Dihydroxyacetone phosphate (DHAP)-dependent aldolases have been widely used for organic synthesis. The major drawback of DHAP-dependent aldolases is their strict donor substrate specificity toward DHAP, which is expensive and unstable. Here we report the development of an in vivo selection system for the directed evolution of the DHAP-dependent aldolase, L-rhamnulose-1-phosphate aldolase (RhaD), to alter its donor substrate specificity from DHAP to dihydroxyacetone (DHA). We also report preliminary results on mutants that were discovered with this screen. A strain deficient in the L-rhamnose metabolic pathway in Escherichia coli (DeltarhaDAB, DE3) was constructed and used as a selection host strain. Co-expression of L-rhamnose isomerase (rhaA) and rhaD in the selection host did not restore its growth on minimal plate supplemented with L-rhamnose as a sole carbon source, because of the lack of L-rhamnulose kinase (RhaB) activity and the inability of WT RhaD aldolase to use unphosphorylated L-rhamnulose as a substrate. Use of this selection host and co-expression vector system gives us an in vivo selection for the desired mutant RhaD which can cleave unphosphorylated L-rhamnulose and allow the mutant to grow in the minimal media. An error-prone PCR (ep-PCR) library of rhaD gene on the co-expression vector was constructed and introduced into the rha-mutant, and survivors were selected in minimal media with l-rhamnose (MMRha media). An initial round of screening gave mutants allowing the selection strain to grow on MMRha plates. This in vivo selection system allows rapid screening of mutated aldolases that can utilize dihydroxyacetone as a donor substrate.