Point Mutations Identified in Lec8 Chinese Hamster Ovary Glycosylation Mutants That Inactivate Both the UDP-galactose and CMP-sialic Acid Transporters*

Nucleotide-sugar transporters (NSTs) are critical components of glycosylation pathways in eukaryotes. The identification of structural elements that are involved in NST functions provides an important task. Chinese hamster ovary glycosylation mutants defective in nucleotide-sugar transport provide access to inactive transporters that can define such structure/function relationships. In this study, we have cloned the hamster UDP-galactose transporter (UGT) and identified defects in UGT gene transcripts from nine independent Chinese hamster ovary mutants that belong to the Lec8 complementation group. Reverse transcription polymerase chain reaction with primers that span the UGT open reading frame showed that three Lec8 mutants express a full-length open reading frame, while six Lec8 mutants predominantly express truncated UGT gene transcripts. Sequencing identified different single or triplet nucleotide changes in full-length UGT transcripts from three of the mutants. These mutations translate into three different amino acid changes at positions that are highly conserved in all the known mammalian NSTs. Transfection of a cDNA encoding either of the mutations D serine 213 or G281D failed to correct the UDP-galactose transport defect in Lec8 transfectants. Most importantly, in-troducing these same mutations into the homologous region of the murine CMP-sialic


Nucleotide-sugar transporters (NSTs) are critical components of glycosylation pathways in eukaryotes.
The identification of structural elements that are involved in NST functions provides an important task. Chinese hamster ovary glycosylation mutants defective in nucleotide-sugar transport provide access to inactive transporters that can define such structure/function relationships. In this study, we have cloned the hamster UDP-galactose transporter (UGT) and identified defects in UGT gene transcripts from nine independent Chinese hamster ovary mutants that belong to the Lec8 complementation group. Reverse transcription polymerase chain reaction with primers that span the UGT open reading frame showed that three Lec8 mutants express a full-length open reading frame, while six Lec8 mutants predominantly express truncated UGT gene transcripts. Sequencing identified different single or triplet nucleotide changes in full-length UGT transcripts from three of the mutants. These mutations translate into three different amino acid changes at positions that are highly conserved in all the known mammalian NSTs. Transfection of a cDNA encoding either of the mutations ⌬serine 213 or G281D failed to correct the UDP-galactose transport defect in Lec8 transfectants. Most importantly, introducing these same mutations into the homologous region of the murine CMP-sialic acid transporter caused inactivation of this transporter. Thus, identifying point mutations that inactivate UGT in Lec8 mutants resulted in the discovery of amino acids that are critical to the activity of both UGT and CST, the two most divergent mammalian NSTs.
The maturation of glycoconjugates that are either secreted or become constituents of the plasma membrane occurs while passing through the Golgi compartments. Processing enzymes, glycosidases, and glycosyltransferases, are Golgi residents, and their spatial organization in the Golgi stack generally reflects the sequence of biosynthetic steps (1). The nucleotide-sugars, which are substrates for luminal glycosyltransferases, are pro-duced in the cytoplasm or, in the case of CMP-sialic acid, in the cell nucleus (reviewed in Refs. 2 and 3) and are supplied to the Golgi lumen via specific nucleotide-sugar transporters (NSTs). 1 The existence of NSTs, their functional properties, and subcellular locations have been described for many years (4). The isolation of NST genes, however, succeeded only recently by expression cloning using glycosylation-defective mutants (reviewed in Refs. 4 and 5).
Complementation cloning was carried out in Had-1 cells and in S. pombe and identified the human (9) and a truncated form of the yeast (10) UGT genes, respectively. Using PCR-based approaches, the murine UGT (11), a second isoform of the human UGT (12), and a full-length S. pombe UGT cDNA (13) were isolated. Sequence alignments reveal strong conservation of UGTs, since species as distant as human and S. pombe are 40% identical (see Fig. 1C).
Little information exists on the functional modi and architecture of the NSTs. Membrane topology has been determined for the murine CMP-sialic acid transporter (CST) and, in contrast to the theoretically predicted eight transmembrane domains, it was shown to contain 10 transmembrane domains (14). Evidence is accumulating that NSTs in the active state are dimers (15)(16)(17)(18) or higher order complexes (19). In the case of the yeast GDP-mannose transporter, the C-terminal span seems to be essential for the formation of the dimer, while the N-terminal cytosolic tail is required for export from the ER (15,18). In contrast, the cytosolically located N-and C-terminal tails of murine UGT were found to be dispensable. Deletion of both ends affected neither the activity nor the subcellular destination of the UGT (11).
