Machine learning prediction of antibody aggregation and viscosity for high concentration formulation development of protein therapeutics.

Machine learning has been recently used to predict therapeutic antibody aggregation rates and viscosity at high concentrations (150 mg/ml). These works focused on commercially available antibodies, which may have been optimized for stability. In this study, we measured accelerated aggregation rates at 45°C and viscosity at 150 mg/ml for 20 preclinical and clinical-stage antibodies. Features obtained from molecular dynamics simulations of the full-length antibody and sequences were used for machine learning model construction. We found a k-nearest neighbors regression model with two features, spatial positive charge map on the CDRH2 and solvent-accessible surface area of hydrophobic residues on the variable fragment, gives the best performance for predicting antibody aggregation rates (r = 0.89). For the viscosity classification model, the model with the highest accuracy is a logistic regression model with two features, spatial negative charge map on the heavy chain variable region and spatial negative charge map on the light chain variable region. The accuracy and the area under precision recall curve of the classification model from validation tests are 0.86 and 0.70, respectively. In addition, we combined data from another 27 commercial mAbs to develop a viscosity predictive model. The best model is a logistic regression model with two features, number of hydrophobic residues on the light chain variable region and net charges on the light chain variable region. The accuracy and the area under precision recall curve of the classification model are 0.85 and 0.6, respectively. The aggregation rates and viscosity models can be used to predict antibody stability to facilitate pharmaceutical development.

Machine Learning Models of Antibody-Excipient Preferential Interactions for Use in Computational Formulation Design.

Preferential interactions of formulation excipients govern their impact on the stability properties of proteins in solution. The ability to predict these interactions without the need to perform experiments would enable formulation design to begin early in the development of a new antibody therapeutic. With that in mind, we developed a feature set to numerically describe local regions of an antibody's surface for use in machine learning applications. Then, we used these features to train machine learning models for local antibody-excipient preferential interactions for the excipients sorbitol, sucrose, trehalose, proline, arginine·HCl, and NaCl. Our models had accuracies of up to about 85%. We also used linear (elastic net) models to quantify the contribution of antibody surface features to the preferential interaction coefficients, finding that the carbohydrates and proline tend to have similar important features, while the interactions of arginine·HCl and NaCl are governed by charge features. We present several case studies demonstrating how these machine learning models could be used to predict experimental aggregation and viscosity behavior in solution. Finally, we propose an approach to computational formulation design wherein a panel of excipients may be considered while designing an antibody sequence.

Identifying Key Residues That Drive Strong Electrostatic Attractions between Therapeutic Antibodies.

Attractive electrostatic protein-protein interactions (PPI) necessarily involve identifying oppositely charged regions of the protein surface that interact favorably. This cannot be done reliably if one only considers a single protein in isolation unless there are obvious charge "patches" that result in extreme molecular dipoles. Prior work [ J. Pharm. Sci. 2019 , 108 , 120 - 132 ] identified three monoclonal antibodies (MAbs) that displayed experimental behavior ranging from net repulsive to strongly attractive electrostatic interactions. The present work provides a systematic computational approach for identifying the origin of diverse PPI, in terms of which sets of amino acids or individual amino acids are most influential, and determining if there are different patterns of pairwise amino acid interaction "maps" that result in different behaviors. The charge was eliminated computationally, one by one, for each charged residue in the wild-type sequences, which resulted in predicted changes in the second osmotic virial coefficient. The results highlight interaction "maps" that correspond to cases with qualitatively different net electrostatic PPI for the different MAbs and solution conditions, as well as key sets of residues that contribute to strongly attractive PPI. A more computationally efficient method is also proposed to identify key amino acids based on Mayer-weighted interaction energies.

Molecular Computations of Preferential Interaction Coefficients of IgG1 Monoclonal Antibodies with Sorbitol, Sucrose, and Trehalose and the Impact of These Excipients on Aggregation and Viscosity.

