High-Throughput Screening of Formulations to Optimize the Thermal Stability of a Therapeutic Monoclonal Antibody

Abstract

Monoclonal antibodies (mAbs) are an important class of biotherapeutics. Successful development of a mAb depends not only on its biological activity but also on its physicochemical properties, such as homogeneity and stability. mAb stability is affected by its formulation. Among the many techniques used to study the stability of mAbs, differential scanning fluorimetry (DSF) offers both excellent throughput and minimal material consumption. DSF measures the temperature of the protein unfolding transition (T_m) based on the change in fluorescence intensity of the environmentally sensitive dye SYPRO Orange. With DSF adapted to a 96-well plate format, we have shown that low-pH or high-salt concentrations decrease the thermal stability of mAb1, whereas some excipients, such as sucrose, polysorbate 80, and sodium phosphate, increase its stability. The basal fluorescence of SYPRO Orange was enhanced by the presence of detergents, limiting the use of this approach to diluted detergent solutions. Throughput of DSF can be increased further with the use of a 384-well plate. DSF thermograms are in good agreement with the melting profiles obtained by differential scanning calorimetry (DSC). The T_ms determined by DSF and DSC were well correlated, with the former being on average lower by 3 °C.

Introduction

Antibodies are heterodimers composed of 2 heavy and 2 light polypeptide chains linked by disulfide bonds. The framework of an antibody is divided into the Fc region (fragment crystallizable) and Fab region (fragment antigen binding), which are built of folded domains derived from the heavy- and light-chain proteins. The Fc, for example, contains heavy-chain constant regions 2 and 3, denoted C_H2 and C_H3, respectively. ¹ Antibodies are classified into 5 isotypes based on the identity of the heavy-chain constant region: immunoglobulin A (IgA), IgD, IgE, IgG, and IgM. Of these classes, IgGs constitute the majority of the recombinant antibodies currently under development. There are 4 subclasses of IgGs (1 to 4). ¹

The number of therapeutic monoclonal antibodies (mAbs) in development has increased considerably since the approval of the first mAb, muromonab-CD3, in 1986. ¹ The fast growth of the therapeutic antibody market is related to the favorable properties of these macromolecules: They are highly specific and effective at low concentrations, cause few side effects, can be modified by attachment of therapeutic or diagnostic molecules, and can be engineered to have a human framework. ¹ Despite their popularity, many mAbs display poor biophysical properties, such as low chemical and physical stability, resulting in unwanted modifications, degradation, and aggregation during manufacturing, storage, and dosing. ² Poor stability can lead to antibody misfolding, resulting in a low product yield and a substantial fraction of inactive material, as well as degradation during storage.^1–3 Aggregation can lead to hypersensitive immune response, including anaphylaxis, loss of biological activity, and occurrence of nonresponders, due to the formation of neutralizing antibodies against aggregates or degradants. ³ Therefore, successful development of mAbs depends not only on the attainment of the desired biological activity but also on superior physicochemical properties (e.g., homogeneity and physical and chemical stability). Although antibodies share similar tertiary structure, they display notable differences in their primary sequence, which leads to significant differences in mAb susceptibility to deamidation, isomerization, oxidation, denaturation, and aggregation. ¹ These properties can be improved through additional sequence engineering of the mAb and by the development of optimized formulations.^3,4 When developing mAb formulations, one needs to consider the therapeutic Ab dosage form (lyophilized or in solution), mAb concentration, ionic strength of the solution, pH, type of buffering substance, and presence of excipients (salts, sugars, polyols, amino acids, surfactants, polymers, antioxidants, preservatives, etc.).^1,2,4 Because of the complexity of mAb molecules and the multiple variables to be considered, development of protein formulation is generally an empirical process. Moreover, this process could benefit from fast and miniaturized screening approaches. Techniques typically applied to analyze proteins include size exclusion chromatography, capillary electrophoresis, light scattering, circular dichroism, intrinsic fluorescence spectroscopy, mass spectrometry, and thermal analysis: differential scanning calorimetry (DSC) and differential scanning fluorimetry (DSF).^2,3,5,6 Only some of these methods are amenable to high throughput.

