Chinese Hamster Ovary (CHO) Cells May Express Six β4-Galactosyltransferases (β4GalTs)

Six β4-galactosyltransferase (β4GalT) genes have been cloned from mammalian sources. We show that all six genes are expressed in the Gat−2 line of Chinese hamster ovary cells (Gat−2 CHO). Two independent mutants termed Pro−5Lec20 and Gat−2Lec20, previously selected for lectin resistance, were found to have a galactosylation defect. Radiolabeled biantennary N-glycans synthesized by Pro−5Lec20 were proportionately less ricin-bound than similar species from parental CHO cells, and Lec20 cell extracts had a markedly reduced ability to transfer Gal to GlcNAc-terminating acceptors. Northern blot analysis revealed a severe reduction in β4GalT-1 transcripts in Pro−5Lec20 cells. The Gat−2Lec20 mutant expressed β4GalT-1 transcripts of reduced size due to a 311-base pair deletion in the β4GalT-1 gene coding region. Northern analysis with probes from the remaining five β4GalT genes revealed that Gat−2 CHO and Gat−2Lec20 cells express all six β4GalT genes. Unexpectedly, the β4GalT-6 gene is not expressed in either Pro−5 or Pro−5Lec20 cells. Thus, in addition to a deficiency in β4GalT-1, Pro−5Lec20 cells lack β4GalT-6. Nevertheless, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry data of N-glycans released from cellular glycoproteins showed that both the β4GalT-1− (Gat−2Lec20) and β4GalT-1−/β4GalT-6−(Pro−5Lec20) mutants have a similar Gal deficiency, affecting neutral and sialylated bi-, tri-, and tetraantennaryN-glycans. By contrast, glycolipid synthesis was normal in both mutants. Therefore, β4GalT-1 is a key enzyme in the galactosylation of N-glycans, but is not involved in glycolipid synthesis in CHO cells. β4GalT-6 contributes only slightly to the galactosylation of N-glycans and is also not involved in CHO cell glycolipid synthesis. These CHO glycosylation mutants provide insight into the variety of in vivosubstrates of different β4GalTs. They may be used in glycosylation engineering and in investigating roles for β4GalT-1 and β4GalT-6 in generating specific glycan ligands.

To identify in vivo functions of each ␤4GalT, it is important to consider the tissue expression pattern as well as acceptor specificity. For example, ␤4GalT-1 is up-regulated in lactating mammary glands (16), whereas ␤4GalT-2 is not. 2 Furthermore, mice deficient in ␤4GalT-1 do not produce lactose in milk (17,18). The ␤4GalT-1, -3, -4, and -5 genes are ubiquitously expressed, whereas the ␤4GalT-2 and ␤4GalT-6 genes exhibit a more restricted expression pattern (8,10,11). Although ϳ80% mice lacking ␤4GalT-1 die soon after birth, the remainder are viable and fertile (17,18). Serum glycoproteins from ␤4GalT-1 Ϫ/Ϫ mice were found to be galactosylated to ϳ10% compared with those from wild-type mice (19), providing evidence for the existence of other functional ␤4GalTs. However, almost nothing is known of the biological roles of these ␤4GalTs, and their acceptor specificity has not been defined for in vivo substrates.
Preparation of Radiolabeled VSV Glycopeptides-Cells growing in suspension were infected with VSV and subsequently cultured in ␣-medium containing reduced glucose (0.5 mg/ml), 2% Nuserum (Collaborative Research), and 83 Ci of [ 3 H]GlcN/10 ml as described previously (22). Virus was purified by gradient centrifugation and exhaustively digested with Pronase to generate Pronase glycopeptides.