In order to define functional subdomains in NSTs, Aoki et al. (20) constructed chimeric molecules by exchanging segments of human UGT isoform 1 with the corresponding domains of human CST. As expected from previous truncation experiments (11), the cytoplasmic tails were interchangeable. However, exchanges of amino acid stretches that represent potential transmembrane domains inactivated the UGT with a single exception. If the C terminus plus the C-terminal helix of the UGT were replaced by the corresponding segments of human CST, the chimera preserved UGT activity. These data are consistent with experiments of Gao and Dean (18) and suggest that the C-terminal domain of NSTs is involved in the folding of the transport active site but not in determining substrate specificity.
An efficient way to obtain information on structure/function relationships in NSTs is by analyzing molecular defects that cause altered NST activity in mutant cells. Such an analysis in CHO Lec2 cells (defective in Golgi transport of CMP-sialic acid) and in murine Had-1 cells identified two highly conserved amino acids (Gly 189 in the hamster CST and Gly 178 in the murine UGT) that are likely to participate in the folding of active transporters (11,21).
The goal of this study was to identify molecular defects that cause inactivation of the UGT in CHO Lec8 cells. Because galactose is the major acceptor for sialic acid in complex type glycoconjugates (22), the lec8 defect interferes with the maturation of sialylated structures. Among the glycotopes that are not synthesized is polysialic acid (PSA), a specific posttranslational modification of the neural cell adhesion molecule (NCAM) (23). PSA can be detected with high sensitivity and has been shown in recent studies to be an excellent indicator for transformants in which defects in the sialylation pathway are complemented. Expression cloning in combination with PSA detection has been used to identify the polysialyltrans-ferase ST8SiaIV (24,25), the CST from mouse (26) and hamster (27), and the CMP-Neu5Ac synthetase (3). In this study, PSA was a marker in analytical steps carried out to determine the functional consequences of lec8 mutations.
Nine independent Lec8 isolates were analyzed at the molecular level. Six were found to express truncated transcripts, while three mutants possess a UGT with a single amino acid change. Amino acids replaced or deleted in the missense mutants are highly conserved residues in all mammalian transporters. Two of these mutations inactivate both the UGT and CST transporters revealing their importance for the correct folding or functional assembly of NSTs.

EXPERIMENTAL PROCEDURES
Antibodies-Monoclonal antibody (mAb) 12CA5, directed against the hemagglutinin (HA) epitope YPYDVPDYASL, was purchased from Roche Molecular Biochemicals, and mAb M5, directed against the FLAG sequence MDYKDDDDK, was from Sigma. mAb 735, directed against PSA, has been described (28). A rabbit antiserum against the catalytic domain of ␣-mannosidase II was a kind gift of Dr. K. Moremen (University of Georgia, Athens, GA). Secondary antibodies used in this study were anti-mouse Ig-alkaline phosphatase conjugate (Dianova), anti-mouse Ig-Cy3 conjugate (Molecular Probes, Inc., Eugene, OR), and anti-rabbit Ig-Alexa488 conjugate (Molecular Probes).
Isolation of the Hamster UGT-A hamster UGT homologue was isolated by colony hybridization using a digoxigenin-labeled DNA probe generated by PCR from the 3Ј-end (nucleotides 640 -1182) of human  (24). In a first round, 3 ϫ 10 3 colony forming units were plated per 140-mm dish and after overnight incubation were transferred to nitrocellulose filters. 25 positive colonies were selected and plated on 94-mm dishes. Preparation of the probe, hybridization, and detection of positive colonies were performed as described in the DIG System user's guide for filter hybridization (Roche Molecular Biochemicals). In the second hybridization round, 10 colonies gave positive signals on duplicate filters. Positive colonies were picked and amplified in Luria broth (LB) medium supplemented with 100 g/ml ampicillin, and plasmid DNA was isolated using the High Pure Plasmid Isolation Kit (Roche Molecular Biochemicals) according to the manufacturer's instructions. Inserts were released by XhoI digestion, and after separation on a 2% agarose gel and transfer onto nitrocellulose membranes were