Pseudouridine, a non-classical nucleoside present in human urine as a degradation product of RNAs, is one of the few molecules that has a glycosidic C-C bond. Through a data base mining approach involving transcriptomic data, we have molecularly identified two enzymes that are involved in the metabolism of pseudouridine in uropathogenic Escherichia coli, the principal agent of urinary tract infections in humans. The first enzyme, coded by the gene yeiC, specifically phosphorylates pseudouridine to pseudouridine 5'-phosphate. Accordingly, yeiC(-) mutants are unable to metabolize pseudouridine, in contrast to wild-type E. coli UTI89. The second enzyme, encoded by the gene yeiN belonging to the same operon as yeiC, catalyzes the conversion of pseudouridine 5'-phosphate to uracil and ribose 5-phosphate in a divalent cation-dependent manner. Remarkably, the glycosidic C-C bond of pseudouridine is cleaved in the course of this reaction, indicating that YeiN is the first molecularly identified enzyme able to hydrolyze a glycosidic C-C bond. Though this reaction is easily reversible, the association of YeiN with pseudouridine kinase indicates that it serves physiologically to metabolize pseudouridine 5'-phosphate rather than to form it. YeiN is homologous to Thermotoga maritima IndA, a protein with a new fold, which we now show to act also as a pseudouridine-5'-phosphate glycosidase. Data base mining indicates that most eukaryotes possess homologues of pseudouridine kinase and pseudouridine-5'-phosphate glycosidase and that these are most often associated in a single bifunctional protein. The gene encoding this bifunctional protein is absent from the genomes of man and other mammals, indicating that the capacity for metabolizing pseudouridine has been lost late in evolution.