Lectin Affinity Chromatography of Radiolabeled Glycopeptides-VSV glycopeptides desalted on Bio-Gel P-2 were fractionated on a 5-ml column of ConA-Sepharose as described previously (22) into branched and biantennary N-glycans. The ConA-bound fraction was desalted, treated with neuraminidase, and fractionated on a 5-ml column of RCA II -agarose. Lectin chromatography was performed at room temperature or at 4°C. Phosphate-buffered saline containing 100 mM GalNAc or 200 mM lactose was used to elute bound glycopeptides. Samples of 0.5-1.0 ml were mixed with Ecolume at a ratio of ϳ1:10 and counted in a scintillation counter.
Preparation of Cell Extracts-Post-nuclear supernatant from Lec20 and parental CHO cells was prepared as described (23). Briefly, cells (ϳ6 ϫ 10 7 ) were washed two times with saline, followed by one wash with homogenizing buffer (10 mM Tris-HCl, pH 7.4, and 250 mM sucrose), and incubated in 1 ml of homogenizing buffer containing an EDTA-free protease inhibitor tablet on ice. After 20 min, the swollen cells were homogenized using a Balch homogenizer (Industrial Tectonic Inc.) at 4°C. The lysate was centrifuged at 3000 rpm for 30 min at 4°C. Glycerol was added to the supernatant to a final concentration of 20% before storage at Ϫ80°C. For preparation of microsomal membranes, the post-nuclear supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C. Also, cell-free extract was prepared in 1.5% Triton X-100 after cell washing, to which glycerol was added to 20% by volume before storage at Ϫ80°C as described (23).
Release of N-Linked Oligosaccharides by PNGase F-Cells were harvested and washed three times with phosphate-buffered saline. The cell pellet was resuspended in 20 mM Tris-HCl, pH 7.4, to obtain ϳ1 ϫ 10 10 cells/ml, to which an equal volume of 3% Triton X-100 was added. The suspension was mixed well and incubated on ice for 10 min and at room temperature for 10 min. The suspension was vortexed for ϳ2 min and then centrifuged at 5000 rpm for 30 min. The supernatant was removed and stored at Ϫ80°C until further use. The protein concentration of the supernatant was ϳ10 mg/ml. N-Linked oligosaccharides were released by PNGase F treatment of glycoproteins bound to polyvinylidene difluoride membranes using a high-throughput microscale method as described by Papac et al. (25). Released oligosaccharides were passed 2 J. H. Shaper, personal communication. through a 0.6-ml cation-exchange resin (AG-50W-X8 resin, H ϩ form, 100 -200 mesh, Bio-Rad) to remove salt and protein contaminants prior to analysis by mass spectrometry.

Matrix-assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF-MS)-MALDI-TOF-MS was performed on
a Voyager DE Biospectrometry workstation (PerSeptive Biosystems) equipped with delayed extraction. A nitrogen laser was used to irradiate samples with ultraviolet light (337 nm), and an average of 240 scans were taken. The instrument was operated in linear configuration (1.2-m flight path), and an acceleration voltage of 20 kV was used to propel ions down the flight tube after a 60-ns delay. Samples (0.5 l) were applied to a polished stainless steel target to which 0.3 l of matrix was added and dried under vacuum (50 ϫ 10 Ϫ3 torr). Oligosaccharide standards were used to achieve a two-point external calibration for mass assignment of ions (26,27). 2,5-Dihydroxybenzoic acid/5-methoxysalicylic acid and 2,4,6-trihydroxyacetophenone matrices were used in the analysis of neutral and acidic oligosaccharides, respectively.
Generation of ␤4GalT-1 Transfectants-Different amounts of plasmid pSVL DNA containing a bovine ␤4GalT-1 cDNA (a generous gift of Dr. Joel H. Shaper) were mixed with pSV2neo DNA (5 g) separately and transfected into Pro Ϫ 5Lec20 cells using the Polybrene method described previously (28). Transfectants were selected for resistance to G418 (1.5 mg/ml active weight). The transfectants were expanded and tested for lectin resistance to PHA-E and for ␤4GalT activity with GlcNAc as acceptor.
