A deficiency in dolichyl-P-glucose:Glc1Man9GlcNAc2-PP-dolichyl alpha3-glucosyltransferase defines a new subtype of congenital disorders of glycosylation.

The underlying causes of type I congenital disorders of glycosylation (CDG I) have been shown to be mutations in genes encoding proteins involved in the biosynthesis of the dolichyl-linked oligosaccharide (Glc(3)Man(9)GlcNAc(2)-PP-dolichyl) that is required for protein glycosylation. Here we describe a CDG I patient displaying gastrointestinal problems but no central nervous system deficits. Fibroblasts from this patient accumulate mainly Man(9)GlcNAc(2)-PP-dolichyl, but in the presence of castanospermine, an endoplasmic reticulum glucosidase inhibitor Glc(1)Man(9)GlcNAc(2)-PP-dolichyl predominates, suggesting inefficient addition of the second glucose residue onto lipid-linked oligosaccharide. Northern blot analysis revealed the cells from the patient to possess only 10-20% normal amounts of mRNA encoding the enzyme, dolichyl-P-glucose:Glc(1)Man(9)GlcNAc(2)-PP-dolichyl alpha3-glucosyltransferase (hALG8p), which catalyzes this reaction. Sequencing of hALG8 genomic DNA revealed exon 4 to contain a base deletion in one allele and a base insertion in the other. Both mutations give rise to premature stop codons predicted to generate severely truncated proteins, but because the translation inhibitor emetine was shown to stabilize the hALG8 mRNA from the patient to normal levels, it is likely that both transcripts undergo nonsense-mediated mRNA decay. As the cells from the patient were successfully complemented with wild type hALG8 cDNA, we conclude that these mutations are the underlying cause of this new CDG I subtype that we propose be called CDG Ih.


The underlying causes of type I congenital disorders of glycosylation (CDG I) have been shown to be mutations in genes encoding proteins involved in the biosynthesis of the dolichyl-linked oligosaccharide (Glc 3 Man 9 GlcNAc 2 -PP-dolichyl) that is required for protein glycosylation.
Here we describe a CDG I patient displaying gastrointestinal problems but no central nervous system deficits. Fibroblasts from this patient accumulate mainly Man 9 GlcNAc 2 -PP-dolichyl, but in the presence of castanospermine, an endoplasmic reticulum glucosidase inhibitor Glc 1 Man 9 GlcNAc 2 -PP-dolichyl predominates, suggesting inefficient addition of the second glucose residue onto lipid-linked oligosaccharide. Northern blot analysis revealed the cells from the patient to possess only 10 -20% normal amounts of mRNA encoding the enzyme, dolichyl-P-glucose:Glc 1 Man 9 GlcNAc 2 -PP-dolichyl ␣3-glucosyltransferase (hALG8p), which catalyzes this reaction. Sequencing of hALG8 genomic DNA revealed exon 4 to contain a base deletion in one allele and a base insertion in the other. Both mutations give rise to premature stop codons predicted to generate severely truncated proteins, but because the translation inhibitor emetine was shown to stabilize the hALG8 mRNA from the patient to normal levels, it is likely that both transcripts undergo nonsensemediated mRNA decay. As the cells from the patient were successfully complemented with wild type hALG8 cDNA, we conclude that these mutations are the underlying cause of this new CDG I subtype that we propose be called CDG Ih.
Type I congenital disorders of glycosylation (CDG I) 1 are often severe multisystemic diseases characterized by the pres-ence of hypoglycosylated glycoproteins in the serum of affected individuals (1)(2)(3)(4)(5). Although glycoproteins play vital roles in many aspects of human cellular physiology (6), the precise relationship between glycoprotein hypoglycosylation and the clinical symptoms of these diseases that include hypotonia, seizures, failure to thrive, psychomotor retardation, and various dysmorphias is not understood (2).
Hypoglycosylation of glycoproteins bearing N-glycans is caused by either an insufficiency in the biosynthesis of the lipidlinked oligosaccharide (LLO) precursor, Glc 3 Man 9 GlcNAc 2 -PPdolichyl, that is required for protein glycosylation or inefficient transfer of the sugar moiety of this LLO onto nascent glycoproteins in the lumen of the endoplasmic reticulum (ER). Theoretically, mutations in any of the genes encoding for the 30 or so proteins involved in this biosynthetic pathway could lead to glycoprotein hypoglycosylation in CDG I patients. However, mutations in only 7 of the genes encoding proteins of the glycosylation pathway have so far been shown to underlie CDG I, and these 7 cases have been classified as CDG I subtypes a-g (Ia, phosphomannomutase 2 (7,8); Ib, phosphomannose isomerase (9, 10); Ic, dolichyl-P-Glc:Man 9 GlcNAc 2 -PP-dolichyl ␣3-glucosyltransferase (11)(12)(13); Id, dolichyl-P-Man:Man 5 GlcNAc 2 -PP-dolichyl ␣3mannosyltransferase (14); Ie, dolichol-P-Man synthase I (15,16); If, the MPDU1 gene product known to facilitate dolichyl-P-Glc and dolichyl-P-Man utilization (17,18); and Ig, dolichyl-P-Man: Man 7 GlcNAc 2 -PP-dolichyl ␣6-mannosyltransferase (19 -21)). Although there are too few patients representing each subtype of the disease to draw precise genotype/phenotype relationships, CDG Ib (PMI deficiency) generally presents as a disease in which central nervous system defects are absent (9,10). Often, PMI deficiency leads to a less severe form of the disease because as well as the PMI-catalyzed conversion of fructose 6-phosphate to mannose 6-phosphate the cell possesses an alternative route for the generation of the latter intermediate. In fact, serum mannose can be taken up by cells (22,23) and phosphorylated by hexokinase to yield mannose 6-phosphate. In PMI deficiency, the flux through this alternative metabolic route can be augmented by giving patients oral mannose, a treatment that has been shown to reverse serum glycoprotein hypoglycosylation and alleviate the symptoms of this form of the disease (24,25).
