The Lec23 Chinese Hamster Ovary Mutant Is a Sensitive Host for Detecting Mutations in α-Glucosidase I That Give Rise to Congenital Disorder of Glycosylation IIb (CDG IIb)*

Lec23 Chinese hamster ovary cells are defective in α-glucosidase I activity, which removes the distal α(1,2)-linked glucose residue from Glc3Man9GlcNAc2 moieties attached to glycoproteins in the endoplasmic reticulum. Mutations in the human GCS1 gene give rise to the congenital disorder of glycosylation termed CDG IIb. Lec23 mutant cells have been shown to alter lectin binding and to synthesize predominantly oligomannosyl N-glycans on endogenous glycoproteins. A single point mutation (TCC to TTC; Ser to Phe) was identified in Lec23 Gcs1 cDNA and genomic DNA. Serine at the analogous position is highly conserved in all GCS1 gene homologues. A human GCS1 cDNA reverted the Lec23 phenotype, whereas GCS1 cDNA carrying the lec23 mutation (S440F in human) did not. By contrast, GCS1 cDNA with an R486T or F652L CDG IIb mutation gave substantial rescue of the Lec23 phenotype. Nevertheless, in vitro assays of each enzyme gave no detectable α-glucosidase I activity. Clearly the R486T and F652L GCS1 mutations are only mildly debilitating in an intact cell, whereas the S440F mutation largely inactivates α-glucosidase I both in vitro and in vivo. However, the S440F α-glucosidase I may have a small amount of α-glucosidase I activity in vivo based on the low levels of complex N-glycans in Lec23. A sensitive test for complex N-glycans showed the presence of polysialic acid on the neural cell adhesion molecule. The Lec23 Chinese hamster ovary mutant represents a sensitive host for detecting a wide range of mutations in human GCS1 that give rise to CDG IIb.

Lec23 Chinese hamster ovary cells are defective in ␣-glucosidase I activity, which removes the distal ␣(1,2)linked glucose residue from Glc 3 Man 9 GlcNAc 2 moieties attached to glycoproteins in the endoplasmic reticulum. Mutations in the human GCS1 gene give rise to the congenital disorder of glycosylation termed CDG IIb. Lec23 mutant cells have been shown to alter lectin binding and to synthesize predominantly oligomannosyl N-glycans on endogenous glycoproteins. A single point mutation (TCC to TTC; Ser to Phe) was identified in Lec23 Gcs1 cDNA and genomic DNA. Serine at the analogous position is highly conserved in all GCS1 gene homologues. A human GCS1 cDNA reverted the Lec23 phenotype, whereas GCS1 cDNA carrying the lec23 mutation (S440F in human) did not. By contrast, GCS1 cDNA with an R486T or F652L CDG IIb mutation gave substantial rescue of the Lec23 phenotype. Nevertheless, in vitro assays of each enzyme gave no detectable ␣-glucosidase I activity. Clearly the R486T and F652L GCS1 mutations are only mildly debilitating in an intact cell, whereas the S440F mutation largely inactivates ␣-glucosidase I both in vitro and in vivo. However, the S440F ␣-glucosidase I may have a small amount of ␣-glucosidase I activity in vivo based on the low levels of complex N-glycans in Lec23. A sensitive test for complex N-glycans showed the presence of polysialic acid on the neural cell adhesion molecule. The Lec23 Chinese hamster ovary mutant represents a sensitive host for detecting a wide range of mutations in human GCS1 that give rise to CDG IIb.
