Human endo-alpha1,2-mannosidase is a Golgi-resident type II membrane protein.

The cDNA for human endo-alpha1,2-mannosidase was reconstructed using two independent EST-clones and its properties characterized. The 2837 bp cDNA construct contained a 1389 bp open reading frame (ORF) encoding for 462 amino acids and an approximately 53.6 kDa protein, respectively. Hydrophobicity analysis of this amino acid sequence, as well as proteolytic degradation studies, indicate that the enzyme is a type II protein, anchored in the membrane via a 19 amino-acid long apolar sequence close to the N-terminus. Human endo-alpha1,2-mannosidase displays a high degree of sequence identity with the catalytic domain of the homologous rat liver endo-enzyme, but differs substantially in the N-terminal peptide region, which includes the transmembrane domain. No sequence similarity exists with other processing alpha-glycosidases. Based on sequence information provided by the 2837 bp construct, the cDNA consisting of the complete 1389 bp ORF was amplified by RT-PCR using human fibroblast RNA. Incubation of E. coli lysates with this cDNA, previously modified for boost translation by codon optimization, resulted in the synthesis of an approximately 52 kDa protein which degraded [(14)C]Glc(3)-Man(9)-GlcNAc(2) efficiently, indicating that the catalytic domain of the enzyme folds correctly under cell-free conditions. Transfection of the endo-alpha1,2-mannosidase wild-type cDNA into COS 1 cells resulted in a moderate (approximately 1.5-fold) but reproducible increase of activity compared with control cells, whereas >18-fold increase in activity was measured after expression of a chimera containing green-fluorescent-protein (GFP) attached to the N-terminus of the endo-alpha1,2-mannosidase polypeptide. This, together with the observation that GFP-endo-alpha1,2-mannosidase is expressed as a Golgi-resident type II protein, points to enzyme-specific parameters directing folding and membrane anchoring, as well as Golgi-targeting, not being affected by fusion of GFP to the endo-alpha1,2-mannosidase N-terminus.

(Arg)3 within the N-terminal domain of glucosidase I contains ER targeting information but is not required absolutely for ER localization.

Glucosidase I is an endoplasmic reticulum (ER) type II membrane enzyme that cleaves the distal alpha1,2-glucose of the asparagine-linked GlcNAc2-Man9-Glc3 precursor. To identify sequence motifs responsible for ER localization, we prepared a protein chimera by transferring the cytosolic and transmembrane domain of glucosidase I to the luminal domain of Golgi-Man9-mannosidase. The GIM9 hybrid was overexpressed in COS 1 cells as an ER-resident protein that displayed alpha1,2-mannosidase activity, excluding the possibility that the glucosidase I-specific domains interfere with folding of the Man9-mannosidase catalytic domain. After substitution of the Args in position 7, 8, or 9 relative to the N-terminus by leucine, the GIM9 mutants were transported to the cell surface indicating that the (Arg)3 sequence functions as an ER-targeting motif. Cell surface expression was also observed after substitution of Arg-7 or Arg-8 but not Arg-9 in GIM9 by either lysine or histidine. Thus the side chain structure, including its positive charge, appears to be essential for signal function. Analysis of the N-linked glycans suggests that the (Arg)3 sequence mediates ER localization through Golgi-to-ER retrograde transport. Glucosidase I remained localized in the ER after truncation or mutation of the N-terminal (Arg)3 signal, in contrast to comparable GIM9 mutants. ER localization was also observed with an M9GI chimera consisting of the cytosolic and transmembrane domain of Man9-mannosidase and the glucosidase I catalytic domain. ER-specific targeting information must therefore be provided by sequence motifs contained within the glucosidase I luminal domain. This structural information appears to direct ER localization by retention rather than by retrieval, as concluded from N-linked Man9-GlcNAc2 being the major glycan released from the wild-type enzyme.

Processing of N-linked carbohydrate chains in a patient with glucosidase I deficiency (CDG type IIb).

Recently, we reported a novel congenital disorder of glycosylation (CDG-IIb) caused by severe deficiency of the glucosidase I. The enzyme cleaves the alpha1,2-glucose residue from the asparagine-linked Glc(3)-Man(9)-GlcNAc(2) precursor, which is crucial for oligosaccharide maturation. The patient suffering from this disease was compound-heterozygous for two mutations in the glucosidase I gene, a T-->C transition in the paternal allele and a G-->C transition in the maternal allele. This gives rise in the glucosidase I polypeptide to the substitution of Arg486 by Thr and Phe652 by Leu, respectively. Kinetic studies using detergent extracts from cultured fibroblasts showed that the glucosidase I activity in the patient's cells was < 1% of the control level, with intermediate values in the parental cells. No significant differences in the activities of other processing enzymes, including oligosaccharyltransferase, glucosidase II, and Man(9)-mannosidase, were observed. By contrast, the patient's fibroblasts displayed a two- to threefold higher endo-alpha1,2-mannosidase activity, associated with an increased level of enzyme-specific mRNA-transcripts. This points to the lack of glucosidase I activity being compensated for, to some extent, by increase in the activity of the pathway involving endo-alpha1,2-mannosidase; this would also explain the marked urinary excretion of Glc(3)-Man. Comparative analysis of [(3)H]mannose-labeled N-glycoproteins showed that, despite the dramatically reduced glucosidase I activity, the bulk of the N-linked carbohydrate chains (>80%) in the patient's fibroblasts appeared to have been processed correctly, with only approximately 16% of the N-glycans being arrested at the Glc(3)-Man(9-7)-GlcNAc(2) stage. These structural and enzymatic data provide a reasonable basis for the observation that the sialotransferrin pattern, which frequently depends on the type of glycosylation disorder, appears to be normal in the patient. The human glucosidase I gene contains four exons separated by three introns with exon-4 encoding for the large 64-kDa catalytic domain of the enzyme. The two base mutations giving rise to substitution of Arg486 by Thr and Phe652 by Leu both reside in exon-4, consistent with their deleterious effect on enzyme activity. Incorporation of either mutation into wild-type glucosidase I resulted in the overexpression of enzyme mutants in COS 1 cells displaying no measurable catalytic activity. The Phe652Leu but not the Arg486Thr protein mutant showed a weak binding to a glucosidase I-specific affinity resin, indicating that the two amino acids affect polypeptide folding and active site formation differently.

Lysine can be replaced by histidine but not by arginine as the ER retrieval motif for type I membrane proteins.

The OST48 subunit of the oligosaccharyltransferase complex is a type I membrane protein containing three lysines in its cytosolic domain. The two lysines in positions 3 and 5 from the C-terminus are able to direct protein localisation within the endoplasmic reticulum (ER) by COPI-mediated retrieval. Substitution of these lysines by arginine resulted in cell-surface expression of OST48, whereas ER residency was maintained when either Lys-5 or Lys-3 but not both was replaced with arginine. Localisation of OST48 was not affected by substitution of the two lysines by histidine, indicating that a His-Xaa-His sequence, in contrast to Arg-Xaa-Arg, contains ER-specific targeting information. These differences show that simple charge interactions are not sufficient for ER retention and that other structural factors also play a role. The His-Xaa-His sequence could represent a new and independent signal for directing ER localisation differing from both the arginine motif in type II proteins and the lysine motif in type I proteins. Our data do not exclude, however, that the histidine sequence simply mimics the lysine motif as a sorting signal, being recognised by and interacting with the same receptor subunit(s) in COP-I-coated vesicles. Conclusions arising from this assumption involving the conformation of lysine at the putative COP-I binding site and the failure of Arg-Xaa-Arg to mediate ER localisation for type I proteins are discussed.