Developing a bioformulation stability profile

Executive summary

The stability and oligomeric state of proteins is highly sensitive to a wide range of parameters such as temperature and chemical environment during preparation, storage condition, and buffer choice. Understanding the impact of environmental parameters on a protein is essential for both the production and maintenance of high quality and safe proteins. Light scattering provides a suite of dilute solution properties that can be used to develop a stability profile for screening biotherapeutic candidates. The ability to handle both low sample volumes and high sample concentrations makes this technology ideal for moving the biotherapeutic screening process, both in terms of prediction and confirmation, to earlier points in the development pipeline.

Here, we describe the development of a bioformulation stability profile, derived from sub-micron light scattering analysis.

Introduction

Light scattering is a staple technology within the biopharmaceutical industry, routinely used to screen biotherapeutic candidates and formulations. Historically, light scattering technologies, such as static light scattering (SLS) , dynamic light scattering (DLS) , and electrophoretic light scattering (ELS), have primarily occupied a monitoring role within formulation development. Classic examples include the use of DLS to test for the appearance of aggregates during stress testing, or the use of ELS to monitor changes in a sample’s zeta potential and colloidal stability with environmental changes, such as pH and ionic strength. 

Malvern Panalytical’s Zetasizer Advance system combines all these light scattering technologies, providing information on particle size distribution, polydispersity trend, and interaction parameter from DLS, the second virial coefficient from SLS, and the electrophoretic mobility and effective charge from ELS. DLS exploits the link between hydrodynamic particle size and diffusion rate, calculated from measurements of time-dependent light fluctuations. ELS assesses the velocity of particles in an electric field, allowing for accurate calculations of zeta potential (an important parameter for understanding colloidal stability), and SLS involves the measurement of time-averaged scattering intensity at a range of sample concentrations to allow calculation of colloidal stability parameters. Each of these properties, measured under dilute conditions, is empirically “predictive” of aggregation propensity, solubility, and increased viscosity under the high concentration conditions typical of biotherapeutic formulations. Together, all light scattering technologies provided by Zetasizer Advance deliver complementary insights into the factors influencing protein quality.

Complementing these colloidal stability measurements by SLS, DLS, and ELS, dynamic light scattering-based aggregation temperature highlights the beginning of protein oligomer formation along the unfolding-aggregation pathways. Aggregation temperature, T_agg, from DLS or SLS, is the onset temperature at which light scattering intensity and apparent hydrodynamic radius start to increase, indicating the formation of oligomers/aggregates. Because scattering is very sensitive, T_agg is often detected as soon as a small fraction of molecules misfold or partially unfold and start to self-associate.

The typical stability parameter for unfolding events is T_m, the melting temperature, which in calorimetry is defined as the temperature at which half of the protein population is unfolded, i.e., the thermodynamic midpoint of the major unfolding transition. Differential scanning calorimetry (DSC), a powerful and label-free technique for accessing the structural stability of protein therapeutics, can measure the unfolding event directly from heat capacity.  

When comparing both parameters to each other, T_agg from DLS is typically a few to tens of degrees lower than T_m from DSC, because aggregation can start from partially unfolded intermediates, long before global unfolding is complete. Within a single protein and a narrow formulation space, a positive correlation between T_m and T_agg may be observed, while across different proteins or broad formulation ranges, both parameters might be decoupled. However, T_agg, via the light scattering technique, is a good first indicator for understanding protein stability. 

When compared, T_agg from DLS is typically a few to several tens of degrees lower than Tm from DSC, because aggregation can begin from partially unfolded intermediates long before global unfolding is complete. Within a single protein and a narrow formulation space, a positive correlation between T_m and T_agg may be observed, while across different proteins or broad formulation ranges, both parameters might be decoupled. However, T_agg via light scattering technique is a good first indicator for understanding protein stability. 

By combining T_agg obtained from light scattering, with T_m from DSC, a comprehensive bioformulation stability profile may be derived for protein therapeutic candidates, for use during early formulation development screening.