Preferential interactions of formulation excipients govern their overall interactions with protein molecules, and molecular dynamics simulations allow for the examination of the interactions at the molecular level. We used molecular dynamics simulations to examine the interactions of sorbitol, sucrose, and trehalose with three different IgG1 antibodies to gain insight into how these excipients impact aggregation and viscosity. We found that sucrose and trehalose reduce aggregation more than sorbitol because of their larger size and their stronger interactions with high-spatial aggregation propensity residues compared to sorbitol. Two of the antibodies had high viscosity in sodium acetate buffer, and for these, we found that sucrose and trehalose tended to have opposite effects on viscosity. The data presented here provide further insight into the mechanisms of interactions of these three carbohydrate excipients with the antibody surface and thus their impact on excipient stabilization of antibody formulations.

Understanding the Role of Preferential Exclusion of Sugars and Polyols from Native State IgG1 Monoclonal Antibodies and its Effect on Aggregation and Reversible Self-Association.

PURPOSE: To investigate differences in the preferential exclusion of trehalose, sucrose, sorbitol and mannitol from the surface of three IgG1 monoclonal antibodies (mAbs) and understand its effect on the aggregation and reversible self-association of mAbs at high-concentrations. METHODS: Preferential exclusion was measured using vapor pressure osmometry. Effect of excipient addition on accelerated aggregation kinetics was quantified using size exclusion chromatography and on reversible self-association was quantified using dynamic light scattering. RESULTS: The doubling of excipient concentration in the 0 to 0.5 m range resulted in a doubling of the mAb transfer free energy for all excipients and antibodies tested in this study. Solution pH and choice of buffering agent did not significantly affect the magnitude of preferential exclusion. We find that aggregation suppression for trehalose, sucrose and sorbitol (but not mannitol) correlates with the magnitude of their preferential exclusion from the native state of the three IgG1 mAbs. We also find that addition of sugars and polyols reduced the tendency for reversible self-association in two mAbs that had weakly repulsive or neutral self-interactions in the presence of buffer alone. CONCLUSIONS: The magnitude of preferential exclusion for trehalose, sucrose and sorbitol correlates well with their partial molar volumes in solution. Mannitol is excluded to a greater extent than that expected from its partial molar volume, suggesting specific interactions of mannitol that might be different than the other sugars and polyols tested in this study. Local interactions play a role in the effect of excipient addition on the reversible self-association of mAbs. These results provide further insights into the stabilization of high-concentration mAb formulations by sugars and polyols.

Electrostatically Mediated Protein-Protein Interactions for Monoclonal Antibodies: A Combined Experimental and Coarse-Grained Molecular Modeling Approach.

Electrostatically mediated protein-protein interactions (PPI) can influence key product properties such as solubility, solution viscosity, and aggregation rates. Predictive models would allow for candidates/formulations to be screened with little or no protein material. Three monoclonal antibodies that display qualitatively different experimental PPI were evaluated at a range of pH and ionic strength conditions that are typical of product formulations. PPI parameters (kD, B22, and G22) were obtained from static and dynamic light scattering measurements and spanned from strongly repulsive to strongly attractive net interactions. Coarse-grained (CG) molecular simulations of PPI (specifically, B22) were compared against experimental PPI parameters across multiple pH and salt conditions, using a CG model that treats each amino acid explicitly. Predicted B22 values with default model parameters matched experimental B22 values semiquantitatively for some cases; others required parameter tuning to account for effects such as ion binding. Experimental PPI values were also analyzed for each monoclonal antibody within the context of single-protein properties such as net charge, and domain-based and global dipole moments. The results show that PPI predicted qualitatively and semiquantitatively by CG molecular modeling of B22 can be an effective computational tool for molecule and formulation assessment.

Kirkwood-Buff-Derived Alcohol Parameters for Aqueous Carbohydrates and Their Application to Preferential Interaction Coefficient Calculations of Proteins.

The CHARMM36 carbohydrate parameter set did not adequately reproduce experimental thermodynamic data of carbohydrate interactions with water or proteins or carbohydrate self-association; thus, a new nonbonded parameter set for carbohydrates was developed. The parameters were developed to reproduce experimental Kirkwood-Buff integral values, defined by the Kirkwood-Buff theory of solutions, and applied to simulations of glycerol, sorbitol, glucose, sucrose, and trehalose. Compared to the CHARMM36 carbohydrate parameters, these new Kirkwood-Buff-based parameters reproduced accurately carbohydrate self-association and the trend of activity coefficient derivative changes with concentration. When using these parameters, preferential interaction coefficients calculated from simulations of these carbohydrates and the proteins lysozyme, bovine serum albumin, α-chymotrypsinogen A, and RNase A agreed well with the experimental data, whereas use of the CHARMM36 parameters indicated preferential inclusion of carbohydrates, in disagreement with the experiment. Thus, calculating preferential interaction coefficients from simulations requires using a force field that accurately reproduces trends in the thermodynamic properties of binary excipient-water solutions, and in particular the trend in the activity coefficient derivative. Finally, the carbohydrate-protein simulations using the new parameters indicated that the carbohydrate size was a major factor in the distribution of different carbohydrates around a protein surface.

