduces significant development and process costs and can result in additional stability concerns during the lyophilization process. The utilization of LLPS phenomenon may offer a pathway to highly concentrated protein solutions with less bioprocessing steps 49.

We observed several consistent conformational changes in all three mAbs when comparing their protein-rich and protein-poor solutions of phase-separated antibodies. Tryptophan emission and vibrational spectra suggest conformational changes of proteins in dynamic clusters. Disulfide bond isomerized to higher energy conformations further confirmed changes in the dynamic nature for all three immunoglobulins in the protein-rich phase. These changes persisted at temperatures well above the coexistence curve but not above the Tm indicating that thermal unfolding disrupts the interactions governing phase separation. The hydrogen bond network of water was significantly more ordered in the protein-rich phase indicating a possible role for water in the dynamic clusters. Our measurements also highlight the utility of techniques which can characterize protein phases directly without the need for dilution. Protein aggregation was slowed in the protein-rich phase relative to the protein-poor phase for mAb’s H and Z while aggregation rates were equivalent for mAb I. These studies suggest that LLPS may be a useful phenomenon in producing stable highly concentrated biotech products. However, it also suggests that the stability and reversibility of rich LLPS phases should be carefully evaluated.

Supplementary Material

1

Acknowledgements

We would like to the thank the Vaccine Analytics and Formulation Center for access to equipment and resources, along with Drs. David Volkin and Sangeeta Joshi. Portions of this work has been published previously as part of the PhD dissertation for NL.40