Biotherapeutic screening parameters

Analytical instrumentation can be used to monitor and predict sample properties. Light scattering techniques are routinely used to monitor a variety of properties during bioformulation stress testing, checking for the presence of aggregates or sub-micron particles. A formulation property measured under one set of conditions can be used to predict behavior under a different set of conditions – for example, screening biotherapeutic candidates for aggregation propensity at high concentrations based on second virial coefficient values measured under dilute solution conditions. 

In addition to the second virial coefficient, other light scattering properties being integrated into screening assays include size distribution and polydispersity trends, the DLS interaction parameter, and effective protein charge. Platforms that combine these measurements with DSF-based structural stability data in a single, high-throughput workflow are emerging as a compelling direction for the field.

Size distribution

Dynamic light scattering (DLS) is a low-resolution technique that requires a threefold difference in particle size to achieve baseline resolution in the particle size distribution (PSD). Scattering intensity, however, varies with the sixth power of the hydrodynamic radius. So, while DLS cannot resolve protein oligomers, its high sensitivity to larger particles can be exploited to monitor subtle changes in the distribution of protein oligomers. Consider, for example, the DLS size distributions shown in Figure 1 for IgG diluted in two buffers, where RS is the Stokes radius measured in the limit of infinite dilution. As the concentration increases, buffer 1 (left) shows an increase in the width of the distribution, suggesting the formation of oligomeric species with increasing sample concentration. These results are indicative of reversible self-association. However, or buffer 2 (right),  an increase in sample concentration leads to an apparent downward shift in the size distribution, with no significant change in the distribution width. These results are indicative of electrostatic repulsion, with the apparent decrease in size being attributable to an electrostatically driven increase in the diffusion coefficient.

Figure 1: Concentration-dependent size distributions for IgG in buffer 1 (left), with Tween stabilizer, and buffer 2 (right) with aspartic acid stabilizer.

Polydispersity trend

The mechanistic interpretation of the results in Figure 1 (reversible self-association vs. electrostatic repulsion) is easier to visualize in the graphical format shown in Figure 2. The DLS polydispersity (Pd) is the width or standard deviation of the size distribution and is represented by error bars in the concentration trends shown below. Both the mean size and width of the IgG oligomeric distribution decline with decreasing concentration in buffer 1, indicative of reversible self-association. or buffer 2, however, the width of the distribution is independent of sample concentration and remains constant, whereas the mean size at higher concentrations is “smaller” than R_S. As R_S is defined as the physical size of the antibody in the absence of restricted diffusion (viscosity) or particle interaction effects, the only explanation for a protein appearing smaller than it is, is if the particles are moving faster than the diffusion velocity. In the absence of a temperature increase, electrostatic repulsion is the only force available to increase the diffusion velocity.

Figure 2: Mean size and polydispersity trend with concentration for IgG in buffer 1 and buffer 2, with the polydispersity represented by error bars.

DLS interaction parameter

The qualitative information contained within the size distribution and Pd trend can be quantitatively expressed with the DLS interaction parameter (k_D). k_D is extracted from the slope of the concentration dependence of the mutual diffusion coefficient (D) as indicated in the expression shown below, where D₀ is the self-diffusion coefficient in the limit of zero concentration (C) from which the Stokes radius is derived, B₂₂ is the 2nd virial coefficient, M_W is the molecular weight, k_f is the 1st order concentration coefficient of the friction coefficient, and υ is the partial specific volume.

Empirically, k_D has been shown to be correlated with both aggregation rate and formulation viscosity. Samples exhibiting larger positive k_D values exhibit a lower propensity for self-association and aggregation, as well as lower viscosities at the high concentrations typical of biotherapeutic formulations. Based on these empirical predictions, it would be expected that buffer 1 from the earlier example, which exhibited reversible self-association, would have a smaller k_D value compared to buffer 2, which exhibited stabilizing electrostatic repulsion. Figure 3, which shows the concentration dependence of the diffusion coefficient for IgG in buffers 1 and 2, confirms that expectation.

Figure 3: Concentration dependence of the measured diffusion coefficient for IgG in buffers 1 and 2.