Reverse Transcription (RT)-PCR-For reverse transcription, 2 g of poly(A) ϩ RNA, 0.5 g of oligo(dT) 12-18 primer, and 0.05 g of random hexamer were heated to 70°C for 10 min and slowly cooled to room temperature before adding 200 -400 units of Superscript II reverse transcriptase together with first strand buffer (Life Technologies, Inc.), 0.5 mM dATP, 0.5 mM dCTP, 0.5 mM dGTP, 0.5 mM dTTP, 10 mM dithiothreitol, and 1 unit/l RNasin (Promega). Reactions were incubated for 50 min at 42°C, heated at 70°C for 15 min, and stored at Ϫ20°C.
PCR for ␤4GalT-1 gene sequencing was performed using the forward primer 5Ј-GTAGCCCACMCCCYTCTTAAAGC-3Ј and the reverse primer 5Ј-AATGAGAGGGACCAGCCCAG-3Ј. The primers were designed on the basis of proximal 5Ј-and 3Ј-untranslated region sequences of the human, bovine, and mouse ␤4GalT-1 genes. The PCR mixture contained 15 pmol of primers, 2 l of reverse transcription product, 1 l of 10 mM dNTPs, 0.5 l of Taq DNA polymerase, 5 l of 10ϫ PCR buffer, and 3 l of 25 mM MgCl 2 in a total volume of 50 l. The mixtures were heated at 94°C for 2 min, followed by 94°C for 1 min, annealing at 65°C for 1 min, and elongation at 72°C for 2 min through 40 cycles. PCR products were purified using a QIAquick gel extraction kit (QIAGEN Inc.) and sequenced, either directly or after subcloning into the pCR2.1 vector using the Original TA cloning kit from Invitrogen.
For RT-PCR in Fig. 5B, the forward primer for CHO ␤4GalT-1 was 5Ј-TCACAGCCCCGGCACATTTCT-3Ј from exon III. The reverse primer in exon VI was 5Ј-TATCTTGGTGTCCCGATGTC-3Ј. For ␤4GalT-6, the forward primer 5Ј-ATGTCTGCGCTCAAGCGGAT-3Ј corresponded to the 5Ј-end of the coding sequence of CHO ␤4GalT-6, and the reverse primer 5Ј-GTCTTCGATTGGAGCTAACTC-3Ј corresponded to the 3Ј-end of the coding sequence. Each sample was subjected to one cycle at 94°C for 1 min, followed by 30 cycles at 94°C for 1 min, 57°C for 2 min, and 72°C for 3 min.
Glycolipid Extraction and Purification-About 10 9 exponentially growing cells were washed twice with cold phosphate-buffered saline. Glycolipids were extracted with chloroform/methanol (2:1) as described (29). The extract was evaporated to dryness, redissolved in chloroform/ methanol (1:1) at 10 8 cell eq/ml, and stored at Ϫ20°C. For purification, 0.9 ml of this preparation were evaporated to dryness and saponified with 1 M sodium hydroxide in methanol at 40°C for 1 h. After being neutralized with 1 M acetic acid, the solution was evaporated to dryness, and the residue was dissolved in 2 ml of methanol and 1.6 M aqueous sodium acetate (1:1). The solution was applied to a Sep-Pack C 18 cartridge, and the flow-through was collected and reapplied twice. The cartridge was washed with 40 ml of water, and glycolipids were eluted with 2 ml of methanol, followed by 10 ml of chloroform/methanol (2:1). The eluant was evaporated to dryness, dissolved in 1 ml of chloroform/ methanol (2:1), and stored at Ϫ20°C.