Footnotes

Declaration of Interests

RE and CKK are employees of AstraZeneca and may own stock or stock options. The remaining authors have no interests to disclose related to this work.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  • 1.Hyman AA, Weber CA, Julicher F. Liquid-liquid phase separation in biology. Annu Rev Cell Dev Biol 2014;30:39–58. [DOI] [PubMed] [Google Scholar]
  • 2.Brangwynne CP, Tompa P, Pappu RV. Polymer Physics of Intracellular Phase transitions. Nat Phys 2015;11(11):899–904. [Google Scholar]
  • 3.Schurtenberger P, Chamberlin RA, Thurston GM, Thomson JA, Benedek GB. Observation of critical phenomena in a protein-water solution. Phys Rev Lett 1989;63(19):2064–2067. [DOI] [PubMed] [Google Scholar]
  • 4.Wang Y, Lomakin A, Latypov RF, Laubach JP, Hideshima T, Richardson PG, Munshi NC, Anderson KC, Benedek GB. Phase transitions in human IgG solutions. J Chem Phys 2013;139(12):121904. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Dijkstra M, van Roij R, Evans R. Phase diagram of highly asymmetric binary hard-sphere mixtures. Phys Rev E Stat Phys Plasmas Fluids Relat Interdiscip Topics 1999;59(5 Pt B):5744–5771. [DOI] [PubMed] [Google Scholar]
  • 6.Kastelic M, Vlachy V. Theory for the Liquid-Liquid Phase Separation in Aqueous Antibody Solutions. J Phys Chem B 2018;122(21):5400–5408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Raut AS, Kalonia DS. Pharmaceutical Perspective on Opalescence and Liquid-Liquid Phase Separation in Protein Solutions. Mol Pharm 2016;13(5):1431–1444. [DOI] [PubMed] [Google Scholar]
  • 8.Raut AS, Kalonia DS. Liquid-Liquid Phase Separation in a Dual Variable Domain Immunoglobulin Protein Solution: Effect of Formulation Factors and Protein-Protein Interactions. Mol Pharm 2015;12(9):3261–3271. [DOI] [PubMed] [Google Scholar]
  • 9.Wang Y, Lomakin A, Latypov RF, Benedek GB. Phase separation in solutions of monoclonal antibodies and the effect of human serum albumin. Proc Natl Acad Sci U S A 2011;108(40):16606–16611. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Trilisky E, Gillespie R, Osslund TD, Vunnum S. Crystallization and liquid-liquid phase separation of monoclonal antibodies and fc-fusion proteins: screening results. Biotechnol Prog 2011;27(4):1054–1067. [DOI] [PubMed] [Google Scholar]
  • 11.von Bulow S, Siggel M, Linke M, Hummer G. Dynamic cluster formation determines viscosity and diffusion in dense protein solutions. Proc Natl Acad Sci U S A 2019;116(20):9843–9852. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Kastelic M, Kalyuzhnyi YV, Hribar-Lee B, Dill KA, Vlachy V. Protein aggregation in salt solutions. Proc Natl Acad Sci U S A 2015;112(21):6766–6770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Godfrin PD, Zarraga IE, Zarzar J, Porcar L, Falus P, Wagner NJ, Liu Y. Effect of Hierarchical Cluster Formation on the Viscosity of Concentrated Monoclonal Antibody Formulations Studied by Neutron Scattering. J Phys Chem B 2016;120(2):278–291. [DOI] [PubMed] [Google Scholar]
  • 14.Yearley EJ, Godfrin PD, Perevozchikova T, Zhang H, Falus P, Porcar L, Nagao M, Curtis JE, Gawande P, Taing R, Zarraga IE, Wagner NJ, Liu Y. Observation of small cluster formation in concentrated monoclonal antibody solutions and its implications to solution viscosity. Biophys J 2014;106(8):1763–1770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Porcar L, Falus P, Chen W-R, Faraone A, Fratini E, Hong K, Baglioni P, Liu Y. Formation of the Dynamic Clusters in Concentrated Lysozyme Protein Solutions. The Journal of Physical Chemistry Letters 2010;1(1):126–129. [Google Scholar]
  • 16.Kastelic M, Dill KA, Kalyuzhnyi YV, Vlachy V. Controlling the viscosities of antibody solutions through control of their binding sites. J Mol Liq 2018;270:234–242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Roberts CJ. Therapeutic protein aggregation: mechanisms, design, and control. Trends Biotechnol 2014;32(7):372–380. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Whitaker N, Xiong J, Pace SE, Kumar V, Middaugh CR, Joshi SB, Volkin DB. A Formulation Development Approach to Identify and Select Stable Ultra-High-Concentration Monoclonal Antibody Formulations With Reduced Viscosities. J Pharm Sci 2017;106(11):3230–3241. [DOI] [PubMed] [Google Scholar]
  • 19.Wegmann S, Eftekharzadeh B, Tepper K, Zoltowska KM, Bennett RE, Dujardin S, Laskowski PR, MacKenzie D, Kamath T, Commins C, Vanderburg C, Roe AD, Fan Z, Molliex AM, Hernandez-Vega A, Muller D, Hyman AA, Mandelkow E, Taylor JP, Hyman BT. Tau protein liquid-liquid phase separation can initiate tau aggregation. Embo j 2018;37(7). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Walter H Consequences of phase separation in cytoplasm. Int Rev Cytol 2000;192:331–343. [DOI] [PubMed] [Google Scholar]
  • 21.Broide ML, Berland CR, Pande J, Ogun OO, Benedek GB. Binary-liquid phase separation of lens protein solutions. Proceedings of the National Academy of Sciences of the United States of America 1991;88(13):5660–5664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Newville M, Stensitzki T, Allen DB, Rawlik M, Ingargiola A, Nelson A. 2016. LMFIT: Non-linear least-square minimization and cruve-fitting for python. ed.: 10.5281/zenodo.598352. [DOI]
  • 23.Wei Y, Xiong J, Larson NR, Iyer V, Sanyal G, Joshi SB, Volkin DB, Middaugh CR. Effect of 2 Emulsion-Based Adjuvants on the Structure and Thermal Stability of Staphylococcus aureus Alpha-Toxin. J Pharm Sci 2018;107(9):2325–2334. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Wei Y, Larson NR, Angalakurthi SK, Russell Middaugh C. Improved Fluorescence Methods for High-Throughput Protein Formulation Screening. SLAS Technol 2018;23(6):516–528. [DOI] [PubMed] [Google Scholar]
