The aim of the present study was to identify functional antisense oligodeoxynucleotides (ODNs) against the rat glutathione S-transferase Mu (GSTM) isoforms, GSTM1 and GSTM2. These antisense ODNs would enable the study of the physiological consequences of GSTM deficiency. Because it has been suggested that the effectiveness of antisense ODNs is dependent on the secondary mRNA structures of their target sites, we made mRNA secondary structure predictions with two software packages, Mfold and STAR. The two programs produced only marginally similar structures, which can probably be attributed to differences in the algorithms used. The effectiveness of a set of 18 antisense ODNs was evaluated with a cell-free transcription/translation assay, and their activity was correlated with the predicted secondary RNA structures. Four phosphodiester ODNs specific for GSTM1, two ODNs specific for GSTM2, and four ODNs targeted at both GSTM isoforms were found to be potent, sequence-specific, and RNase H-dependent inhibitors of protein expression. The IC50 value of the most potent ODN was approximately 100 nM. Antisense ODNs targeted against regions that were predicted by STAR to be predominantly single stranded were more potent than antisense ODNs against double-stranded regions. Such a correlation was not found for the Mfold prediction. Our data suggest that simulation of the local folding of RNA facilitates the discovery of potent antisense sequences. In conclusion, we selected several promising antisense sequences, which, when synthesized as biologically stable oligonucleotides, can be applied for study of the physiological impact of reduced GSTM expression.