Reduced Galactosylation of N-Glycans in Pro
Ϫ 5Lec20 CHO Cells-Independently isolated Lec20 CHO mutants are resistant to the Gal-binding lectins PHA-E, PHA-L, and ricin (21), consistent with a reduction in cell-surface Gal residues. To rapidly determine if N-glycans synthesized in Lec20 cells lack Gal residues, uniformly labeled Pronase glycopeptides of the G-glycoprotein of VSV grown in parental (Pro Ϫ 5 CHO) or Pro Ϫ 5Lec20 cells were subjected to serial lectin affinity chromatography. ConA-Sepharose chromatography showed no difference in the proportion of branched (ϳ20%) and biantennary (ϳ80%) complex N-glycans between parental and mutant-derived VSV glycopeptides. However, when the ConA-bound, biantennary population of N-glycans was fractionated on RCA IIagarose at room temperature, a marked difference between parental and mutant glycopeptides was revealed. Whereas 46% of the Pro Ϫ 5 CHO/VSV biantennary N-glycans bound to RCA IIagarose, consistent with the presence of 2 Gal residues/Nglycan (30), there were no RCA II -bound glycopeptides among the Pro Ϫ 5Lec20/VSV biantennary species (Fig. 1A). This could be due to increased sialylation or decreased galactosylation. After neuraminidase treatment, ϳ71% of the Pro Ϫ 5Lec20/VSV biantennary glycopeptides bound to RCA II -agarose at 4°C (Fig.  1B) and were eluted with GalNAc, consistent with the presence of only 1 Gal residue/biantennary N-glycan (30). Proof that this binding was due to terminal Gal was obtained by ␤-galactosidase treatment, after which no Lec20/VSV glycopeptides bound to RCA II -agarose (Fig. 1C). Reduced RCA II -agarose binding of desialylated biantennary N-glycans was also found with [ 3 H]Gal-labeled glycopeptides from Pro Ϫ 5Lec20 cellular glycoproteins (data not shown). The combined data suggest that Pro Ϫ 5Lec20 cells have a defect in the addition of Gal residues to complex N-glycans.
Lec20 Mutants Have Reduced ␤4GalT Activity-␤4GalT enzyme assays were performed with detergent cell extracts and GlcNAc or biantennary GlcNAc-terminating glycopeptide (GnGn) as acceptor. Pro Ϫ 5Lec20 and Gat Ϫ 2Lec20 cell extracts had Յ10% galactosyltransferase activity compared with parental cells (Table I). Mixing equal amounts of parental and mutant cell extracts yielded one-half the level of ␤4GalT activity (Table I), showing that the reduced activity in Lec20 cells is not due to the presence of an inhibitor.
The coding region of ␤4GalT-1 cDNAs from Gat Ϫ 2 CHO cells was sequenced (Fig. 3A). The sequence predicts a polypeptide of 393 amino acids, and hydropathy analysis (31) revealed a single hydrophobic membrane-spanning domain of 20 amino acids near the N terminus, which predicts the type II transmembrane topology typical of Golgi glycosyltransferases (32). The sequence also predicts one putative N-glycosylation site. Clust-alW analysis showed that Gat Ϫ 2 CHO ␤4GalT-1 is 90.2% identical to mouse, 83.2% to human, 76.7% to bovine, and 61.9% to chicken ␤4GalT-1 at the amino acid level. The number and positions of all 7 Cys residues are conserved in Gat Ϫ 2 CHO ␤4GalT-1 (Fig. 3A).
Comparison of the Gat Ϫ 2 CHO and Gat Ϫ 2Lec20 ␤4GalT-1 sequences revealed that the Lec20 mutant was identical except for a 311-bp deletion that results in the production of a truncated protein of 214 amino acids derived from exons I, II, and V (Fig. 3B). This deletion includes a significant portion of the catalytic domain and therefore appears to be responsible for the lectin resistance phenotype and reduced ␤4GalT activity of the Gat Ϫ 2Lec20 mutant (Table I). The marked reduction of ␤4GalT-1 gene transcripts in Pro Ϫ 5Lec20 cells gives rise to an essentially identical galactosylation-defective phenotype.