Charge-mediated Fab-Fc interactions in an IgG1 antibody induce reversible self-association, cluster formation, and elevated viscosity.

Concentration-dependent reversible self-association (RSA) of monoclonal antibodies (mAbs) poses a challenge to their pharmaceutical development as viable candidates for subcutaneous delivery. While the role of the antigen-binding fragment (Fab) in initiating RSA is well-established, little evidence supports the involvement of the crystallizable fragment (Fc). In this report, a variety of biophysical tools, including hydrogen exchange mass spectrometry, are used to elucidate the protein interface of such non-covalent protein-protein interactions. Using dynamic and static light scattering combined with viscosity measurements, we find that an IgG1 mAb (mAb-J) undergoes RSA primarily through electrostatic interactions and forms a monomer-dimer-tetramer equilibrium. We provide the first direct experimental mapping of the interface formed between the Fab and Fc domains of an antibody at high protein concentrations. Charge distribution heterogeneity between the positively charged interface spanning complementarity-determining regions CDR3H and CDR2L in the Fab and a negatively charged region in CH3/Fc domain mediates the RSA of mAb-J. When arginine and NaCl are added, they disrupt RSA of mAb-J and decrease the solution viscosity. Fab-Fc domain interactions between mAb monomers may promote the formation of large transient antibody complexes that ultimately cause increases in solution viscosity. Our findings illustrate how limited specific arrangements of amino-acid residues can cause mAbs to undergo RSA at high protein concentrations and how conserved regions in the Fc portion of the antibody can also play an important role in initiating weak and transient protein-protein interactions.

Molecular Investigation of the Mechanism of Non-Enzymatic Hydrolysis of Proteins and the Predictive Algorithm for Susceptibility.

A number of potential degradation routes can limit the shelf life of a biotherapeutic. While these are experimentally measurable, the tests to do so require a substantial investment in both time and material, resources rarely available early in the drug development process. To address the potential degradation route of non-enzymatic hydrolysis, we performed a molecular modeling analysis, together with an experimental study, to gain detailed insight into the reaction. On the basis of the mechanism, an algorithm for predicting the likely cleavage sites of a protein has been created. This algorithm measures four key properties during a molecular dynamics simulation, which relate to the key steps of the hydrolysis mechanism, in particular the rate-determining step (which can vary depending on the local environment). The first two properties include the secondary structure and the surface exposure of the amide bond, both of which help detect if the addition of the proton to the amide bond is possible. The second two properties relate to whether the side chain can cyclize and form a furane ring. These two properties are the orientation of the side chain relative to the amide bond and the number of hydrogen bonds between the side chain and the surrounding protein. Overall, the algorithm performs well at identifying reactive versus nonreactive bonds. The algorithm correctly classifies nearly 90% of all amide bonds following an aspartic or glutamic acid residue as reactive or nonreactive.

Mitigation of reversible self-association and viscosity in a human IgG1 monoclonal antibody by rational, structure-guided Fv engineering.

Undesired solution behaviors such as reversible self-association (RSA), high viscosity, and liquid-liquid phase separation can introduce substantial challenges during development of monoclonal antibody formulations. Although a global mechanistic understanding of RSA (i.e., native and reversible protein-protein interactions) is sufficient to develop robust formulation controls, its mitigation via protein engineering requires knowledge of the sites of protein-protein interactions. In the study reported here, we coupled our previous hydrogen-deuterium exchange mass spectrometry findings with structural modeling and in vitro screening to identify the residues responsible for RSA of a model IgG1 monoclonal antibody (mAb-C), and rationally engineered variants with improved solution properties (i.e., reduced RSA and viscosity). Our data show that mutation of either solvent-exposed aromatic residues within the heavy and light chain variable regions or buried residues within the heavy chain/light chain interface can significantly mitigate RSA and viscosity by reducing the IgG's surface hydrophobicity. The engineering strategy described here highlights the utility of integrating complementary experimental and in silico methods to identify mutations that can improve developability, in particular, high concentration solution properties, of candidate therapeutic antibodies.

