A Low-Volume Screening Tool to Optimize Colloidal Stability for Highly Concentrated Protein Formulations

Abstract

This work presents a screening tool for the optimization of colloidal stability for highly concentrated protein formulations (HCF). One main challenge in the formulation development of HCF is physical degradation, in particular aggregation. Because aggregation is triggered by protein-protein interactions (PPI), colloidal stability can be optimized by determination and optimization of PPI during formulation development. An innovative, fast, and protein-saving method based on static light scattering is presented to determine PPI for different highly concentrated formulations. Thus, the procedure facilitates the development of a physically stable protein formulation without the need for execution of time- and protein-consuming long-term stability studies on a multitude of formulations.

Introduction

HCF are frequently required for subcutaneous or intramuscular administration. Subcutaneous administration, for instance, is a well-established approach for the application of therapeutic proteins in autoimmune and inflammatory diseases, since self-administration at home improves compliance for chronic disease. For this reason, HCF become more and more important for novel therapeutic approaches in those therapeutic areas.^1,2 Besides solubility and administration issues associated with HCF, physical stability is one of the most important challenges during HCF development. This is due to the fact that aggregation is often a concentration-dependent process.^3,4 Predictive parameters—such as the second virial coefficient (B₂₂) measured by static light scattering, or the interaction parameter (k_D) measured by dynamic light scattering—are well known in the formulation development of low concentrated protein formulations (up to about 10 mg/mL).^5,6 Both parameters characterize bimolecular interactions between molecules or particles and therefore can be used to predict colloidal stability and solubility of protein formulations at lower concentrations.^5,7 There are multiple methods to determine B₂₂ values, for example, by self-interaction chromatography, size-exclusion high performance liquid chromatography with dual light scattering, and concentration detection, composition gradient static light scattering, or analytical ultracentrifugation.^6,8-10 B₂₂ has also been reported to predict aggregation propensity, solubility, and viscosity behavior at higher concentrations.^11-13 However, general validity of the reported results has been controversially discussed. In several cases, it has been shown that the prediction is not valid for highly concentrated formulations since aggregation at high concentrations can be triggered by another mode of action, compared to lower concentrations. As a general rule, all values measured at highly diluted solutions assume ideal thermodynamic behavior. Therefore, they clearly fail to reflect HCF conditions since they typically show non-ideal behavior.^7,14

Bajaj, et al. showed that B₂₂ values did not correlate with the long-term stability of a mAb and ovalbumin in solution.¹⁵ Up to now, B₂₂ values and PPI in respect of HCF have only been investigated dependent on different pH and ionic strength conditions.^14,16,17 This represents mainly electrostatic PPI, but Kumar, et al. stated that short-range hydrophobic interactions can outweigh long-range electrostatic forces and play an important role in aggregation mechanisms at high concentrations.¹⁸ To predict the behavior of HCF, it is therefore imperative to have a method determining all kinds of PPI and represent conditions at increased concentrations instead of only measuring in diluted solution. This has also been demonstrated for analytical techniques to avoid measuring of artifacts, particularly in terms of potential reversible aggregation.^3,19 An important prerequisite for the applicability during HCF development is high-throughput and minimal material requirements. An increased number of excipients have to be examined in comparison to lower concentrated solutions because of the specific challenges with HCF.^13,20 The need for a valid, predictive, and robust high-throughput and protein-saving method to determine PPI for HCF is obvious.

Theoretical Concept

This work describes a simple screening method based on static light scattering to characterize PPI for HCF. The procedure enables the screening of PPI in different formulations regarding pH and excipients, and thus, determine the formulation conditions with the highest colloidal stability. Figure 1 shows a possible mode of action in protein aggregation. Native or denaturized protein molecules diffuse in solution due to Brownian molecular motion. In absence of PPI in solution, all protein molecules are randomly distributed and the encounter of 2 protein molecules occurs solely on a stochastic basis. If attractive interactions are present, the probability of 2 or more bodies colliding is elevated and the formation of aggregates during, for example, storage is likely. Repulsive interactions, on the other hand, decrease the probability of molecule collisions, and thus, subsequent aggregate formation, because all protein molecules endeavor to have maximum distance from each other. This behavior can be investigated by static light scattering. Figure 2 shows the light scattering intensity at different protein concentrations in the absence and presence of PPI. 

Figure 1. Simplified scheme of colloidal instability in protein solutions.

Figure 2. Schematic of light scattering signals as a function of protein concentration. Red line: attractive interactions, blue line: no interactions, green line: repulsive interaction.