Bovine ␤4GalT-1 Corrects the Phenotype of Lec20 Cells-To confirm that the reduced ␤4GalT activity and the lectin resistance phenotype of Lec20 cells result from an absence of ␤4GalT-1, bovine ␤4GalT-1 cDNA was transfected into Pro Ϫ 5Lec20 cells. Transfectants were obtained by selection with G418 and tested for their ability to bind the lectin PHA-E, for which the Lec20 mutant shows 7-fold resistance (21), and for their ␤4GalT activity. All transfectants were more sensitive to the toxicity of PHA-E and had increased ␤4GalT activity (Fig. 4). Two transfectants reverted almost to the parental phenotype. These results support the conclusion that the loss of ␤4GalT-1 is the cause of the Lec20 mutant phenotype.
Pro Ϫ 5 CHO Cells Lack ␤4GalT-6 Transcripts-The absence of functional ␤4GalT-1 in Lec20 CHO mutants clearly does not lead to a complete loss of ␤4GalT activity in cell extracts (Table  I). Thus, it was important to determine which of the other five mammalian ␤4GalT genes are expressed in CHO and Lec20 cells. Two Northern blots were prepared with poly(A) ϩ RNA from parental and mutant cells and hybridized with probes of ϳ1 kb derived by RT-PCR from the coding region of the corresponding murine ␤4GalT sequence. The results in Fig. 5A show that Gat Ϫ 2 CHO and Gat Ϫ 2Lec20 cells express the six ␤4GalT genes at similar levels. ␤4GalT-1 transcripts were the only ones altered in size in Gat Ϫ 2Lec20 cells. By contrast, Pro Ϫ 5 CHO and the Pro Ϫ 5Lec20 mutant were missing ␤4GalT-6 transcripts. Both also had a somewhat reduced level of ␤4GalT-3 transcripts (Fig. 5A). A complete lack of ␤4GalT-6 transcripts in Pro Ϫ 5 CHO and Pro Ϫ 5Lec20 cells was confirmed by RT-PCR (Fig. 5B). Thus, the Pro Ϫ 5 CHO cell, considered a "wild-type" CHO cell, is actually a "mutant" lacking ␤4GalT-6. Pro Ϫ 5Lec20 is a double mutant, essentially missing ␤4GalT-1 (transcripts were detected by the sensitive RT-PCR experiment in Fig. 5B, but not by Northern analysis in Fig. 2) and completely lacking  ]Glucosamine-labeled Pro Ϫ 5 CHO/VSV (E) and Pro Ϫ 5Lec20/VSV (q) Pronase glycopeptides that bound to and were eluted from ConA-Sepharose were desalted and fractionated on RCA II -agarose at room temperature (A). These biantennary N-glycans were also treated with neuraminidase (NANЈase) (B) or neuraminidase and ␤-galactosidase (␤-Gal'ase) (C) and fractionated on RCA II -agarose at 4°C. Bound glycopeptides were eluted with 100 mM GalNAc (open arrow), followed by 200 mM lactose (closed arrow). In A, lactose was added at slightly different fractions for Pro Ϫ 5Lec20 (solid arrow) and Pro Ϫ 5 CHO (dashed arrow).

FIG. 3.
A, ClustalW alignment of ␤4GalT-1 from CHO cells, mouse, human, bovine, and chicken. A potential N-glycosylation site (*) and conserved ␤4GalT-1 cysteine residues (q) are marked. B, the ␤4GalT-1 gene deletion in Gat Ϫ 2Lec20. Shown is a schematic diagram of a ␤4GalT-1 cDNA and protein product. Exons I-VI are based on the human ␤4GalT-1 gene (9). The Gat Ϫ 2Lec20 ␤4GalT-1 cDNA lacked nucleotides reflecting deletion of exon III and all of exon IV except for the last A residue. Shaded bars represent translated protein with the first and last amino acids.