DSF is a biophysical technique that is suitable for high-throughput screening with a minimal protein requirement. DSF estimates the temperature of the protein unfolding transition (T_m) by measuring the change in the fluorescence intensity of an environmentally sensitive fluorophore added to the solution. ⁷ Such molecules display low fluorescence in a hydrophilic environment. In most folded proteins, hydrophobic regions are sequestered from the solution within the protein core. During denaturation, proteins lose their tertiary and secondary structures leading to exposure of the hydrophobic regions. These regions can bind an environmentally sensitive dye leading to an increase in its fluorescence.^5,7

Here, we analyzed the impact of formulation on the thermal stability of mAb1 with DSF adapted to a 96-well plate format using SYPRO Orange dye and a real-time PCR system. With this setup, up to 32 conditions could be simultaneously screened in triplicate in less than 75 min. Throughput of DSF could be further increased with a 384-well plate format. Effects of formulation components revealed by DSF are in agreement with general trends for physical stability of Abs reported in the literature.

Material and Methods

Reagents

Reagents were prepared with chemicals of the highest quality commercially available and water treated with a Milli-Q purification system (Millipore, Bedford, MA). SYPRO Orange fluorophore was purchased from Life Technologies (Grand Island, NY) as a 5000× concentrated stock solution in DMSO. Polysorbate 80 (PS80), polysorbate 20 (PS20), sucrose, histidine, sodium chloride, sodium succinate, citrate, and sodium phosphate (di- and monobasic) were obtained from Sigma-Aldrich (St. Louis, MO).

Monoclonal Antibody

Antibodies were manufactured by Gilead Sciences, Inc. (Foster City, CA). The human monoclonal antibody Ab1 (mAb1, subtype IgG4) was purified by protein A capture, low pH elution, viral inactivation, cation and anion exchange chromatography, and viral filtration, and concentrated and filtrated by ultrafiltration and diafiltration to 10 mg/mL in 10 mM sodium phosphate buffer, pH 5.8. Mouse monoclonal antibodies Ab2 (mAb2, IgG2b) and Ab3 (mAb3, IgG1) were purified by MabSelect SuRe (GE Healthcare, Piscataway, NJ) protein A affinity chromatography, followed by concentration by ultrafiltration to 10 mg/mL and dialysis to 20 mM sodium phosphate, pH 6.5, 140 mM sodium chloride, 0.01% PS20. The molecular weights and isoelectric points (pIs) of mAb1, mAb2, and mAb3, calculated based on the amino acid composition, were 145 kDa and 7.9, 146 kDa and 7.7, and 144 kDa and 7.9, respectively.

Monoclonal Antibody Samples Preparation

Samples of mAb1 were prepared in various formulations by dialysis throughout a period of 16 h with 2 buffer changes each with volume exceeding 1000 times the volume of the antibody sample. After dialysis, concentration of mAb1 was measured by ultraviolet-visible spectroscopy using a calculated extinction coefficient of 1.45 (mg/mL)⁻¹. When necessary, samples were concentrated with the use of Amicon Ultra-15 centrifugal filter devices, and their concentrations were adjusted to 10 mg/mL with dialysis buffer.

Differential Scanning Fluorimetry

DSF was carried out in a ViiA7 real-time PCR instrument (Life Technologies) in the respective formulation buffer. Each sample was measured in triplicate. Unless otherwise specified, 96-well MicroAmp Fast reaction plates (Life Technologies) were used with 25 µL sample per well. All concentrations are final after mixing. MAbs were diluted to 1 mg/mL concentration in the respective formulation buffer, unless otherwise stated. SYPRO Orange was diluted 1000-fold from the 5000× concentrated stock to the working dye solution in the appropriate formulation buffer prior to addition to the mAb samples. To prevent bleaching, the working solution of SYPRO Orange was added to the reaction mixture just prior to the experiment. Thermal denaturation was carried out by increasing the temperature from 25 °C to 95 °C at a rate of 0.017 °C per second. Fluorescence intensity (excitation at 490 nm and emission with the use of a ROX filter at 600 to 630 nm) was collected at 0.07 °C intervals and analyzed with Protein Thermal Shift Software (Life Technologies), using the first derivative approach to calculate T_m. In this method, T_m is the temperature corresponding to the maximum value of the first derivative of the DSF melting curve. Delta T_m values (ΔT_m) for various formulation buffers were calculated by subtracting the T_m value obtained for mAb1 in a reference formulation from the T_m value obtained for mAb1 in the tested formulation.