There are a growing number of human diseases that arise from altered glycosylation of glycoproteins or lipid-linked oligosaccharides and are termed congenital disorders of glycosylation (CDG) 1 (1,2). In addition, there are a number of myop-athies due to defective O-glycan synthesis on ␣-dystroglycan (3,4) and diseases that disrupt the synthesis of proteoglycans (5). As disease mutations are identified, it becomes important to characterize the altered biochemical properties of mutant genes and to investigate ways in which enzyme activity may be modulated. Such studies are best performed in a cell that is null for the activity under investigation. Cells that lack a glycosylation enzyme activity may be derived from a particular patient or a mutant mouse or mutant cells selected in culture. The latter are most desirable because fibroblasts isolated from humans or mice often have a finite lifetime, are not easy to grow, or are difficult to transfect. By contrast, cultured Chinese hamster ovary (CHO) cells are well characterized, are easy to culture and transfect, and have many characterized glycosylation mutants (6 -8). In this paper we describe the molecular basis of the Lec23 CHO mutant and show that it is an excellent host for characterizing mutations in ␣-glucosidase I that give rise to the human congenital disorder of glycosylation CDG IIb.
Lec23 cells were obtained following a single step selection for lectin resistance from a mutagenized population of CHO cells (9). The cells are highly resistant to the leukoagglutinin from Phaseolus vulgaris (L-PHA) and to wheat germ agglutinin (WGA) and are hypersensitive to concanavalin A (Con A). Complementation analysis showed that the lec23 mutation is recessive, and biochemical analysis revealed defective ␣-glucosidase I activity (10). ␣-Glucosidase I initiates the processing of N-glycans by removing the distal ␣(1,2)Glc residue from Glc 3 Man 9 GlcNAc 2 Asn in the endoplasmic reticulum (11,12). The G glycoprotein of vesicular stomatitis virus obtained from Lec23 mutant cells has predominantly oligomannosyl N-glycans with three terminal glucoses (10). Lec23 cells have been used to study endoplasmic reticulum chaperones, such as calnexin, calreticulin, and ERp57, that recognize monoglucosylated oligomannosyl N-glycans and aid in the folding of glycoproteins (13)(14)(15)(16)(17)(18). Lec23 cells have also been useful in characterizing Golgi endomannosidase (19,20) and the biological roles of N-glycans in human immunodeficiency virus infection (21,22). Thus it was important to determine the molecular basis of the mutation in Lec23 cells that might be in the Gcs1 gene that encodes ␣-glucosidase I or in a regulatory gene. In either case, Lec23 cells would be useful in understanding mutations that give rise to an ␣-glucosidase I deficiency and CDG IIb (23,24). Here we report that Lec23 CHO cells express a point mutation in the Gcs1 gene that generates a missense mutation and largely inactivates ␣-glucosidase I. Lec23 cells are shown to be sensitive hosts for characterizing mutations in human GCS1 that give rise to CDG IIb.

EXPERIMENTAL PROCEDURES
Materials-Mouse monoclonal anti-␣2,8-polysialic acid (PSA) IgG 2a antibody 735 (25) was a gift from Dr. Rita Gerardy-Schahn (Zentrum Biochemie). N-CAM13 (BD Biosciences) is a monoclonal IgG 2a antibody recognizing the extracellular domain of mouse N-CAM. FITC-conjugated anti-mouse IgGϩM antibody (Zymed Laboratories Inc.) was used as a secondary antibody. FLAG-tagged proteins were detected by anti-FLAG M2 antibody from Sigma. Lectins including Con A, WGA, L-PHA, and ricin with and without biotinylation were purchased from Vector Laboratories (Burlingame, CA). Restriction enzymes and buffers were from Roche Diagnostics GmbH (Mannheim, Germany), Invitrogen, and New England Biolabs (Beverly, MA). Hybond Nϩ membrane was from Amersham Biosciences. DNA probe was labeled with Prime-It II random primer labeling kits (Stratagene) with [␣ 32 P]dCTP (PerkinElmer Life Sciences). Synthetic oligonucleotides, Superscript II, and dNTPs were purchased from Invitrogen. 3 H-Labeled Glc 3 Man 9 GlcNAc 2 Asn in Glc residues was prepared as described (10). Other chemicals and reagents were from Sigma and Fisher.
