Part II: Enabling Freeze-Thaw Stability of PBS-Based Formulations of a Monoclonal Antibody

JUAN GARTNER/getty images

Part II: Long-Term Frozen Storage

Article submitted: Oct. 20, 2015.
Article accepted: Apr. 12, 2016.

Abstract

This study evaluates the long-term frozen storage stability of a monoclonal antibody in phosphate buffered saline (PBS)-based formulations as a continuation of a single freeze-thaw study reported by Mezhebovsky et al. in Part I of this series. Similar to the single freeze-thaw stress findings, polyol-containing formulations rendered the highest stabilization against aggregation or change in submicron particle formation during long-term frozen storage. For non-polyol-containing formulations stored for 24 months at -70 °C or -80 °C, the extent of degradation was essentially the same as that observed upon a single freeze-thaw stress, suggesting that degradation occurred primarily upon freezing and thawing rather than long-term storage. Frozen storage at -20 °C differed in that no degradation was observed after a single freeze-thaw stress while erratic increase in aggregation of varying levels occurred after an initial lag phase of two to three months, indicative of kinetic events. Low salt and/or low protein content exacerbated frozen storage instability at -20 °C, while high salt and slow freezing exacerbated instability at -80 °C. This study offers a strategy for stabilization of biotherapeutics for long-term frozen storage in PBS-based formulations.

Stability on frozen storage is important in biotherapeutics lifecycle and supply management (1, 2). Routine practice in bioprocessing involves storage of frozen bulk drug substance as well as storage of frozen reference standard (3, 4).

In addition, many protein reagents used in diagnostics and various assays in biotechnology industry are stored frozen and often subjected to multiple freeze-thaw cycles (5).

As a physiologically compatible solution, phosphate buffered saline (PBS) has been used in research and biotherapeutics development despite being notorious for freeze-thaw instability (6-8). Instability of proteins to freeze-thaw stress in PBS is primarily attributed to the steep temperature dependence of phosphate buffer pKa and the high propensity of sodium dibasic phosphate to crystallization (8); the latter can cause up to a few units drop in pH upon freezing (7). The detrimental effect of slow freezing and thawing on protein stability has also been extensively reported (9, 10). In Part I of this series, the authors presented the results of a single freeze-thaw-induced aggregation for 12 PBS-based formulations of a monoclonal antibody (mAb) to evaluate critical factors driving the protein aggregation in this vehicle (11). The results demonstrated that slow freezing and high salt were the most detrimental factors and that polysorbate-80 displayed mild protective effect; only addition of polyols completely abrogated aggregation of the mAb by a single freeze-thaw stress. In this article, the authors present a complementary study where the same PBS-based formulations of this mAb were frozen using the same experimental freezing protocols and subjected to long-term frozen storage of up to 24 months.

Materials and methods

Materials
A recombinant mAb manufactured in Chinese hamster ovary (CHO) cells at Morphotek was used in these studies. Study reagents and materials were described in Part I (11). Long-term storage was conducted in the following units:

• CoolCell cell freezing container (Biocision) for highly reproducible (1 °C/min) slow cooling; as well as plastic and cardboard cryoboxes for flash freezing in nitrogen vapors and in the designated freezers with subsequent frozen storage, respectively.

• Baker UF7556 -80 °C freezer with REES constant temperature monitoring

• Revco ULT 2586-9SI-A38 -70 °C freezer

• -20 °C walk-in room with REES constant temperature monitoring

• Forma Environmental Chamber, temperature controlled at 2-8 °C.

JMP 10.0.0 software package from 2010 SAS Institute was used for statistical data analysis.

Methods

Formulations. All formulations contained 10 mM Na phosphate, pH 7.2. The formulation matrix is listed in Table IA sorted by composition. Samples in 2-mL round bottom cryogenic vials were prepared as described in Part I (11). A few samples were frozen in 1.2-mL conical cryogenic tubes to determine the effect of container type. Duplicate samples were prepared for each time point. All samples were frozen and thawed as described previously (11).

Table IA: Formulation compositions.