Differential Scanning Calorimetry

DSC measurements were performed on a Microcal VP-Capillary DSC platform (GE Healthcare). The mAb1 samples were diluted in respective formulation buffer to a concentration of 1 mg/mL. Matched formulation buffer was used as a reference. The samples were scanned from 20 °C to 95 °C at a rate of 1 °C/min using a filtering period of 8 s and 15 min of pre-equilibration prior to the start of each run. Data were first normalized for protein concentration, then baseline corrected and buffer subtracted using Origin 7.0 AutoSampler DSC software (GE Healthcare). Melting transitions were irreversible and were analyzed with cursor-initiated DSC peak fit function using the non-two-state unfolding model within the Origin 7.0 software.

Results and Discussion

Applicability of DSF for Studying the Thermostability of mAb

One of the important properties of successful mAbs is their conformational stability (i.e., the ability to retain their native secondary, tertiary, and quaternary structures).^1,2 Conformational stability can be assessed by measuring thermal stability with the use of techniques sensitive to changes in protein folding, such as circular dichroism, fluorescence spectroscopy, light scattering, DSC, and DSF. A representative thermal unfolding profile (thermogram) of mAb1 obtained by DSF is depicted in Figure 1A . At lower temperatures, at which antibody remains folded, fluorescence intensity was very low, in agreement with the presence of mostly hydrophilic amino acid residues on the surfaces of immunoglobulins under native conditions. Increase in the temperature higher than 55 °C was accompanied by a sharp increase in fluorescence intensity that reached a plateau and ultimately decreased with further increase in the temperature. The initial rise in fluorescence reflects greater exposure of the hydrophobic areas of mAb1 because it undergoes thermal unfolding. Quenching of the fluorescence in the terminal phase of the experiment was shown to be a nonspecific transition associated with a gradual decrease in the inherent fluorescence of SYPRO Orange in a hydrophobic environment at high temperatures. ⁵ The thermogram for mAb1 displays 2 transitions, which are better visualized in the plot of the first derivative of the fluorescence melting curve (dF/dT; Fig. 1B ). Temperature of the unfolding (melting) transition (T_m) was determined for mAb1 from the value of the local maximum of the dF/dT curve. At 1 mg/mL protein concentration in formulation containing 10 mM sodium phosphate, pH 5.8, 140 mM sodium chloride, 0.02% PS80, the lower and the higher temperature transitions occurred with T_m1 and T_m2 values of 64.9 °C and 71.5 °C, respectively. These multiple temperature-dependent transitions result from the fact that individual domains of the monoclonal antibodies have distinct conformational stabilities and display various levels of the intramolecular interactions with other fragments of the protein, which leads to the differences in cooperativity of domains unfolding. ⁸

Figure 1.

Comparison of thermal unfolding transitions and melting temperatures (T_ms) determined for monoclonal antibody Ab1 (mAb1) by (A, B) differential scanning fluorimetry (DSF) and (C, D) differential scanning calorimetry (DSC). Experiments were performed at 1 mg/mL of mAb1 [an immunoglobulin-4 (IgG4) subtype] in buffer containing 10 mM sodium phosphate, pH 5.8, 140 mM sodium chloride, 0.02% polysorbate 80. (A) DSF melting curve. (B) The first derivative curve: dF/dT. T_m in (A) is at the inflection point of the protein unfolding transition, which corresponds to the peak of dF/dT in (B). DSF thermogram for mAb1 displays 2 transitions: first with a lower (T_m1) and second with a higher (T_m2) melting temperature (64.9 °C and 71.5 °C, respectively). The first transition corresponds to unfolding of the C_H2 domain, and the second to melting of the more stable Fab and C_H3 domains. (C) DSC thermogram for mAb1 also displays 2 melting transitions (67.6 °C and 71.2 °C). Raw data (solid blue line), first and second transition peaks (dotted red line), and their sum (dotted blue line) are plotted. (D) Comparison of the T_m values obtained for mAb1, mAb2 (IgG2b), and mAb3 (IgG1) in various formulations (Suppl. Table S1) with the use of DSC (Suppl. Fig. S1) and DSF. The correlation between T_m values obtained from DSF and DSC is very good (R² = 0.92). The average difference in the absolute values of T_m between the 2 techniques was 3.3 ± 0.7 °C with identical dependency trends.

