Members of the lipocalin superfamily share a common structural fold, but differ from each other with respect to the molecules with which they interact. They all contain eight beta-strands (A-H) that fold to form a well-defined beta-barrel, which harbours a binding pocket for hydrophobic ligands. These strands are connected by loops that vary in size and structure and make up the closed and open ends of the pocket. In addition to binding ligands, some members of the family interact with other macromolecules, the specificity of which is thought to be associated with the variable loop regions. Here, we have investigated whether the macromolecular-recognition properties can be transferred from one member of the family to another. For this, we chose the prototypical lipocalin, the plasma retinol-binding protein (RBP) and its close structural homologue the epididymal retinoic acid-binding protein (ERABP). RBP exhibits three molecular-recognition properties: it binds to retinol, to transthyretin (TTR) and to a cell-surface receptor. ERABP binds retinoic acid, but whether it interacts with other macromolecules is not known. Here, we show that ERABP does not bind to TTR and the RBP receptor, but when the loops of RBP near the open end of the pocket (L-1, L-2 and L-3, connecting beta-strands A-B, C-D and E-F, respectively) were substituted into the corresponding regions of ERABP, the resulting chimaera acquired the ability to bind TTR and the receptor. L-2 and L-3 were found to be the major determinants of the receptor- and TTR-binding specificities respectively. Thus we demonstrate that lipocalins serve as excellent scaffolds for engineering novel biological functions.