To transfect expression constructs, cells were cultured on a 24-well plate for 1 day to ϳ90% confluency, washed once with Opti-MEM I medium (Invitrogen), and incubated in 0.5 ml Opti-MEM I medium containing 3% fetal calf serum. Plasmid DNA (2 g) and Lipo-fectAMINE 2000 from Invitrogen (2 l) were mixed with 50 l of Opti-MEM I in separate tubes. After 5 min at room temperature, plasmid DNA and LipofectAMINE 2000 were mixed, incubated for 20 min at room temperature, and added to cells. After 12-24 h at 37°C, cells were detached with trypsin/EDTA and cultured in ␣ minimum Eagle's medium containing 10% fetal calf serum. Transient transfectants were analyzed 48 -72 h post-transfection.
Lectin Toxicity Assay-Cells (2,000/well of a 96-well dish) were incubated with increasing concentrations of Con A, WGA, or L-PHA in complete medium for 3-4 days at 37°C. Cells that remained attached were stained using methylene blue, and the concentration of lectin that killed 90% of the cells was determined.
Flow Cytometric Analysis-To compare CHO and Lec23 cells for lectin binding, 10 6 cells were incubated with FITC-WGA (Vector, 25 g/ml), FITC-L-PHA (Vector, 10 g/ml), FITC-Con A (Vector, 10 g/ml), Matrix-assisted Laser Desorption Ionization Time-of-flight Mass Spectrometry-Glycoproteins were extracted from parent and Lec23 CHO cells in 1.5% Triton X-100 containing protease inhibitors (Roche Applied Science), and 500 g of protein was treated with peptide:Nglycosidase F (PNGase F) or endoglycosidase H (Endo H) (New England Biolabs) as described (28). Released N-glycans were methylated under conditions that specifically methylated sialic acid residues only so they could be detected in the positive ion mode (29). Subsequent sample purification was by ion exchange chromatography as described (29), followed by polygraphite carbon column chromatography as described (28). MALDI-TOF mass spectrometry was performed on a Voyager DE Biospectrometry work station (Perseptive Biosystem) equipped with delayed extraction. Oligosaccharide standards of known structure were used for external calibration for mass assignment of ions, and D-arabinosazone (30) was used as the matrix in the positive ion mode for the analysis of neutral oligosaccharides.
Northern Blot Analysis-Total RNA (30 g) was extracted using TRIzol (Invitrogen) and separated in a 1% agarose-denaturing gel. After transfer to a Hybond Nϩ membrane, a 1.5-kb probe was PCR-amplified from pCR-CHO-GCS1(3Ј) plasmid using CHO-GCS2F (5Ј-atggccggaagtcctactaccagaag-3Ј) and CHO-GCS4R (5Ј-aggggctccagagatggcggctgtc-3Ј) primers and purified from a 1% agarose gel. The PCR products were radiolabeled using the Random-It priming kit (Stratagene). The membrane was incubated in prehybridization solution (0.05 M PIPES, 0.1 M NaCl, 0.05 M NaPO 4 (pH 7.0), 10 mM EDTA, 5% SDS, 0.1 mg/ml denatured salmon sperm DNA) at 60°C for 2 h before adding the probe, and the incubation was continued at 60°C overnight. Washing was performed twice in 2ϫ SSC, 0.1% SDS and twice in 0.2ϫ SSC, 0.1% SDS before exposure to x-ray film for 2 days at Ϫ80°C. For stripping, the membrane was soaked in boiled 0.5% SDS in diethyl pyrocarbonatetreated water, cooled to room temperature, and stored at Ϫ20°C before probing with a glyceraldehyde-3-phosphate dehydrogenase probe to determine relative loading and RNA integrity.
Cloning of CHO Gcs1 cDNA-To clone Chinese hamster Gcs1, subpools of 5,000 cDNA clones were made from a CHO-K1 cell library in pSPORT (a gift from Drs. Osamu Kuge and Masahiro Nishijima, Institute of Infectious Diseases, Tokyo, Japan) and screened by PCR with primers PS673 (5Ј-gctggagagtgaccgtagagcctcagg-3Ј, forward) and PS674 (5Ј-gaaaggggctttcagtgtcacctgctg-3Ј, reverse), which were designed from mouse and rat Gcs1-coding sequences. PCR was performed at 94°C (30 s), 58°C (30 s), and 72°C (1 min) for 35 cycles using platinum Taq polymerase (Invitrogen). From a subpool showing positive bands by PCR, a 5Ј fragment of hamster Gcs1 was amplified with an upstream vector primer and PS674. The longest fragment, which covered a putative second exon of the Gcs1 gene, was purified from a 1% agarose gel and sequenced. A 3Ј fragment was amplified with PS673 and downstream vector primer, sequenced, and found to contain poly(A) signals. This fragment of 2.1 kb was cloned into the pCR2.1 vector (Invitrogen) and named pCR-CHO-Gcs1(3Ј).
