An adapted consensus protein design strategy for identifying globally optimal biotherapeutics

Abstract

Biotherapeutic optimization, whether to improve general properties or to engineer specific attributes, is a time-consuming process with uncertain outcomes. Conversely, Consensus Protein Design has been shown to be a viable approach to enhance protein stability while retaining function. In adapting this method for a more limited number of protein sequences, we studied 21 consensus single-point variants from eight publicly available CD3 binding sequences with high similarity but diverse biophysical and pharmacological properties. All single-point consensus variants retained CD3 binding and performed similarly in cell-based functional assays. Using Ridge regression analysis, we identified the variants and sequence positions with overall beneficial effects on developability attributes of the CD3 binders. A second round of sequence generation that combined these substitutions into a single molecule yielded a unique CD3 binder with globally optimized developability attributes. In this first application to therapeutic antibodies, adapted Consensus Protein Design was found to be highly beneficial within lead optimization, conserving resources and minimizing iterations. Future implementations of this general strategy may help accelerate drug discovery and improve success rates in bringing novel biotherapeutics to market.

Keywords: Consensus protein design; bioinformatics; biotherapeutics; data science; developability; molecular modeling.

Figure 1.

(a) The bispecific antibody format used in assessing the biophysical properties and function of the CD3 variants was generated using knob-in-hole technology. The Fab arm corresponding to a tumor-associated antigen was inserted on the knob side. The anti-CD3 scFv was formatted in the V_H:V_L orientation with a 20 amino acid (GGGGS)₄ linker between the heavy and light chains. The scFv was appended to the N-terminus of the hole with a 5 amino acid GGGGS linker spacer between the scFv and Fc. (b). Multiple sequence alignment of the amino acid sequences of eight publicly available CD3 binders. V_L domains are aligned in the top panel and the alignment of V_H domains is shown in the bottom panel. Note that V_L sequences of PUBs 4–6 are 100% identical. These sequences were counted only once while deriving the first consensus sequence, CON1.

Figure 2.

Comparison of computed molecular descriptors for the Fv regions of the publicly available (PUBs) and consensus (CONs) CD3 binders. Black triangles show the PUBs, diamonds show the CONs, and filled/open circles show the optimal (CON22 (blue, best in biophysical properties), CON23 (purple, best in in silico descriptors) and CON24 (green, best overall))/de-optimized (CON25–CON27 (black circles)) combinations of the consensus variants (CombiCONs).

Figure 3.

Comparison of experimental data for the PUBs and CONs obtained using the bispecific molecular format shown in Figure 1(a). Black triangles show the PUBs, diamonds show the CONs, and filled/open circles show the optimal (CON22 (blue, best in biophysical properties), CON23 (purple, best in in silico descriptors) and CON24 (green, best overall))/de-optimized (CON25–CON27 (black circles)) combinations of the consensus variants (CombiCONs). Three kinds of data are shown in this figure. Titer and quality via %Monomer after Protein A purification represent the production data. Melting temperature (T_m), aggregation onset temperature (T_agg), analytical HIC retention time (aHIC RT), high-molecular-weight species (HMWS) after 5 weeks of storage at 40°C (%HMWS) and Diffusion Interaction parameter (k_D) represent biophysical data. Dissociation Constant (KD) from in vitro CD3 binding assays and EC₅₀ from T-cell cytotoxicity assays represent the functional data. Note that the CombiCONs (CON22–CON27) were not tested for function because we found the 40-fold range of affinity values for the PUB and CON molecules resulted in highly similar EC₅₀ values.

Figure 4.

Serum stability assessment of the PUBs and the CONs. Serum stability was measured as % of binding to target (CD3) antigen after 48 h incubation in mouse serum at 37°C. Black triangles represent the PUBs, diamonds show the CONs, and filled/open black circles show the CombiCONs (CON22 (blue biophysical properties), CON23 (purple, best in in silico descriptors), CON24 (green, best overall) and CON25-27 (black circles)). The error bars were derived from two replicas. A subset of samples was retested at a different time point with a highly similar outcome.

Figure 5.

Ridge regression coefficients of (a) computational and (b) experimental attributes for the 21 consensus molecules. Each bar encodes a specific position/amino acid combination in scFv. High values indicate that these combinations have a positive impact on the developability attributes. Note that mutations at several positions such as L77 in the V_L, and A49, L80 and S87 in V_H show opposite trends in computed descriptors versus experimental measurements. This may be due to the use static homology model used here and points to the need to molecular dynamic simulations, which were not included in this work.