Here we report on a patient presenting clinical symptoms similar to those observed in CDG Ib patients but whose PMI levels were found to be normal. We demonstrate that cells derived from this patient accumulate Glc 0 -1 Man 9 GlcNAc 2 -PPdolichyl and display dramatically reduced levels of mRNA encoding dolichyl-P-glucose:Glc 1 Man 9 GlcNAc 2 -PP-dolichyl ␣3glucosyltransferase (hALG8p). Sequencing of the hALG8 genomic DNA from the patient revealed each allele to possess mutations that generate premature stop codons.

EXPERIMENTAL PROCEDURES
Western Blot-Western blotting of serum transferrin was performed as described previously (26) using a rabbit polyclonal anti-transferrin antibody. Phosphomannomutase (27,28) and phosphomannose isomerase (29) activities were assayed as described previously.
Cells, Cell Culture, and Metabolic Radiolabeling of Cells-Skin biopsy fibroblasts, obtained from patient M. P. and a patient diagnosed with CDG Ic (25), were grown in Dulbecco's modified Eagle's medium containing 2 g/liter glucose, 10% fetal calf serum, and 1% penicillin/ streptomycin. Primary skin fibroblasts were immortalized by Dr. Thierry Levade (INSERM U466), as reported previously (30), and cultivated as described above. EBV-transformed lymphoblasts were generated from peripheral blood mononuclear cells isolated using a Ficoll-Paque Plus gradient and were grown in RPMI 1640 medium supplemented as described above. Confluent fibroblasts in 25-cm 2 tissue culture flasks or EBV-transformed lymphoblasts were pulse-radiolabeled for 30 min with 100 Ci of [2-3 H]mannose (23.9 Ci/mmol, PerkinElmer Life Sciences) in 1 ml of glucose-free Dulbecco's modified Eagle's medium or RPMI 1640 supplemented with 0.5 mM glucose and 5% dialyzed fetal calf serum. Where appropriate, the glycosidase inhibitors castanospermine (CST, Cambridge Research Biochemicals, Northwich, UK) and kifunensin (KIF, Toronto Research Chemicals Inc.) were added to the cells 30 min prior to the onset of the radiolabeling period, at concentrations of 2 mM and 100 M, respectively.
Analysis of LLO Glucosylation in Streptolysin-O-permeabilized Lymphoblasts-EBV-transformed lymphoblasts (2 ϫ 10 8 cells) were pulseradiolabeled with 200 Ci of [2-3 H]mannose as described above. The cells were then permeabilized with streptolysin-O (SLO, Bacto-Streptolysin O, product reference 0482, BD Biosciences) as described previously (31). Briefly, cells were washed into cell permeabilization buffer (PB: 130 mM K ϩ /glutamate, 10 mM NaCl, 2 mM EGTA, 1 mM CaCl 2 , 2 mM MgCl 2 , 5 mM HEPES/KOH, pH 7.3, containing 2 mg/ml bovine serum albumin and 1 mM dithiothreitol) and incubated with 5 units of SLO (dissolved in PB) for 20 min at 4°C. After washing twice with ice-cold PB, the cells were warmed to 37°C for 5 min and then placed on ice for 30 min. The cells were again washed in PB prior to being resuspended in the same buffer, dispatched into 1.5-ml centrifuge tubes, and incubated for different times at 37°C in the absence or presence of 2 mM UDP-glucose and/or 4 mM CST.
Isolation and Analysis of Lipid-linked and N-Linked Oligosaccharides-Radiolabeled cells and SLO-permeabilized cells were extracted with organic solvents as described previously (19). Briefly, cells were rinsed with ice-cold phosphate-buffered saline and then suspended in MeOH, 100 mM Tris/HCl, pH 7.4, containing 4 mM MgCl 2 , 2:1. Finally, an equal volume of CHCl 3 was added before vigorous shaking of the cell suspension. After centrifugation, the lower (CHCl 3 ) and upper (methanolic) phases were removed and kept. The interphase proteins were washed once with MeOH and then with H 2 O, and then again with MeOH prior to being extracted twice with 2 ml of CHCl 3 /MeOH/H 2 O, 10:10:3. The lower (CHCl 3 ) and 10:10:3 phases were dried, and hydrolyzed with 0.02 N HCl (32) in order to release oligosaccharides from LLO. The 10:10:3-extracted protein pellet was dried and digested with Pronase prior to being incubated with endo H in order to release polymannose-type oligosaccharides from glycopeptides (33). After desalting on AG-1/AG-50 columns, all oligosaccharide mixtures were resolved by TLC on silica-coated plastic sheets (Merck) in n-propyl alcohol/acetic acid/H 2 O, 3:3:2 for 36 -48 h. Radioactive components were detected by fluorography after spraying the dried plates with En 3 hance® (PerkinElmer Life Sciences). Standard oligosaccharides were generated as follows: Man 5 GlcNAc 2 , Man 5 GlcNAc 2 -P obtained from [2-3 H]mannose-labeled Thy Ϫ1 mouse lymphocytes was treated with alkaline phosphatase (34); a mixture of Glc 1-3 Man 9 GlcNAc 2 oligosaccharides was obtained by mild acid hydrolysis of LLO recovered from [2-3 H]mannose-labeled HepG2 hepatocellular carcinoma cells; Man 9 -8 GlcNAc 2 oligosaccharides were obtained from the cytosolic frac-tion of SLO-permeabilized HepG2 hepatocellular carcinoma cells as described previously (31).