Computational tool for the early screening of monoclonal antibodies for their viscosities.

Highly concentrated antibody solutions often exhibit high viscosities, which present a number of challenges for antibody-drug development, manufacturing and administration. The antibody sequence is a key determinant for high viscosity of highly concentrated solutions; therefore, a sequence- or structure-based tool that can identify highly viscous antibodies from their sequence would be effective in ensuring that only antibodies with low viscosity progress to the development phase. Here, we present a spatial charge map (SCM) tool that can accurately identify highly viscous antibodies from their sequence alone (using homology modeling to determine the 3-dimensional structures). The SCM tool has been extensively validated at 3 different organizations, and has proved successful in correctly identifying highly viscous antibodies. As a quantitative tool, SCM is amenable to high-throughput automated analysis, and can be effectively implemented during the antibody screening or engineering phase for the selection of low-viscosity antibodies.

A systematic multitechnique approach for detection and characterization of reversible self-association during formulation development of therapeutic antibodies.

In addition to controlling typical instabilities such as physical and chemical degradations, understanding monoclonal antibodies' (mAbs) solution behavior is a key step in designing and developing process and formulation controls during their development. Reversible self-association (RSA), a unique solution property in which native, reversible oligomeric species are formed as a result of the noncovalent intermolecular interactions has been recognized as a developability risk with the potential to negatively impact manufacturing, storage stability, and delivery of mAbs. Therefore, its identification, characterization, and mitigation are key requirements during formulation development. Considering the large number of available analytical methods, choice of the employed technique is an important contributing factor for successful investigation of RSA. Herein, a multitechnique (dynamic light scattering, multiangle static light scattering, and analytical ultracentrifugation) approach is employed to comprehensively characterize the self-association of a model immunoglobulin G1 molecule. Studies herein discuss an effective approach for detection and characterization of RSA during biopharmaceutical development based on the capabilities of each technique, their complementarity, and more importantly their suitability for the stage of development in which RSA is investigated.

Effects of salts from the Hofmeister series on the conformational stability, aggregation propensity, and local flexibility of an IgG1 monoclonal antibody.

This work examines the effect of three anions from the Hofmeister series (sulfate, chloride, and thiocyanate) on the conformational stability and aggregation rate of an IgG1 monoclonal antibody (mAb) and corresponding changes in the mAb's backbone flexibility (at pH 6 and 25 °C). Compared to a 0.1 M NaCl control, thiocyanate (0.5 M) decreased the melting temperatures (Tm) for three observed conformational transitions within the mAb by 6-9 °C, as measured by differential scanning calorimetry. Thiocyanate also accelerated the rate of monomer loss at 40 °C over 12 months, as monitored by size exclusion chromatography. Backbone flexibility, as measured via H/D exchange mass spectrometry, increased in two segments in the CH2 domain with more subtle changes across several additional regions. Chloride (0.5 M) caused slight increases in the Tm values, small changes in aggregation rate, and minimal yet consistent decreases in flexibility across various domains with larger effects noted within the VL, CH1, and CH3 domains. In contrast, 0.5 M sulfate increased Tm values, had small stabilizing influences on aggregate formation over time, yet substantially increased the flexibility of two specific regions in the CH1 and VL domains. While thiocyanate-induced conformational destabilization of the mAb correlated with increased local flexibility of specific regions in the CH2 domain (especially residues 241-251 in the heavy chain), the stabilizing anion sulfate did not affect these CH2 regions.

Correlating excipient effects on conformational and storage stability of an IgG1 monoclonal antibody with local dynamics as measured by hydrogen/deuterium-exchange mass spectrometry.