verified by a second hybridization step with the digoxigenin-labeled probe described above. Positive clones of Ͼ1 kb were sequenced by the dideoxy chain termination method using a DNA sequencing kit (AbiPrism). Only one clone, pABE-SO1, contained the full-length hamster UGT sequence in inverse orientation. Northern Blot Analysis-Total RNA was isolated from CHO and Lec8 cells using the RNeasy Midi Kit (Qiagen). Polyadenylated RNA (Poly(A) ϩ RNA) was selected from total RNA using the QuickPrep Micro FIG. 1. Cloning and analysis of the hamster UGT cDNA. A, schematic representation of the strategy used to isolate the hamster UGT by a homology approach. A digoxigenin-labeled probe comprising 543 nucleotides of the human UGT coding region was used to screen a cDNA library from CHO cells. Clone pABE-SO1 was identified, which contains the entire ORF and the 3Ј-untranslated region of the hamster UGT cDNA. 5Ј-Rapid amplification of cDNA ends was used to confirm the translational start site contained in pABE-SO1. A fragment of 295 bp was amplified, but no additional ATG was found. The arrows indicate the position and direction of primers used to amplify the UGT ORF from wild type and mutant cell lines. B, primary sequence comparison of UGTs cloned from different species. Invariant amino acid positions are shown in black boxes, and amino acids conserved in four transporter sequences are shaded. Accession numbers are AF299335 (hamster UGT), D88454 (human UGT1), D88146 (human UGT2), AB027147 (murine UGT), and AB023425 (S. pombe UGT). C, sequence identity between UGTs of different species are expressed as percentages. D, Northern blot analysis of CHO K1 and Lec8. mRNA (2 g/lane) was electrophoresed, transferred to nylon membrane, and hybridized with a digoxigenin-labeled RNA antisense probe, transcribed from the 5Ј-end of the hamster UGT cDNA. Hybridization signals of 2.4 and 1.4 kb are visible in CHO K1. In the Lec8 mutant, only the higher molecular weight band was visible and at a drastically reduced level. To confirm equal loading, the blot was developed with a digoxigenin-labeled probe for glyceraldehyde-3-phosphate dehydrogenase.
Construction of Epitope-tagged Transporters-To generate N-terminally FLAG-tagged hamster UGT, the insert contained in the vector pABE-SO1 was amplified with oligonucleotide primers SO23 and SO24, which include EcoRI and XbaI restriction sites, respectively (Table II). PCR products were treated with EcoRI and XbaI, purified, and ligated into the vector pME8.1, which is a derivative of the eukaryotic expression vector pcDNA3 containing the FLAG-sequence (MDYKDDDDKEF) 5Ј of the EcoRI site. The resulting plasmid pcDNA3-FLAG-SO1 directs the expression of an N-terminally FLAG-tagged UGT under the control of the cytomegalovirus promoter. The nonapeptide underlined in the FLAGsequence is recognized by the anti-FLAG mAb M5. The construction of the FLAG-tagged murine CST has been described previously (21).
To generate C-terminally HA-tagged transporters, coding sequences of the hamster UGT and murine CST were amplified by PCR using oligonucleotides SO18 and SO19 or ME41 and ME42, respectively, which introduce BamHI restriction sites upstream and downstream of the coding sequence (Table II). After digestion with BamHI, PCR products were purified and ligated into the BamHI sites of the eukaryotic expression vector pEVRF0-HA (29). The resulting plasmids direct the expression of C-terminally HA (GSYPYDVPDYASL)-tagged transporters under the control of the cytomegalovirus promoter. The sequence motif recognized by the anti-HA mAb 12CA5 is underlined. All constructs were confirmed by sequencing.
Reverse Transcription Polymerase Chain Reaction (RT-PCR)-1 g of poly(A) ϩ RNA was reverse transcribed using 200 units of Superscript II RNase H Ϫ reverse transcriptase (Life Technologies, Inc.) and the oligonucleotide SO47 annealing to nucleotides 1273-1291 in the hamster UGT cDNA. The cDNA was then amplified by nested PCR using the primer pair SO45 and SO46 in the first amplification round and the primer pair SO48 and SO11 in the second amplification round. PCR conditions were as follows: 30 s at 57°C, 90 s at 72°C, and 15 s at 94°C for 40 cycles. Products of RT-PCR reactions were purified on agarose gels (GFX-PCR DNA and Gel Band Purification Kit; Amersham Pharmacia Biotech) and subcloned into the pGEM-T vector (Promega) prior to sequencing.
