L-threo-3-Hydroxyhomoserine phosphate, derived from the antimetabolites L-threo-3-hydroxyaspartate and L-threo-3-hydroxyhomoserine [Shames, S. L., Ash, D. E., Wedler, F. C., and Villafranca, J. J. (1984) J. Biol. Chem. 258, 15331-15339], is a classical competitive inhibitor of threonine synthase (Ki = 6 microM) with structural elements of both substrate and product. L-2-Amino-5-phosphonovaleric acid also inhibits the enzyme competitively with a Ki (31 microM), comparable to Km for L-homoserine phosphate. In contrast, a structural analogue of Hse-P, L-2-amino-3-[(phosphonomethyl)thio]propionic acid exhibits a Ki = 0.11 microM (ca. 100-fold less than Km for L-Hse-P), along with "slow, tight" inhibition kinetics. Nuclear magnetic resonance was used with these inhibitors to probe for pyridoxal phosphate-catalyzed hydrogen-deuterium exchange reactions characteristic of substrates. With L-threo-3-hydroxy-homoserine phosphate, H-D exchange occurs only at the C-alpha position, but for homoserine in the presence of phosphate and for L-2-amino-5-phosphonovaleric acid and L-amino-3[(phosphonomethyl)thio]propionic acid (APMTP), H-D exchange occurs at C-alpha and stereospecifically at C-beta. For L-homoserine plus phosphate and L-2-amino-5-phosphonovaleric acid, the rate of H-D exchange at C-alpha is 8-45 times faster than at C-beta. For L-2-amino-3-[(phosphonomethyl)thio]propionic acid, the C-alpha to C-beta exchange rate ratio is near unity, due to a 700-fold decrease in the C-alpha rate for the analogue. Taken with information from molecular modeling, these data can be interpreted in terms of the current working hypothesis for the catalytic mechanism. Specifically, the slow, tight inhibition by APMTP results from its being carried further into the catalytic cycle than other analogues prior to forming an intermediate that is blocked from further catalysis.