Western Blot Analysis-Lec23 cells (1.5 ϫ 10 6 ) were incubated in a 6-well plate in complete medium for 12 h. Cells were washed once with Opti-MEM I, and Opti-MEM I with 3% fetal calf serum was added. pME-3FLAG-hGCS1 or hGCS1 mutant cDNA (10 g/well) and 10 l of LipofectAMINE 2000 were mixed with 250 l of Opti-MEM I, respectively. After 5 min at room temperature, diluted plasmids and Lipo-fectAMINE 2000 were mixed together. After 20 min, DNA liposomes were added to cells and cultured for 1 day in a 5% CO 2 incubator. Cells were detached using trypsin/EDTA and cultured in complete medium for 1 day. To detect FLAG-tagged GCS1, cells (5 ϫ 10 6 ) were dissolved in 1 ml of 1% Nonidet P-40 in TNE buffer (50 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA) for 1 h. After centrifugation at 15,000 ϫ g for 10 min, tagged proteins were immunoprecipitated from the supernatant with biotinylated anti-FLAG M2 antibody-conjugated beads (Sigma) for 1 h. After washing three times with 1% Nonidet P-40 in TNE buffer, proteins were released from beads by boiling in sample buffer, separated in 7.5% acrylamide gel by SDS-PAGE, and transferred to polyvinylidene difluoride membrane. Western blot analysis was performed with biotinylated anti-FLAG M2 antibody (1/2000) and horseradish peroxidase-conjugated avidin (Pierce) (1/5000) as described (33).
To detect complex N-glycans on N-CAM, cells (1 ϫ 10 7 ) were extracted in 1 ml of 1% Nonidet P-40 in TNE. Supernatants obtained after 1 h of incubation on ice and centrifugation at 10,000 ϫ g for 5 min were immunoprecipitated with anti-PSA mAb 735 and protein G beads (Amersham Biosciences). After adding sample buffer and boiling, proteins were separated on a 6% acrylamide gel by SDS-PAGE and transferred to polyvinylidene difluoride membrane. N-glycans were detected with biotinylated L-PHA (10 g/ml) (Vector) and horseradish peroxidaseconjugated avidin (1/2000). PSA modified N-CAM was detected on the same membrane after incubation in stripping buffer (62.5 mM Tris (pH 6.8), 2% SDS, 100 mM 2-mercaptoethanol) at 55°C for 30 min followed by washing in 5 mM Tris (pH 8.0) containing 1% Nonidet P-40. PSA on N-CAM was detected with anti-PSA mAb 735 and sheep anti-mouse IgG horseradish peroxidase.
Glycosidase Treatments and Western Blot Analysis-Cell pellets from 2 ϫ 10 6 cells were extracted in 100 l of 5 mM Tris (pH 8.0) with 1% Nonidet P-40 for 15 min on ice. Enzyme treatment was per the manufacturer's instructions (New England Biolabs). Briefly, after centrifugation to remove nuclei, 16 l of cell extract and 2 l of bovine pancreas RNase B (10 g) were mixed with 2 l of 10ϫ glycoprotein denaturation buffer and incubated at 100°C for 5 min. After cooling, 2 l of 10ϫ G7 buffer and 2 l of PNGase F or distilled water, or 2 l of 10ϫ G5 buffer and 2 l of Endo H or distilled water, were added to cell extracts. After incubation at 37°C for 5 h, reactions were stopped by adding 44 l of 4% SDS sample buffer and boiling for 5 min. Proteins were separated in a 10% SDS-polyacrylamide gel, transferred to polyvinylidene difluoride membrane, and incubated with anti-N-CAM antibodies followed by sheep anti-mouse IgG horseradish peroxidase. Bound antibodies were detected by an enhanced chemiluminescence kit and exposure to x-ray film. To detect PSA-modified N-CAM proteins, the same membrane was incubated in stripping buffer for 30 min at 55°C and washed in Trisbuffered saline containing 0.1% Tween. PSA on proteins was detected with anti-PSA mAb 735 followed by sheep anti-mouse IgG horseradish peroxidase.

