Interrogation of side chain biases for oligomannose recognition by antibody 2G12 via structure-guided phage display libraries

Abstract

Monoclonal antibodies (mAbs) are essential reagents for deciphering gene or protein function and have been a fruitful source of therapeutic and diagnostic agents. However, developing anticarbohydrate antibodies to target glycans for those purposes has been less successful because the molecular basis for glycan-mAb interactions is poorly understood relative to protein- or peptide-binding mAbs. Here, we report our investigation on glycan-mAb interactions by using the unique architectural scaffold of 2G12, an antibody that targets oligomannoses on the HIV-1 glycoprotein gp120, as the template for engineering highly specific mAbs to target glycans. We first analyzed 24 different X-ray structures of antiglycan mAbs from the Protein Data Bank to determine side chain amino acid distributions in of glycan-mAb interactions. We identified Tyr, Arg, Asn, Ser, Asp, and His as the six most prevalent residues in the glycan-mAb contacts. We then utilized this information to construct two phage display libraries ("Lib1" and "Lib2") in which positions on the heavy chain variable domains of 2G12 were allowed to vary in restricted manner among Tyr, Asp, Ser, His, Asn, Thr, Ala and Pro to interrogate the minimal physicochemical requirements for oligomannose recognition. We analyzed the sequences of 39 variants from Lib1 and 14 variants from Lib2 following selection against gp120, the results showed that there is a high degree of malleability within the 2G12 for glycan recognitions. We further characterized five unique phage clones from both libraries that exhibited a gp120-specific binding profile. Expression of two of these variants as soluble mAbs indicated that, while specificity of gp120-binding was retained, the affinity of these mutants was significantly reduced relative to WT 2G12. Nonetheless, the results indicate these is some malleability in the identity of contact residues and provide a novel insight into the nature of glycan-antibody interactions and how they may differ from protein-antibody binding interactions.

Fig. 1. Domain-exchange architecture of 2G12

(A) Schematic of domain architecture for canonical IgG molecules (top) and 2G12 (bottom). Canonical IgG architecture results in two antigen binding sites at the interface of light and heavy chain variable domains (VL-VH and VL’-VH’). The VH-VH’ domain exchange in 2G12 creates an extended antigen binding surface that contains up to two additional binding sites at the VH-VH’ interface. (B) Crystal structure of the 2G12 Fab dimer in complex with Man₉GlcNAc₂ (orange) reported by Calarese et al. (PDB ID 1OP5). Domain colors as shown in panel A. (C) Fab monomer of 2G12 showing features that stabilize the domain exchange.

Fig. 2. Phage display of functional 2G12

(A) Coding regions of pHP153-2G12z and pHP153-2G12h. In both cases, the light chain and heavy chain VH-CH fragment were coexpressed in the periplasm using an stII signal sequence. A FLAG epitope was included at the light chain C-terminus. (B and C) Phage ELISA of particles produced from pHP153-2G12z (B) or pHP153-2G12h (C). The anti-FLAG antibody M2 was used to confirm expression of the construct.

Fig. 3. Frequency of specific amino acid-glycan interactions

Observed frequency of direct interactions with glycan antigens for all antigens in the dataset (A) or classified into type of antigen (B–F).

Fig. 4. Limited diversity 2G12 library design and results from selection

(A) Positions that were targeted for mutagenesis involve direct contact residues for glycans at both the primary and secondary sites; blue side chains in the close-up were for Lib1 and orange side chains were for Lib 2. All of the residues were allowed to vary among eight residues (Y/D/S/H/N/T/A/P) using the NMC degenerate codon. For clarity, only one antigen for each is shown. (B and C) Percent occurrence of the eight possible residues following selection against gp120 at each position for Lib1 (B) and Lib2 (C). For reference and to control for expression biases, a parallel selection was performed against M2, an anti-FLAG antibody, and the percent occurrence listed in brackets.

Fig. 5. Characterization of individual clones from limited diversity libraries

(A) Sequences of mutants from Lib1 and Lib2 that were further studied. The WT residue identity at each position is indicated, those positions that were identical to WT are shaded. (B) ELISA of phage displayed 2G12 and mutants against gp120, M2 (anti-FLAG), and BSA. The WT 2G12 phage were tested at a phage titer of ~10¹⁰ iu/mL whereas the variants were tested at ~10¹¹–10¹² iu/mL. (C) ELISA of purified 2G12 mAb, and variants L1–42 and L1–83 against gp120 and BSA. The EC50s for gp120 were 2.3 pM for 2G12, and 1.0 and 1.6 µM for L1–42 and L2–83, respectively.