Second virial coefficient

The second virial coefficient B₂₂ is a thermodynamic parameter that quantifies how pairwise interactions between molecules cause a solution to deviate from ideal, non-interacting behavior. In protein solutions, the osmotic second virial coefficient specifically reflects how strongly protein molecules attract or repel each other on average: negative B₂₂ indicates net attraction (prone to aggregation), while positive B₂₂ indicates net repulsion (better colloidal stability). Because of this, B₂₂ is widely used as a semi-quantitative descriptor of colloidal stability, protein crystallization windows, and aggregation risk. 

There are 2 components to k_D: a thermodynamic component (B₂₂M_W) and a hydrodynamic component (k_f + υ). The predictive power of k_D derives from the thermodynamic component, specifically B₂₂. 

B₂₂ is extracted from the slope of a Debye plot, which represents the concentration dependence of the sample scattering intensity, as indicated in the expression shown below, where K is an optical constant, R_θ is the Rayleigh ratio of scattered to incident light intensity, M_W is the weight average molecular weight, and P(θ) is the shape factor, which is equivalent to 1 for small proteins such as antibodies.

Figure 4 shows a comparison of the Debye plots for the previous IgG in buffers 1 and 2, along with a third buffer utilizing a lactose stabilizer. As seen in this figure, buffers 2 and 3 both exhibit positive virial coefficient values, with buffer 1 exhibiting a negative virial coefficient. Based upon empirical predictions, buffer 2 would be expected to be the most stable and buffer 1 the least stable, with buffer 3 somewhere in between.

Buffer 3 highlights the importance of the hydrodynamic component in k_D. While not shown in Figure 3, IgG in buffer 3 exhibited a k_D value that was more negative than both buffers 1 and 2, with measured k_D values of 31.9 mL/g, -5.2 mL/g, and -9.7 mL/g for buffers 2, 1, and 3, respectively. If the hydrodynamic component were ignored, it might be erroneously concluded that buffer 1 is likely to be more stable than buffer 3. This is disproved by the B₂₂ results. It is for this reason that k_D, while easier to measure than B₂₂, is generally restricted to use as a stability predictor within the k_D > 0 region, with B₂₂ considered the more reliable predictor across the widest range of formulation conditions.

Buffer 3 highlights the importance of the hydrodynamic component in k_D. While not shown in Figure 3, IgG in buffer 3 exhibited a k_D value that was more negative than both buffers 1 and 2. The measurement k_D values were 31.9 mL/g, -5.2 mL/g, and -9.7 mL/g for buffers 2, 1, and 3, respectively. If the hydrodynamic component were ignored, one might erroneously conclude that buffer 3 is  more stable than buffer 1. However, the B₂₂ results demonstrate that this is not the case. It is for this reason that k_D, while easier to measure than B₂₂, is generally restricted to use as a stability predictor within the k_D > 0 region, with B₂₂ considered the more reliable predictor across the widest range of formulation conditions.

Figure 4: Debye plots for IgG in buffers 1, 2, and 3, with theta conditions at B₂₂ = 0 represented by the dashed line. 

Intrinsic properties

Here, we define “intrinsic” properties as those of the protein within the formulation buffer, but in the absence of concentration-dependent effects that impact reported values – including restricted diffusion and/or electrostatic interactions. By determining intrinsic properties in the limit of infinite dilution, these can be minimized.

Oligomeric molecular weight

The weight-averaged molecular weight (M_W) is derived during an ensemble SLS second virial coefficient measurement. As it is mass-weighted, small amounts of oligomeric species can have a large effect on MW. If the monomeric molecular weight is known, as it often is, then comparison to the oligomeric M_W gives a qualitative idea of the degree of oligomerization.

Stokes radius

The Stokes radius (R_S) is derived during the course of an ensemble DLS interaction parameter (k_D) measurement. The Stokes radius is an intensity-weighted average size of the oligomeric distribution. Comparison of the Stokes radius to the known hydrodynamic size of the monomer provides qualitative information regarding the oligomeric distribution, in the absence of concentration-dependent thermodynamic effects.

