(M1230-01-02) Additivity of Transfer Free Energies Enables Description of Complex Protein Formulations in Implicit Solvent Molecular Dynamics Simulations

Purpose: Proteins’ function can be lost upon unfolding or aggregation, and these unwanted phenomena can often be mitigated by the addition of protein-stabilizing osmolytes.^{1, 2} Understanding the interactions between proteins and osmolytes is therefore essential to develop stable protein drug products.³ In this context, the concept of group transfer free energy (GTFE), which represents the change in chemical potential of a protein being transferred from water to an osmolyte solution, has emerged as a powerful tool to compute the energetics involved in protein-osmolyte interaction.^{4, 5} We experimentally determined the GTFEs for the excipients sodium chloride (NaCl), L-arginine hydrochloride (ArgHCl) and polysorbate 20 (PS20), which are frequently used in pharmaceutical protein formulations. Furthermore, we investigated if the free energy cost for transferring a protein to a multi-component formulation can be calculated by summing up the contributions of the individual components in an additive framework. Additionally, we applied GTFEs within implicit solvent molecular dynamics (MD) calculations of granulocyte colony-stimulating factor (G-CSF), for both single- and multi-component matrices. The MD simulation results were validated via a short-term stability study.

Methods: GTFEs were determined based on the change in solubility between water and the selected osmolyte solution. The solubility of all the 20 standard proteinogenic amino acids, and of cyclic glycylglycine (CGG) as backbone model, was determined in pure water, 1 M NaCl, 1 M ArgHCl, and 0.04% (w/v) PS20 solutions via density measurements, as described in Auton et al.⁴ In addition, additivity of GTFEs in multi-component systems was also investigated via density measurements for selected amino acids in mixtures of sucrose, NaCl, ArgHCl, and PS20 (Figure 1). To verify our newly determined GTFEs, and their additivity in multi-component matrices, differential scanning microcalorimetry (µDSC) was employed for 1 mg/mL RNaseA in 10 mM histidine buffer pH 6.0 in presence of these excipients, and their mixtures.
The implicit solvent MD simulations were conducted using the AMBER 22 simulation suite⁶ applying a modified GTFE-MIST approach^{7, 8}, and the ff14SBonlysc force field⁹. G-CSF (pdb 1CD9¹⁰) was simulated at 37 °C, at pH 3.5 and pH 7.0 in different formulations including sucrose, NaCl, ArgHCl, and their mixtures. For validation of the MD simulation results, experimental data were collected in a short-term stability study of the G-CSF formulations for up to 11 days at 37 °C, using nano differential scanning fluorimetry (nanoDSF), and backgrounded membrane imaging (BMI).

Results: Therapeutic protein formulations typically contain multiple excipients that accomplish different functions, such as cryoprotectants, tonifiers, or surfactants.¹¹ Determining GTFEs for excipient mixtures in protein formulations would require extensive work, if solubility measurements had to be repeated for every matrix. In order to address this problem, we tried to verify whether GTFEs for multi-component matrices can be computed by simply adding up the contributions of the separate components, in a linear, additive framework. A very good correlation (Pearson’s r = 0.97, Figure 1) was obtained between experimentally measured GTFE values for multi-component systems, and calculated GTFEs based on additivity considerations. This implies that, at least for the osmolytes studied in this work, GTFEs are additive. Our newly determined GTFE values, as well as their additivity, were validated by µDSC measurements of RNase A. GTFEs are thermodynamically linked to the (de)stabilizing effect of the osmolyte/ osmolytes combination, and a good correlation was indeed obtained between GTFEs and the shift in melting temperature as determined by µDSC (Pearson’s r = 0.97, data not shown). In order to further evaluate the potential of our newly measured GTFEs, MD simulations including the GTFEs in the calculation of the system potential energy were performed. Different G-CSF formulations were simulated using this innovative approach. The simulations indicated that aggregation of G-CSF occurs in presence of salt (NaCl or ArgHCl) or at pH 7, i.e., as indicated by the formation of hydrogen bonds between the G-CSF molecules (Figure 2 and Table 1). These findings were confirmed by the short-term stability data, i.e., by the particle burden obtained by BMI and by the changes in the T_m or T_agg observed in nanoDSF (Table 1), supported by literature data ^12–14.

Conclusion: In the present study, which is not yet published, GTFE values were measured for NaCl, ArgHCl, and PS20. Additionally, it was shown that GTFEs show an additive behavior in multi-component formulations and can therefore be obtained by summing up the contributions of the separate components in a simple, additive framework. We also showed that GTFEs can be incorporated within MD simulations. The MD predictions regarding protein (de)stabilization in different formulations were found to be in good agreement with the results of a short-term stability study performed for G-CSF. The presented approach has the potential to support costly formulation screenings by simulating the effects of osmolytes on protein stability in silico.