Dolichyl-P-glucose Synthase Assay-EBV-transformed lymphoblasts from a normal individual and patient M. P. were harvested, washed in phosphate-buffered saline, and resuspended in 50 mM Tris/HCl, pH 8.0 (buffer A), at 4°C. The cells were homogenized in buffer A by several passages through a syringe fitted with a 24 ϫ 0.5-mm needle. The homogenate was centrifuged at 1,500 ϫ g Av for 5 min at 4°C, and the resulting supernatant was subjected to ultracentrifugation at 84,000 ϫ g Av for 30 min at 4°C. The resulting pellet was resuspended in buffer A, and protein was determined by the bicinchoninic acid (Sigma) method. Dolichyl-P-glucose synthase was measured by incubating microsomal membranes in a buffer containing 1 mM AMP, 25 mM Tris/HCl, pH 8.0, 5 mM MgCl 2 , 0.1% Triton X-100 and, where appropriate, 1 g of dolichyl phosphate, in a final volume of 50 l. Finally, 0.5 Ci of UDP-[ 3 H]glucose (11.5 Ci/mmol, PerkinElmer Life Sciences) was added to the mixtures which were then incubated at 37°C for 20 min. Reactions were terminated by the addition of 150 l of buffer, 400 l of MeOH, and 600 l of CHCl 3 . After vigorous shaking the tubes were centrifuged, and the upper and inter phases were discarded. The lower CHCl 3 phase, containing dolichyl-P-[ 3 H]glucose, was washed twice with new upper phase prior to being assayed for radioactivity by scintillation counting.
Northern Blot-Cells were rapidly disrupted in 4 M guanidine isothiocyanate, and total RNA was isolated (35). RNA (20 g) was denatured, electrophoresed, transferred onto Positive TM Membrane (Appligene, Illkirch, France), and hybridized with specific probes (see Table I) as described previously (36). In some experiments cells were treated with the translation inhibitor emetine (Sigma, product reference E-2375, dissolved in H 2 O) for 8 h before RNA extraction. 18 S rRNA was monitored using a complementary oligonucleotide (37) or, where indicated, methylene blue staining (38).
Mutation Analysis-The different primers used for PCR, sequencing, subcloning, and hybridization are listed in Table I. All sequencing was performed on both strands and on two independent PCRs. PCR products were purified with the QIAquick PCR purification kit (Qiagen SA France) prior to automated sequencing. The genomic DNA from the patient and parents was extracted with Trizol from peripheral blood mononuclear cells isolated using a Ficoll-Paque Plus gradient. Trizol and Ficoll-Paque were used according to the manufacturer's instructions. DNA containing exon 4 from the ALG8 gene from the patient was subcloned from genomic DNA after PCR amplification using the primers described in Table I. The PCR products obtained in this way were subcloned into the pCR3.1 plasmid employing the Eukaryotic TA Expression Bi-directional kit (Invitrogen).
HIV-1-derived Lentiviral Vectors and Transduction of Fibroblasts-The hALG8 cDNA sequence was amplified from the human expressed sequence tag (accession number AJ224875, I.M.A.G.E. Clone ID 265361, obtained from the I.M.A.G.E. Consortium, Livermore, CA (39)), using the primers indicated in Table I, and subcloned into the pSIN.PW.eGFP HIV-1-derived transfer vector as described previously (19). Transduction and radiolabeling of cells were performed as reported before (19).

A Patient with Gastrointestinal Problems but Without Central Nervous System Deficits Presents with Hypoglycosylated
Serum Glycoproteins-A girl, M. P., the first child of unrelated healthy parents, was referred at 4 months of age for edematoascitic syndrome related to severe hypoalbuminemia resulting from protein losing enteropathy. Upon admission, she had no dysmorphic symptoms and normal psychomotor development, but had severe diarrhea and moderate hepatomegaly. Routine blood tests showed severe hypoalbuminemia (9 g/liter), normal aminotransferase activities, increased cephalin kaolin time (3 N), and low factor XI (12%), protein C (21%), and antithrombin III (17%) levels. Abdominal ultrasonography, echocardiography, and cerebral magnetic resonance imaging were normal, but electroretinography showed slight anomalies. The combined presence of coagulation factor anomalies and protein losing enteropathy was suggestive of CDG. This diagnosis was confirmed upon investigation of the glycosylation status of the serum glycoproteins from the patient by Western blot as shown in Fig. 1A. Although the electrophoretic profile of transferrin derived from serum of a normal subject displays a single band, that from the patient reveals three distinct components whose migration positions coincide with transferrin species observed to occur in patients previously diagnosed with type I CDG (Fig.  1A). Initially, the girl required total parenteral nutrition and albumin infusions because of severe digestive complications. Oral mannose treatment was ineffective, but digestive indications improved with a low fat diet in association with essential fatty acid supplementation. After 18 months of dietary treatment, the diarrhea and protein-losing enteropathy were resolved, but despite normal liver function tests there was mild hyperechogen hepatomegaly without portal hypertension, and coagulation anomalies persisted. Psychomotor development continued normally, and electroretinographic observations remained unchanged.