Freezing procedures. The same freezing protocols as those used in Part I of this series (11) were applied. Samples of the 12 formulations listed in Table IA were frozen under conditions listed in Table IB in separate containers for each time point. Separate containers were for duplicate samples prepared for 2, 6, 9, 12, and 24 months of frozen storage. All samples were put on frozen storage on the same day. For each time point, containers were pulled out of the respective freezers and thawed at 2-8 °C on the same day, except for samples subjected to flash freezing in nitrogen vapors that were thawed at ambient temperature (18-25 °C).

Table IB: Freezing protocols.

Sample testing. Product quality was analyzed at t=0 and after 2, 6, 9, 12, and 24 months of frozen storage. Samples were analyzed in duplicates within two weeks of preparation as described previously (11). All samples were analyzed by size exclusion-high-performance liquid chromatography (SE-HPLC) and dynamic light scattering (DLS); samples at 24 months of frozen storage were not analyzed by DLS due to technical issues with the instrument. The data were averaged for each formulation, protocol, and time point.

Data analyses. The data were analyzed for significant trends using statistical software package JMP 11 (2013 SAS Institute). Because sample volume and container type did not significantly affect degradation on freeze thaw stress, these parameters were excluded from further statistical analysis. Analysis of covariance (ANCOVA), an extension of analysis of variance (ANOVA), was chosen for the initial analysis of the entire data array to identify parameters that have significantly different impact on aggregation than others. ANCOVA’s advantage lies in the ability to remove the covariates’ effect in the analysis of dependent variable (12). Partitioning analysis resulted in two splits: in the first split, Protocol P6 landed into a separate group from other freeze-thaw protocols as significantly different; in the second split, tonicifier was separated as the most significant factor.

For further analysis of the rank order of the impact of the remaining parameters on aggregation, protocol P6 was excluded. The screening fit model was applied using the aggregate level as the model output (or response), and protein concentration, salt concentration, presence and type of polyols, freezing protocols (P1, P2, P3, P4, P5, P7), and presence of surfactant as model inputs. This model identified the significant factors affecting aggregation under all conditions, except P6. The significant scaled parameter estimates were identified when probability was > t (t ratio). Scaled estimate examines parameter estimates in a scale-invariant fashion to better understand and compare effect sizes. The factors are scaled to have a mean of zero and a range of two.

Results

Aggregation profile
The size-exclusion HPLC chromatogram of the mAb typically displays three high molecular weight (HMW) species, which were added together (HMW1-3) to assess the impact of different parameters on long term frozen storage (11).

Figure 1: % aggregates in phosphate buffered saline (PBS)-based formulations of a monoclonal antibody (mAb) during long-term frozen storage (baseline level ~2%). X-axis represents formulations F1 through F12, Y-axis represents freeze-thaw protocols P1 through P6.

Figure 1: % aggregates in phosphate buffered saline (PBS)-based formulations of a monoclonal antibody (mAb) during long-term frozen storage (baseline level ~2%). X-axis represents formulations F1 through F12, Y-axis represents freeze-thaw protocols P1 through P6.

Figure 1 displays the aggregate levels in 12 tested formulations (Table IA) using six freezing protocols (Table IB) for up to 24 months of frozen storage. Because it was not possible to fit the data to any kinetic model to determine a rate constant, a global evaluation of the % aggregates at each time point was made.

The non-polyol formulations showed an initial large increase in aggregates at two months of frozen storage for all freezing protocols, except P6, followed by no subsequent significant trend. Protocol P6 was different from all the other protocols in that it showed erratic fluctuations in aggregate levels at different time points after a lag period of approximately two to three months (Figure 1 and Figure 2). Frozen storage under P6 condition resulted in the highest levels of aggregates compared to all other conditions.

For polyol-containing formulations, aggregate levels remained essentially constant throughout the entire 24 months of frozen storage (Figure 1, F9-F12, the last four columns at each time point) under all freezing conditions. This stabilization against aggregation during long-term storage was similar to the results for a single freeze-thaw event (11).

