Design of nanobody targeting SARS-CoV-2 spike glycoprotein using CDR-grafting assisted by molecular simulation and machine learning

Abstract

The design of proteins capable effectively binding to specific protein targets is crucial for developing therapies, diagnostics, and vaccine candidates for viral infections. Here, we introduce a complementarity-determining region (CDR) grafting approach for designing nanobodies (Nbs) that target specific epitopes, with the aid of computer simulation and machine learning. As a proof-of-concept, we designed, evaluated, and characterized a high-affinity Nb against the spike protein of SARS-CoV-2, the causative agent of the COVID-19 pandemic. The designed Nb, referred to as Nb Ab.2, was synthesized and displayed high-affinity for both the purified receptor-binding domain protein and to the virus-like particle, demonstrating affinities of 9 nM and 60 nM, respectively, as measured with microscale thermophoresis. Circular dichroism showed the designed protein's structural integrity and its proper folding, whereas molecular dynamics simulations provided insights into the internal dynamics of Nb Ab.2. This study shows that our computational pipeline can be used to efficiently design high-affinity Nbs with diagnostic and prophylactic potential, which can be tailored to tackle different viral targets.

Fig 1

(A) Schematic representation of the design pipeline encompassing A. Hot spots (shown as red spheres) mapping by combining enhanced sampling simulations and alanine scanning; B. Hot spots grafting by generating a library containing the Nb CabBCII-10 framework and multiple different CDRs; C. Protein design. (B) and (C) Structural comparison of conformations as a function of the projection on the first two PCA vectors of the trajectory obtained from conventional cMD (1 μs) and metaD (3 replicates of 100 ns), respectively.

Fig 2

(A) Important interactions between CR3022 and the receptor-binding domain (RBD) of SARS-CoV-2 [31]. The heavy chain of CR3022 is depicted in orange, the light chain in yellow, and the SARS-CoV-2 RBD in cyan. Dashed lines indicate the presence of hydrogen bonds(B) Computational mutagenesis by alanine scanning for the residues in the interaction interface. Each bar denotes the predicted binding $\mathrm{\Delta }\mathrm{\Delta }$G for a given residue upon alanine mutation. A threshold of 1 REU, shown as a dashed line, was chosen as cut-off to predict destabilizing effects (red bars), suggesting importance for binding. Blue bars represent stabilizing or no effect on the $\mathrm{\Delta }\mathrm{\Delta }$G upon alanine replacement.

Fig 3. Machine learning models for the binary classification between high and low/moderate Nb-antigen complex affinities.

(A) Linear discriminant value calculated for each sample in the dataset, demonstrating that the explained variance considering only one LD is 100% (B) Contribution of each feature used to calculate the LDs values (C) Assessment of the classification models’ performance through their characteristic ROC curves. Five different curves are shown, each representing one of the 5-fold cross-validations (CVs). The shaded gray area represents the standard deviation across the 5-folds, while the blue line denotes the mean ROC AUC.

Fig 4. (A) Structural alignment between the designed structure (green) and the AF3 model (purple).

(B) Time-series properties obtained from MD simulations. top: Root mean square deviation (RMSD) as a function of time between the alpha carbons from the simulated structural ensemble and the crystallographic structure (CaBCII-10, PDB ID: 3DWT) and the modeled structure (Ab.2). Bottom: per-residue root mean square fluctuation (RMSF) for the alpha carbons calculated for the last 0.5 μs of simulation. The shaded gray area represents the CDRs 1-3 regions. The blue line corresponds to CaBCII-10, while the green line corresponds to Nb Ab.2. (C) Structural alignment between the designed complex model (green), AF3 (purple) and ClusPro (orange) predicted models. (D) RMSD for the designed complex Nb Ab.2 and SARS-CoV-2 RBD over time, represented by the green line. All the structural models are depicted in cartoon representation.

Fig 5. Biophysical characterization of Nb Ab.2.

(A) Circular Dichroism (CD) spectra of Nb Ab.2 displaying characteristic signals of $\beta$-structured immunoglobulin (IgG) folded domains. (B) Estimation of secondary structure components derived from the CD data using BeStSel. (C, D) left: MST binding curve illustrating the interaction between recombinant SARS-CoV-2 RBD and Nb Ab.2. right: MST binding curve demonstrating the interaction between SARS-CoV-2 VLPs and Nb Ab.2. (E) Solvent-accessible surface area obtained from different snapshots of 200-ns molecular dynamics simulations of the RBD binding site considering the free RBD (orange) and within the fully glycosylated spike protein (blue). The snapshots were recorded at every 1 ns.