A large number of promising enzyme reactions, such as diastereospecific carbon-carbon bond formation by aldolases, suffers from an unfavorable position of the reaction equilibrium. Combining continuous chromatography and enzymatic reactions should allow for a new in situ or online product recovery process to achieve high reaction productivity and yield and make these biocatalysts economically more attractive. The integration imposes a series of constraints on the chromatographic separation, mainly on the applicable solvent, which is at the same time the reaction medium for the enzymatic reaction. We exploit this concept for a model process, the integrated biocatalytic production of L-allo-threonine from glycine and acetaldehyde. Of crucial importance for this process is the separation of the two physicochemically similar amino acids glycine and threonine, in particular in the presence of additional compounds such as the second starting material and enzyme cofactors. This separation was first investigated on a lab-scale simulated moving bed (SMB) unit under enzyme compatible conditions. After triangle theory-based identification of SMB operating points, the two amino acids could be efficiently separated, applying aqueous eluents with minor content of organic co-solvent at neutral pH on a weak cation exchanger resin. Remarkably, the separation performance with respect to the two amino acids was only slightly reduced by coupling the SMB to a continuously operated enzyme membrane reactor, whose efflux contained, in addition to the amino acids, acetaldehyde and the cofactor pyridoxal-5-phosphate. This represents an important step to the future design of even further integrated biocatalytic reaction-separation schemes. (c) 2008 Elsevier Ltd. All rights reserved.