ANCOVA partition analysis of all 990 data points in JMP software revealed P6 to have approximately 10-fold different LogWorth than all the other six protocols (Table II). The LogWorth is a (-log10[p-value]). Large effects of very small p-values are transferred to large LogWorths, while non-significant p-values have low LogWorths. A LogWorth of zero corresponds to a non-significant p-value of 1. Any LogWorth above 2 corresponds to a p-value below 0.01 (13). An additional level of partitioning analysis showed that for all protocols within the two groups from the first split, polyol-containing formulations are significantly different from non-polyol containing formulations at any time point or composition of other components or protocols used (Table II).

Table II: Analysis of covariance (ANCOVA)-based partitioning analysis of the effect of freezing protocols on aggregation of an antibody during long-term frozen storage. Std Dev is standard deviation.

Table II: Analysis of covariance (ANCOVA)-based partitioning analysis of the effect of freezing protocols on aggregation of an antibody during long-term frozen storage. Std Dev is standard deviation.

With P6 excluded, screening analysis in JMP identified the significant factors 
affecting stability of the mAb during frozen storage at -80 °C (Table III). The most significant detrimental factors were high NaCl concentration, lack of tonicifier, and slow freezing (P2). The significant stabilizing effects were addition of polyols and high protein concentration. Polysorbate-80 showed no significant impact.

DLS results

CLICK FIGURE TO ENLARGE Figure 3: Time course of % polydispersity for frozen storage of a monoclonal antibody (mAb).

Although DLS does not quantify submicron particles (intensity of scattered light increases non-linearly with increase in the scattering particle size), % polydispersity (% Pd) in the single exponential decay approximation of the autocorrelation function is a surrogate because it reflects the degree of homogeneity of the single pool of particles. A higher % Pd corresponds to a higher heterogeneity of the corresponding particle pool. The observed time course of % Pd as a function of frozen storage time is presented in Figure 3. At baseline, non-polyol containing formulations displayed lower % Pd than polyol-containing formulations; however, % Pd increased on long-term frozen storage of non-polyol formulations, while it remained constant for polyol-containing formulations. Additionally, the % Pd values trended lower for high salt and high protein polyol-containing formulations than for low salt and low protein-containing formulations (see Figure 3). At baseline, % Pd was lower for non-polyol formulations (approximately 3-7%) compared to 20% for sucrose-containing formulations (20%) and 12% for sorbitol-containing formulation with low protein and low salt concentrations. Non-polyol formulations, however, displayed progressive increase in the heterogeneity of the sub-visible particle pool during frozen storage at -20 ºC (% Pd doubled to approximately 6-16% by 12 months), while % Pd remained constant for frozen polyol-containing formulations. The sorbitol-containing formulation with high-protein and high-salt concentrations appeared optimal in terms of % Pd as it had as low a polydispersity (5%) as the non-polyol formulations at baseline, and remained constant during 24 months of frozen storage.

Discussion

The current study of time course of frozen storage is complementary to the stability study following a single freeze-thaw event (11). Long-term frozen storage at -70 °C and -80 °C resulted in no additional aggregation after the first time point tested, suggesting that degradation occurs most likely during the freezing and thawing event rather than long-term storage. High NaCl content and slow freezing rate showed major adverse impact on aggregation during long-term frozen storage as also previously observed for a single freeze-thaw event. By contrast, low NaCl, low protein-containing formulations showed the highest level of aggregation at -20 °C with erratic behavior over time. The data suggest different degradation mechanisms for these two different types of freezing protocols. Addition of polyols (5% sucrose or 3% sorbitol) resulted in effective stabilization of the mAb against aggregation for long-term frozen storage under all tested conditions, likely due to preferential exclusion mechanism (14). A higher protein concentration showed a stabilizing effect for all protocols, excluding P6 (Table III), similar to the trend observed after a single freeze-thaw stress (11). Polysorbate-80 rendered statistically insignificant protective effect for all protocols (Table III).

Table III: Screening for major factors affecting aggregation of a monoclonal antibody during longterm frozen storage (with protocol P6 excluded).

Table III: Screening for major factors affecting aggregation of a monoclonal antibody during longterm frozen storage (with protocol P6 excluded).

