We have used a phage-display-based system to investigate the effect of simultaneous substitutions within the C5a effector domain. Two different libraries were constructed. In library I, known binding positions 67, 68, 72 and 74 of human complement C5a (hC5a) and in library II, positions 69-73 of hC5a without C-terminal Arg74 (des-Arg74-C5a) were randomly mutated. In more than 80% (position 72) or 90% (positions 68 and 74) of all cases, the original residues of hC5a were selected from library I, demonstrating that the phage system can be used to define binding points within the C5a molecule. Surprisingly, in more than 90% of all clones, a Phe residue was enriched at position 67 instead of the original His residue which, however, did not affect the binding affinity or the signalling activity. In library II, Leu was preferentially selected at positions 70-72 and Tyr at position 73, while no enrichment of an individual amino acid was observed at position 69. Mutants with (a) Leu in positions 71 and 72 (b) Ser or Leu in position 70 and (c) Arg or Tyr in position 73, showed a 4-10-fold higher binding affinity as compared to des-Arg74-[Ala27, Phe67]C5a. The binding affinity was indistinguishable from that of hC5a. In consequence, not only position 72 but also positions 70, 71 and 73 are able to interact with the C5a receptor, whereas position 69 is not. Intriguingly, one mutant with a high binding affinity but without signalling activity was selected. Thus, random mutagenesis of phage-displayed C5a was proven to be a powerful strategy to define receptor-binding points and to select C5aR antagonists based on the structure of the natural ligand.