We report the primary structures of human and rabbit brush border membrane beta-glycosidase complexes (pre-pro-lactase-phlorizin hydrolase, or pre-pro-LPH, EC 220.127.116.11-62), as deduced from cDNA sequences. The human and rabbit primary translation products contain 1927 and 1926 amino acids respectively. Based on the data, as well as on peptide sequences and further biochemical data, we conclude that the proteins comprise five domains: (i) a cleaved signal sequence of 19 amino acids; (ii) a large 'pro' portion of 847 amino acids (rabbit), none of which appears in mature, membrane-bound LPH; (iii) the mature LPH, which contains both the lactase and phlorizin hydrolase activities in a single polypeptide chain; (iv) a membrane-spanning hydrophobic segment near the carboxy terminus, which serves as membrane anchor; and (v) a short hydrophilic segment at the carboxy terminus, which must be cytosolic (i.e. the protein has an Nout-Cin orientation). The genes have a 4-fold internal homology, suggesting that they evolved by two cycles of partial gene duplication. This repetition also implies that parts of the 'pro' portion are very similar to parts of mature LPH, and hence that the 'pro' portion may be a water-soluble beta-glycosidase with another cellular location than LPH. Our results have implications for the decline of LPH after weaning and for human adult-type alactasia, and for the evolutionary history of LPH.