Lec23 Mutants Are Readily Differentiated from CHO Cells by
Lectin Binding-Lec23 cells were previously distinguished from CHO cells by a lectin resistance cytotoxicity test as shown in Fig. 1A (9). They are highly resistant to L-PHA and WGA, lectins that bind to complex N-glycans, and are hypersensitive to Con A, a lectin that binds best to oligomannosyl N-glycans (34). To determine whether resistance to lectins reflects a reduction in lectin binding and hypersensitivity reflects increased lectin binding, Lec23 and CHO cells were examined by fluorescence-activated cytometry (FACS) using fluorescencetagged lectins (Fig. 1B). Compared with parental CHO cells, Lec23 mutants showed much less binding of L-PHA and WGA and more binding of Con A. The reduction in L-PHA binding was ϳ6-fold greater than the reduction in WGA binding, consistent with differences observed with the lectin resistance assay. Lec23 cells also bound less ricin than parent CHO, as expected if complex N-glycans are reduced (Fig. 1B).
Endogenous Lec23 Glycoproteins Carry Oligomannosyl Nglycans Terminating with Three Glc Residues-Previous studies using 1 H NMR spectroscopy showed that vesicular stomatitis virus G glycoprotein produced from infected Lec23 cells carried predominantly oligomannosyl N-glycans terminating in three Glc residues with Glc 3 Man 7 GlcNAc 2 being the major species (10). By contrast, vesicular stomatitis virus G from parent CHO cells carries only complex N-glycans at both Nglycosylation sites (35,36). To determine the proportion of N-glycans that are affected by the lec23 mutation, Lec23 cell glycoprotein extracts were treated with PNGase F to release all N-glycans or with Endo H to release oligomannosyl and hybrid N-glycans, and the mildly methylated N-glycans were analyzed by MALDI-TOF mass spectrometry in the positive ion mode. The conditions of methylation were such that only sialic acid residues should be methylated, thereby rendering them detectable in the positive ion mode (29). It can be seen in Fig. 2A and Table I that the N-glycans released from Lec23 glycoproteins by PNGase F were predominantly oligomannosyl. Glc 3 Man 7 GlcNAc 2 was the major species, and there were no sialylated species. There were small amounts of biantennary species and truncated triantennary species as well as Man 8 GlcNAc 2 lacking Glc. Therefore the mutant ␣-glucosidase I in Lec23 cells must have a small amount of activity, or these N-glycans reflect the existence of an alternative processing pathway. The major N-glycan released by Endo H was Glc 3 Man 7 GlcNAc 1 ( Fig. 2B and Table I), and there were minor species lacking Glc, which indicates a small amount of ␣-glucosidase I activity in Lec23 or an alternative pathway. Although Man 8 GlcNAc 1 lacking Glc might theoretically arise from the action of a Golgi endomannosidase (37), there is no evidence that CHO or Lec23 cells have endomannosidase activity (20,38). (10,20). To investigate the expression of the Gcs1 gene that encodes ␣-glucosidase I, Northern blot analysis of total RNA was performed. Lec23 and CHO cells had Gcs1 transcripts of similar size (Fig. 3A). Therefore, a mutant Gcs1 gene in Lec23 cells may have a small deletion, insertion, or point mutation, or the mutation might lie in a regulatory region or regulatory gene.