  • 25.Stoner MR, Fischer N, Nixon L, Buckel S, Benke M, Austin F, Randolph TW, Kendrick BS. Protein-solute interactions affect the outcome of ultrafiltration/diafiltration operations. J Pharm Sci 2004;93(9):2332–2342. [DOI] [PubMed] [Google Scholar]
  • 26.Mason BD, Zhang-van Enk J, Zhang L, Remmele RL, Jr., Zhang J. Liquid-liquid phase separation of a monoclonal antibody and nonmonotonic influence of Hofmeister anions. Biophys J 2010;99(11):3792–3800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Wu G, Wang S, Tian Z, Zhang N, Sheng H, Dai W, Qian F. Elucidating the weak protein-protein interaction mechanisms behind the liquid-liquid phase separation of a mAb solution by different types of additives. Eur J Pharm Biopharm 2017;120:1–8. [DOI] [PubMed] [Google Scholar]
  • 28.Sun G, Wang Y, Lomakin A, Benedek GB, Stanley HE, Xu L, Buldyrev SV. The phase behavior study of human antibody solution using multi-scale modeling. J Chem Phys 2016;145(19):194901. [DOI] [PubMed] [Google Scholar]
  • 29.Wang Y, Lomakin A, Latypov RF, Benedek GBJPotNAoS. Phase separation in solutions of monoclonal antibodies and the effect of human serum albumin. 2011;108(40):16606–16611. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Liu H, May K. Disulfide bond structures of IgG molecules: structural variations, chemical modifications and possible impacts to stability and biological function. MAbs 2012;4(1):17–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Van Wart HE, Lewis A, Scheraga HA, Saeva FD. Disulfide bond dihedral angles from Raman spectroscopy. Proc Natl Acad Sci U S A 1973;70(9):2619–2623. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Clark NJ, Zhang H, Krueger S, Lee HJ, Ketchem RR, Kerwin B, Kanapuram SR, Treuheit MJ, McAuley A, Curtis JE. Small-angle neutron scattering study of a monoclonal antibody using free-energy constraints. J Phys Chem B 2013;117(45):14029–14038. [DOI] [PubMed] [Google Scholar]
  • 33.Wen ZQ. Raman spectroscopy of protein pharmaceuticals. J Pharm Sci 2007;96(11):2861–2878. [DOI] [PubMed] [Google Scholar]
  • 34.Verma PK, Kundu A, Puretz MS, Dhoonmoon C, Chegwidden OS, Londergan CH, Cho M. The Bend+Libration Combination Band Is an Intrinsic, Collective, and Strongly Solute-Dependent Reporter on the Hydrogen Bonding Network of Liquid Water. J Phys Chem B 2018;122(9):2587–2599. [DOI] [PubMed] [Google Scholar]
  • 35.Newberry RW, Raines RT. The n→π* Interaction. Acc Chem Res 2017;50(8):1838–1846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Boulet-Audet M, Byrne B, Kazarian SG. High-throughput thermal stability analysis of a monoclonal antibody by attenuated total reflection FT-IR spectroscopic imaging. Anal Chem 2014;86(19):9786–9793. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Lakowicz JR. 2006. Principles of Fluorescence Spectroscopy. ed., New York: Springer. [Google Scholar]
  • 38.Ioerger TR, Du C, Linthicum DS. Conservation of cys-cys trp structural triads and their geometry in the protein domains of immunoglobulin superfamily members. Mol Immunol 1999;36(6):373–386. [DOI] [PubMed] [Google Scholar]
  • 39.Ionescu RM, Vlasak J, Price C, Kirchmeier M. Contribution of Variable Domains to the Stability of Humanized IgG1 Monoclonal Antibodies. Journal of Pharmaceutical Sciences 2008;97(4):1414–1426. [DOI] [PubMed] [Google Scholar]
  • 40.Larson NR. 2020. Development and Application of Spectroscopic Techniques for Characterizing Turbid Solutions of Biopharmaceuticals. PhD Dissertation: University of Kansas, Lawrence, KS. p 109–159. https://kuscholarworks.ku.edu/handle/1808/31620 [Google Scholar]
  • 41.Vlasak J, Ionescu R. Fragmentation of monoclonal antibodies. MAbs 2011;3(3):253–263. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Liu C, Asherie N, Lomakin A, Pande J, Ogun O, Benedek GB. Phase separation in aqueous solutions of lens gamma-crystallins: special role of gamma s. Proc Natl Acad Sci U S A 1996;93(1):377–382. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Kheddo P, Bramham JE, Dearman RJ, Uddin S, van der Walle CF, Golovanov AP. Investigating Liquid-Liquid Phase Separation of a Monoclonal Antibody Using Solution-State NMR Spectroscopy: Effect of Arg·Glu and Arg·HCl. Mol Pharm 2017;14(8):2852–2860. [DOI] [PubMed] [Google Scholar]
  • 44.Majumdar A, Dogra P, Maity S, Mukhopadhyay S. Liquid–Liquid Phase Separation Is Driven by Large-Scale Conformational Unwinding and Fluctuations of Intrinsically Disordered Protein Molecules. The Journal of Physical Chemistry Letters 2019;10(14):3929–3936. [DOI] [PubMed] [Google Scholar]
  • 45.Li Y, Roberts CJ. Lumry-Eyring nucleated-polymerization model of protein aggregation kinetics. 2. Competing growth via condensation and chain polymerization. J Phys Chem B 2009;113(19):7020–7032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Pellarin R, Caflisch A. Interpreting the aggregation kinetics of amyloid peptides. J Mol Biol 2006;360(4):882–892. [DOI] [PubMed] [Google Scholar]
  • 47.Ghosh R, Calero-Rubio C, Saluja A, Roberts CJ. Relating Protein-Protein Interactions and Aggregation Rates From Low to High Concentrations. J Pharm Sci 2016;105(3):1086–1096. [DOI] [PubMed] [Google Scholar]
  • 48.Turner MR, Balu-Iyer SV. Challenges and Opportunities for the Subcutaneous Delivery of Therapeutic Proteins. J Pharm Sci 2018;107(5):1247–1260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Bramham JE, Davies SA, Podmore A, Golovanov AP. Stability of a high-concentration monoclonal antibody solution produced by liquid-liquid phase separation. MAbs 2021;13(1):1940666. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1

RESOURCES