% Polydispersity

A third intrinsic property that can be used to provide qualitative information regarding the oligomeric distribution is the % polydispersity. This value is derived during the course of a DLS k_D measurement (%Pd = Pd/size). While DLS cannot resolve oligomers, the width of the intensity-weighted size peak is very sensitive to the presence of oligomeric species. The rule of thumb for monomeric protein samples is that %Pd should be less than 15-20%. A %Pd value greater than this is a clear indication of the presence of oligomeric components.

Net charge

The net charge (Z) is calculated from the electrophoretic mobility (µ_E), measured using ELS, and the Stokes radius (R_S), measured using DLS, as indicated in the expression shown below, where h is the viscosity, κ is the inverse Debye length, f(κR_S) is the Henry function, and Z_Eff is the effective charge at the slipping or interaction plane.

The net charge is the principal driver of colloidal stability. The greater the net charge, the greater the electrostatic repulsion between like particles. For antibodies and other proteins, the net charge is particularly important, due to the heterogeneity of the surface charge, which can lead to attractive dipole-dipole interactions at the higher concentrations typical of biotherapeutics. For antibodies exhibiting large dipole moments, the net charge must be large enough to counter these attractive interactions, otherwise, aggregation and increased viscosity at high sample concentration are probable. As a general rule of thumb, an effective charge of 4-6 is typically indicative of good colloidal stability for mAb formulations, with an effective charge > 6 being indicative of excellent stability.

Table 1 below shows a comparison of the net and effective charges of the previous IgG in buffers 1, 2, and 3 as an example. These samples are ranked in order of decreasing stability, as predicted by the second virial coefficient. A fourth buffer (PBS with no stabilizer) noted as “#4”, is included for discussion purposes.

Table 1: Colloidal Stability Predictors

As seen in Table 1, the effective charge predictions are generally consistent with the B₂₂ predictions – only with buffer 2 does the sample exhibit a large positive B₂₂, indicating good colloidal stability. At lower values, however, (buffers 1 and 4, for example), the charge is a less effective predictor, as dipole-dipole interactions become a factor in this realm. These dipole-dipole effects are accounted for in thermodynamic B₂₂ values, but not in net or effective charge values.

Melting and aggregation temperature

The protein aggregation onset temperature (T_agg) is the minimum temperature required to induce aggregation in a protein formulation. As a consequence of the R6 dependence of the scattering intensity, subtle changes in the aggregation state of a protein formulation are easily detected with light scattering techniques. T_agg is determined by light scattering techniques as the temperature where aggregation becomes clearly detectable – for example, the point where the scattering intensity/size signal crosses a defined slope or threshold. Figure 5 shows the DLS thermal ramps for the previous IgG in buffers 1, 2, 3, and 4.

Figure 5: DLS thermal ramps for IgG in buffers 1, 2, 3, and 4, showing the sharp increase in the Z average size at T_onset, indicative of the onset of aggregation.

As evident in Figure 5, aggregation leads to a sharp increase in the Z average size, clearly defining T_agg for the aggregation for IgG in buffers 1, 3, and 4. Buffer 2, however, shows no evidence of aggregation up to a temperature of 90 °C.

The melting temperature (T_m) is the temperature at which protein unfolding occurs, usually measured by DSC, where half the protein population is unfolded. T_m is then indicative of the inherent structural stability of the folded protein, and in the absence of denaturation, T_agg would be indicative of the colloidal stability. However, T_agg tends to follow T_m, and is usually a direct consequence of the T_m denaturation, rather than a loss of colloidal stability. In fact, it’s only when T_agg precedes T_m that it can be used as a stability predictor, although the use of T_m as a predictor of inherent structural stability is routine within bioformulations.

Traditionally, these two parameters have been derived from separate measurement techniques: DLS or SLS for T_agg, and DSC for T_m. DSC enables label-free measurement of T_m via changes in heat capacity.