Accumulation of Hypoglucosylated Dolichyl-linked Oligosaccharides in Skin Biopsy Fibroblasts Obtained from Patient M. P.-Further diagnosis was performed by assaying phosphomannomutase and phosphomannose isomerase (PMI) enzyme activities that are known to be deficient in CDG Ia and Ib, respectively. However, both these activities were found to be normal (phosphomannomutase, 4.1 units/g total protein (normal Ͼ3.4 units/g total protein), and PMI, 8.9 units/g total protein (normal Ͼ 5.5 units/g total protein)). Next, skin biopsy fibroblasts from the patient were subjected to metabolic radiolabeling with [2-3 H]mannose in order to examine LLO biosynthesis in these cells. After mild acid hydrolysis the oligosaccharide moieties of LLO from the patient and normal cells were resolved by TLC. A preliminary experiment revealed that although the control cells yielded predominantly glucosylated LLO, accumulations of LLO whose oligosaccharide structures comigrated with Man 9 GlcNAc 2 , and to a lesser extent Glc 1 Man 9 GlcNAc 2 , were apparent in the cells from the patient (Fig. 1B). Similar results were observed when cells from a CDG Ic patient (deficiency in dolichyl-P-glucose:Man 9 GlcNAc 2 -PPdolichyl ␣3-glucosyltransferase) were examined, but in these fibroblasts the monoglucosylated structure was less apparent (Fig. 1B). These observations suggested that in patient M. P. there is inefficient addition of glucose residues onto the growing LLO in the lumen of the ER. However, enzymic assay of dolichyl-P-glucose synthase (hALG5p) revealed that this enzyme that is responsible for the synthesis of the glucose donor molecules required for LLO glucosylation was not impaired in the fibroblasts from the patient (Fig. 1C). Finally, the dolichyl-P-glucose:Man 9 GlcNAc 2 -PP-dolichyl glucosyltransferase (hALG6) gene from the patient was sequenced, but no mutations were found. Glc 1 Man 9 GlcNAc 2 -PP-dolichyl Accumulates upon Treatment of Fibroblasts from the Patient with the Glucosidase Inhibitor Castanospermine-When normal and M. P. fibroblasts were radiolabeled in the presence of the ER glucosidase I and II inhibitor CST, we noted that LLO profiles were different from those generated in the absence of this agent. Thus, as demonstrated in Fig. 2A, whereas the CHCl 3 phase derived from organic solvent extraction of normal fibroblasts yields LLO containing mainly fully mannosylated oligosaccharides bearing between zero and three glucose residues (Glc 0 -3 Man 9 GlcNAc 2 ), the chloroform/methanol/H 2 O (10:10:3) phase was observed to comprise mainly an LLO possessing the fully glucosylated structure (Glc 3 Man 9 GlcNAc 2 ) that is known to be efficiently transferred onto glycoprotein in the lumen of the ER. Similar examination of the oligosaccharide species derived from LLO recovered from CST-treated normal cells revealed the almost exclusive appearance of fully glucosylated Glc 3 Man 9 GlcNAc 2 in both organic phases. By contrast, CST treatment of the cells from the patient reduced the accumulation of Man 9 GlcNAc 2 but brought about substantial increases of Glc 1 Man 9 GlcNAc 2 as well as Glc 3 Man 9 GlcNAc 2 (Fig. 2B). When a similar experiment was conducted on fibroblasts derived from a patient diagnosed as having CDG Ic (13,14), the glucosidase inhibitor had no effect on the accumulation of Man 9 GlcNAc 2 , and furthermore, no Glc 1 Man 9 GlcNAc 2 was apparent under these conditions (results not shown). These results can be explained by the glucosyltransferase-glucosidase shuttle proposed by Spiro and co-workers (40 -42). According to this mechanism, represented in Fig. 2B, there are two ways in which Man 9 GlcNAc 2 -PP-dolichyl can be formed in mammalian cells. These authors showed that in addition to nascent glucose-containing glycoproteins in the lumen of the ER, glucosylated LLO are also susceptible to trimming by ER glucosidases, and it was hypothesized that the ability to both add and remove glucose residues LLOs recovered from the CHCl 3 phase were treated with mild acid, and the released oligosaccharides were analyzed by TLC as described under "Experimental Procedures." Radioactive components were visualized by fluorography, and the migration positions of standard oligosaccharides are indicated: M 5 , Man 5 GlcNAc 2 ; M 9 , Man 9 GlcNAc 2 ; G 1 M 9 , Glc 1 Man 9 GlcNAc 2 ; G 2 M 9 , Glc 2 Man 9 GlcNAc 2 ; G 3 M 9 , Glc 3 Man 9 GlcNAc 2 . C, lymphoblasts from a normal subject (N) and patient M. P. (MP) were assayed for dolichyl-P-glucose synthase activity, in both the absence (ϪDol-P) and presence (ϩDol-P) of dolichyl phosphate as described under "Experimental Procedures." from LLO allows the cell to regulate the pool size of mature triglucosylated LLO. Thus, the accumulation of Man 9 GlcNAc 2 -PP-dolichyl observed in patient M. P. may be due to the deglucosylation, by ER glucosidase II, of a pool of Glc 1 Man 9 GlcNAc 2 -PP-dolichyl that has arisen due to inefficient addition of the second glucose residue by the hALG8 gene product (dolichyl-P-glucose:Glc 1 Man 9 GlcNAc 2 -PP-dolichyl glucosyltransferase, see Fig. 2B).
