In the search for quinazoline thymidylate synthase inhibitors that are not subject to intracellular polyglutamation, a class of dipeptide analogues of the diglutamate of 2-desamino-2-methyl-N10-propargyl-5,8-dideazafolic acid (ICI 198583-gamma-L-glu) has been evaluated for their stability to in vivo hydrolysis. Replacement of the second glutamate with another amino acid, e.g. alanine, prevented polyglutamation in vitro but such compounds were subject to hydrolysis when injected into mice. The extent of hydrolysis was measured in plasma, liver and kidney by HPLC analysis of tissue removed from mice 1 hr after i.p. injection. The enzyme responsible for this hydrolysis is thought to be a gamma-glutamyl hydrolase which hydrolyses the amide bond, releasing ICI 198583 which may then be polyglutamated. Development of stable dipeptide compounds was achieved by structural modification in two principal ways: either by replacement of the second amino acid (e.g. glutamate or alanine) with its D-enantiomer or removal of the carboxyl on the alpha-carbon of the second amino acid (alpha'-COOH). In this second approach two series of compounds were investigated. Monocarboxylate-derived dipeptides, e.g. ICI 198583-gamma-L-phenylalanine or ICI 198583-gamma-phenylglycine, resulted in stable compounds after removal of the alpha'-COOH (to give -ethylamide and -benzylamide derivatives, respectively). However, for the dicarboxylic amino acids a less clear picture emerged. Although removal of the alpha'-COOH from ICI198583-gamma-L-glutamate to give ICI 198583-gamma-gamma-aminobutyric acid resulted in a stable compound, the corresponding aspartate analogue (-beta-alanine) was subject to hydrolysis.