Differences in the representative structures of dipeptide epimerases, six of which are superposed here, give rise to divergent substrate specificities. Reprinted with permission from PNAS. 1
The recent surge in genomic sequencing has swollen protein databases with sequences that lack functional annotation. Large enzyme superfamilies such as the enolases are particularly challenging to annotate because they can include members that encode widely disparate functions. Babbit, Almo, Gerlt, Jacobsen and colleagues of the Enzyme Function Initiative (EFI) tackled enolase substrate specificity using a computational approach to target a representative subset of enzymes for in vitro biochemical and structural studies.
Within the enolase superfamily, over 700 proteins have two Lys acid/base catalysts and an Asp-x-Asp motif at the end of three β strands in the barrel domain that act to epimerize dipeptides. The authors generated homology models and docking simulations spanning the space of all possible dipeptides for 66 of the epimerases available at the time that computational predictions were undertaken. Two well-characterized L-Ala-D/L-Glu epimerases (AEEs) from Escherichia coli and Bacillus subtilus that are thought to process peptidoglycans anchored the analysis.
In addition to confirming L-Ala-L-Glu specificity of the known AEEs, modeling predicted a new phylogenetically distinct group of AEEs as well as novel dipeptide specificity classes. These included a small set of epimerases specific to positively charged dipeptides and a number of groups specific for hydrophobic dipeptides.
To test the predictions, 17 modeled proteins were purified and subjected to biochemical analysis against a panel of dipeptides—a mass spectrometry assay that uses incorporated deuterium to detect epimerization—and kinetic experiments. Six epimerases were also used to produce 18 crystal structures, including five bound to identified substrates.
The results largely validated computational predictions, including the identification of a new class of AEEs with somewhat relaxed specificity at the N-terminal position, represented by an epimerase from Bacteroides thetaiotamicron; the surprising existence of epimerases such as from Methylococcus capsulatus that are specific for positively charged dipeptides; and the classification of hydrophobic dipeptide groups that include most of the plant epimerases. The combined use of computational prediction and targeted experimental validation should allow large-scale functional assessment of enzyme superfamilies.