(M1030-03-19) In Silico Modeling Helps Predict Antibody Solution Behavior with Charged Cargo Moieties

Purpose: Interactions between antibodies in solution are partially driven by their surface properties such as surface charge and hydrophobicity patterns. In silico modeling tools can provide an early screening assessment of antibody surface properties that may estimate the propensity to aggregate in the presence of other charged cargo moieties, such as peptides, nucleic acids, of opposite charge to the antibody surface patches.

Methods: We used in silico modeling to predict the surface properties of 8-10 monoclonal antibodies (mAbs) having varying isoelectric point (pI) values. All mAb structures were obtained using homology modeling using ‘Model Full Length Antibody Structure’ in BIOVIA Discovery Studio 2020 software. All IgG4 mAb sequences were grafted on 5dk3.pdb. IgG1 sequences were grafted using a modified structure of 1hzh.pdb to incorporate the disulfide bonds in the hinge region which are not present in the original X-ray crystallography structure. The modified 1hzh structure was then energy minimized and equilibrated for 70 ns using molecular dynamics (MD) simulations in the NPT ensemble, where the number of particles (N), pressure (P) and temperature (T) are maintained constant during the simulation. All mAb homology modeled structures were then energy minimized and equilibrated using MD for 5 ns to equilibrate the side chain orientation. The energy minimization and the equilibration steps were performed using Gromacs 2020.2. The surface properties were then obtained on the final conformation from the MD simulation using ‘Calculate Protein Formulation Properties’ in Discovery Studio software. Surface charge was obtained using ‘Calculate Protein Ionization and Residue pK’ and ‘Calculate Electrostatic Potential’ at pH 6.0, 7.4 and 8.0 again in Discovery Studio software. In silico results were compared to experimental data for solution solubility observations with charged cargo moieties at different pH conditions (pH 5 – 11).

Results: MAbs selected for this study have their theoretical sequence-based pI values calculated using Henderson-Hasselbalch acid dissociation model[1] ranging from ~7 – 10. In silico results were compared with experimental solutions of each mAb with a negatively charged biological moiety. MAbs that displayed higher net positive surface charge (mean net charge of mAbs at pH 6.0 ~42e, at pH 7.4 ~25e, at pH 8.0 ~17e) and higher dipole moment (mean dipole moment at all pH ranges from 1800 to 2600 D), referred to as group A mAbs, displayed precipitation when mixed, in a wet lab, with another negatively charged biological moiety in solution. This precipitation was observed across a range of different pH conditions (pH 5 – 11). These mAbs also had generally higher theoretical sequence-based pI values (range 9.2 – 9.5) and 3D structure-based pI calculated in Discovery Studio (range 9.0 – 9.4). MAbs that had lower theoretical pI (7.0 – 7.8) and pI based on 3D structure from Discovery Studio (range 8.6 – 8.7), displayed generally lower net surface charge (mean net charge at pH 6.0 ~32e, at pH 7.4 ~15e, at pH 8.0 ~8e) and a lower dipole moment (mean dipole moment at all pH ranges from 650 to 1100 D), referred to as group C mAbs. One mAb (group B) exhibited a weaker correlation with the trend of higher pI values precipitating with negatively charged moieties, which may be explained by its in silico calculated surface properties. The lower dipole of intermediate-case (group B) provides a rationale for why it did not precipitate. This mAb demonstrated a higher net surface charge of 37.5e at pH 6.0 as compared to the other two group C mAbs that did not precipitate (mean ~32e) but also lower than those group A mAbs that did precipitate (mean ~42e) at pH 6.0.  In silico analysis determined that the dipole of this group B mAb at all pH was ranging from 230 – 540 D, which is significantly lower than those that precipitated (group A). Figure 1 displays a representative surface charge map for each group as described above based on if they precipitated or not when mixed with negatively charged moieties. The lower dipole moment indicates that the surface charge patches are distributed more evenly over the antibody surface. This distribution results in a more neutral overall surface that enhances the stability of the antibody when interacting with negatively charged moieties.  In contrast, antibodies that precipitated exhibited greater segregation of positive and negative charges (dipole moment) on their surfaces, creating surface patches that are prone to aggregation with oppositely charged moieties.

Conclusion: Early-stage computational screening serves as a valuable predictive tool for assessing the critical balance of net surface charge and dipole moment, enabling the identification of antibodies that will remain stable in solution when co-formulated or conjugated with oppositely charged moieties, while distinguishing those prone to aggregation and precipitation.

References: [1] Amino Acids, the Henderson-Hasselbalch Equation, and Isoelectric Points. (2024, September 30).

Surface charge map for a representative monoclonal antibody (mAb) from each group based on their net charge and dipole moment properties and if they precipitated when mixed, in a wet lab, with another negatively charged biological moiety in solution. Red and blue represent negatively and positively charged patches, respectively.