Point Mutation of Lec23 Mutant Cells in ␣-Glucosidase I-Previous studies showed that Lec23 mutant cells have no detectable ␣-glucosidase I activity in vitro
To isolate Gcs1 cDNAs for sequencing, a CHO cDNA library was screened by PCR using degenerate primers designed from mouse and rat Gcs1 3Ј-coding regions. A partial hamster sequence was obtained that corresponded to exons 2, 3, and 4 of mammalian Gcs1 genes (Fig. 3B). The hamster Gcs1 cDNA sequence was 84, 90, and 90% identical to human, mouse, and rat cDNA, respectively. The deduced amino acid sequence was 91% identical to mouse Gcs1. Unfortunately, attempts to obtain the 5Ј CHO Gcs1 coding sequence were not successful for two reasons. First, we could not amplify exon 1 from the pSPORT PCR library using vector and 5Ј-coding region primers. It is likely that the 5Ј end of the Gcs1 cDNA is not represented in the library because it was generated following digestion with NotI for cloning purposes, and human, mouse, and rat Gcs1 sequences contain one or two NotI sites in exon 1. Second, attempts at 5Ј rapid amplification of cDNA ends were probably not successful because of the high GC content of Gcs1 5Ј-coding regions.
Reverse transcription-PCR was used with primers spanning the partial CHO cDNA to obtain Gcs1 cDNAs from CHO and Lec23 total RNA. Direct sequencing of these PCR products identified a single nucleotide change of C to T at nucleotide 964 in the partial hamster Gcs1. We found the same mutation in a PCR fragment amplified from Lec23 genomic DNA. This change causes a missense mutation and a replacement of Ser with Phe. This position corresponds to Ser-440 in the human GCS1 sequence as noted in Fig. 3A. Sequencing of reverse transcription-PCR products from Lec23 cells gave no evidence for the presence of Gcs1 transcripts with the parental CHO  (19,20). Observed m/z are compared with predicted m/z in Table I  sequence. In addition, only the mutant sequence was present in genomic DNA as determined by sequencing of genomic DNA PCR products. These data are consistent with previous biochemical studies showing that Lec23 CHO cells do not have ␣-glucosidase I activity that could arise from a wild-type allele (10,20).
Reduction of ␣-Glucosidase I Activity by the Lec23 Gcs1 Mutation-A sensitive test of ␣-glucosidase I activity is to convert the reduced L-PHA binding of Lec23 cells to parental levels (Fig. 1). This was accomplished by transient transfection of a human GCS1 cDNA into Lec23 cells (Fig. 4). About 30% of the transfected cells bound L-PHA over a wide range, which was consistent with varying levels of expression in the transfected population. A small fraction of cells bound more L-PHA than CHO control cells, indicating the effects of overexpression of ␣-glucosidase I. No increase in L-PHA binding was observed with vector control. By contrast, none of the Lec23 cells expressing a mutant GCS1 cDNA carrying the lec23 mutation (S440F) had increased binding of L-PHA, although the amount of enzyme expressed was shown by Western blot analysis to be equivalent in both cases (Fig. 4A). Therefore an overexpressed human ␣-glucosidase I cDNA with the S440F mutation did not rescue the Lec23 phenotype.