Assuming Rayleigh scattering, in very simple terms, the following relationship may be expressed (Eq. 1):

I_LS ~ c * d⁶                   Eq.1

where I_LS represents the light scattering intensity, c is the protein concentration, and d is the molecule diameter.

Without interactions between the protein molecules and a random distribution in solution, Eq. 1 may be rearranged (Eq. 2):

with d = const, I_LS ~ c            Eq.2

This relationship is not valid in the presence of PPI, due to the increased or decreased probability of dimerization and non-random distribution in solution. Thus, Eq. 2 must be expressed as follows:

with d ≠ const, I_LS ≁ c

This results in a concentration-dependent, non-proportional increase or decrease of the static light scattering signal indicating attractive or repulsive PPI. The measurement can be done at high protein concentrations (eg, 100 mg/mL), but also at lower concentrations (eg, 20 mg/mL), if just a small amount of material is available. The suitable concentration range is dependent on protein properties such as size or interaction potential, and has to be evaluated once. PPI becomes stronger with increasing concentrations. Therefore, increased screening concentrations facilitate a more distinct discrimination between different formulations and also a better estimation of strengths of PPI. For this reason, screening at higher concentrations is recommended.

Method Development

All measurements were done in triplicate on a micro-cuvette static light scattering device combined with a fluorescence spectrometer. As model systems, an Fc-fusion protein and a monoclonal antibody (mAb) were selected. At first, a suitable concentration range was investigated. Therefore the signal-to-noise ratio was determined for lower concentrations with different formulations containing salts, sugars, surfactants, and the 2 different proteins. For reliable sample detection, scattering intensity should be at least 500 counts/s which is equal to a signal-to-noise ratio of 2. This is equivalent to an antibody concentration of about 1.5 mg/mL. The upper limit is defined by the detection limit of the instrument. However, it is recommended to select at least an upper concentration at which PPI can be reliably detected in order to allow predictive extrapolation to HCF formulations. Figure 3 shows light scattering data for an Fc fusion protein at 2 pH values. The curves are comparable to the theoretical model introduced in Figure 1. To determine the nature and strength of PPI, 40 mg/mL seems sufficient, and was selected as the highest concentration for further investigations. The same applies for the mAb. For the formulation screening, the proteins were prepared in the required formulations, and their concentration was maintained constant at 40 mg/mL. Starting with the 40-mg/mL stock solutions, a dilution series was prepared with 2, 4, 6, 8, and 40 mg/mL. For reliable results, at least 3 different concentrations up to 10 mg/mL and one concentration in the upper concentration range (cf. section Data Processing) are needed. The light scattering intensity of all concentrations was determined at 473 nm and 25°C with a sample volume of only 9 μL. With this procedure, up to 24 formulations per day can be screened with only 2 mg protein/formulation. Note that the instrument is also able to measure the light scattering intensity at 266 nm, but for the selected model proteins, the signal at 473 nm was more reproducible and assures only Rayleigh scattering was measured (d ≤ λ/20). However, a suitable wavelength might be dependent on the protein properties and it may be recommended to individually confirm this parameter prior to further studies. Figure 3A shows the light scattering data of different protein concentrations for an Fc-fusion protein at 2 pH values. Nonlinearity of I_LS at higher concentrations is due to the presence of PPI. To extract valid results, statistical evaluations and further data processing are necessary.

Figure 3. Example of data processing procedure for an Fc-fusion protein at two different pH values. Red data: pH 7.0, green data: pH 5.5. 3A: Raw data in line with the theoretical model. 3B: Residual analysis with non-linearity ≥20 mg/mL and linearity up to 8 mg/mL (inlet). 3C: Ratio of measured I LS divided by I_LS from regression analysis to compare level of PPI in different formulations; values ≥1 indicate attractive PPI, values ≤1 indicate repulsive PPI. 4D: Relative percental deviation of I_LS in comparison to linearity and no PPI; black lines mark significant deviation from linearity and therefore presence of PPI.

