Aldoxime dehydratase is a heme-containing enzyme that utilizes the ferrous rather than the ferric ion to catalyze the synthesis of nitriles by dehydration of the substrate. We report a theoretical study of this enzyme aimed at elucidating its catalytic mechanism and understanding this oxidation state preference (Fe(2+) versus Fe(3+)). The uncatalyzed dehydration reaction was modeled by including three and four water molecules to assist in the proton transfer, but the computed barriers were very high at both the DFT (B3LYP) and coupled cluster CCSD(T) levels. The enzymatic dehydration of Z-acetaldoxime was explored through QM/MM calculation using two different QM regions and covering all three possible spin states. The reaction starts by substrate coordination to Fe(2+) via its nitrogen atom to form a six-coordinated singlet reactant complex. The ferrous heme catalyzes the N-O bond cleavage by transferring one electron to the antibond in the singlet state, while His320 functions as a general acid to deliver a proton to the leaving hydroxide, thus facilitating its departure. The key intermediate is identified as an Fe(III)(CH(3)CH═N(•)) species (triplet or open-shell singlet), with the closed-shell singlet Fe(II)(CH(3)CH═N(+)) being about 6 kcal/mol higher. Subsequently, the same His320 residue abstracts the α-proton, coupled with electron transfer back to the iron center. Both steps are calculated to have feasible barriers (14-15 kcal/mol), in agreement with experimental kinetic studies. For the same mode of substrate coordination, the ferric heme does not catalyze the N-O bond cleavage, because the reaction is endothermic by about 40 kcal/mol, mainly due to the energetic penalty for oxidizing the ferric heme. The alternative binding option, in which the anionic aldoxime coordinates to the ferric ion via its oxyanion, also results in a high barrier (around 30 kcal/mol), mainly because of the large endothermicity associated with the generation of a suitable base (neutral His320) for proton abstraction.