The Lec23 Mutant Reveals Activity of GCS1 Mutant Alleles Identified in CDG IIb Patients-The Lec23 Gcs1 mutation is located in exon 4, the exon thought to encode the catalytic activity of ␣-glucosidase I (40,41). Two mutations in the human GCS1 gene (R486T and F652L) are also located in exon 4. The only CDG IIb patient so far reported was a compound heterozygote that carried both of these mutant alleles (23). Biochemical analyses of these mutations in fibroblasts or introduced into a human GCS1 cDNA and transfected into COS A, Northern blot analysis in which total RNA (30 g) purified from CHO and Lec23 cells was separated on a denaturing gel and probed with a 1.5-kb hamster Gcs1 cDNA fragment that contained exons 3 and 4. After transferring to the membrane, the blot was exposed to x-ray film for 3 days (upper panel). After stripping, the blot was probed for glyceraldehyde-3-phosphate dehydrogenase (GAPDH). B, the deduced amino acid sequence of the partial hamster Gcs1 cDNA (GenBank TM accession no. AB115503) is compared with human GCS1 (X87237), rat (AF087431), mouse (NM_020619), Saccharomyces cerevisiae (U35669.1), Arabidopsis (NM_105416.3), and Caenorhabditis elegans (NM_069652.1). The numbering of the amino acids is based on the human cDNA. Positions of the Lec23 mutation (S440F in human) and two mutations from a CDG IIb patient (R486T and F652L) are indicated by asterisks. Exon boundaries are marked by triangles. Each mutation occurs in exon 4. cells indicated that both mutations reduce enzyme activity Ն97% (24). Therefore a human cDNA carrying either of these ␣-glucosidase I mutations would be expected to behave like the Lec23 mutant cDNA following transfection into Lec23 cells. Surprisingly, however, when cDNAs carrying these mutations were introduced into Lec23 cells, a significant level of L-PHA binding was restored (Fig. 4B). Thus ␣-glucosidase I with either of the known CDG IIb mutations was able to rescue the phenotype of Lec23 almost as well as wild-type enzyme. Nevertheless, when cell extracts were prepared from stable transfectants expressing the same mutant cDNAs, their ␣-glucosidase I activity was extremely low in vitro (Table II) as reported previously (23). Thus human GCS1 with the mutation R486T or F652L has good ␣-glucosidase I activity in vivo in the Lec23 CHO cellular environment but almost none in vitro, whereas GCS1 with the S440F mutation has no ␣-glucosidase I activity in vitro and also very little in vivo (10) (Fig. 2). However, we discovered that a few complex N-glycans are synthesized by Lec23 cells and can be readily detected on N-CAM.
Lec23 Cells Express PSA on N-CAM-The MALDI-TOF mass spectrometric analyses of the N-glycans on Lec23 glycoproteins identified small amounts of complex N-glycans ( Fig. 2A) as was observed previously on vesicular stomatitis virus G glycoprotein from Lec23 cells (10). The synthesis of complex N-glycans by Lec23 cells was also revealed by investigating the presence of PSA on N-CAM by Western blot analysis. PSA is a polymer of ␣(2,8)-linked sialic acids attached to the terminal sialic acid of complex N-glycans in mammals (reviewed in Ref. 42). Previous experiments have shown that N-CAM is essentially the only glycoprotein in CHO cells that can be detected with anti-PSA antibodies (43,44).
To determine whether Lec23 cells make PSA on N-CAM, proteins extracted in 1% Nonidet P-40 were mixed with bovine pancreas RNase B and treated with PNGase F or Endo H. After 5 h, aliquots were separated on a 6% SDS-polyacrylamide gel, and the portion of the gel with proteins with a molecular mass of Ͼ25 kDa was transferred to the membrane. Coomassie Blue staining of the remaining portion of the gel showed that the RNase B included as a control in each sample was digested by both PNGase F and Endo H to a species of ϳ13 kDa. Under these SDS-PAGE conditions, CHO N-CAM migrated predominantly as an ϳ140-kDa species that was sensitive to PNGase F but resistant to digestion with Endo H. Lec23 N-CAM had a slightly lower molecular mass of ϳ130 kDa and was sensitive to  When this blot was stripped and reprobed with anti-PSA mAb 735, it could be seen that only a small fraction of N-CAM in CHO cells was modified by PSA. This species of ϳ240 kDa was sensitive to digestion with PNGase F but not to Endo H in both CHO and Lec23 extracts. Therefore, the PSA in Lec23 N-CAM was attached to complex N-glycans. Interestingly, the amount of Lec23 PSA-N-CAM was similar to that in parent CHO cells, suggesting that the few complex N-glycans synthesized in Lec23 were disproportionately expressed on N-CAM, perhaps for the purpose of making PSA-N-CAM. A similar result was obtained by FACS analysis (Fig. 5), although in this case, CHO cells clearly bound more anti-PSA mAb than Lec23 cells. Lec2 CHO cells that are deficient in CMP-sialic acid transport expressed almost no PSA as shown previously (27,45,46). DISCUSSION Here we identify the molecular basis of the lec23 mutation in CHO cells as a missense mutation in the Gcs1 gene coding region. Lec23 cells express only the mutant Gcs1 gene. The single base change that converts Ser to Phe occurs in a highly conserved region of exon 4. When the analogous mutation S440F was engineered into a full-length human GCS1 cDNA, the mutant cDNA was unable to rescue the Lec23 phenotype, and transfectants had almost no detectable ␣-glucosidase I activity in vitro despite overexpression of the cDNA. The Lec23 mutant is therefore useful for studying the properties of mutant GCS1 genes, for investigating alternative pathways of N-linked glycosylation, and for identifying the consequences to the cell of maintaining Glc residues on glycoproteins and oligosaccharides that exit the endoplasmic reticulum and traffic through the cell.
