2,4-Diaminoquinazoline antifolates with a lipophilic side chain at the 5-position, and in one case with a classical (p-aminobenzoyl)-L-glutamate side chain, were synthesized as potentially selective inhibitors of a site-directed mutant of human dihydrofolate reductase (DHFR) containing phenylalanine instead of leucine at position 22. This mutant enzyme is approximately 100-fold more resistant than native enzyme to the classical antifolate methotrexate (MTX), yet shows minimal cross resistance to the nonclassical antifolates piritrexim (PTX) and trimetrexate (TMQ). Although they were much less potent than trimetrexate and piritrexim, the lipophilic 5-substituted analogues were all found to bind approximately 10 times better to the mutant DHFR than to the wild-type enzyme. The potency of the analogue with a classical (p-aminobenzoyl)-L-glutamate side chain was similarly diminished in comparison with MTX, but the difference in its binding affinity to the two DHFR species was only 5-fold. Thus, by making subtle structural changes in the antifolate molecule, it may be possible to attack resistance due to mutational alterations in the active site of the target enzyme. Also, to test the hypothesis that DHFR from Pneumocystis carinii and Toxoplasma gondii may have a less sterically restrictive active site than the enzyme from mammalian cells, inhibition assays using several of the lipophilic analogues in the series were carried out against the P. carinii and T. gondii reductases in comparison with the enzyme from rat liver. In contrast to their preferential binding to mutant versus wild-type human DHFR, binding of these analogues to the P. carinii and T. gondii enzymes was weaker than binding to rat enzyme. It thus appears that, if the active site of the DHFR from these parasites is less sterically restrictive than the active site of the mammalian enzyme, this difference cannot be successfully exploited by moving the side chain from the 6-position to the 5-position.