A thermogram for mAb1 was also obtained by DSC ( Fig. 1C ). Similar to the DSF unfolding profile, the DSC melting curve displayed 2 transitions with T_m1 and T_m2 equal to 67.4 °C and 71.3 °C, respectively (1 mg/mL, in 10 mM sodium phosphate, pH 5.8, 140 mM sodium chloride, 0.02% PS80). Studies have shown that in DSC, mAbs display distinctive unfolding profiles as a result of melting transitions of their individual domains: C_H2, C_H3, and Fab. The thermal stability of IgG molecules is affected by all of these domains.^8–11 DSC curves reported for antibodies of IgG4 subtype display 2 clearly discernible transitions: a lower temperature representing unfolding of the C_H2 domain (T_m1), and a higher temperature representing melting of C_H3–Fab domains (T_m2).^9,10 Fab melting transition is generally well defined with a peak up to 3 times greater than that of C_H2 or C_H3, which was also true in the case of mAb1.^8–10

The temperature for the first transition of mAb1 obtained by DSC was slightly higher than T_m1 determined by DSF. The same trend was observed for mAb1 and other antibodies tested in several additional formulations ( Fig. 1D , Suppl. Fig. S1, and Suppl. Table S1) and for unrelated antibodies studied by others. ⁵ On average, the T_m1 determined by DSF was approximately 3 °C lower than the corresponding DSC melting temperature. The reason for this observation is not fully understood. It is possible that SYPRO Orange slightly destabilizes mAbs by virtue of its preference for hydrophobic regions of the protein, which are present mainly in a partially or fully unfolded protein. Moreover, an extrinsic fluorescent probe could be more sensitive to the changes in mAb stability than DSC because any conditions causing a change in the gross hydrophobicity of the protein surface will be reflected in the increase of the SYPRO Orange fluorescence, even if they do not lead to a complete unfolding of an Ab domain. ⁵ A good linear correlation (R² = 0.92) was seen, however, between the T_ms determined by both methods. DSF and DSC reported the same relative stabilities of antibody domains in different formulations^5–7,12 ( Fig. 1D ), and the unfolding profile of mAb1 obtained with the use of DSF resembles closely a DSC thermogram with 2 melting transitions reflecting unfolding of C_H2 and C_H3–Fab. Therefore, DSF using a SYPRO Orange reporter appears to be a reliable method for determination of the relative stabilities of mAbs. DSC is well established for measuring the thermodynamic stability of mAbs. However, DSC requires a substantial amount of mAbs, has a limited range of applicable protein concentration, and allows analysis of only 1 formulation at a time. Even with the recent introduction of capillary calorimetry (e.g., MicroCal VP-Capillary DSC) equipped with a robotic arm and adapted for a 96-well plate format, the throughput does not exceed 50 samples in 24 h, ¹³ with some of the sample slots being used by buffer controls and cell-cleaning steps. In contrast, DSF is easy to set up, uses common laboratory instrumentation (a real-time quantitative PCR instrument) and an inexpensive fluorescent dye, requires a minimal amount of protein, and allows for simultaneous analysis of all plate wells, offering speed and flexibility in formulation screening.

Adaptation of DSF for mAb Formulation Screening

To evaluate the performance and robustness of T_m determination by DSF, the thermal stability of mAb1 was studied in the presence of increasing concentrations of SYPRO Orange, at various concentrations of antibody, and in the presence of increasing amounts of PS80 ( Fig. 2 ). SYPRO Orange is supplied as a 5000-fold concentrated stock solution in 100% DMSO. Using the dye at concentrations from 1.25- to 10-fold did not change the values of the T_ms for mAb1 ( Fig. 2A ). Increasing the concentration of SYPRO Orange increased the fluorescence intensity in the plateau of the melting profile and in the peak of the first derivative curve, resulting in a higher signal-to-noise ratio and better defined transitions. The background fluorescence increased only marginally at the highest SYPRO Orange concentration (data not shown). For the remainder of this study, SYPRO Orange was used at 5-fold concentration.

Figure 2.