Site-directed Mutagenesis-Missense mutations identified in Lec8derived UGT cDNAs as well as other point mutations described in this study were introduced into the wild type sequences of both epitopetagged forms of the hamster UGT and the murine CST. Site-directed mutations were introduced by PCR using Pfu polymerase (Stratagene) followed by fusion PCR. Oligonucleotide primers used to introduce point mutations and to generate full-length constructs by fusion PCR are listed in Table II. In order to generate HA-tagged UGT variants in the vector pEVRF0-HA, the primer combinations SO18/SO85 and SO19/ SO86, SO18/SO97 and SO19/SO98, and SO18/SO105 and SO19/SO106 were used to introduce the mutations Y122H, ⌬S213, and G281D, respectively. DNA fragments resulting from these PCRs were purified on 2% agarose gels using the GFX-PCR Kit (Amersham Pharmacia Biotech), corresponding fragments were combined, and fusion PCRs were carried out with the primers SO18/SO19, which add BamHI restriction sites to the 3Ј-and 5Ј-ends of the DNA sequence (see Table II). To generate N-terminally FLAG-tagged UGT variants in the vector pcDNA3, the reactions described above were repeated, but SO18 and SO19 were replaced by SO23 and SO24, which add EcoRI and XbaI restriction sites, respectively, to the DNA sequences. Mutations in the murine CST were introduced with the primer combinations SO114/ SO115 for Y98H, SO116/SO117 for ⌬S188, and SO118/SO119 for G256D. Flanking primers used in the mutagenesis step were ME41 (used together with SO114, SO116, and SO118) and ME42 (used together with SO115, SO117, and SO119). ME41 and ME42, which add BamHI restriction sites to the 3Ј-and 5Ј-ends of the DNA sequences (see Table II) were also used in the fusion reaction. All constructs were confirmed by sequencing.
Transient Transfections-Recipient cells for transient transfections were the Lec8 subclone Lec4.8.7A and the Lec2 subclone 6B2 (21). Expression was monitored by Western analysis, and UGT function was measured by the ability to produce sialylated NCAM detected by anti-PSA antibody (22). While Lec2 cells do not produce spontaneous PSApositive revertants at a detectable level (even after several months of culture), spontaneous Lec8 revertants arose at a significant frequency. Thus, it was necessary to deplete revertants by panning on anti-PSAantibody 735 every 3-5 days (for details, see Ref. 3). Of the nine Lec8 subclones used in this study, Lec4.8.7A cell populations produced the fewest revertants, and therefore this cell line was routinely used as the recipient in transfection experiments.
Transfections were carried out with 5 ϫ 10 5 cells seeded in 6-cm Petri dishes (or on glass coverslips placed into 6-cm Petri dishes) 24 h prior to transfection. Cells were washed twice with Opti-MEM and transfected with 1 g of cDNA mixed with 6 l of LipofectAMINE in 1 ml of Opti-MEM (Life Technologies). Cells were cultured for 6 -8 h in the presence of the transfection mixture. Transfections were stopped with 2 volumes of the normal culture medium supplemented with 10% fetal calf serum. After an additional 24 -72 h of culturing, analytical steps were performed.
Indirect Immunofluorescence-Cells grown overnight on glass coverslips were washed with PBS and fixed in 4% paraformaldehyde in PBS for 20 min. After three washing steps with PBS, the cells were permeabilized for 30 min with 0.2% saponin in PBS containing 0.1% bovine

Molecular Cloning and Expression of the Hamster UGT-To
isolate a hamster UGT cDNA, a cDNA library from CHO K1 cells was screened using colony hybridization with a digoxigenin-labeled probe derived from the 3Ј part (nucleotides 640 -1182) of the coding region of isoform 1 of the human UGT (UGT1; Ref. 9; accession number D84454). One of 10 colonies that gave positive signals on duplicate filters pABE-SO1 harbored an insert of sufficient size (ϳ1.4 kb) to encode a fulllength UGT. Sequencing revealed an open reading frame (ORF) of 1197 bp, predicting a protein of 398 amino acids with a calculated molecular mass of 41.5 kDa. Since only nine nucleotides upstream of the translation initiation codon were present (see Fig. 1A), 5Ј-rapid amplification of cDNA ends was used to search for additional start sites. 295 nucleotides were analyzed, but no in frame ATG was found. The 3Ј-noncoding region contained a polyadenylation signal AATAAA starting 23 bp upstream of the poly(A) tail.
In Fig. 1B, the amino acid sequence deduced from the newly cloned gene is aligned to UGT sequences from human (9, 12), mouse (11), and yeast (13). The high conservation between the mammalian genes (overall identity Ͼ93%; see Fig. 1, B and C) strongly suggests that the new cDNA represents the hamster homologue of UGT. In humans, two UGT cDNAs that result from alternative splicing have been isolated. The two predicted UGT proteins differ only at their C termini (see Fig. 8) (12). Interestingly, the hamster UGT is similar to human UGT2, while the recently cloned murine UGT resembles human UGT1 (see Fig. 1B). RT-PCR was used in this study to search for alternative UGT splice variants in CHO cells, but none were isolated.
Northern blot analysis was used to examine UGT gene transcripts in the Lec8 mutant (Fig. 1D). A digoxigenin-labeled cRNA transcribed from the 5Ј-coding region (nucleotides 32-503) of the hamster UGT was used as a probe. Bands of about 2.4 and 1.4 kb represent UGT gene transcripts in CHO K1 cells. Only the higher molecular weight band could be detected in the Lec8 mutant, and this was present at drastically reduced intensity, as revealed by the glyceraldehyde-3-phosphate dehydrogenase control. These data suggest reduced production or stability of UGT gene transcripts in Lec8.
