A designer quest for the Achilles’ heel of influenza

Summary

In this issue of Structure, Sevy at al. describe a computationally designed cyclic peptide that mimics the CDRH3 loop of the C05 antibody. The cyclic peptide recognizes a more conserved part of the paratope on hemagglutinin and consequently binds to a wider variety of influenza strains than C05 mAb.

Influenza is an acute and recurring respiratory disease, with high morbidity rates for children and seniors, and for patients with chronic illnesses. Antigenic drifts and shifts of these viruses hinder the development of effective prevention and treatment strategies that include vaccination and antivirals.

Computer aided drug design (CADD) approaches have been aiding drug discovery efforts since the 1980s, with such notable successes as the rational design of small molecule inhibitors to HIV protease (e.g. Saquinavir), to neuraminidase (Tamiflu), to carbonic anhydrase II (Dorzolamide) or to thymidylate synthase (Ralitrexed) among many others, for the treatment of various infectious diseases and cancer. The field of recombinant therapeutic agents started with insulin in 1981 and a few years later the first monoclonal antibody was approved by the FDA in 1986 (Muromonab-CD3) for preventing kidney transplant rejection. Recombinant therapeutic agent discovery has seen a surge of drug development successes, with 178 such drugs approved as of today [1]. More recently, monoclonal antibodies are considered a viable therapeutic modality for infectious disease targets, but only a handful of these have been approved by the FDA. This is partly because the remarkable specificity of mAb that makes them so powerful against drug targets also presents practical limitations. A highly specific antibody can quickly become ineffective if the target protein displays mutations, which is typical in viral species, or if multiple strains of the same virus are to be neutralized. Additional limitations are connected to the large size of mAb, which makes them relatively expensive to produce and problematic to store due to degradation. Finally, mAbs have limited cell penetration, which restricts their use to cell surface or extracellular targets.

The work of Sevy et al. in this issue of STRUCTURE makes a progress to address several of these issues. First, it actively engages CADD technology to design a protein-like molecule. While designing an inhibitor is more challenging than identifying a mAb against a specific target, it does bring the possibility of engineering a molecule with desired features. In this case, a designed molecule with broadly neutralizing activity against multiple strains of influenza, something that is atypical for a mAb. The work of Sevy et al. also attempts to circumvent the problems associated with the large size of mAb, as it focuses on a cyclic peptide design that recognizes only part of the epitope of a corresponding antibody. The epitope recognition part of an antibody consist of 3 complementarity determining regions (CDRs) in the variable domains of each the heavy and light chains, which correspond to the connecting variable loop regions in the beta sandwich structure of the immunoglobulin fold. The most variable of these three loops is CDR3 of the heavy chain (CDRH3), and most often it is also the longest one. The hypothesis behind the work is that CDRH3 is often recognizing a more conserved paratope, while CDR1 and CDR2 loops make contacts with more variable parts of the antigen. A newly designed cyclic peptide mimicking CDRH3 alone thus would be expected to deliver a more broadly neutralizing antibody across various strains. This proof of principle study focused on improving breadth of binding of the C05 antibody, which had been shown to bind to both groups of influenza hemagglutinin.

The computational design of this cyclic peptide, mimicking the CDRH3 of C05, utilizes the Rosetta software package and extensive molecular dynamics simulations. The computationally isolated 26 residue long CDRH3 fragment and a truncated alternative version of it were cyclized by adding Cys residues at its termini. The sequences were redesigned with Rosetta Software for the purpose of recapitulating the original conformation of CDRH3. Subsequently, several residues important for binding were re-introduced in the cyclic peptide designs if they satisfied stability criteria and increased affinity. After examining 1000 simulated designs, 8 were selected for experimental testing. Biolayer Interferomety (BLI) and ELISA experiments were used to assess binding to recombinant influenza HA. Two designs showed 124 and 506 nM binding, respectively, compared with the wild type C05 mAb which binds with 88nM affinity. These designed cyclic peptides share 39 and 55% sequence identify with the wild type CDRH3 of C05. The designed cyclic peptides not only retained binding to the original strains but had an increased breadth compared to the wild type C05 IgG mAb, which bound to, and neutralized, 8 out of 24 tested strains with at least 100 nM binding affinity. The two cyclic peptide designs showed 100nM binding to at least 11 and 10 strains, respectively. However, when the biological activity of these peptides was tested in influenza HN1 hemagglutinin neutralization assay, these designed peptides showed no activity. Among several factors, the authors attribute the lack of avidity in neutralizing essays to the monomeric nature of the cyclic peptide designs, which can be addressed in the future by using a scaffold protein.

Although the newly designed reagents in this study would require additional modifications before biological activity is achieved, this work was successful in delivering cyclic peptides that assume biologically active conformations and confirmed the hypothesis that by focusing on the more conserved part of the paratope, a molecule can be created that binds to a wider range of influenza strains. This work showcases the benefit and viability of rationally designing new biologics for desired therapeutics purposes.