Coronaviruses have been the causative agent of three epidemics and pandemics in the past two decades, including the ongoing COVID-19 pandemic. A broadly-neutralizing coronavirus therapeutic is desirable not only to prevent and treat COVID-19, but also to provide protection for high-risk populations against future emergent coronaviruses. As all coronaviruses use spike proteins on the viral surface to enter the host cells, and these spike proteins share sequence and structural homology, we set out to discover cross-reactive biologic agents targeting the spike protein to block viral entry. Through llama immunization campaigns, we have identified single domain antibodies (VHHs) that are cross-reactive against multiple emergent coronaviruses (SARS-CoV, SARS-CoV-2, and MERS). Importantly, a number of these antibodies show sub-nanomolar potency towards all SARS-like viruses including emergent CoV-2 variants. We identified nine distinct epitopes on the spike protein targeted by these VHHs. Further, by engineering VHHs targeting distinct, conserved epitopes into multi-valent formats, we significantly enhanced their neutralization potencies compared to the corresponding VHH cocktails. We believe this approach is ideally suited to address both emerging SARS-CoV-2 variants during the current pandemic as well as potential future pandemics caused by SARS-like coronaviruses.

Spontaneous post-translational modifications (PTMs) like Asn deamidation and Asp isomerization are common in therapeutic antibodies, impacting their structure and function. Predicting these PTMs could accelerate antibody therapeutic development. This presentation details a study that combines QM-based proton-affinity calculations and microsecond Molecular Dynamics to elucidate the significance of the n+1 residue in influencing isomerization and deamidation reactions in proteins. By analyzing key structural and conformational attributes directly linked to the reaction pathway on the Asp/Asn residues in 131 mAbs, a minimalistic physics-based model is proposed to predict isomerization and deamidation reactions. This approach facilitates the development and enhanced design of therapeutic antibodies with minimized chemical risks

Rapidly evolving influenza A viruses (IAVs) and influenza B viruses (IBVs) are major causes of recurrent lower respiratory tract infections. We discovered a neuraminidase (NA)-targeting monoclonal antibody (mAb), FNI9, that potently inhibits the enzymatic activity of all IAVs and IBVs including contemporary immune-evading H3N2 strains. Here, we performed molecular dynamics (MD) simulations to obtain a more detailed understanding of the binding of FNI9 and the clonally related FNI17 mAb that lost potency against modern H3N2 strains. While static structure analysis shows similar contact energies for the two complexes, the dynamic contact energy from MD shows that FNI9 is a stronger binder than FNI17 for NA (H3N2 A/Tanzania/205/2010), in line with binding affinity measurements. The interactions in the FNI9-NA complex are balanced across more residues than in the FNI17-NA complex both in the dynamic paratope and in the dynamic epitope, providing a mechanistic explanation for the superior binding of FNI9 and superior resilience to immune-evading influenza strains.

In this work, we present a method for modeling antibodies and performing pH-dependent conformational sampling, which can enhance property calculations. Structure-based charge descriptors are evaluated for their predictive performance on recently published antibody pI, viscosity, and clearance data. From this, we devised four rules for therapeutic antibody profiling which address developability issues arising from hydrophobicity and charged-based solution behavior, PK, and the ability to enrich for those that are approved by the U.S. Food and Drug Administration. Differences in strategy for optimizing the solution behavior of human IgG1 antibodies versus the IgG2 and IgG4 isotypes and the impact of pH alterations in formulation are discussed.

Proteins mediate their functions through chemical interactions; modeling these interactions, which are typically through sidechains, is an important need in protein design. However, constructing an all-atom generative model requires an appropriate scheme for managing the jointly continuous and discrete nature of proteins encoded in the structure and sequence. We describe an all-atom diffusion model of protein structure, Protpardelle, which instantiates a “superposition” over the possible sidechain states, and collapses it to conduct reverse diffusion for sample generation. When combined with sequence design methods, our model is able to co-design all-atom protein structure and sequence. This capability can potentially enable conditional generation of de novo protein structures for functional design.

Antibody thermostability is challenging to predict from sequence and/or structure. This difficulty is likely due to the absence of direct entropic information. Herein, we present AbMelt where we model the inherent flexibility of homologous antibody structures using molecular dynamics (MD) simulations at three temperatures and learning the relevant descriptors to predict the temperatures of aggregation (T_agg), melt onset (T_m,on), and melt (T_m). We observed that the radius of gyration deviation of the complementarity determining regions (CDRs) at 400K is the highest correlated descriptor with aggregation temperature (r_p = -0.68) and the deviation of internal molecular contacts at 350K is the highest correlated descriptor with both T_m,on (r_p = -0.74) as well as T_m (r_p = -0.69). Moreover, after descriptor selection and machine learning (ML) regression, we predict on a held-out test set containing both internal and public data and achieve robust performance for all endpoints compared to baseline models (T_agg R² = 0.57 ± 0.11, T_m,on R² = 0.56 ± 0.01, and T_m R² = 0.60 ± 0.06). Additionally, the robustness of the AbMelt MD methodology is demonstrated by only training on <5% of the data and significantly outperforming more traditional ML models trained on the entire dataset of more than 500 internal antibodies. Users can predict thermostability measurements for Fvs by collecting descriptors and using AbMelt, which we are making available at http://github.com/merck/AbMelt.

Authors: Sepehr Dehghani-Ghahnaviyeh, Michael Smith, Yan Xia, Athanasios Dousis, Alan Grossfield, Sreyoshi Sur

Lipid nanoparticles (LNPs) containing ionizable aminolipids are among the leading platforms for successful delivery of nucleic acid-based therapeutics, including messenger RNA (mRNA). The two recently FDA-approved COVID-19 vaccines developed by Moderna and Pfizer/BioNTech belong to this category. Ionizable aminolipids, cholesterol, and DSPC lipids are among the key components of such formulations, crucially modulating physico-chemical properties of these formulations, and consequently, the potency of these therapeutics. Despite the importance of these components, the distribution of these molecules in LNPs containing mRNA is not clear. In this study, we used all-atom molecular dynamics (MD) simulations to investigate the distribution and effects of the Lipid-5 (apparent pKa of the lipid nanoparticle = 6.56), a rationally designed and previously reported ionizable aminolipid by Moderna, on lipid bilayers, Mol. Ther., 2018, 26, 1509-1519.

The simulations were conducted with half of the aminolipids charged and half neutral, to be approximately close to the expected ionization in the microenvironment of the LNP surface.

In all five simulated systems in this work, the cholesterol content was kept constant, whereas the DSPC and Lipid-5 concentrations were changed systematically.

We found that at higher concentrations of the ionizable aminolipids, the neutral aminolipids form a disordered aggregate in the membrane interior that preferentially includes cholesterol.

The rules underlying the lipid redistribution could be used to rationally choose lipids to optimize LNP function.

Multispecific antibodies hold immense potential in addressing unmet medical needs unachievable by monoclonal antibodies, yet they cannot be predictably manufactured. We look at promising protein building blocks which can be used to assemble multispecifics, discuss advancements in computational protein design that have been used to overcome challenges, and what it will take to enable rapid and reliable multispecific design via machine learning.