Development of a differential scanning fluorimetry (DSF) thermostability assay to measure the unfolding of monoclonal antibody Ab1 (mAb1). The DSF assay was performed for mAb1 at 1 mg/mL concentration in buffer containing 10 mM sodium phosphate, pH 5.8, 140 mM sodium chloride, 0.005% polysorbate 80 (PS80), and 5× SYPRO Orange (SO) unless otherwise stated. (A) Effect of SO concentration on the robustness of the signal (fluorescence intensity) and the values of T_m. The first and second unfolding transitions (T_m1 and T_m2, respectively) are independent of SO concentration and equal to 63.5 ± 0.2 °C and 69.9 ± 0.2 °C, respectively (mean and standard deviation from assays performed at all 5 SYPRO Orange concentrations in triplicate). (B) Effects of mAb1 concentration on T_m1 and T_m2. The DSF assay was performed in solutions containing increasing concentrations of mAb1. Interpretable thermograms can be obtained at mAb1 concentrations as low as 5 µg/mL. (C, D) Effect of PS80 on the profile of DSF thermograms and ability to determine T_m. (C) DSF thermal unfolding curves were collected in the absence of sodium chloride and in the presence of increasing amounts of PS80. An elevated fluorescence background is observed at PS80 concentrations >0.02% and masks T_m1. (D) Dependence of measured T_m1 and T_m2 on the concentration of PS80. T_m1 cannot be determined in formulations containing >0.04% of PS80.

The study depicted in Figure 2B confirms that DSF can be performed at a wide range of protein concentrations, from 10 mg/mL to as low as 5 µg/mL. At very low protein concentrations (lower than 20 µg/mL), the first melting transition was not well defined, and its midpoint could not be accurately determined; however, the second transition (associated with an overall larger fluorescence change) remained well defined. This shows one of the strengths of DSF: Unlike DSC, DSF can be successfully applied even with limited amounts of material. Dynamic light scattering (DLS) and analytical ultracentrifugation (AUC) showed that mAb1 did not aggregate in the protein concentration range used in these studies and that the main state of the protein is monomeric (Suppl. Material and Suppl. Fig. S2A–C). Both T_m1 and T_m2 for mAb1 decreased (by around 5 °C and 4.5 °C, respectively) with increasing protein concentration. A similar trend has been previously observed for an unrelated mAb in various formulations by DSF and DSC.^5,12,14 This decrease in conformational stability can be explained by an increased probability of reversible and irreversible self-associations between antibody molecules in a crowded antibody solution on exposure to the elevated temperatures, rather than by the decrease in the intrinsic structural stability of mAb1.^5,12,14 These interactions can be mediated by electrostatic contacts. The associated antibodies could be more prone to structural perturbations, exposure of hydrophobic surfaces, and further bimolecular self-association. ¹⁴ Therefore, concentrated antibody solutions can lead to higher levels of aggregation under stress conditions.

Some antibody formulations contain surfactants, such as Poloxamer 188/407, polysorbate 20/40/80, or sodium lauryl sulfate, to prevent protein aggregation, surface denaturation, and surface adsorption. ⁴ Surfactants possess hydrophobic groups to which SYPRO Orange could bind, contributing to an increase in fluorescence background. We tested the effect of surfactant on the performance of the DSF assay by varying PS80 from 0.0013% to 0.16% ( Fig. 2C ). With an increase in PS80 concentration, the initial fluorescence (background) increased and the profile of the mAb1 thermal unfolding became distorted ( Fig. 2C ). Nevertheless, the T_m for unfolding transitions could be still determined reliably from the first derivative of the thermograms (Suppl. Fig. S3A) up to around 0.04% and 0.16% PS80 for T_m1 and T_m2, respectively ( Fig. 2D ). In this formulation (10 mM sodium phosphate, pH 5.8, 140 mM sodium chloride, 1 mg/mL of antibody), PS80 in the concentration range from 0.0013% to 0.16% did not have any effect on the thermal stability of either C_H2 or C_H3–Fab, with an average T_m1 and T_m2 equal to 66.2 ± 0.4 °C and 71.0 ± 0.1 °C, respectively ( Fig. 2D ).

