The pH-sensing mechanism of antibody recycling by the neonatal Fc receptor revealed using free energy perturbation calculations

The immune system produces antibodies to recognize and provide protection against infection. The immunoglobulin G (IgG) antibody isotype is present at high serum concentrations and has a longer half-life than other isotypes due to the interaction between its fragment crystallizable (Fc) region with the neonatal Fc receptor (FcRn). This Fc-FcRn interaction, which takes place in many cell types throughout the cardiovascular system, mediates pH-dependent formation of the IgG-FcRn complex and leads to the rescue of IgG from eventual degradation via transport from the low-pH early endosome back to the cell surface for release into serum at pH 7.4. Because this process is the primary determinant of IgG antibody half-life, and because the Fc region is common to all antibodies of the same subtype, the Fc-FcRn system has been a target of numerous antibody design and engineering studies. Indeed, several engineered Fcs have been reported with extended serum half-lives. These novel Fc variants, however, have generally been the result of extensive experimental screening and combinations of individual Fc mutations with known biophysical properties; there are few reports of predominantly structure-based rational Fc design. Notably, simply increasing Fc binding affinity for FcRn at low pH does not appear to be sufficient to achieve the largest increases in half-life (and in some cases, very high affinity results in reduced serum half-life). Most of these engineered Fcs have increased affinity not only at low pH (~6.0), but also at pH 7.4. The longest-lived Fc variant known to date, however, with mutations L309D/Q311H/N434S (“DHS”), has only a modest 5-fold increase in binding affinity compared to wt Fc at low pH, but also exhibits negligible binding to the receptor at pH 7.4 (Lee et al., 2019). This is consistent with previous reports that identify efficient release at physiologic serum pH to be critical to FcRn-mediated half-life extension. Thus, while engineering for affinity at low pH, it is also important to optimize the pH dependence of binding for optimal release at serum pH. The rational design process requires a detailed understanding of the structural and functional details of the interaction, which for a pH-dependent complex like Fc-FcRn must also include an accurate model of the pH-sensing mechanism. Unfortunately, the only publicly available crystal structure of a human Fc-FcRn complex is of the M252Y/S254T/T256E (“YTE”) variant, and was determined only to a relatively low 3.8 Å resolution, leaving the atomic positions of many sidechains, and even regions of the protein backbone, subject to substantial uncertainty. Furthermore, the widely accepted conventional mechanism of pH sensing, involving protonation of key histidine residues on Fc at low pH due to the assumed histidine pKa of 6.5 being within the range of interest (pH 6.0-7.4), is thermodynamically impossible. In this thesis I present an extensive analysis of the Fc-FcRn system, including the generation of all-atom models of human wild-type (wt) and variant complexes and the rat wt complex, and assignment of dominant protonation states at pH 6.0, at which most binding experiments are performed. I validate these models using retrospective molecular dynamics (MD)-based free-energy perturbation (FEP) calculations to compare to a large dataset of wt and mutant binding affinities. During this validation process I identify a residue on FcRn, glutamic acid 133, which adopts a highly unusual configuration in the complex and, due to quantum mechanical electronic polarization effects, is not described well by the fixed-charge molecular mechanics force field used by the FEP calculations, resulting in systematic errors for mutations that affect its hydrogen bonding network. I also identify a new variant, with a V308P mutation in a YTE background (“YTEP”), which induces a previously unreported conformational change that accounts for its high binding affinity compared to YTE and wt. To address the problem of the pH-sensing mechanism, I describe a general method for calculating the pH-sensing free energy of binding for any complex, based on a study of the pH dependence of protein unfolding free energies (Yang and Honig, 1993). The key observation underlying this method is that pH-dependent complex formation must be accompanied by a change in the pKa of one or more titratable groups between the unbound and bound states. Furthermore, the change in binding energy between two pHs can be directly calculated based on those pKas alone. As there are no experimental pKa measurements available for the Fc-FcRn interface residues, I perform these pH-sensing free energy calculations using FEP-based calculated pKas to quantitatively assess which residues at the interface are involved in sensing pH over the physiologically relevant pH range, and present a residue-level model for pH sensing in the Fc-FcRn system. Finally, I present some preliminary work toward the rational design of modified Fc regions with both increased affinity at low pH, and increased pH dependence of binding, using FEP calculations to guide experiment. This type of approach, of computational screening of a large number of different variants, followed by more limited experimental testing of promising leads, has the potential to streamline Fc design efforts and provide further insight into the structural basis of function for the Fc-FcRn system.