Previous studies of ␣-glucosidase I from rat mammary gland have shown that it is an ϳ85-kDa glycoprotein with a single transmembrane domain and a luminally oriented catalytic domain (40). A sulfhydryl group has been identified in the active site (47), and the C-terminal 39 kDa of the enzyme contains the catalytic activity (40,41). The expression level of ␣-glucosidase I is regulated during lactation and by hormone treatments (39). The S440F mutation identified in Lec23 cells is a novel ␣-glucosidase I-inactivating mutation. The only other GCS1 mutations were identified in a CDG IIb patient (23). This patient was a compound heterozygote and carried both R486T and F652L mutant alleles. The ␣-glucosidase I activity of patient fibroblasts was less than 3% of normal activity (24). The pathological consequences were severe and included hepatomegaly, hypoventilation, and feeding problems that led to death 70 days after birth. Thus it was surprising to find that a GCS1 cDNA carrying either mutation was able to almost fully rescue the Lec23 L-PHA-binding phenotype following transfection (Fig. 4). The fact that the mutant cDNAs expressed in Lec23 did not have significant ␣-glucosidase I activity in vitro (Table  II) agrees with previous results from transfected COS cells and from patient fibroblasts (24). The discrepancy might reflect the sensitivity of the L-PHA binding assay or a real difference in activity in the in vitro enzyme assay versus the activity in an intact cell. This point might theoretically be addressed by examining lectin binding of CDG IIb patient fibroblasts. However, studies of N-glycans from the fibroblasts of patients revealed only ϳ16% of N-glycans arrested at the Glc 3 Man 9 -7 GlcNAc 2 stage (24). This appears to be the result of Golgi endomannosidase activity, because the urine of the patient had a high concentration of the endomannosidase product (24).
It is interesting that CHO cells and the Lec23 mutant have no detectable Golgi endomannosidase activity when assayed in vitro or by the generation of oligomannosyl N-glycans from labeled Glc 3 Man 9 GlcNAc 2 in vivo (19). Yet it is clear from structural analyses that Lec23 glycoproteins acquire some oligomannosyl N-glycans lacking Glc and also complex N-glycans (Fig. 2). In addition, when it comes to the expression of PSA on the complex N-glycans of N-CAM, Lec23 cells have levels similar to parent CHO cells (Fig. 5). If Lec23 cells actually possess a small amount of Golgi endomannosidase that is not easily detected by the sensitive assays tested previously, the Man 8 GlcNAc 2 isomer on Lec23 glycoproteins should be the predicted product of endomannosidase action. However, it would not explain the small amount of Man 9 GlcNAc 2 detected after treatment of Lec23 glycoproteins with Endo H (Fig. 2B). It is possible that Lec23 cells generate complex N-glycans by a heretofore unknown pathway that bypasses the need for ␣-glucosidase I. For example, ␣-glucosidase II may be able to remove the terminal ␣(1,2)Glc of Glc 3 Man 9 GlcNAc 2 to a small degree in vivo, thereby generating its physiological substrate and allowing all three Glc residues to be removed. Alternatively, a novel activity may be responsible for removing the terminal Glc of Glc 3 Man 9 GlcNAc 2 in the absence of ␣-glucosidase I.