Starting from the structure of known beta-lactam covalent human cytomegalovirus (HCMV) protease inhibitors and from the knowledge of the residues implicated in the active site of this enzyme, we designed a series of phenylalanine-derived 2-azetidinones bearing a 4-carboxylate moiety that could be apt for additional interactions with the guanidine group of the Arg165/Arg166 residues of the viral protease. Some compounds within this series showed anti-HCMV activity at 10-50 muM, but rather high toxicity. The presence of aromatic 1-acyl groups and a certain hydrophobic character in the region of the 4-carboxylate were stringent requirements for anti-HCMV activity. To go a step ahead into the search for effective HCMV medicines, we then envisaged a series of noncovalent inhibitors by simple deletion of the carbonyl group in the beta-lactam derivatives to provide the corresponding azetidines. This led to low micromolar inhibitors of HCMV replication, with 17 and 27 being particularly promising lead compounds for further investigation, although their toxicity still needs to be lowered.