Data Processing

Data exposition as shown in Figure 3A allows one to make a reliable decision whether attractive or repulsive interactions are present. Ranking of different formulations with PPI of the same nature is, however, difficult. Additional data processing is required to enable a final decision on the representativeness of results. For both proteins in all tested formulations, linearity of I_LS to concentrations up to 8 mg/mL could be proved by residual analysis, and a nonlinear increase or decrease in I_LS did not occur until concentrations ≥10 mg/mL (Figure 3B). This observation is used for further data evaluation, whereby a linear regression in the concentration range up to 8 mg/mL was performed and extrapolated to higher concentrations. Unfortunately, I_LS in the linear range was shown to be not reproducible between all formulations. Whereas linearity could be confirmed for all formulations, slopes of regression lines differ significantly between different formulations. Hence, it is advisable to conduct a dilution series for every single formulation that should be investigated using the data processing procedure mentioned in this section. After linear extrapolation to higher concentrations, the ratio of measured, actual I_LS and theoretical I_LS from regression analysis is calculated in terms of attractive PPI (Figure 3C). For a better comparability between the individual strengths of PPI, the inverted ratio has to be calculated for repulsive PPI leading to values ≥1 in both attractive and repulsive PPI cases. A ratio of 1 at higher concentrations means actual I_LS and theoretical I_LS from regression are equal and indicates no PPI. To define a significant level at which PPI are present, the standard deviation (SD) of all ratios at all concentrations up to 8 mg/ mL was calculated. This concentration range is proved to be linear and therefore to show no PPI. Three times SD was set as the limit for the presence of PPI (black lines in Figure 3C and 3B). Finally, the relative deviation from theoretical linearity (Δ%) is calculated at a defined concentration with negative values for repulsive and positive values for attractive PPI (Figure 3D). Reproducibility was tested for one formulation with 3 different preparations, dilution series, and experiments in triplicate, resulting in a total of 27 measurements. The overall SD of the method at 40 mg/mL for an Fc-fusion protein could be determined as 0.15 (ratio) or 15% (relative deviation).

Results and Discussion

In Figure 4, the relative percental deviation from linearity (Δ%) for different formulations of an Fc-fusion protein and a mAb are provided. The formulations are classified for increased colloidal stability from left to right. A clear discrimination between the 2 pH values is possible, with pH 5.5 suggesting repulsive and pH 7.0 suggesting attractive interactions for the Fc-fusion protein. For the mAb, pH 5.0 indicates repulsive interactions in comparison to pH 6.5 where data indicate no interactions. The Fc-fusion protein shows an experimentally determined pI value of 9.0, whereas the mAb has a pI of 7.0. It is therefore intuitive that at pH values near the pI attractive hydrophobic and no electrostatic repulsive PPI are dominant. In contrast, at pH values far from the pI, electrostatic repulsive PPI are expected. After addition of 100 mM sodium chloride, only for the Fc-fusion protein at pH 5.5 PPI could be determined. Note that the investigated PPI were much lower than those without salt. At pH 7 for the Fc-fusion protein, or in case of the mAb, neither attractive nor repulsive PPI could be determined after the addition of salt. It has been reported that salt “shields” the charge on protein surfaces and therefore minimizes electrostatic repulsion or attraction. Surprisingly, a similar effect was found for 8% trehalose. In the presence of trehalose, attractive and repulsive PPI decrease as well for both proteins, probably also because of shielding effects on the charged surfaces. So, all results measured are valid due to the expectations according to electrostatic charged surfaces and their shielding. But the prediction will be more difficult if hydrophobic or steric interactions are dominant or interactions are affected by addition of specific excipients.

Figure 4. Results for A: Fc-Fusion protein and B: Monoclonal antibody. Relative percental deviation from linearity of 4 different formulations. Black lines mark “no-interaction-zone.” Formulations arranged in order of colloidal stability from left to right. Repulsive PPI Δ% ≤0, attractive PPI Δ% ≥0.

Conclusion

This set of preliminary experiments suggested that the developed screening method delivers robust and valid results, and can be used for the optimization of colloidal stability for HCF. With the novel method, different formulation parameters and excipients (such as pH values, salt, or sugar) can be examined and optimized in terms of colloidal stability. This simple screening tool enables the prediction of relative colloidal stability even in complex protein formulations. It might therefore be a promising tool for future HCF formulation development challenges, avoiding long-term, time- and protein-consuming stability studies. In comparison to other predictive parameters used to date, the screening can be done directly in highly concentrated solution and might be used to close a wide gap within the formulation development of HCF.

