Galacto-oligosaccharides (GOS) are generally enzymatically synthesized with β-galactosidases. GOS are of interest because of their prebiotic effects on human health. They are mainly applied in infant nutrition, because of their resemblance to human milk oligosaccharides, but they are also applied in e.g. dairy products and beverages. β-Galactosidases synthesize GOS from lactose through transgalactosylation: instead of only using water as acceptor (as in hydrolysis), they can use carbohydrates as acceptor. In this way, GOS with a degree of polymerization up to ten can be formed. The ratio of hydrolysis over transgalactosylation depends on the substrate concentration, temperature, and the source of the enzyme. A β-galactosidase preparation from Bacillus circulans, called Biolacta N5, is known to produce high GOS yields compared to enzymes from other sources. The aim of this thesis was to obtain more insight on the mechanism of GOS production with Biolacta N5 and to investigate how the GOS production process can be optimized. Biolacta N5 consists of four β-galactosidase isoforms, β-gal-A, β-gal-B, β-gal-C, and β-gal-D, which were purified and characterized in chapter 2. At low substrate concentrations, these isoforms differ in hydrolysis and transgalactosylation activity. β-Gal-D seems the best isoform for GOS production, followed by β-gal-C and β-gal-B, and β-gal-A showed the least GOS formation. By studying the thermodynamics of lactose conversion with isothermal titration calorimetry (ITC), the differences in behavior were confirmed, although the interpretation of the results of ITC was quite difficult dealing with a complex mixture of reactions. In contrast to the selectivity at low substrate concentrations, the selectivity of the isoforms hardly differed at high lactose concentrations. These conditions are usually used for industrial GOS production. Only β-gal-A produced slightly more galactose. The initial GOS formation rates indicated that β-gal-A and β-gal-B are the best isoforms for GOS production. In chapter 3, the effect of high concentrations was further studied on the behavior of the complete Biolacta N5 preparation. High concentrations of reacting and non-reacting carbohydrates were added to the oNPG activity assay with Biolacta N5. Small carbohydrates were found to act as acceptor in the reaction, which resulted in an increased reaction rate. The rate of the limiting step of the reaction, i.e. the binding of the galactose residue with the acceptor, is increased, and therewith the release of the product is faster. At the same time, the additives cause molecular crowding, which results in a higher affinity between the enzyme and the substrate. In chapter 4, a kinetic model was developed to quantify the effects of lactose, glucose, galactose, and oligosaccharides on the oNPG converting activity of the β-galactosidases from B.circulans, Aspergillus oryzae and Kluyveromyces lactis. Using multiple substrates simultaneously yields more information than using only lactose or oNPG, because of the competition between the substrates. Three main differences were found that explain why Biolacta N5 produces higher GOS yields than other β-galactosidases: (i) it had a higher reaction rate constant of using lactose or oligosaccharides as substrate relative to water as acceptor (so it had a very low relative hydrolysis rate); and (ii) it also had a high reaction rate with galactose as acceptor, whereas (iii) the other two enzymes are strongly inhibited by galactose. The reaction rate constants indicate that β-gal-A is the most active isoforms in GOS production; however, also its hydrolysis rate is highest. Many of the rate constants increase with increasing molecular weight of the isoforms. Chapter 5 reports on the stability of Biolacta N5 at various temperatures in buffer, and in systems with initially 5.0 and 30% (w/w) lactose. Samples were taken in time and analyzed for oNPG converting activity. The oNPG converting activity was corrected for the presence of lactose, glucose, galactose, and oligosaccharides with the mechanistic model from chapter 4. The stability, expressed with the half-life time, of the enzyme was found to strongly increase with initial lactose concentrations. At high substrate concentration, higher temperatures can be used for GOS production than was presumed feasible based on stability measurements in diluted solutions. Biolacta N5 is still active after one batch run of GOS production, but in a batch process the enzyme is wasted after the reaction. For this reason, the use of immobilized enzyme in a continuous packed bed reactor (PBR) was investigated in terms of productivity in chapter 6. The carbohydrate composition of the product in both systems was comparable. The half-life time of the immobilized enzyme at a lactose concentration of 33% (w/w) and 50ºC was approximately 90 days. The enzymatic productivity using immobilized enzyme in a PBR may be six times higher than that using free enzyme in a batch reactor. When striving for an equal volumetric productivity of both systems, the volume of a PBR can be much smaller than that of a batch reactor, depending on the enzyme dosage and running time of the one batch. Chapter 7 discusses various alternatives for process optimization. One option for a higher GOS productivity is to use an enzyme preparation that contains only β-gal-A and β-gal-B. A somewhat higher oligosaccharide yield can be obtained when initially using a mixture of lactose with a better acceptor molecule. This results in a changed oligosaccharide composition and less lactose in the final product. The sustainability of GOS production in a PBR with immobilized enzyme and 33% (w/w) lactose seems to be similar in terms of exergy to that in a batch reactor with free enzyme and 60% (w/w) lactose.