All jawed vertebrates have humoral immune systems that are capable of recognizing a nearly limitless number of potential antigens, yet an individual organisms genome contains a limited number of genes. In order to achieve this stunning adaptability, these limited number of germline gene segments are combined to form the primary antibody repertoire. Further expansion of affinity and specificity is achieved by somatic hypermutation, a process which usually requires T cell help. Carbohydrate antigens are customarily unable to elicit such T cell help and are thus more dependent on the primary germline gene repertoire. Carbohydrate specific antibodies thus make an excellent model to scrutinize how these germline gene segments are capable of balancing the need for specificity, while maintaining the capability of adapting to new antigenic challenges. This thesis explores the structural basis of this adaptability and specificity using carbohydrate-specific antibodies derived from two model systems. The first model system exploits antibodies generated against Chlamydial inner core lipopolysaccharide LPS) carbohydrates, while the second model system is an antibody specific for the Tn antigen Thr/Ser-GalNAc). The crystallization of several homologous Chlamydia-specific antibodies that differ in their fine specificity revealed a conserved Kdo binding pocket and an adaptive binding groove. Small changes in the sequences of CDR H2 and CDR H3 within the binding groove enable these remarkable antibodies to distinguish among varying Chlamydial LPS epitopes. The crucial role of CDR H3 in defining antibody specificity was highlighted by the examination of this series of antibodies of generally similar sequence but with varying CDR H3. The structure of CDR H3 was found to play an important role in defining antibody promiscuity or specificity in binding, and was found to code for both redundant and differential antigen recognition. The structures of these antibodies revealed how specificity is maintained by coding for a conserved Kdo binding pocket, while D and J gene re-arrangements combined with somatic hypermutation allow differential recognition of Chlamydial epitopes, thus providing an important means of antigen adaptability. The structure of the Tn antigen-specific 237mAb revealed a unique mechanism of imposing specificity by having the combining site require the interaction of both a sugar moiety in a binding pocket composed of germline gene residues) and a peptide moiety in a long surface groove). This structure revealed how the immune system can create a highly specific antibody by forcing different regions of the combining site to both contribute to binding, thus preventing cross-reactivity with similar epitopes. Determination of the structures of antibodies from these two models demonstrates how a binding pocket that provides the base specificity of the germline gene segments, combined with sequence variation in CDR H3 allows for adaptability to modifications on the core epitopes. In addition to examining adaptability and specificity in mAbs, this thesis examined the carbohydrate specificity of the toxin aerolysin. Aerolysin is a bacterial channel-forming toxin produced by Aeromonas species. The toxin and its inactive precursor proaerolysin both bind to the conserved glycan core of glycosylphosphatidylinositol GPI)-anchored proteins with high affinity. Here, I report the high resolution structure of proaerolysin in complex with mannose-6-phosphate, which is a component of the GPI anchor core glycan. The structure reveals unambiguous electron density for the monosaccharide and for the residues involved in binding. Trp-127, Arg-323, Trp-324 and Arg-336 all form hydrogen bonds to mannose-6-phosphate and there are electrostatic interactions between the phosphate moiety and the side chain of Arg-336. Trp-127 forms hydrophobic stacking interactions with the mannose ring, which result in the expulsion of water from the binding site. Examination of the carbohydrate specificity of the toxin by surface plasmon resonance SPR) revealed that the toxin did not bind to all GPI anchor structures equally. SPR also suggested that mannose-6-phosphate unlikely represents the true binding determinant of the toxin, but rather only a partial low affinity ligand.