References

1. Liu J, Nguyen MDH, Andya JD, Shire SJ. Reversible selfassociation increases the viscosity of a concentrated monoclonal antibody in aqueous solution. J Pharm Sci. 2005;94(9):1928-1940.
2. Harn N, Allan C, Oliver C, Middaugh CR. Highly concentrated monoclonal antibody solutions: Direct analysis of physical structure and thermal stability. J Pharm Sci. 2007;96(3):532-546.
3. Shire SJ, Shahrokh Z, Liu J. Challenges in the development of high protein concentration formulations. J Pharm Sci. 2004;93(6):1390-1402.
4. Frokjaer S, Otzen DE. Protein drug stability: a formulation challenge. Nat Rev Drug Discov. 2005;4(4):298-306.
5. Saluja A, Fesinmeyer RM, Hogan S, Brems DN, Gokarn YR. Diffusion and Sedimentation Interaction Parameters for Measuring the Second Virial Coefficient and Their Utility as Predictors of Protein Aggregation. Biophysi J. 2010;99(8):2657-2665.
6. Ruppert S, Sandler SI, Lenhoff AM. Correlation between the Osmotic Second Virial Coefficient and the Solubility of Proteins. Biotechnol Progr. 2001;17(1):182-187.
7. Valente JJ, Payne RW, Manning MC, Wilson WW, Henry CS. Colloidal Behavior of Proteins: effects of the Second Virial Coefficient on Solubility, Crystallization and Aggrrgation. Curr Pharm Biotechnol. 2005;6(6):427-436.
8. Lehermayr C, Mahler HC, Mäder K, Fischer S. Assessment of net charge and protein-protein interactions of different monoclonal antibodies. J Pharm Sci. 2011;100(7):2551-2562.
9. Tessier PM, Lenhoff AM, Sandler SI. Rapid Measurement of Protein Osmotic Second Virial Coefficients by Self-Interaction Chromatography. Biophys J. 2002;82(3):1620-1631.
10. Bajaj H, Sharma VK, Kalonia DS. Determination of Second Virial Coefficient of Proteins Using a Dual-Detector Cell for Simultaneous Measurement of Scattered Light Intensity and Concentration in SEC-HPLC. Biophys J. 2004;87(6):4048-4055.
11. Saito S, Hasegawa J, Kobayashi N, Kishi N, Uchiyama S, Fukui K. Behavior of Monoclonal Antibodies: Relation Between the Second Virial Coefficient (B 2) at Low Concentrations and Aggregation Propensity and Viscosity at High Concentrations. Pharm Res. 2012;29(2):397-410.
12. Neergaard MS, Kalonia DS, Parshad H, Nielsen AD, MØller EH, van de Weert M. Viscosity of high concentration protein formulations of monoclonal antibodies of the IgG1 and IgG4 subclass − Prediction of viscosity through protein−protein interaction measurements. Eur J Pharm Sci. 2013;49(3):400-410.
13. Connolly B, Petry C, Yadav S et al. Weak Interactions Govern the Viscosity of Concentrated Antibody Solutions: High-Throughput Analysis Using the Diffusion Interaction Parameter. Biophys J. 2012;103(1):69-78.
14. Scherer TM, Liu J, Shire SJ, Minton AP. Intermolecular Interactions of IgG1 Monoclonal Antibodies at High Concentrations Characterized by Light Scattering. J Phys Chem B. 2010;114(40):12948-12957.
15. Bajaj H, Sharma V, Badkar A, Zeng D, Nema S, Kalonia D. Protein Structural Conformation and Not Second Virial Coefficient Relates to Long-Term Irreversible Aggregation of a Monoclonal Antibody and Ovalbumin in Solution. Pharm Res. 2006;23(6):1382-1394.
16. Sahin E, Grillo AO, Perkins MD, Roberts CJ. Comparative effects of pH and ionic strength on protein-protein interactions, unfolding, and aggregation for IgG1 antibodies. J Pharm Sci. 2010;99(12):4830-4848.
17. Yadav S, Shire SJ, Kalonia DS. Viscosity behavior of high-concentration monoclonal antibody solutions: Correlation with interaction parameter and electroviscous effects. J Pharm Sci. 2012;101(3):998-1011.
18. Kumar V, Dixit N, Zhou L, Fraunhofer W. Impact of short range hydrophobic interactions and long range electrostatic forces on the aggregation kinetics of a monoclonal antibody and a dual-variable domain immunoglobulin at low and high concentrations. Intl J Pharm. 2011;421(1):82-93.
19. Matheus S, Friess W, Mahler HC. FTIR and nDSC as Analytical Tools for High-Concentration Protein Formulations. Pharm Res. 2006;23(6):1350-1363.
20. Aucamp JP, Cosme AM, Lye GJ, Dalby PA. High-throughput measurement of protein stability in microtiter plates. Biotechnol Bioeng. 2005;89(5):599-607