The ORF Contained in pABE-SO1 Encodes a Functional UGT-To test functional activity, the ORF contained in pABE-SO1 was amplified by PCR and subcloned into pcDNA3-FLAG and pEVRF0-HA, that direct the expression of N-terminally FLAG (FLAG-UGT) and C-terminally HA-tagged (UGT-HA) translation products, respectively. cDNAs were transiently expressed in Lec4.8.7A cells, and complementation was monitored by reappearance of the polysialylated form of NCAM

FIG. 3. Expression of UGT mRNA in CHO wild type cells and Lec8 mutants.
RT-PCR analysis of mRNA samples isolated from CHO K1 and Lec8 mutants is shown. Poly(A) ϩ RNA was reverse transcribed using the primer SO47, which is complementary to nucleotides 1273-1291 in the 3Ј-untranslated region of hamster UGT (Fig. 1A). Nested PCR was carried out with the primer pairs SO48/SO46 in the first and SO48/SO11 in the second round (primers are shown in Fig. 1A and Table II). A product of the expected molecular mass of the UGT ORF (ϳ1.2 kb) was obtained in CHO K1 and the mutants Lec8.5H, Lec8.1C, and LEC10.Lec8. Multiple bands of lower molecular masses were amplified from mRNA of all clones including CHO K1. Sequencing demonstrated that all arise from aberrant splicing events.  (PSA-NCAM), which is present in CHO wild type cells (Fig. 2) (34). The defect in UDP-galactose transport in Lec8 mutants precludes the addition of galactose to glycoconjugates and consequently also prevents the addition of sialic acid, resulting in a PSA-negative phenotype in Lec8 cells (see Fig. 2). Both the epitope-tagged constructs and the UGT without an epitope tag rescued the lec8 mutation, leading to reappearance of PSA, which was specifically detectable with mAb 735 (28). To demonstrate specificity, the FLAG-UGT construct was also expressed in Lec2 cells (6B2; see Fig. 2). Due to a defect in CST that severely reduces Golgi import of CMP-sialic acid (35), Lec2 cells are PSA-negative. In accordance with the high substrate specificity of NSTs, 6B2 cells could not be rescued with the hamster UGT cDNA (see Fig. 2). In all samples, the specificity of the PSA signal was shown by EndoNE digestion, which degrades PSA and abolishes binding of mAb 735 (36). Expression levels of the recombinant epitope-tagged proteins were monitored by Western blot analysis using the anti-FLAG mAb M5 and the anti-HA mAb 12CA5. The data presented in Fig. 2 (bottom panel) demonstrate that, while epitope-tagged transporters were expressed in both the Lec8 clone Lec4.8.7A and the Lec2 clone 6B2, rescue of PSA expression by transfected UGT occurred only for the Lec8 phenotype, while the Lec2 phenotype was not corrected.

Molecular Analysis of UGT in Independent Lec8
Mutants-To investigate the molecular basis of UGT defects in nine independent Lec8 CHO mutants, UGT gene expression was examined by RT-PCR in wild type and mutant cells. Reverse transcription was performed with the gene-specific primer SO47 that annealed to the 3Ј-untranslated region (see Fig. 1A). Nested PCR was performed with primer pairs SO45/ SO46 in the first and SO48/SO11 in the second round to obtain the ϳ1.2-kb UGT ORF. A strong band of ϳ1.2 kb was present in CHO K1 wild type cells and in the mutants Lec8.5H, Lec8.1C, and LEC10.Lec8. These mutants also expressed the ϳ2.4-kb UGT transcript by Northern analysis (not shown). The largest product from the Lec8 mutant migrated slightly faster, and the largest product from Lec3.2.8, Lec3.2.8.1, and Lec4A.Lec8 migrated at about 1 kb. Several RT-PCR products of lower molecular weight were observed in all cells including wild type and represent aberrant UGT splice products as discussed below.
