Man9‐mannosidase, a processing enzyme found in the endoplasmic reticulum (ER), catalyses the removal of three distinct mannose residues from peptide‐bound Man9‐GlcNAc2 oligosaccharides producing a single Man6 isomer [Bause, E., Breuer, W., Schweden, J., Roesser, R. & Geyer, R. (1992) Eur. J. Biochem. 208, 451–457]. We have isolated four Man9‐mannosidase‐specific clones from a human kidney cDNA library and used these to construct a full‐length cDNA of 3250 base pairs. A single open reading frame of 1875 nucleotides encodes a protein of approximately 71 kDa, consistent with data from immunological studies. Analysis of the coding sequence predicts that Man9‐mannosidase is a type II transmembrane protein consisting of a short cytoplasmic polypeptide tail, a single transmembrane domain acting as a non‐cleavable signal sequence and a large luminal catalytic domain. This domain architecture closely resembles that of other ER and Golgi‐located processing enzymes, pointing to common structural motifs involved in membrane insertion and topology. The protein sequence of the Man9‐mannosidase contains three potential N‐glycosylation sites of which only one site is used. The amino acid sequence of several peptide regions, including a calcium‐binding consensus sequence, bears striking similarities to an ER α‐1,2‐mannosidase from yeast, whereas, by contrast, no sequence similarity was detectable with rat liver ER α‐mannosidase and Golgi α‐mannosidase II. This finding may indicate that the mammalian α‐mannosidases, which differ significantly in their substrate specificity, are coded for by evolutionarily unrelated genes, providing an attractive means of regulation and fine‐tuning oligosaccharide processing, not only at the enzymic but also at the transcriptional level.