Glycosidic compounds are chemical structures that consist of sugars molecules or that have a sugar attached to another molecule. They are now already used in various industries and have the potential to serve an even much wider range of applications. They can be synthesised chemically, but with the right enzymes, production processes can often be much cleaner and more efficient. Glycoside phosphorylases are such enzymes that can be used to synthesise glycosides and in that respect sucrose phosphorylase is highly interesting. Sucrose phosphorylase is namely not restricted to its wild-type substrates. It can also transfer a sugar moiety to a variety of different acceptor molecules. In addition, it uses sucrose as a cheap, renewable and reactive donor substrate, which makes it an attractive biocatalyst for the production of special sugars and glucoconjugates. Activities on relevant components are unfortunately poor and in many cases not even exceeding the unwanted hydrolytic side activity. Moreover, for industrial processes increased stability is also desired. It is thus clear that sucrose phosphorylase offers many possibilities, but that still some improvements are required to fully exploit its potential. Therefore, in this thesis enzyme engineering was applied to alter the specificity and enhance the stability of sucrose phosphorylase.Knowing how specificity is controlled, would allow to optimise alternative activities in a more efficient way. Therefore, a map of the acceptor site of the SP from Bifidobacterium adolescentis was first created by substituting each residue by alanine and analysing the influence on the affinity for both the natural (inorganic phosphate and fructose) and alternative acceptors (D-arabitol and pyridoxine). All residues examined were found to contribute to the specificity for phosphate (Arg135, Leu343, Tyr344), fructose (Tyr132, Asp342) or both (Pro134, Tyr196, His234, Gln345). Alternative acceptors that are glycosylated rather efficiently e.g. d-arabitol were found to interact with the same residues as fructose, whereas poor acceptors like pyridoxine did not seem to make any specific interactions with the enzyme. Furthermore, it was shown that SP is already optimised to outcompete water as an acceptor substrate, meaning that it will be very difficult to lower its hydrolytic activity any further. Consequently, increasing the transglycosylation activity towards alternative acceptors seems to be the best strategy, although that would probably require a drastic remodelling of the acceptor site in most cases.Regioselectivity can most likely be engineered more easily than completely shifting specificity, but it can nevertheless be equally important. Kojibiose (Glc-α1,2-Glc) for instance is a very expensive and hardly available compound with prebiotic properties, while its regioisomer maltose (Glc-α1,4-Glc) is cheap and has little added value. Natural sucrose phosphorylases produce a mixture of both and a variant that only makes kojibiose would of course be of great interest. To that end, ten positions in the acceptor site were randomised individually and screened with a high performance anion exchange chromatography (HPAEC) based screening procedure. Several improved mutants were obtained from this first round of mutagenesis and the best mutant L341I displayed a selectivity of 79%. Rational combination as well as combinations predicted by a statistical ProSAR model could further improve the selectivity, although the former approach yielded the best mutants. Two double (L341I_Q345S and L341I_Q345N) and one triple mutant (L341I_Y344A_Q345N) were obtained with selectivities of 93 to 95%. Activities were only slightly lower than the wild-type enzyme and therefore the variants created in this work will allow the development of a cost-effective and scalable process for the enzymatic synthesis of kojibiose from the readily available and low-cost substrates sucrose and glucose.As discussed above, stable biocatalysts are a real benefit for industrial processes. Thermophilic organisms are a rich source of stable enzymes, but for sucrose phosphorylase this area has not yet been explored. Hence, in this study, the putative sucrose phosphorylase from the thermophile Thermoanaerobacterium thermosaccharolyticum was recombinantly expressed and fully characterised. The enzyme showed significant activity on sucrose (optimum at 55°C), and with a melting temperature of 79°C and a half-life of 60 hours at the industrially relevant temperature of 60°C, it is far more stable than known sucrose phosphorylases. Substrate screening and detailed kinetic characterisation revealed however a preference for sucrose 6’-phosphate over sucrose. The enzyme can thus be considered as a sucrose 6’-phosphate phosphorylase, a specificity not yet reported to date. Homology modelling and mutagenesis pointed out particular residues (Arg134 and His344) accounting for the difference in specificity. Moreover, phylogenetic