The effects of sucrose and arginine on the conformational and storage stability of an IgG1 monoclonal antibody (mAb) were monitored by differential scanning calorimetry (DSC) and size-exclusion chromatography (SEC), respectively. Excipient effects on protein physical stability were then compared with their effects on the local flexibility of the mAb in solution at pH 6, 25°C using hydrogen/deuterium-exchange mass spectrometry (H/D-MS). Compared with a 0.1 M NaCl control, sucrose (0.5 M) increased conformational stability (T(m) values), slowed the rate of monomer loss, reduced the formation of insoluble aggregates, and resulted in a global trend of small decreases in local flexibility across most regions of the mAb. In contrast, the addition of arginine (0.5 M) decreased the mAb's conformational stability, increased the rate of loss of monomer with elevated levels of soluble and insoluble aggregates, and led to significant increases in the local flexibility in specific regions of the mAb, most notably within the constant domain 2 of the heavy chain (C(H)2). These results provide new insights into the effect of sucrose and arginine on the local dynamics of IgG1 domains as well as preliminary correlations between local flexibility within specific segments of the C(H)2 domain (notably heavy chain 241-251) and the mAb's overall physical stability.

A systematic multitechnique approach for detection and characterization of reversible self-association during formulation development of therapeutic antibodies.

In addition to controlling typical instabilities such as physical and chemical degradations, understanding monoclonal antibodies' (mAbs) solution behavior is a key step in designing and developing process and formulation controls during their development. Reversible self-association (RSA), a unique solution property in which native, reversible oligomeric species are formed as a result of the noncovalent intermolecular interactions has been recognized as a developability risk with the potential to negatively impact manufacturing, storage stability, and delivery of mAbs. Therefore, its identification, characterization, and mitigation are key requirements during formulation development. Considering the large number of available analytical methods, choice of the employed technique is an important contributing factor for successful investigation of RSA. Herein, a multitechnique (dynamic light scattering, multiangle static light scattering, and analytical ultracentrifugation) approach is employed to comprehensively characterize the self-association of a model immunoglobulin G1 molecule. Studies herein discuss an effective approach for detection and characterization of RSA during biopharmaceutical development based on the capabilities of each technique, their complementarity, and more importantly their suitability for the stage of development in which RSA is investigated.

Local dynamics and their alteration by excipients modulate the global conformational stability of an lgG1 monoclonal antibody.

A molecular understanding of excipient effects on the interrelationship(s) between dynamics and conformational stability of proteins, such as monoclonal antibodies (mAbs), can be important for their pharmaceutical development. The current study examines stabilizing and destabilizing effects of excipients on the conformational stability and local dynamics of distinct solvent-exposed regions within an IgG1 monoclonal antibody (mAb-B). The principles of site-selective photoselection upon red-edge excitation, accompanied by acrylamide quenching of tryptophan fluorescence were employed in this study. The initiation of mAb-B thermal unfolding occurs by structural alterations in the more solvent-exposed regions of the antibody, which subsequently leads to a cascade of structural alterations in its relatively more solvent-shielded regions. In addition, an increase in internal dynamics of solvent-shielded regions made mAb-B more susceptible to thermally induced structural perturbations resulting in its global destabilization. Sucrose and arginine exert their stabilizing and destabilizing effects by predominantly influencing the conformational stability of solvent-exposed regions in mAb-B. The complex molecular effects of sucrose and arginine on local dynamics of different regions in mAb-B and their correlation with the protein's conformational stability are described within the pretransition range, at the onset temperature (T(onset)) and at the thermal melting temperature (T(M)).

Excipients differentially influence the conformational stability and pretransition dynamics of two IgG1 monoclonal antibodies.

Since immunoglobulins are conformationally dynamic molecules in solution, we studied the effect of stabilizing and destabilizing excipients on the conformational stability and dynamics of two IgG1 monoclonal antibodies (mAbs; mAb-A and mAb-B) using a variety of biophysical approaches. Even though the two mAbs are of the same IgG1 subtype, the unfolding patterns, aggregation behavior, and pretransition dynamics of these two antibodies were strikingly different in response to external perturbations such as pH, temperature, and presence of excipients. Sucrose and arginine were identified as stabilizers and destabilizers, respectively, on the basis of their influence on conformational stability for both the IgG1 mAbs. The two excipients, however, had distinct effective concentrations and different effects on the conformational stability and pretransition dynamics of the two mAbs as measured by a combination of differential scanning calorimetry, high-resolution ultrasonic spectroscopy, and red-edge excitation shift fluorescence studies. Stabilizing concentrations of sucrose were found to decrease the internal motions of mAb-B, whereas arginine marginally increased its adiabatic compressibility in the pretransition region. Both sucrose and arginine did not influence the pretransition dynamics of mAb-A. The potential reasons for such differences in excipient effects between two IgG1 mAbs are discussed.

