Random association of VL and VH repertoires contributes considerably to antibody diversity. The diversity and the affinity are then increased by hypermutation in B cells located in germinal centres. Except in the case of 'heavy chain' disease, naturally occurring heavy-chain antibodies have not been described, although antigen binding has been demonstrated for separated heavy chains or cloned VH domains. Here we investigate the presence of considerable amounts of IgG-like material of M(r) 100K in the serum of the camel (Camelus dromedarius). These molecules are composed of heavy-chain dimers and are devoid of light chains, but nevertheless have an extensive antigen-binding repertoire, a finding that calls into question the role of light chains in the camel. Camel heavy-chain IgGs lack CH1, which in one IgG class might be structurally replaced by an extended hinge. Heavy-chain IgGs are a feature of all camelids. These findings open new perspectives in the engineering of antibodies.

Functional heavy-chain ^imnuinoglobulins lacking light chains occur naturally in Camelidae. We now show the feasibility of immunising a dromedary, cloning the repertoire of the variable domains of its heavy-chain antibodies and panning, leading to the successful identification of minimum sized antigen binders. The recombinant binders are expressed well in E. coli, extremely stable, highly soluble, and react specifically and with high affinity to the antigens. This approach can be viewed as a general route to obtain small binders with favourable characteristics and valuable perspectives as modular building blocks to manufacture multispecific or multifunctional chimaeric proteins.

The Camelidae is the only taxonomic family known to possess functional heavy-chain antibodies, lacking light chains. We report here the 2.5 A resolution crystal structure of a camel VH in complex with its antigen, lysozyme. Compared to human and mouse VH domains, there are no major backbone rearrangements in the VH framework. However, the architecture of the region of VH that interacts with a VL in a conventional FV is different from any previously seen. Moreover, the CDR1 region, although in sequence homologous to human CDR1, deviates fundamentally from the canonical structure. Additionally, one half of the CDR3 contacts the VH region which in conventional immunoglobulins interacts with a VL whereas the other half protrudes from the antigen binding site and penetrates deeply into the active site of lysozyme.