Examination of LLO Glucosylation in Intact and SLO-permeabilized Lymphoblasts Derived from Patient M. P.-Analysis of LLO biosynthesis in EBV-transformed lymphoblasts from patient M. P. revealed similar results to those obtained from fibroblasts, except that the addition of CST to these cells caused a less dramatic redistribution of glucosylated LLO species when compared with that observed for fibroblasts (results not shown). Therefore, EBV-transformed lymphoblasts from this subject were permeabilized with the plasma membrane pore-forming reagent SLO in order to perform in vitro analyses of LLO glucosylation. Permeabilized cells were incubated in either the absence or presence of UDP-Glc as shown in Fig. 3A.
Under these conditions there is no transfer of oligosaccharide from LLO onto protein (43), and after these brief 10-min incubations less than 10% of the LLO fraction is lost (probably as free oligosaccharides (43)). After permeabilization, LLO in both the patient's and normal cells are found to be partially deglucosylated so that in both cases Man 9 GlcNAc 2 -PP-dolichyl is a major species. Upon incubation of permeabilized cells with UDP-Glc, LLO is glucosylated such that the majority of LLO possesses three glucoses after 5 min. However, it is evident that successful triglucosylation of LLO is less efficient in the cells from the patient and that the abnormality in LLO glucosylation is at the level of Glc 1 Man 9 GlcNAc 2 -PP-dolichyl. Crucially, these experiments revealed that the initial quantity of Man 9 GlcNAc 2 -PP-dolichyl declines at the same rate in both cell types, indicating that addition of the first glucose residue is not  Fig. 1B. B, the glucosyltransferase/glucosidase shuttle which has been proposed by Spiro and coworkers (40 -42) to operate in mammalian cells. In human cells the biosynthetic steps are thought to be carried out by the three glucosyltransferases which are now known to be encoded by the human orthologs of the yeast ALG6, ALG8, and ALG10 loci. The degradative steps have been proposed to be catalyzed by ER glucosidases I and II (GLS1, and -2 gene products, respectively). The permeabilized cells were then incubated in either the absence or presence of UDP-Glc for the indicated times at 37°C. Subsequently, the cells were extracted with organic solvents, and LLO recovered from the CHCl 3 , and 10:10:3 phases were pooled. 40,000 cpm of the oligosaccharides liberated from LLO by mild acid hydrolysis were resolved by thin layer chromatography as described in the legend to Fig. 1. The abbreviations used are as described for Fig. 1. B, a similar experiment was performed in either the presence of UDP-Glc alone (left panel) or UDP-Glc and 4 mM CST (right panel). After resolution of oligosaccharides released from LLO, the indicated radioactive species were eluted from the thin layer chromatography plate and quantitated by scintillation counting. The recovery of each oligosaccharide is expressed as the % of the total (M 9 ϩ G 1 M 9 ϩ G 2 M 9 ϩ G 3 M 9 ). Open circles, cells from control subject; closed circles, cells from patient M. P. limiting in M. P. cells. Furthermore, when the same experiments were performed in the presence of CST, we were still able to detect the Glc 1 Man 9 GlcNAc 2 -PP-dolichyl intermediate, demonstrating that this component is not generated by ER glucosidase I/II action on fully glucosylated LLO (Fig. 3B). Finally, in both cell populations LLO glucosylation also occurs to a lesser extent in the absence of UDP-Glc and is probably driven by an endogenous pool of dol-P-Glc.
Reduced hALG8 mRNA Expression in Cells from Patient M. P.-In yeast, the ALG8 gene thought to encode the dolichyl-P-glucose:Glc 1 Man 9 GlcNAc 2 -PP-dolichyl ␣3-glucosyltransferase has been cloned (44), and the complete sequence of the putative human ortholog of this gene (EMBL: BC001133 and AJ224875) is available (45). This information allowed us to create a probe in order to examine hALG8 mRNA in the cells from the patient by Northern blot. As demonstrated in Fig. 4A, there was a dramatic reduction in the quantity of the hALG8 message in the cells from the patient when compared with that observed in normal fibroblasts. By contrast, the level of the hALG6 message was similar in the two cell populations. As the quantity of the hALG8 message in the cells from the patient was less than 10% of that observed in control cells, it is apparent that both of the transcripts of the alleles from the patient at this locus are affected (Fig. 4A).

Identification of Two Mutations Leading to Premature Stop Codons in the ALG8 Genomic DNA from the Patient-
The availability of the hALG8 cDNA sequence allowed us to find five partial genomic sequences from chromosome 11q14. By using these data, we were able to define the structure of the gene that comprised 13 exons (Fig. 4B). All the exon/intron boundaries follow the AG/GT rule, with the exception of intron 6 which starts with GC instead of GT. This observation was confirmed in the four available genomic sequences comprising exon 6 and the entire 276 bp of intron 6. By using the hALG8 gene structure, we designed intronic primers (see Table I) in order to amplify the different exons. The ALG8 alleles from the patient were sequenced and compared with those obtained from the genomic DNA from the parents. As shown in Fig. 5A, two mutations were found in exon 4 of the ALG8 sequence from the patient: the allele originating from the father of the patient contained a deletion (413 del C) and that originating from the mother contained an insertion (396 ins A). In order to read the patient's sequence between these two mutations, we subcloned this region from genomic DNA. Twenty independent clones corresponding to either of the two alleles were sequenced and found to occur in equal proportions, and additional mutations in this region of the gene were not found. The 396 ins A and 413 del C mutations generated premature stop codons (underlined bases in Fig. 5A) whose translation is predicted to generate severely truncated polypeptides (Fig. 5B). We also found a G665A variation in the sequence of exon 6; the father was found to be heterozygous at this position, but taking into account that both the mother and the patient possess only A at this position, and that this variation occurs after the two premature stop codons, we believe that this variation does not lead to the hALG8p deficiency and may correspond to a polymorphism.