Effects of pH, Buffer Type, and Excipients on Thermal Stability of mAb Revealed by DSF

The effect of sodium chloride concentration on the stability of mAb1 is shown in Figure 3A . As the concentration of sodium chloride was increased from 10 mM to 1000 mM, the unfolding transition temperatures shifted to lower values, indicating destabilization of the antibody structure. The first unfolding transition, T_m1, was more sensitive than the second transition, T_m2, to the increase in salt concentration. Increasing sodium chloride to 1000 mM resulted in a decrease in T_m1 and T_m2 by 6.3 ± 0.4 °C and 1.9 ± 0.3 °C, respectively. At low salt concentrations, the thermal unfolding transitions of domains C_H2 and C_H3–Fab of mAb1 were more separated, suggesting that unfolding of antibody domains is less cooperative (Suppl. Fig. S3B). DLS measurements revealed that high concentrations of sodium chloride did not induce aggregation of mAb1 at room temperature (Suppl. Fig. S2D). The destabilizing effect of high concentrations of sodium chloride shown in our study agrees with previous studies performed with various techniques. Salts neutralize charge that is present on the surface of an antibody, which reduces repulsion between macromolecules, leading to enhanced interactions between exposed hydrophobic surfaces of unfolding intermediates that become propagated under stress conditions such as increased temperature. Therefore, salts increase the propensity of antibodies to unfold and aggregate.^4,6

Figure 3.

Effects of (A) sodium chloride concentration and (B–D) pH on melting temperatures determined for monoclonal antibody Ab1 (mAb1) with the use of differential scanning fluorimetry (DSF). (A) The melting temperature of first and second unfolding transition (T_m1 and T_m2, respectively) for 0.5 mg/mL mAb1 in 10 mM sodium phosphate, pH 5.8, as a function of sodium chloride concentration. High sodium chloride concentrations destabilized mAb1. Stability of the C_H2 domain (T_m1) is more affected than stability of C_H3–Fab domains (T_m2). (B) T_m1 and T_m2 values for mAb1 at 1 mg/mL in 10 mM sodium phosphate as a function of pH. T_m1 (C_H2) is very sensitive to the pH changes (increases with pH increase). T_m2 (C_H3–Fab) decreases slightly with increasing pH. ΔT_m1 for mAb1 at pH 7.5 versus pH 5.0 is 6.2 ± 0.1 °C (mean and standard deviation of 3 determinations). (C) The effect of buffering substance identity and pH on T_m1 for mAb1. Seventeen formulations were analyzed: 10 mM histidine (pH 6.0, 6.5, 7.0, and 7.5), 10 mM sodium phosphate (pH 5.5, 6.0, 6.5, 7.0, and 7.5), 10 mM citrate (pH 5.0, 5.5, 6.0, and 6.5), and 10 mM sodium succinate (pH 5.0, 5.5, 6.0, and 6.5). The higher the pH, the greater the T_m1 is in all buffering substances. For each buffer at its highest pH, protein stability increased in the order of citrate < sodium succinate < histidine ~ sodium phosphate. (D) T_m1 values from (C) were transformed into ΔT_m for better visualization of the dependencies of mAb1 stability on the formulation by subtracting a T_m1 obtained for mAb1 in a reference formulation (10 mM sodium phosphate, pH 5.9, 150 mM sodium chloride) from a T_m1 found in each tested formulation. Error bars represent standard deviations of mean values determined from 3 repeats.

The pH of the formulation also had a substantial influence on the thermal stability of mAb1, especially its C_H2 domain. At 1 mg/mL protein concentration in 10 mM sodium phosphate, T_m1 increased by 6.2 ± 0.1 °C and T_m2 decreased by 0.8 ± 0.1 °C as pH increased from 5.0 to 7.5 ( Fig. 3B and Suppl. Fig. S3C). Formulation’s pH had the largest effect on the thermal stability of mAb1 at pH values lower than pH 6.5. Formulation pH values higher than pH 6.5 had only incremental influence, and mAb1 stability seemed to plateau at pH levels higher than 7.0. Within the pH range from 6.0 to 6.5, sodium phosphate and sodium succinate seemed to provide more thermostability than citrate and histidine buffer ( Fig. 3C ). To better visualize the dependencies of the mAb1 stability on the pH and the type of the buffering substance in the formulations, a ΔT_m plot was constructed by subtracting the T_m1 values obtained for mAb1 in the reference formulation (10 mM sodium phosphate, pH 5.9, 150 mM sodium chloride) from the T_m1 obtained in each tested formulation ( Fig. 3D ). The detrimental effects of low pH on an antibody thermal stability has been reported by others with the use of calorimetry and other protein analysis methods, and attributed to the increased susceptibility of protein to aggregate at low pH.^4–6