The magnitude of light scattered from a sample is proportional to the square of the molecular weight and the square of the refractive index increment (dn/dC), which is proportional to the molecule density or the partial specific volume. As proteins unfold, they expand, which leads to an increase in size and a decrease in dn/dC. Therefore, light scattering can also be used to monitor thermal-induced denaturation. Consider Figure 6, for example, which shows expanded views of the DLS thermal ramps for IgG in buffers 1 and 4, along with the simultaneously measured scattering intensity.

Figure 6: Expanded views of DLS thermal ramps for IgG in buffers 1 and 4, along with the simultaneously measured scattering intensity

In buffer 1, a sharp increase in hydrodynamic size is observed at 66 °C, which indicates a thermal-induced aggregation. Prior to this event however, a small increase in size accompanied by a sharp decrease in scattering intensity is observed at the aggregation temperature of T_agg = 56 °C. The small increase in size could be a result of either self-association or unfolding (either fully or partially). In buffer 2 in comparison, there is as well a sharp increase in hydrodynamic size at 65 °C (T_agg), while very little changes in size around 55 °C are observed. Here, the small increase in hydrodynamic size is accompanied by an increase in scattering intensity, unlike in buffer 1. This increase suggests an increase in molecular weight, likely arising from thermally induced self-association before denaturation and subsequent agglomeration at the 70 °C.

The combination of in-depth data of changes in hydrodynamic radius and scattering intensity by DLS can help to identify the temperatures, where small changes in aggregation behaviour occur, and helps to differentiate different aggregation mechanisms, being complementary and supportive in the analysis and understanding of further DSC data. 

The bioformulation stability profile

The stability profile of a biotherapeutic candidate is a collection of stability predictors and descriptors that provides insights into the developability and manufacturability of the bioformulation. The resultant stability profile for the IgG in buffers 1 and 2 examined in this report is shown in Table 2. The green, yellow, and red coding indicates “go”, “caution,” and “no go”, while “RSA” and “CR” represent reversible self-association and charge repulsion. While specific targets depend on the company and screening program, the values indicated in this example are reasonably representative of typical target values. That said, the developability and manufacturability conclusions are somewhat arbitrary in this example. 

Table 2. Bioformulation Stability Profiles - IgG in Buffer 1 and 2.

Although not exhaustive, the parameters included in Table 2 are readily extracted from parameters measured using light scattering technology. These parameters form the initial basis of a predictive stability profile that is readily enhanced by integrating with, or correlating to, additional parameters such as aggregation rates, DSC results, high-order structure, and other properties relevant to a particular target biotherapeutic formulation profile.

Stability profile over different storage conditions

To investigate the effect of storage conditions on IgG, freshly prepared antibody samples were treated in the following ways: stored for 35 days at 4 °C; stored for 31 days at 25 °C, then for four days at 4 °C; and subjected to five freeze/thaw cycles then stored at 4 °C. All samples were prepared at a concentration of 0.8 mg/ml in the same buffer.

All three samples show a peak at around 10 nm (Figure 7), the expected diameter of the antibody, but the relative width peak and the size distribution are different for each sample. With those samples stored under non-ideal conditions, there is evidence of large aggregates, particularly when subjected to freeze/thaw cycles. The “quality” of each sample is evident from z-average diameter and polydispersity index data, which show that freeze/thaw cycling is particularly detrimental (Table 3).

Figure 7. Size distribution (by percent volume) from antibody samples stored under different conditions: 4 °C (blue), 25 °C then 4 °C (red) and 5x frozen/thawed (green).

Table 3. Size and polydispersity index are influenced by storage conditions

DLS is a powerful technique for detecting the presence of very small numbers of relatively large particles, which can give an early indication of stability issues during protein storage or during a treatment process.

DLS can assess protein stability under different storage conditions and reveal condition-dependent changes in aggregation behavior. These observations can be independently supported by DSC profiles, which capture the corresponding thermal unfolding transition of the protein, and more importantly, reveal the total amount of folded protein molecules left under storage conditions. Thus, the change in calorimetric enthalpy  ΔH_cal reveals the less favorable storage conditions, and the amount of active protein left in the formulation. Together, DLS and DSC provide complementary evidence that links storage-induced stress and thus changes in colloidal stability with alterations in conformational stability.