Effect of Translation Inhibition on the Quantity of ALG8 mRNA Recovered from Normal and M. P. Lymphoblasts-It is known that the presence of premature stop codons can lead to atypeofmRNAdegradationthatisaccomplishedbyatranslationdependent process known as nonsense-mediated mRNA decay (NMD) (46,47). In order to examine the possibility that the low expression of ALG8 mRNA in cells from patient M. P. is the result of such a process, we treated the cells from the patient and the control with different concentrations of the translation inhibitor emetine, as has been described previously (48). As shown in Fig. 6, Northern blot analysis reveals that emetine provokes a concentration-dependent increase of ALG8 mRNA in both control and M. P. lymphoblasts suggesting that the message is stabilized in both cell lines. However, whereas the message is only stabilized 3.5-fold in the control cells treated with 100 g/ml emetine, a 20-fold increase is observed in the cells from the patient.
Transduction of M. P. Fibroblasts with hALG8 cDNA-In order to demonstrate unambiguously that a deficiency in hALG8p is the cause of the accumulation of underglucosylated LLO in cells from patient M. P., we have transduced immortalized M. P. fibroblasts with hALG8 cDNA using HIV-1-de- Total RNA was extracted from preconfluent fibroblasts. Aliquots of the same total RNA preparation were loaded in duplicate on the gel. After blotting, the membrane was cut in two. One membrane was hybridized with the ALG8 probe and the other with the ALG6 probe. Whereas the membrane that was hybridized with the ALG8 probe was stripped and rehybridized with the 18 S probe, the membrane that was hybridized with the ALG6 probe was stained with methylene blue. The blot revealed identical coloring of the 18 S region for all the lanes (not shown). The probes used for visualizing hALG6 and hALG8 mRNA were generated by 32 P labeling of the PCR products amplified by primer couple 6S/6AS or 1BisS/7TerAS (see Table I), respectively. B, genomic organization of the human ALG8 gene was determined by LFASTA (61) comparison of two mRNA sequences (BC001133 and AJ224875) with 5 partial genomic sequences from chromosome 11q14 (GenBank TM accession numbers AP001805, AC023532, AP002520, AP000571, and AP001447). Exon (uppercase letters) and intron (lowercase letters) boundaries follow the ag/gt (in bold) rule except for intron 6. The numbering of the cDNA sequence starts with ATG. rived lentiviral transfer vectors. Results presented in Fig. 7 show that whereas a vector harboring GFP cDNA alone had little effect on the distribution of LLO in CST-treated M. P. fibroblasts, a vector containing both hALG8 and GFP cDNA markedly reduced the appearance of underglucosylated LLO in the cells from the patient and restored the distribution of LLO to a pattern similar to that observed in normal, GFP-transduced, fibroblasts treated with CST. In this experiment it was noted that the distribution of LLO between the CHCl 3 and 10:10:3 organic phases was different from that usually observed. Although incubation of cells with CST favors the recovery of triglucosylated LLO in the 10:10:3 phase (see Fig. 2A), we noted much reduced quantities of LLO in the CHCl 3 organic phase in this experiment. At present the reason for this is not clear, but it is noteworthy that there is significant variation in the distribution of LLO between the two organic phases depending on the cell type and treatment (M. P. cells transduced with GFP; 31% total LLO in CHCl 3 phase, M. P. cells transduced with GFP/ALG8; 10% total LLO in CHCl 3 phase).
Transfer of Underglucosylated Oligosaccharides from LLO onto Glycoprotein in M. P. Fibroblasts-Finally, we examined the nature of oligosaccharides that are transferred from LLO onto polypeptides in the lumen of the ER. Cells were treated with CST and KIF, an inhibitor of ER mannosidase I and Golgi mannosidase. When fibroblasts were pulse-radiolabeled with [2-3 H]mannose under these conditions, N-linked oligosaccharides remain untrimmed and may reflect the structures that are transferred from LLO onto polypeptides. Accordingly, whereas in uninhibited control fibroblasts the predominant N-linked oligosaccharides are Glc 0 -1 Man 9 GlcNAc and Man 8 GlcNAc, in inhibited cells Glc 3 Man 9 GlcNAc is the predominant species detected (Fig. 8, A and B). Similar observations were made for the cells from the patient, but, in inhib-ited cells, we noted a modest increase in the proportion of an oligosaccharide migrating as Glc 1 Man 9 GlcNAc, when compared with that occurring in CST ϩ KIF-treated control cells, indicating that either this structure is transferred directly from LLO onto polypeptide or that Man 9 GlcNAc is transferred onto polypeptide prior to being post-translationally monoglucosylated by UDP-glucose:glycoprotein glucosyl transferase (UGGT).

