Lessons learned from merging wet lab experiments with molecular simulation to improve mAb humanization

Abstract

Humanized monoclonal antibodies (mAbs) are among the most promising modern therapeutics, but defined engineering strategies are still not available. Antibody humanization often leads to a loss of affinity, as it is the case for our model antibody Ab2/3H6 (PDB entry 3BQU). Identifying appropriate back-to-mouse mutations is needed to restore binding affinity, but highly challenging. In order to get more insight, we have applied molecular dynamics simulations and correlated them to antibody binding and expression in wet lab experiments. In this study, we discuss six mAb variants and investigate a tyrosine conglomeration, an isopolar substitution and the improvement of antibody binding towards wildtype affinity. In the 3D structure of the mouse wildtype, residue R94h is surrounded by three tyrosines which form a so-called 'tyrosine cage'. We demonstrate that the tyrosine cage has a supporting function for the CDRh3 loop conformation. The isopolar substitution is not able to mimic the function appropriately. Finally, we show that additional light chain mutations can restore binding to wildtype-comparable level, and also improve the expression of the mAb significantly. We conclude that the variable light chain of Ab2/3H6 is of underestimated importance for the interaction with its antigen mAb 2F5.

Fig. 1

Amino acid sequence of (A) variable heavy chain and (B) variable light chain of su3H6, su3H6 variants (BM), wt3H6 variants (TR), BM07/vL variants and mouse antibody Ab2/3H6 (wt3H6). Kabat numbering scheme is used and the CDR regions are defined by Kabat using the abYsis tool (Kabat et al., 1991; Swindells et al., 2017). CDRs are highlighted in gray. Corresponding light chain for variants in panel A is su3H6, respectively, except for TR02/TR03 which have a wt3H6 light chain. BM07/vL variants in panel B use the BM07 heavy chain.

Fig. 2

Snapshots (1 per ns) of arginine or lysine at position 94h visualizing the flexibility of residue 94h in the variants. (A) black, gray = wt3H6, BM07; (B) blue, purple = BM09, BM10; (C) red = BM11 and (D) green, blue, pink = BM07/vL01, BM07/vL02, BM07/vL03. The cartoon and sticks are from the wt3H6 crystal structure.

Fig. 3

Clusters of CDRh3 loop structure for wt3H6, su3H6, TR02, TR03, BM07, BM09-11 and BM07/vL01-03 including only clusters with overall occurrence >2%.

Fig. 4

The CMS structures of CDRh3 loop clusters. (A–F) In dark gray on top: 2F5, in light gray below: 3H6, both from the crystal structure. The CDRh3 loop is at the center of the picture, left of it is vH, right of it is vL. In (A) the first five CMS structures are overlaid on top of the crystal, in (B–F) CDRh3, FRh3 and FRh4 regions of the five CMS structures are shown individually. The ‘tyrosine cage’, R/K/T94h and F46lL + Q49lS are shown as sticks. Colors correspond to the CDRh3 clusters in Fig. 3.

Fig. 5

K _D measurements of Ab2/3H6 variants: real-time bio-layer interferometry (BLI) sensorgram aligned to baseline crude culture supernatant of (A) expressed heavy chain BM variants and (B) light chain BM07/vL variants detected with 2F5 IgG.