We test, using site-directed mutagenesis, predictions based on the X-ray structure of N-acetyl-L-glutamate kinase (NAGK), the paradigm of the amino acid kinase protein family, about the roles of specific residues on substrate binding and catalysis. The mutations K8R and D162E decreased V([sustrate]= infinity ) 100-fold and 1000-fold, respectively, in agreement with the predictions that K8 catalyzes phosphoryl transfer and D162 organizes the catalytic groups. R66K and N158Q increased selectively K(m)(Asp) three to four orders of magnitude, in agreement with the binding of R66 and N158 to the C(alpha) substituents of NAG. Mutagenesis in parallel of aspartokinase III (AKIII phosphorylates aspartate instead of acetylglutamate), another important amino acid kinase family member of unknown 3-D structure, identified in AKIII two residues, K8 and D202, that appear to play roles similar to those of K8 and D162 of NAGK, and supports the involvement of E119 and R198, similarly to R66 and N158 of NAGK, in the binding of the amino acid substrate, apparently interacting, respectively, with the alpha-NH(3)(+) and alpha-COO(-) of aspartate. These results and an improved alignment of the NAGK and AKIII sequences have guided us into 3-D modelling of the amino acid kinase domain of AKIII using NAGK as template. The model has good stereochemistry and validation parameters. It provides insight into substrate binding and catalysis, agreeing with mutagenesis results with another aspartokinase that were not considered when building the model.AKIII is homodimeric and is inhibited by lysine. Lysine may bind to a regulatory region that is C-terminal to the amino acid kinase domain. We make a C-terminally truncated AKIII (AKIIIt) and show that the C-region is involved in intersubunit interactions, since AKIIIt is found to be monomeric. Further, it is inactive, as demanded if dimer formation is essential for activity. Models for AKIII architecture are proposed that account for these findings.