DISCUSSION
In the present work we demonstrate that cells from a child displaying serum glycoprotein hypoglycosylation, and some clinical symptoms suggestive of type I CDG, reveal an inefficiency in their ability to add the second glucose residue onto LLO. Under normal radiolabeling conditions the predominant LLO is Man 9 GlcNAc 2 -PP-dolichyl. However, we were unable to detect any changes in the expression, or mutations, in the gene that encodes dolichyl-P-Glc:Man 9 GlcNAc 2 -PP-dolichyl glucosyltransferase, nor were we able to detect a reduction in dolichyl-P-glucose synthase activity in the cells from this patient. In fact, striking accumulations of monoglucosylated LLO only became apparent when M. P. fibroblasts were treated with CST, the ER glucosidase I and II inhibitor. Although the LLO species recovered from the 10:10:3 organic phase are generally more heavily glucosylated than those species recovered from the CHCl 3 fraction, our experience with primary human fibroblasts has shown that the proportion of total LLO that is glucosylated is quite variable. At present the factors that lead to this variability are not understood. However, where control pulse-radiolabeling experiments do yield large quantities of Man 9 GlcNAc 2 -PP-dolichyl, parallel incubations conducted in the presence of CST lead to a block in the appearance of this structure and a concomitant appearance of triglucosylated LLO, suggesting that the glucosyltransferase/glucosidase shuttle functions under certain cellular growth/stress conditions. Indeed it has been shown that the deglucosylating reactions occur when tissue slices or cells are deprived of oxygen (41), and it has been suggested that this stress leads to inefficient glucosylation of LLO due to reduced availability of dolichyl-Pglucose (41). More evidence for an ALG8p deficiency in M. P. cells was obtained by examining LLO glucosylation in SLOpermeabilized lymphoblasts. In this in vitro system we were able to demonstrate that whereas M. P. cells could efficiently add the first glucose to LLO, the addition of the second glucose residue was slow when compared with that observed in control cells. It was noted that in the absence of exogenously added UDP-Glc, the cells from the patient accumulated predominantly Man 9 GlcNAc 2 -PP-dolichyl, whereas in the presence of the sugar nucleotide Glc 1 Man 9 GlcNAc 2 -PP-dolichyl was more abundant than Man 9 GlcNAc 2 -PP-dolichyl. Thus, successful triglucosylation of LLO in cells from the patient may be particularly sensitive to cellular UDP-Glc levels that are known to fluctuate during hypoxia (49) and glucose insufficiency (50). Interestingly, examination of LLO biosynthesis in a yeast strain deficient in ALG8p reveals the accumulation of Glc 1 Man 9 GlcNAc 2 -PP-dolichyl and not Man 9 GlcNAc 2 -PP-dolichyl as has been observed here in M. P. fibroblasts (51). It is not clear why yeast and mammalian cells deficient in ALG8p should behave differently in this respect, but although mammalian glucosidases have been shown to be active toward glucosylated LLO (52,53), we have been unable to find any data in the literature indicating that the yeast glucosidases are active toward this substrate. In fact, in a yeast strain deficient in both ALG8p and ER glucosidase I, the structure Glc 2 Man 9 GlcNAc 2 was detected N-linked to protein (51). In that study it was proposed that the elevated levels of Glc 1 Man 9 GlcNAc 2 -PP-dolichyl, caused on the one hand by the absence of ALG8p and on the other by the slow transfer of Glc 1 Man 9 GlcNAc 2 onto protein, allowed ALG10p to "cap" the monoglucosylated LLO with an ␣2-linked glucose residue. Apparently then, in yeast and mammalian cells accumulated Glc 1 Man 9 GlcNAc 2 -PP-dolichyl may have different fates, and in the latter cell type, under certain conditions, this structure can be deglucosylated, whereas in yeast it can be glucosylated by ALG10p.
Whatever the significance of the glucosyltransferase/glucosidase shuttle, it is apparent that treatment of cells from the patient with CST gave us an early clue as to the underlying defect in this patient. Indeed, incubations performed in the FIG. 5. The genomic DNA from the patient contains two mutations that lead to premature stop codons in the hALG8 gene. A, the region of hALG8 exon 4 in which two mutations were found (numbering of the cDNA sequence starts with ATG). The mutated allele from the father contained a deletion (413 del C), and the mutant allele from the mother contained an insertion (396 ins A). Both mutations induced frame shifts that led to premature stop codons (underlined). B, amino acid sequences were generated from normal hALG8, and the two mutant alleles and Kyte-Doolittle (62) hydropathy plots were derived. A hydropathy index greater than 2 often indicates the presence of a series of amino acids capable of forming a transmembrane region. New peptide sequences between the frameshift and the premature stop codon are underlined.  Fig. 4. B, ALG8 mRNA was quantitated by densitometric scanning, and the results were normalized with respect to the quantity of 18 S ribosomal RNA detected in the same gel lanes.
presence of the glucosidase inhibitor allowed us to identify Glc 1 Man 9 GlcNAc 2 -PP-dolichyl as the primary accumulating LLO intermediate in this case, and it is likely that, in other CDG Ix cases where Man 9 GlcNAc 2 -PP-dolichyl is seen to accumulate, the use of CST will help pinpoint which of the three glucosyltransferases (hALG6p, hALG8p, or hALG10p) is at fault.