Non-polyol formulations displayed progressive increase in the heterogeneity of the sub-visible particle pool, whereas polydispersity remained constant during long-term frozen storage in polyol-containing formulations. While 5% sucrose was optimal for preventing aggregation and stabilizing the level of sub-micron particles, the latter was higher than non-polyol formulations, pointing to the complex nature of factors that affect protein size distribution. The sorbitol-containing formulation with high-protein and high-salt concentrations appeared optimal in terms of polydispersity as it had as low a polydispersity as the non-polyol formulations at baseline, and remained constant over 12 months of frozen storage, event at -20 ºC. Note that the % Pd was higher at low-mAb and low-salt concentrations in both sucrose- and sorbitol-containing formulations. While there are concerns with immunogenicity of protein aggregates and particulates (15, 16), a safety limit for sub-micron particles has not been established. Sub-micron particle range of 5-15% Pd is within reasonable limits observed in pharmaceuticals. The authors’ unpublished data confirmed that sucrose at 1-5% and sorbitol in the range of 1-3% were similarly effective in preventing freeze-thaw induced degradation of the model mAb. Thus, the type and concentration of polyol, the mAb concentration, and other potential excipients are amenable to fine tuning for mitigating freeze-thaw instability of biotherapeutics in PBS.

Since all samples in the long-term frozen stability study are subject to one freeze-thaw cycle for analysis, stabilization against events occurring during freeze-thaw is a pre-requisite to develop a stable frozen protein formulation. The freezing process involves formation of crystalline ice and also crystals of those excipients that crystallize (e.g., phosphate and NaCl). Crystal formation has a lag time until “seeds” are formed, the rate of which is dependent on factors such as the presence of particles in the solution or primary container wall surface irregularities that could vary from one container type to another, as well as from one container to another. The rate of crystallization affects the size, and hence, the surface area of crystals onto which protein denaturation and subsequent aggregation would occur. Additionally, during freezing, a freeze-concentrate of protein and excipients forms that can also result in protein denaturation and aggregation. The randomness of aggregate levels in phosphate-buffered, NaCl-containing formulations is likely associated with the randomness of ice and excipient crystal formation and duration of freeze-concentration prior to transition to an immobilized protein in the solid. While surfactants such as polysorbate-80 often protect against protein denaturation at ice-crystal surfaces and liquid-ice interfaces, they do not suppress excipient crystallization (17, 18).

The study findings suggest that degradation of the mAb during freezing in polyol-free formulations at all freezing protocols (except P6) likely occurs predominantly in freeze concentrates but not on ice water or ice air interfaces (19, 20). Differential scanning calorimetry (DSC) analysis has revealed a glass transition of no higher than -74.1 °C in PBS (21); above this glass transition temperature, a concentrated amorphous solution of protein and other excipients is formed as a “freeze-concentrate” with adequate protein mobility to support aggregation. Thus, storage of a mAb in non-modified PBS at -70 °C and -80 °C may be thermodynamically challenging.

Figure 2:

Figure 2: Time course of aggregation at -20 °C.

DSC studies have also revealed that addition of polyols results in a significant increase in transition temperature. For instance, Tg* values changed from -73.6 °C to -49.2 °C and -58.7 °C for F9 and F10, respectively (21), likely due to the blending effect of miscible components (22). Tg* was determined by slow heating of the fast frozen sample to enhance the magnitude of glass transition. Although Tg* values are close to Tg’ values, they are higher than glass transition temperatures (21). The stabilizing effect of polyols is thus in part due to thermodynamically stabilized composition protecting against protein denaturation in the freeze-concentrate (23). Other stabilizing effect of polyols comes from suppression of excipient crystallization, and protection against protein denaturation at the crystal interface. Storage at -20 °C is significantly above glass transition temperature for all the tested formulations and close to the crystallization temperature of salt (21). Since significant molecular mobility is expected above glass transition temperature, proteins stored at -20 °C in the tested formulations are kinetically unstable. Formulations containing low salt and/or low protein levels showed the greatest and most erratic increase in aggregation (Figure 2), typical for kinetic processes (15). Figure 2 also shows a dramatic increase in aggregation after an initial lag period of approximately two months, which may be due to a slow molecular motion in the frozen state above Tg’. The observed lag phase at -20 °C is consistent with the lack of change in aggregate level upon a single freeze-thaw event (11).