and sequence analysis suggest that glycoside hydrolase 13 subfamily 18 might harbour even more specificities. In addition, the second gene residing in the same operon as sucrose 6’-phosphate phosphorylase was identified as well, and found to be a phosphofructokinase. The concerted action of these both enzymes implies a new pathway for the breakdown of sucrose, in which the reaction products end up at different stages of the glycolysis.In addition to the development of applications, techniques like ancestral reconstruction, correlated mutation analysis and in silico prediction of stability were evaluated for their general utility in protein engineering. Present-day enzymes have evolved from a common ancestor that is believed to be promiscuous and to have an appropriate mutational background to evolve. Ancestral enzymes would accordingly be ideally suited to create new (related) specificities by directed evolution. To investigate whether an ancestral enzyme could be a good starting point for mutagenesis and what the role of promiscuity is, a full-length ancestor of sucrose phosphorylase (SP) and sucrose 6’-phosphate phosphorylase (SPP) enzymes was reconstructed. The ancestor behaved like a sucrose phosphorylase, but its activity was one to three orders of magnitude lower than present-day SP’s. This was mainly due to a low kcat (<1 s-1) and a high KM for sucrose and fructose (>200 mM). In addition, promiscuity was very similar in qualitative terms, but activities were in general lower. Swapping SP specific residues with their SPP counterpart could furthermore not introduce SPP activity in the ancestor. However, the same mutations could enable a specificity switch in a thermostable present-day SP, showing that for directed evolution a stable template might be more appropriate than a reconstructed ancestor.Another aspect of evolution that could be useful for enzyme engineering, is coevolution of specific positions in related enzymes. Statistical analysis of large multiple sequence alignments can reveal such correlated positions, i.e. positions that house specific combinations of residues more prevalent than expected from random distribution. They often form complex networks and are likely to be involved in activity, specificity and stability. Their exact role is however currently still obscure. Therefore, in this work we compared correlated mutations in two related specificities (sucrose and sucrose 6’-phosphate phosphorylase) and experimentally evaluated hypotheses on different aspects of the correlation network. Distribution across the structure suggests that correlated positions might be involved in supporting the different active site topologies, rather than directly being involved in substrate contact. Retracing the evolution showed that correlated positions that are close to each other in the structure or that are strongly correlated often not evolve simultaneously. Introduction of the active site of sucrose 6’-phosphate phosphorylase in sucrose phosphorylase enable this latter enzyme to catalyse the former reactions. The activity was however very low en therefore correlated positions were mutated as well. This caused a decrease in activity (up to fifty- fold) in most cases, but in one double mutant the activity could selectively be increased by a factor of two. Mutation of correlated positions can thus alter the specificity, but general guidelines on which positions to mutate and which residues to introduce, could unfortunately not be derived and will need further research.Stable enzymes are important as already discussed previously. They can be obtained from natural diversity, but not every specificity is available and therefore stable variants need to be created by engineering existing enzymes. In that respect, in silico prediction of stabilising mutations could greatly reduce time, cost and effort compared to current in vitro screening procedures. Here, we evaluated the potential of the FoldX algorithm. All possible point mutations were introduced in a sucrose phosphorylase and those predicted to improve the free energy of folding or the dimer interaction energy were visually inspected for unreasonable mutations. Two thirds were rejected during this manual selection, mostly because hydrophobic residues became too solvent exposed. From the remaining mutants, the nine most promising were experimentally tested. Unfortunately none proved to be more stable: four appeared to be neutral or nearly neutral, while five were deleterious. Remarkably, the mutant predicted to be the most stabilising drastically impaired stability. To explain these findings, molecular dynamics (md) simulations were performed. Root mean square fluctuations (rmsf), representing local flexibility, were found to be bad predictors of stability. Examining if the interactions responsible for the predicted stability are maintained over a period of time in contrast could indeed explain the observed effects in many cases.