Chalcone synthases (CHSs) and acridone synthases (ACSs) belong to the superfamily of type III polyketide synthases (PKSs) and condense the starter substrate 4-coumaroyl-CoA or N-methylanthraniloyl-CoA with three malonyl-CoAs to produce flavonoids and acridone alkaloids, respectively. ACSs which have been cloned exclusively from Ruta graveolens share about 75-85% polypeptide sequence homology with CHSs from other plant families, while 90% similarity was observed with CHSs from Rutaceae, i.e., R. graveolens, Citrus sinensis and Dictamnus albus. CHSs cloned from many plants do not accept N-methylanthraniloyl-CoA as a starter substrate, whereas ACSs were shown to possess some side activity with 4-coumaroyl-CoA. The transformation of an ACS to a functional CHS with 10% residual ACS activity was accomplished previously by substitution of three amino acids through the corresponding residues from Ruta-CHS1 (Ser132Thr, Ala133Ser and Val265Phe). Therefore, the reverse triple mutation of Ruta-CHS1 (mutant R2) was generated, which affected only insignificantly the CHS activity and did not confer ACS activity. However, competitive inhibition of CHS activity by N-methylanthraniloyl-CoA was observed for the mutant in contrast to wild-type CHSs. Homology modeling of ACS2 with docking of 1,3-dihydroxy-N-methylacridone suggested that the starter substrates for CHS or ACS reaction are placed in different topographies in the active site pocket. Additional site specific substitutions (Asp205Pro/Thr206Asp/His207Ala or Arg60Thr and Val100Ala/Gly218Ala, respectively) diminished the CHS activity to 75-50% of the wild-type CHS1 without promoting ACS activity. The results suggest that conformational changes in the periphery beyond the active site cavity volumes determine the product formation by ACSs vs. CHSs in R. graveolens. It is likely that ACS has evolved from CHS, but the sole enlargement of the active site pocket as in CHS1 mutant R2 is insufficient to explain this process.