The effect of excipients such as 0.01% PS80, 10% sucrose, or both on the thermostability of mAb1 is summarized in Supplemental Figure S4. The C_H2 and C_H3–Fab transitions are clearly visible in all tested formulations (Suppl. Fig. S4A). The addition of excipients to mAb1 in 10 mM sodium phosphate, pH 5.8, stabilized the conformation of all domains of the antibody, shifting their T_ms to higher values. This effect was apparent not only in sodium phosphate buffer but also in formulations containing histidine, sodium succinate, and citrate (Suppl. Fig. S4B). ΔT_m values were determined by subtracting the T_ms obtained in 10 mM sodium phosphate, pH 5.9, 150 mM sodium chloride, from T_ms obtained in formulations with additional buffering substances and excipients. Sucrose and sucrose with PS80 supported higher thermal stability of mAb1 better than buffer alone or buffer with PS80 (Suppl. Fig. S4B). Excipients like sugars and surfactants are thought to work by stabilizing the native state of the protein, therefore protecting the antibody from physical degradation pathways such as denaturation, aggregation, precipitation, and surface adsorption.^1,3,4 This protective effect results from various types of protein surface solute interactions. ⁴ For example, those agents, on binding, might prevent attainment of conformations that are prone to aggregate formation or reduce the frequency of surface contacts that promote aggregation.^3,4 The stabilizing effects of sucrose and PS80 revealed in our study are consistent with previous studies performed with the use of DSF and other techniques.^1–3,5,6,15 Therefore, the parameters determined with the use of DSF can predict the effects of formulation components on the conformational stability of mAb and on their propensity to form aggregates. With the use of real-time PCR instrumentation, DSF can be conveniently used for the parallel screening of multiple formulation conditions. For example, a 96-well plate setup enables the testing of 32 conditions in triplicate in less than 75 min, with a rate of temperature increase of 1 °C per minute. Experimental throughput in combination with minimal protein consumption facilitate the rapid screening of different candidate molecules and formulation conditions early in the biologics discovery process and throughout development.

Extension of DSF Assay Throughput

The throughput of antibody formulation screening using DSF could be further increased by use of a 384-well plate format. With a 384-well plate setup, up to 128 conditions can be simultaneously screened in triplicate, in less than 75 min, without compromising the sensitivity and reproducibility of the assay. Figure 4 shows correlations between T_m1 and T_m2 measured for mAb1 in 17 different formulations by DSF in 96-well and 384-well plates. Formulations components and the position of the samples in plates are depicted in Supplemental Table S2 and Figure S5, respectively. Melting temperatures determined in both assay formats correlated very well (R² = 0.89 and 0.73 for T_m1 and T_m2, respectively) with the average difference in the T_m values obtained, with the 2 formats being within experimental error of the technique (0.5 °C to 1 °C for multiple independent determinations among multiple days). An additional increase in the assay throughput could be accomplished by automation of the reagent handling and integration of the ViiA™ 7 with an Applied Biosystems Twister II Robot plate stacker. High-throughput capabilities are of great importance because of the number of buffer components that need to be analyzed before the best formulation for supporting mAb stability is identified. Additives that stabilize mAb under a particular condition might not support stability in combination with other components. Therefore, formulation development is typically a long and iterative process.

Figure 4.

Correlation between melting temperatures (T_ms) measured with the use of differential scanning fluorimetry (DSF) in 96-well and 384-well plate formats. Seventeen formulations (Suppl. Table S2) were analyzed in both formats. The exact positions of the samples on 96-well and 384-well plates are depicted in Supplemental Figure S5A and S5B, respectively. The correlation between T_ms obtained in both formats is very good (R² = 0.89 and 0.73 for T_m1 and T_m2, respectively).

In summary, our study ratifies that DSF is a viable tool for antibody characterization, offering the advantages of rapid method development, a simple format, high-throughput, and minimal sample consumption. The parameters measured by DSF correlate well with those from more established techniques like DSC. DSF can be used productively at multiple stages in the mAb development process, from the screening of early leads in search of the most robust scaffold to the late stage of formulation optimization.