References: 1. Kuhlman B, Bradley P. Advances in protein structure prediction and design. 2019. Nat Rev Mol Cell Biol 20: 681–697.
2. Mahler H-C, Friess W, Grauschopf U, Kiese S. Protein aggregation: pathways, induction factors and analysis. 2009. J Pharm Sci 98: 2909–2934.
3. Kamerzell TJ, Esfandiary R, Joshi SB, et al. Protein-excipient interactions: mechanisms and biophysical characterization applied to protein formulation development. 2011. Advanced drug delivery reviews 63: 1118–1159.
4. Auton M, Bolen DW. Application of the transfer model to understand how naturally occurring osmolytes affect protein stability. 2007. Methods Enzymol 428: 397–418.
5. Auton M, Holthauzen LMF, Bolen DW. Anatomy of energetic changes accompanying urea-induced protein denaturation. 2007. Proceedings of the National Academy of Sciences of the United States of America 104: 15317–15322.
6. Pearlman DA, Case DA, Caldwell JW, et al. AMBER, a package of computer programs for applying molecular mechanics, normal mode analysis, molecular dynamics and free energy calculations to simulate the structural and energetic properties of molecules. 1995. Computer Physics Communications 91: 1–41.
7. Arsiccio A, Ganguly P, Shea J-E. A Transfer Free Energy Based Implicit Solvent Model for Protein Simulations in Solvent Mixtures: Urea-Induced Denaturation as a Case Study. 2022. The journal of physical chemistry. B 126: 4472–4482.
8. Pisano R, Arsiccio A, Collins V, et al. Understanding Glucagon Aggregation: In Silico Insights and Experimental Validation. 2024. Mol Pharm 21: 3815–3823.
9. Nguyen H, Maier J, Huang H, et al. Folding simulations for proteins with diverse topologies are accessible in days with a physics-based force field and implicit solvent. 2014. Journal of the American Chemical Society 136: 13959–13962.
10. Aritomi M, Kunishima N, Okamoto T, et al. Atomic structure of the GCSF-receptor complex showing a new cytokine-receptor recognition scheme. 1999. Nature 401: 713–717.
11. Gervasi V, Dall Agnol R, Cullen S, et al. Parenteral protein formulations: An overview of approved products within the European Union. 2018. Eur J Pharm Biopharm 131: 8–24.
12. Chi EY, Krishnan S, Kendrick BS, et al. Roles of conformational stability and colloidal stability in the aggregation of recombinant human granulocyte colony-stimulating factor. 2003. Protein Sci 12: 903–913.
13. Krishnan S, Chi EY, Webb JN, et al. Aggregation of granulocyte colony stimulating factor under physiological conditions: characterization and thermodynamic inhibition. 2002. Biochemistry 41: 6422–6431.
14. Ko SK, Berner C, Kulakova A, et al. Investigation of the pH-dependent aggregation mechanisms of GCSF using low resolution protein characterization techniques and advanced molecular dynamics simulations. 2022. Comput Struct Biotechnol J 20: 1439–1455.

Acknowledgements: The authors acknowledge the possibility to run simulations on the HPC of the University of Ghent, as well as the support team there for their help.

Figure 1. Additivity of GTFEs for multi-component solutions. Correlation between GTFEs obtained via linear combinations of single-component GTFEs vs. experimentally determined GTFEs. 1 M sucrose + 0.5 M NaCl is shown in black, 0.5 M sucrose + 0.5 M ArgHCl in red, 0.5 M sucrose + 0.5 M ArgHCl + 0.04% PS20 in blue, and 0.5 M NaCl + 0.5 M ArgHCl in green.

Figure 2. MD simulation results for the G-CSF formulation screening. The distributions of α-helical content and number of intermolecular hydrogen bonds are shown for different formulations at pH 3.5 and 7.0. Representative proteins conformation as extracted from the MD trajectories are also shown in each panel.

Table 1. G-CSF formulation screening results shown as heat map. Red indicates protein destabilization, whereas blue suggests protein stabilization. The first column (hydrogen bonds) summarizes the mean number of intermolecular hydrogen bonds from the MD simulations. The subsequent columns are the experimental data, i.e., the onset temperature of aggregation (Tagg), the apparent melting temperature (Tm) from nanoDSF, and the particle burden obtained by BMI at T0 and after 11 days incubation at 37 °C. For the simulation data (i.e., number of hydrogen bonds) errors were estimated by block averaging. Briefly, the equilibrated trajectories were divided into four blocks and the standard deviation computed over the average values in each of the blocks. For the experimental data, values are provided as mean ± standard deviation from n=3 measurements. Standard deviations for Tm and Tagg were for all measurements below 0.1 °C. Abbreviations: not detected (n.d.); not available (n.a).