We went on to demonstrate that in cells derived from this patient, there is a dramatically reduced expression of the gene (hALG8) which encodes the enzyme that attaches the second glucose onto growing LLO. First, the patient possesses very low levels of hALG8 mRNA, and second, the gene from the patient contains two mutations in exon 4 which lead to PSCs. The reduced expression of both the ALG8 alleles from the patient could result from either a reduced transcription rate or a decrease in message stability. NMD is responsible for the degradation of mRNA containing PSCs, but this process only takes place if the PSC is ϳ50 bp from the last intron/exon junction of the gene in question (46,47), which is the case for both the PSCs found in the patient. As NMD is known to be a translationdependent process, we treated EBV-transformed lymphoblasts from patient M. P. with emetine, a translation inhibitor known to stabilize certain mRNA transcripts containing PSCs. Indeed, our results show that this reagent stabilizes the patient's ALG8 transcript such that its level is the same as that of its normal counterpart found in control cells treated with the same concentration of the drug. We noted that emetine also increases the expression of ALG8 mRNA in normal cells. In fact, in yeast many of the ALG gene mRNAs behave like the transcripts of "early growth-response" genes that are known to be stabilized in the presence of protein synthesis inhibitors (54 -56). To date, this is the first time that NMD has been shown to be operative in cells from patients with CDG. In other diseases it has been shown that when NMD of PSC-containing mRNA is operational, the symptoms are less severe than in those patients for which the mutant mRNA is stable (46,47). This is probably due to the fact that NMD can clear the cell of PSC containing open reading frames, which, if translated would lead to the accumulation of potentially deleterious, dominant negative, truncated proteins (46,47). If any mRNA is translated in patient M. P., it is apparent that only the first ϳ20% of the protein, containing only two potential transmembrane regions, is produced rendering it highly unlikely that these two alleles could give rise to active proteins. The dolichyl-P-monosaccharide requiring mannosyltransferases and glucosyltransferases of the LLO pathway are extremely hydrophobic enzymes comprising 10 -14 transmembrane regions (45). The importance of the transmembrane regions for enzyme function is attested to by the observation that several of the disease causing mutations in CDG I are found in or near the membrane spanning regions that occur along the entire length of the defective glycosyltransferases. By , or presence (ϩ) of the glycosidase inhibitors castanospermine and kifunensin (CϩK). The cellular glycoproteins were sequentially treated with Pronase and endo H, and the released oligosaccharides were resolved by thin layer chromatography as described under "Experimental Procedures." B, after visualization by fluorography, the oligosaccharide components derived from the glycosidase-inhibited (CϩK) cells were eluted from the chromatography plate and quantitated by scintillation counting. The recovery of the individual components is expressed as a percentage of the total amount of oligosaccharide recovered for each of the two cell populations. The abbreviations used are: M 8 , Man 8 GlcNAc; M 9 , Man 9 GlcNAc; G 1 M 9 , Glc 1 Man 9 GlcNAc; G 3 M 8 , Glc 3 Man 8 GlcNAc; G 3 M 9 , Glc 3 Man 9 GlcNAc. taking into account the paucity of hALG8 mRNA, the above observations strongly suggest that patient M. P. has little or no ALG8p activity.
The paradox raised by our work is that despite the probable low leakiness of the hALG8 mutations presented here, the patient's disease presents as a CDG with a less severe clinical picture than that generally associated with CDG Ic in which a deficit in LLO glucosylation is also the underlying cause. It is clear that cells from this patient manage to synthesize substantial quantities of fully glucosylated LLO. In fact, incorporation of [2-3 H]mannose into glycoproteins extracted from M. P. fibroblasts was as efficient as that observed in control fibroblasts. Furthermore, although substantial amounts of underglucosylated LLO are generated when M. P. cells are treated with CST, the specificity of oligosaccharyltransferase ensures that fully glucosylated oligosaccharides are preferentially transferred from LLO onto polypeptides in the lumen of the ER (Fig. 8). Surprisingly, although we noted that when the fibroblasts from the patient are treated with CST there are increased amounts of triglucosylated LLO, we were unable to show that this allowed greater transfer of radioactive oligosaccharides from LLO onto glycoprotein (results not shown).
The fact that fibroblasts and lymphoblasts from patient M. P. are able to synthesize triglucosylated LLO suggests that one or both of the alleles can give rise to an active transferase, or an as yet unidentified glucosyltransferase, with little or no similarity to already described glucosyltransferases, is capable of carrying out this reaction. Alternatively, when the Glc 1 Man 9 GlcNAc 2 -PP-dolichyl pool is elevated (sufficient endogenous UDP-Glc, or addition of CST to cells), this monoglucosylated LLO may be glucosylated again by either ALG6p or ALG10p. Should this be the explanation, it is more likely that the second glucose is added by ALG6p. First, biochemically, it has not been possible to distinguish the enzymes (ALG6p and ALG8p) that add the first two ␣3-linked glucoses onto LLO (40,57,58), suggesting that they may have very similar properties. Second, ALG10p adds an ␣2linked glucose to the LLO, and this enzyme has been shown to have distinct biochemical properties to those of ALG6p and ALG8p. Third, analysis of the amino acid sequences of the three glucosyltransferases reveals that ALG6p and ALG8p are more closely related to each other (59) than either of them are related to ALG10p (45,60).
In conclusion, we have identified a new subtype of CDG I, which we suggest be called CDG Ih, in which there is a deficiency in hALG8p, the enzyme which adds the second glucose onto growing LLO. The gene encoding this enzyme was found to contain mutations that generate premature stop codons in both of the alleles from the patient. Further work is underway in order to understand how cells from this patient synthesize fully glucosylated LLO despite possessing such apparently drastic mutations, and why the clinical picture of this patient is so mild when compared with that observed in patients with CDG I subtypes a and c-g.