To identify UGT mutations in Lec8 mutants, UGT cDNAs were excised from the gel and subcloned, and plasmid DNA was sequenced. To eliminate PCR artifacts, the steps from RT-PCR to sequencing were carried out in triplicate. Four mutations identified by sequencing are listed in Table III. Loss of UGT activity in the Lec8 mutant is caused by a deletion of 100 bp in the coding region and the consequent introduction of a premature stop codon. A potential translation product (UGT-E92stop) would consist of 92 amino acids and terminate after a newly introduced glutamic acid residue. In clone Lec8.5H, the in frame elimination of a triplet (nucleotides 636 -638) is responsible for the excision of serine 213 (⌬S213). The transitions G844A and T364C cause replacement of glycine 281 by aspartic acid (G281D) in Lec8.1C and tyrosine 122 by histidine (Y122H) in LEC10.Lec8, respectively. Since bands displayed in Fig. 3 hybridized to a UGT cDNA probe (data not shown), the 1-kb product and many of the faster migrating PCR products were isolated. Sequencing demonstrated that all contained UGTcoding sequence with different internal deletions. These products were therefore not further characterized.
Expression of Mutant UDP-Galactose Transporter Proteins-In order to find out whether the single amino acid changes identified in cDNAs encoding the UGT ORF of Lec8.5H, Lec8.1C, and LEC10.Lec8 interfere with the expression of stable proteins, site-directed mutagenesis was used to introduce each change into the wild type UGT sequence. Both FLAG-and HA-tagged hamster UGT constructs were subjected to site-directed mutagenesis and transiently expressed in COS-7 cells. UGT expression was analyzed by Western blotting 48 h post-transfection using the anti-HA mAb 12CA5 (Fig. 4B) and the anti-FLAG mAb M5 (Fig. 4A). Wild type and mutant UGTs (⌬S213, G281D, Y122H) migrated with an apparent molecular mass of about 42 kDa. Other bands detected by mAb 12CA5 were not related to the HA tag, since they were present in all lanes including control. Interestingly, the truncated protein FLAG-UGT-E92stop isolated from clone Lec8 was also stable and migrated with an apparent molecular mass of 22 kDa. Fig. 5, the three segments of the hamster UGT sequence that harbor a lec8 missense mutation are aligned with the corresponding homologous region of other NSTs. It can be seen that the amino acid residues mutated in Lec8 mutants are invariant in all mammalian transporters, irrespective of their substrate specificity. Moreover, the three positions appear to be conserved in the UGT from S. pombe, and two are present in the GDP-mannose transporter isolated from Saccharomyces cerevisiae, but different residues occur in the UDP-GlcNAc transporter from Kluyveromyces lactis or the GDP-mannose transporter from Leishmania donovanii.

Lec8 Mutations Occur at Highly Conserved Amino Acid Residues-In
The high conservation among mammalian transporters of different substrate specificities strongly suggests that lec8 mutations affect structural elements that fulfill essential functions in this protein family. To test this hypothesis, the Lec8 UGT mutations were introduced into a transporter of different substrate specificity, the HA-tagged murine CST. The functional activity of CST mutants was tested in the Lec2 mutant 6B2 (21) by monitoring PSA reexpression after transfection (Fig. 6). Both Lec4.8.7A and 6B2 mutants are PSA-negative due to their respective NST defect, but each expresses PSA after transfection with the appropriate epitope-tagged NST cDNA (wild type). Deletion of the invariant serine-residue (⌬S213 in UGT; ⌬S188 in CST) and exchange of the invariant glycine to aspartic acid (G281D in UGT; G256D in CST) abolished the ability of these cDNAs to rescue PSA expression, and thus transport of both UDP-galactose and CMP-sialic acid was inhibited by these mutations. By contrast, the amino acid exchange Y122H identified in the LEC10.Lec8 mutant did not lead to an inactivated UGT; nor did the corresponding Y98H change lead to inactivation of CST (Fig. 6A).
The T364C mutation was the only change in sequence in LEC10.Lec8 UGT cDNAs, and the same mutation was shown to be present in genomic DNA of LEC10.Lec8 cells. When 100 base pairs surrounding this site were amplified by PCR of genomic DNA from LEC10.Lec8 and control cells, LEC10.Lec8 DNA gave two products; about 50% carried the wild type se-quence with T at nucleotide position 364, which was also present in genomic DNA of CHO K1 and the LEC10 parent of LEC10.Lec8, and about 50% carried the point mutation T364C, which causes the exchange Y122H. Therefore, the mutation T364C is present in the genomic DNA of LEC10.Lec8 cells and must represent the active UGT allele, since no cDNAs with T at position 364 were found by RT-PCR of LEC10.Lec8 mRNA.
