A group of trimethoprim (TMP) analogues containing 3,5-dialkyl(or halo)-4-alkoxy, -hydroxy, or -amino substitution were analyzed in terms of their inhibitory activities against four dihydrofolate reductase (DHFR) isozymes. Although selectivities were lower than with TMP, the activities against vertebrate DHFR were usually at least 2 orders of magnitude less than against enzymes from microbial sources. However, the profiles of activity were remarkably similar for rat, Neisseria gonorrhoeae, and Plasmodium berghei enzymes in all three series, although somewhat different for Escherichia coli DHFR, leading to the conclusion that the hydrophobic pockets are similar for the first three isozymes. Optimal substitution was reached with 3,5-di-n-propyl or 3-ethyl-5-n-propyl groups. Branching of chains at the alpha-carbon, which resulted in increased substituent thickness, was detrimental to E. coli DHFR inhibition in particular. MR is an inadequate parameter for use in correlating such substituent effects. Conformational changes of the more bulky inhibitors can be invoked to explain some differences in inhibitory pattern. Although log P explains simple substituent effects with the vertebrate DHFRs very well, it is insufficient in the more complex cases described here, where shape is clearly involved as well. Solvent-accessible surface areas were measured for TMP in E. coli and chicken DHFRs, where the coordinates are now known. The environment is more hydrophobic in the latter case; this can also be postulated for rat DHFR, which has a very similar activity profile. As with the mammalian isozymes, N. gonorrhoeae DHFR contains an active site phenylalanine replacing Leu-28 of E. coli DHFR, thus creating a more hydrophobic pocket. A similar replacement may also occur in the P.berghei isozyme. Selectivity for bacterial DHFR is dependent on the nature of the 4-substituent, being low for polar 4-hydroxy compounds but high for polar 4-amino analogues, possibly as a result of solvation differences. With complex substituents, the environment of each atom in the active site must be taken into account to adequately explain structure-activity relationships.