␤4GalT-6. Changes in galactosylation of glycoproteins and glycolipids in the three CHO ␤4GalT mutants must be interpreted on this basis.

MALDI-TOF-MS Analysis of N-Glycans in CHO Cells
Lacking ␤4GalT-1, ␤4GalT-6, or Both-MALDI-TOF-MS has been used for both structural characterization (33) and relative quantitation (34) of neutral and sialylated N-glycans in a mixture. To examine in vivo galactosylation of the spectrum of N-glycans in CHO glycoproteins, N-glycans were released from parental and mutant CHO cellular glycoproteins by PNGase F and analyzed by MALDI-TOF-MS. Because most N-glycans derived from mammalian glycoproteins are composed of only a few monosaccharides and generate structures with unique masses that are of the oligomannosyl or bi-, tri-, or tetraantennary complex type, the nature of the species released by PNGase F may be deduced from their molecular mass in the context of known N-glycan structures (25)(26)(27).
The mass spectrometry of neutral N-glycans revealed markedly increased complexity for the ␤4GalT-1 mutants Gat Ϫ 2Lec20 and Pro Ϫ 5Lec20 compared with the ␤4GalT-6 mutant Pro Ϫ 5 CHO and wild-type Gat Ϫ 2 CHO cells (Fig. 6), consistent with the synthesis of a range of immature, undergalactosylated N-glycans in Lec20 cells. Analysis of these spectra is presented in Table II, in which the numbered peaks in Fig. 6 are identified based on the observed mass of [M ϩ Na] ϩ ions. It can be seen that each CHO cell line synthesizes a similar complement of oligomannosyl structures. Therefore, a lack of one or two ␤4GalTs did not significantly alter the proportion of these species, as expected. By contrast, both cell lines lacking a functional ␤4GalT-1 had an increased proportion of all the possible forms of undergalactosylated bi-(peaks 4, 6, 7, and 9), tri-(peaks 11, 12, 15, and 18), and tetraantennary (peaks 16, 19, 21, and 22) N-glycans. They also made significantly less fully galactosylated bi-(peaks 10 and 14) and triantennary (peak 20) N-glycans (Fig. 6), consistent with their reduced in vitro activities with exogenous acceptors (see Table IV). Nevertheless, fully galactosylated tetraantennary N-glycans (peak 23) were equivalently represented in wild-type and Lec20 cells (Fig. 6). Most striking was the fact that fully galactosylated N-glycans of each branched type, including biantennary, were synthesized in the absence of functional ␤4GalT-1. Although radiolabeled, desialylated VSV biantennary G-glycopeptides from Lec20 did not contain 2 Gal residues (see Fig. 1), highperformance anion-exchange chromatography with pulsed amperometric detection analysis of the reduced proportion of  5. A, shown is the expression of ␤4GalT-2, -3, -4, -5, and -6 in CHO and Lec20 cells. Two separate Northern blots (blot-1 and blot-2) containing 7 g of poly(A) ϩ RNA from each cell line were hybridized to an ϳ1-kb murine probe and subsequently to a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe of a 250-bp PCR product as described under "Experimental Procedures." B, RT-PCR was performed with specific primers for CHO ␤4GalT-6 and CHO ␤4GalT-1 as described under "Experimental Procedures." The ␤4GalT-6 cDNA generated from Gat Ϫ 2 CHO RNA was confirmed by digestion with NarI. The predicted 656-and 478-bp products were generated. [ 3 H]Gal-labeled, desialylated biantennary glycopeptides from Lec20 cellular glycoproteins revealed biantennary glycopeptides that eluted in the position of a digalactosylated species (data not shown), consistent with the mass spectrometry data obtained from total N-glycanase-released species.