Formulation development of therapeutic monoclonal antibodies using high-throughput fluorescence and static light scattering techniques: role of conformational and colloidal stability.

In this work, we describe the application of two different high-throughput screening (HTS) techniques that can be used to determine protein stability during early formulation development. Differential scanning fluorescence (DSF) and differential static light scattering (DSLS) are used to determine the conformational and colloidal stability of therapeutic monoclonal antibodies (mAbs) during thermal denaturation in a high-throughput fashion. DSF utilizes SYPRO Orange, a polarity-sensitive extrinsic fluorescent probe, to monitor protein unfolding. We found that melting temperatures determined by DSF have a linear correlation with melting temperatures of the first domain unfolding determined by differential scanning calorimetry, establishing DSF as a reliable method for measuring thermal stability. The DSLS method employs static light scattering to evaluate protein stability during thermal denaturation in a 384-well format. Overall comparison between mAb aggregation under typical accelerated stress conditions (40°C) and the thermal stability obtained by DSF and DSLS is also presented. Both of these HTS methods are cost effective with high-throughput capability and can be implemented in any laboratory. Combined with other emerging HTS techniques, DSF and DSLS could be powerful tools for mAb formulation optimization.

Buffer-dependent fragmentation of a humanized full-length monoclonal antibody.

During storage stability studies of a monoclonal antibody (mAb) it was determined that the primary route of degradation involved fragmentation into lower molecular weight species. The fragmentation was characterized with size-exclusion high performance liquid chromatography (SE-HPLC), SDS-PAGE, and matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) mass spectrometry. Fragmentation proceeded via hydrolysis, likely catalyzed by trace metal ions, of a peptide bond in the hinge region of the mAb's heavy chain, which produced two prominent low molecular weight species during storage: a single, free Fab fragment and a Fab + Fc fragment. The fragmentation is observed in phosphate-buffered solutions at two ionic strengths but not in histidine-buffered solutions at identical ionic strengths. Chaotrope-induced and thermally induced unfolding studies of the mAb indicated differences in the unfolding pathways between the two buffer solutions. The folding intermediate observed during chaotrope-induced unfolding was further characterized by intrinsic fluorescence quenching, which suggested that a small portion of the molecule is resistant to chaotrope-induced unfolding in histidine buffer systems. The thermally induced unfolding indicates a reduction in cooperativity of the unfolding process in the presence of histidine relative to phosphate. A relationship between the histidine-induced effects on unfolding pathway and the relative resistance to fragmentation is suggested.

Understanding and modulating opalescence and viscosity in a monoclonal antibody formulation.

Opalescence and high viscosities can pose challenges for high concentration formulation of antibodies. Both phenomena result from protein-protein intermolecular interactions that can be modulated with solution ionic strength. We studied a therapeutic monoclonal antibody (mAb) that exhibits high viscosity in solutions at low ionic strength ( approximately 20 cP at 90 mg/mL and 23 degrees C) and significant opalescence at isotonic ionic strength (approximately 100 nephelometric turbidity units at 90 mg/mL and 23 degrees C). The intermolecular interactions responsible for these effects were characterized using membrane osmometry, static light scattering, and zeta potential measurements. The net protein-protein interactions were repulsive at low ionic strength ( approximately 4 mM) and attractive at isotonic ionic strengths. The high viscosities are attributed to electroviscous forces at low ionic strength and the significant opalescence at isotonic ionic strength is correlated with attractive antibody interactions. Furthermore, there appears to be a connection to critical phenomena and it is suggested that the extent of opalescence is dependent on the proximity to the critical point. We demonstrate that by balancing the repulsive and attractive forces via intermediate ionic strengths and by increasing the mAb concentration above the apparent critical concentration both opalescence and viscosity can be simultaneously minimized.