This study illustrates a strategy for optimizing frozen storage stability of biotherapeutics in PBS-based formulations. Such optimization requires use of orthogonal analytical methods to monitor aggregation and particulates, glass transitions (21), freeze-thaw stress, as well as select long-term frozen stability studies tailored to each specific case.

Conclusion

Of all conditions and formulations tested, the addition of polyols had the largest positive impact on the stability of the model mAb in PBS-based formulations on long-term frozen storage. Formulating the mAb in PBS containing 3% sorbitol was optimal for preventing frozen storage aggregation and sub-micron particles. While 5% sucrose was optimal for preventing aggregation and stabilizing the level of sub-micron particles, the level of sub-micron particles was higher than non-polyol formulations, the mechanism for which is not currently understood. As such, optimization of PBS-based formulation for freeze-thaw stability by addition of polyols should be performed under thorough particulate control in the entire range from soluble aggregates to visible particulates.

References

1. J. Patel et al., Bioprocess Tech., January: 20-31 (2011).
2. B.S. Bhatnagar et al., Pharm. Dev. Technol., 12:505-523 (2007).
3. W. Wang, Intern. J. Pharm., 289: 1-30 (2005).
4. L.A. Kueltzo et al., J. Pharm. Sci., 97 (5) 1801-1812 (2008).
5. Abcam, www.abcam.com/protocols/antibody-storage-guide, accessed Aug. 16, 2016.
6. K.A. Pikal-Cleland et al., Arch. Biochem. Biophys., 384 (2) 398-406 (2000).
7. G. Gomez et al., Pharm. Res., 18: 90-97 (2001). 
8. A. Pyne, K. Chatterjee, and R. Suryanarayanan, Pharm. Res., 20 (5) 802-803 (2003).
9. M.A. Miller et al., J. Pharm. Sci., 102 (4) 1194-1208 (2013).
10. E. Cao et al., Biotechnol. Bioeng., 82 (6) 684-690 (2003).
11. T. Mezhebovsky et al., BioPharm Int., 29 (8) 33-39 (2016). 
12. Laerd Statistics, accessed Aug. 16, 2016.
13. JMP, accessed Aug. 16, 2016.
14. T. Arakawa and S.N. Timasheff, Biochemistry, 21: 6536-6544 (1982).
15. W. Brostow et al., Materials Letters, 62: 3152-3155 (2008).
16. S.H. Jiang and S.L. Nail, Eur. J. Pharm. Biopharm., 45: 249-257 (1998). 
17. B.S. Chang, B.S. Kendrick, and J.F. Carpenter, J. Pharm. Sci., 85 (12) 1325-1330 (1996).
18. A. Hillgren et al., Int. J. Pharmaceut., 237: 57-69 (2002).
19. K. Izutsu, Pharm. Res., 17 (10) 1316-1322 (2000).
20. L.M. Her et al., Pharm. Sci., 12 (5) 768-772 (1995).
21. L. Thomas et al., Biopharm. Int., manuscript accepted.
22. A. Mire-Sluis et al., Bioprocess Int., 9 (10) 38-47 (2011). 
23. S. Singh et al., Pharm. Res., 28: 873-885 (2011).

About the Authors

Tatyana Mezhebovsky is principal scientist,BioFormulations Development, Sanofi, 1 The Mountain Rd., Framingham, MA 01701; Eric Routhier is director, Pharmaceutical Development, Alexion Pharmaceuticals, 100 College St., New Haven, CT 06510; Philip Sass is president and CEO, Liquid Biotech USA, 1903 Black Hawk Circle, Audubon, PA 19403; and *Zahra Shahrokh is chief development officer, STC Biologics, 763 D Concord Ave, Cambridge, MA 02138, zshahrokh7@gmail.com.
*To whom all correspondence should be addressed.

Article Details

BioPharm International
Vol. 29, No. 9
Pages: 38–47

Citation

When referring to this article, please cite as T. Mezhebovsky et al., "Enabling Freeze-Thaw Stability of PBS-Based Formulations of a Monoclonal Antibody," BioPharm International 29 (9) 2016.