Because UGT-122H in the LEC10.Lec8 genome is present on the LEC10 mutant background, it seemed possible that the dominant mutation LEC10, which causes the de novo expression of ␤-4-N-acetylglucosaminyltransferase III and introduces a bisecting GlcNAc into the core of complex N-glycans (37), might somehow cause the UGT-122H mutation not to be phenotypically expressed. However, when LEC10.Lec8 cells were used as hosts to express the mutant UGT-122H cDNA, correction of the PSA-negative phenotype was again observed, demonstrating that complementation by UGT-122H also occurred in this background (Fig. 6B). The lower PSA signal obtained after transient transfection of LEC10.Lec8 cells is caused by LEC10.Lec8 mutant populations generating PSA-positive revertants with considerable frequency and requiring daily panning to avoid background staining, which interferes with optimal transfection conditions. Nevertheless, Western analysis showed readily detectable expression of each transfected construct in all cells in Fig. 6 (see bottom panels). It is therefore apparent that expression of EndoNE-sensitive PSA did not depend on differential levels of expression of transfected plasmids. Thus, it seems likely that the ability of UGT-122H to function is due to overexpression in transfectants. Whereas overexpression of the two NST mutants (⌬S213 in UGT; ⌬S188 in CST and G281D in UGT; G256D in CST) did not obscure their inability to transport nucleotide-sugar, overexpression of the Y122H mutation allowed this mutant UGT-122H (and the corresponding CST-98H) to function. By contrast, endogenous levels of the UGT-122H expressed in LEC10.Lec8 cells are clearly inactive (Fig. 6B). A similar expression-dependent activity has most recently been described for variants of the yeast GDP-mannose transporter that exhibit single amino acid exchanges in a highly conserved primary sequence motif (38).

Subcellular Localization of Wild Type and Mutant
NSTs-Biochemical studies carried out on isolated organelles have demonstrated that UDP-galactose and CMP-sialic acid transport activity are strictly associated with Golgi vesicles (39). Later studies confirmed these results by demonstrating Golgi localization of epitope-tagged UGT (40) and CST (26). To find out whether UGT and CST carrying a lec8 mutation are correctly targeted inside the cell, indirect immunofluorescence was used to compare the subcellular localization of HA-tagged wild type and mutant proteins. Constructs were transiently expressed in CHO K1, and 30 h later, cells were fixed in paraformaldehyde and permeabilized with saponin, and the localization of epitopes was determined by indirect immunofluorescence with the anti-HA mAb 12CA5. Simultaneously, the cells were stained with an antiserum directed against ␣-mannosidase II, a known marker for the Golgi apparatus (33). In accordance with the earlier data (26,40) Golgi localization was found for the wild type transporters (Fig. 7), and the same strict co-localization with ␣-mannosidase II was observed for UGT and CST proteins with a missense mutation. Lec8 missense mutations do not interfere with the signals that are responsible for Golgi destination and do not appear to significantly affect the folding of UGT. In contrast, the FLAG-tagged mutant UGT-E92stop isolated from subclone Lec8 showed an ER-like distribution. DISCUSSION In order to define structure/function relationships in the NST family of proteins, we have identified mutations that weaken or inactivate mammalian CST (21) and UGT (this study). The hamster UGT gene was cloned from a CHO cDNA library, and epitope-tagged versions of the encoded protein were shown to complement the Lec8 CHO glycosylation mutant (see Fig. 2). The conservation between UGT sequences of different animal species is high (Ͼ93% amino acid identity; see Fig. 1, B and C). However, while two UGT isoforms exist in humans (12) only one form was found in CHO cells (this study) and in mice (11). Moreover, the mouse UGT resembles human UGT1, while the CHO gene encodes a protein most homologous to UGT2. Differences in human UGT isoforms are confined to the extreme C terminus (see Fig. 1B) and result from alternative splicing as schematically shown in Fig. 8. The human UGT gene contains five exons, and the ORF contains stop codons in exons 4 and 5. Translation of exons 1-4 results in isoform 1. If a splice consensus motif contained in exon 4 is used, exon 5 becomes fused to the 5Ј part of exon 4, and isoform 2 can be translated. PCR of hamster genomic DNA revealed that splice consensus sites in exons 4 and 5 are conserved between the two species (data not shown). Therefore, the fact that CHO cells express only UGT2 suggests that differences in isoform expression may reflect tissue-and/or species-specific regulation processes.
Out of nine independent Lec8 clones, three (Lec8.5H, Lec8.1C, and LEC10.Lec8) synthesize cDNAs encoding a fulllength UGT ORF, each of which contains a single amino acid change. By contrast, a UGT cDNA from the Lec8 mutant contains a deletion of 100 bp, which causes the introduction of a premature stop codon. The resulting protein, UGT-E92stop, is stable but localizes to the endoplasmic reticulum. The lack of the C-terminal region in the UGT of Lec8 cells explains a recent observation made by Ishida and co-workers (11). Using a polyclonal antibody directed against the C-terminal peptide of the murine UGT, the authors were able to precipitate the UGT from CHO K1 but not from Lec8 cells (11).
