Packing motifs as predictors of the propensity of antibody fragments to crystallize

Abstract

A recurring theme in the crystallization of antibody fragments in our laboratory has been a packing pattern involving formation of intermolecular, antiparallel β-pleated sheets across two-fold axes. The most common motif is the antiparallel stacking of constant (C) domains of light (L) chain dimers or Fab molecules. Here, cross-molecule six-stranded sheets are produced by hydrogen-bonding interactions of three-residue polypeptide segments (triads), in the i, i+2 and i+4 positions of the final strands (designated 3–3) of the three-chain layers from two adjacent molecules. In the variable (V) domains the triads are supplied by the first strands (4–1) of the four-chain layers and the resulting cross-molecule sheets contain eight strands. The latter type of packing is more likely to be seen in crystals of Fv fragments (V domains only) than in those of L chain dimers or Fabs. Amongst the triads from either the V or C domains, there are on average four sets of backbone carbonyl and amide groups within hydrogen bonding distance (<3.2 Å) of each other. In at least one example, the adjacent antiparallel strands are sterically aligned, but only two of the appropriate sets of atoms are sufficiently close to meet the distance criteria for intermolecular hydrogen bonding. These observations have been used to construct a list of rules for predicting which types of L chain dimers, Fab and Fvs are likely to crystallize in these packing patterns.