CD30, the so-called Reed-Sternberg antigen, constitutes a promising cell-specific target for the treatment of Hodgkin's lymphoma. Starting from the previously characterized cognate HRS3 mouse monoclonal antibody, the bacterially produced functional Fab fragment was humanized by grafting the CDRs from the mouse antibody framework on to human immunoglobulin consensus sequences. This procedure led to a 10-fold decreased antigen affinity, which surprisingly was found to be mainly due to the VH domain. To improve the antigen-binding activity, an in vitro evolution strategy was employed, wherein random mutations were introduced into the humanized VH domain by means of error-prone PCR, followed by a filter sandwich Escherichia coli colony screening assay for functional Fab fragments using a recombinant extracellular domain of the CD30 antigen. After three cycles of in vitro affinity maturation, the optimized Fab fragment huHRS3-VH-EP3/1 was identified, which carried four exchanged residues within or close to the VH CDRs and had an affinity that was almost identical with that of the murine HRS3 Fab fragment. The resulting humanized Fab fragment was fully functional with respect to CD30 binding both in ELISA with the recombinant antigen and in FACS experiments with CD30-positive L540CY cells. In the light of the previously successful clinical application of an alphaCD30 x alphaCD16 bispecific mouse quadroma antibody derived from HRS3, the humanized Fab fragment comprises an important step towards the construction of a fully recombinant therapeutic agent. The combination of random mutagenesis and colony filter screening assay that was successfully applied here should be generally useful as a method for the rapid functional optimization of humanized antibody fragments.