The point mutations identified in three Lec8 UGTs fall at invariant amino acid residues that are part of homology cassettes conserved in all mammalian transporters cloned so far (see Fig. 5). It is therefore unlikely that these lec8 mutations interfere with UGT-specific functions and probable that they mark positions essential for the formation of transport activity. This speculation is supported by our finding that two lec8 mutations inactivate not only UGT but also the functional activity of murine CST. Since CST and UGT are the most distant relatives in the known family of mammalian NSTs (41), it seems highly likely that the conserved amino acids identified by these point mutations are of general importance.
To our surprise, the mutant UGT-122H and its counterpart CST-98H were active following transfection and were able to complement the defect in Lec8 and Lec2 cells, respectively. The point mutation T364C, which gives rise to UGT-122H, is present in the genomic DNA of LEC10.Lec8 cells, and the rest of the UGT sequence is identical to the UGT from three other CHO lines. Since LEC10.Lec8 cells are PSA-negative (see Fig. 6B) and express normal levels of UGT-122H mRNA (not shown), the loss of PSA in Pro Ϫ 5LEC10.Lec8 cells must be caused by the point mutation Y122H. Defective UGT transport activity is, however, manifest only if the mutant gene is expressed at endogenous levels. The gain of function in transfected cells is most likely due to overexpression, which may improve folding processes, resulting in stabilized transport units. Support for this interpretation comes from a very recent study by Gao et al. (38). The authors identified variants of the yeast GDP-mannose transporter (VRG4), whose GDP-mannose transport activity was severely reduced. However, if the mutant proteins were expressed at higher levels by using strong promoter sequences, transport activity could be restored, and yeast cells grew normally. In the in vitro situation, the transport defect could also be compensated for by increased GDP-mannose concentration, indicating a reduced affinity for the nucleotide-sugar. Mutants described by Gao et al. (38) map to a region conserved in both confirmed and putative GDP-mannose transporter ORFs but not in NSTs of a different specificity. The analysis of these mutants allowed the authors to identify a sequence motif that is involved in the binding of GDP-mannose.
In mutant Lec8.5H, the in frame excision of the triplet 636 -638 removes the highly conserved serine at position 213 in hamster UGT. This serine is the starting point of a homology box consisting of 13 amino acids. 11 positions are invariant in all mammalian transporters cloned so far (see Fig. 1B). The importance of this region was previously identified in lec2 CST mutations, since the mutant CST-G189E (G189 in CST is identical to G214 in the hamster UGT) was shown to inactivate the CST (21). These inactive transporters with a point mutation were correctly targeted to the Golgi, demonstrating that the mutations do not grossly interfere with protein folding (Ref. 21 and this study). Moreover, alanine is tolerated at both conserved positions (21). 2 A potential explanation, therefore, is that the highly conserved stretch of amino acids participates in motile processes that are required to translocate nucleotidesugars across the Golgi membrane. Additional kinetic and mutational studies are required to substantiate this model.
Five clones (Lec8.2B, Lec3.2.8, Lec3.2.8.1, Lec4.8.7A, and Lec4A.Lec8) contained few UGT gene transcripts that were not detected by Northern analysis but were revealed by RT-PCR (Fig. 3). Sequencing of UGT-specific sequences from these cells revealed extended internal deletions in the UGT coding region. Lec8 mutations in these clones seem to affect splice donor or acceptor sites and not elements that regulate gene expression. Aberrant splice products were present in all cell lines investigated including wild type (CHO K1; see Fig.  3) and have recently been described to occur also in Had-1 clones (11). Their ubiquitous presence may have some functional significance.
It has been known for a long time that NSTs play a central role in the maturation of glycoconjugates (for review, see Ref. 4), and the lack of a clinical manifestation that could be explained by the inactivation of a NST was explained by the likelihood of a lethal phenotype. However, very recently a new type of carbohydrate-deficient glycoprotein syndrome has been described, which can be attributed to a defect in the GDP-fucose transport system (42). The patient is severely affected and suffers from various neurological, immunological, and psychomotor defects (43). Oral supplementation of fucose has been shown to partially compensate for the clinical manifestations (44). Cloning the human GDP-fucose transporter and identifying its molecular defect will demonstrate whether the substrate binding motif identified by Gao et al. (38) is concerned in these patients. Moreover, the observation that individual fucosyltransferases are differentially influenced by the transport defect further supports earlier studies, in which NSTs have been suggested to be involved in the metabolic control of the glycosylation process (45,46). Understanding structure-function relationships in this recently identified protein family is therefore not only of medical importance but is of interest also with respect to biotechnological processes.