Recently, we reported on a transaldolase B variant (TalB F178Y) that is able to use dihydroxyacetone (DHA) as donor in aldol reactions. In a second round of protein engineering, we aimed at improving the affinity of this variant towards nonphosphorylated acceptor aldehydes, that is, glyceraldehyde (GA). The anion binding site was identified in the X-ray structure of TalB F178Y where a sulfate ion from the buffer was bound in the active site. Therefore, we performed site-directed saturation mutagenesis at three residues forming the putative phosphate binding site, Arg181, Ser226 and Arg228. The focused libraries were screened for the formation of D-fructose from DHA and d,l-GA by using an adjusted colour assay. The best results with respect to the synthesis of D-fructose were achieved with the TalB F178Y/R181E variant, which exhibited an at least fivefold increase in affinity towards d,l-GA (K(M)=24 mM). We demonstrated that this double mutant can use D-GA, glycolaldehyde and the L-isomer, L-GA, as acceptor substrates. This resulted in preparative synthesis of D-fructose, D-xylulose and L-sorbose when DHA was used as donor. Hence, we engineered a DHA-dependent aldolase that can synthesise the formation of polyhydroxylated compounds from simple and cheap substrates at preparative scale.