Many antifolates are known to inhibit dihydrofolate reductase from murine Pneumocystis carinii, with 50% inhibitory concentrations (IC50s) ranging from 10(-4) to 10(-11) M. The relationship of the potency against isolated enzyme to the potency against intact murine P. carinii cells was explored with 17 compounds that had proven selectivity for or potency against P. carinii dihydrofolate reductase. Pyrimethamine and one analog were inhibitory to P. carinii in culture at concentrations two to seven times the IC50s for the enzyme, suggesting that the compounds may enter P. carinii cells in culture. Methotrexate was a potent inhibitor of P. carinii dihydrofolate reductase, but the concentrations effective in culture were more than 1,000-fold higher than IC50s for the enzyme, since P. carinii lacks an uptake system for methotrexate. Analogs of methotrexate in which chlorine, bromine, or iodine was added to the phenyl ring had improved potency against the isolated enzyme but were markedly less effective in culture; polyglutamation also lowered the activity in culture but improved activity against the enzyme. Substitution of a naphthyl group for the phenyl group of methotrexate produced a compound with improved activity against the enzyme (IC50, 0.00019 microM) and excellent activity in culture (IC50, 0.1 microM). One trimetrexate analog in which an aspartate or a chlorine replaced two of the methoxy groups of trimetrexate was much more potent and was much more selective toward P. carinii dihydrofolate reductase than trimetrexate; this analog was also as active as trimetrexate in culture. These studies suggest that modifications of antifolate structures can be made that facilitate activity against intact organisms while maintaining the high degrees of potency and the selectivities of the agents can be made.