The lack of ␤4GalT-6 in Pro Ϫ 5 CHO cells had a very small effect on the spectrum of galactosylated neutral N-glycans (Fig.  6). The only truncated N-glycans that appeared to be slightly increased in cells lacking ␤4GalT-6 were peaks 3, 9, and 18 ( Fig. 6 and Table II). There were almost no partially galactosylated tri-or tetraantennary N-glycans in the absence of ␤4GalT-6. Remarkably, although partially galactosylated biantennary N-glycans were found in wild-type Gat Ϫ 2 CHO and in the absence of ␤4GalT-6 (Pro Ϫ 5 CHO cells), there was no evidence of incomplete tri-or tetraantennary structures in glycoproteins from either of these cells. It appears that once a third or fourth GlcNAc is added to the trimannose core of an N-glycan, galactosylation goes to completion. By contrast, there were many partially galactosylated tri-and tetraantennary N-glycans in the absence of ␤4GalT-1. Most interestingly, fully galactosylated complex N-glycans (peaks 10, 14, 20, and 23 in Table II) are well represented in the absence of functional ␤4GalT-1 and ␤4GalT-6. Thus, it is clear that the combined activities of ␤4GalT-2, -3, -4, and -5 cause Gal to be added to all antennae. It can be concluded that ␤4GalT-6 plays an insignificant role in the galactosylation of neutral N-glycans and that ␤4GalT-1 plays an important role in the efficient completion of their galactosylation.
In the case of sialylated N-glycans, the picture is somewhat different (Fig. 7). First, there were significant differences between the MALDI-TOF-MS spectra from Gat Ϫ 2 CHO and Pro Ϫ 5 CHO N-glycans, indicating some clear consequences of the lack of ␤4GalT-6 in Pro Ϫ 5 CHO cells (Fig. 7). Pro Ϫ 5 CHO cells had a reduced proportion of sialylated bi-, tri-, and tet-raantennary species (see particularly peaks 7, 8, 13, 15, 16, and 18 in Fig. 7), suggesting that ␤4GalT-6 is involved in the galactosylation of all sialylated N-glycans, including those with  Fig. 6 (observed). Among different spectra, species with the same peak number had a mass within 0.5 units of that given in the table.
b N-Glycans were predicted on the basis of the mass and composition of known neutral N-glycan structures (G, galactose; Gn, N-acetylglucosamine; M, mannose; F, fucose). c (ϩ), peaks were present, but in low amounts.  (Table III). Second, it is clear that ␤4GalT-1 is involved in synthesizing all classes of sialylated N-glycans and to a much greater extent than ␤4GalT-6. Most of the sialylated complex N-glycans present in Gat Ϫ 2 CHO cells (peaks 11 and 13-19 in Table III) were missing in Gat Ϫ 2Lec20 and Pro Ϫ 5Lec20 cells. However, most of the small population of sialylated N-glycans that were present in Lec20 cells contained a full complement of Gal (peaks 1/2, 6, 9, and 12 in Table III). In particular, it is notable that equivalent peaks of SG 2 Gn 2 M 3 Gn 2 F (where S is sialic acid, G is galactose, Gn is N-acetylglucosamine, M is mannose, and F is fucose) (peaks 1/2) were present in the wild type and cells lacking ␤4GalT-1, ␤4GalT-6, or both ( Fig. 7 and Table III). In addition, few of the many partially galactosylated neutral species generated in the absence of ␤4GalT-1 (Table II) were found as sialylated N-glycans ( Fig. 7 and Table III). The only exceptions were peaks 3 and 4 in Fig. 7 (SG 2 Gn 3 M 3 Gn 2 and SG 1 Gn 4 M 3 Gn 2 in Table III). Interestingly, these species were equally represented in cells lacking functional ␤4GalT-1, ␤4GalT-6, or both. The combined data suggest that fully galactosylated N-glycans are sialylated more efficiently than Nglycans with a partial Gal complement.
CHO Mutants Lacking ␤4GalT-1, ␤4GalT-6, or Both Have a Normal Complement of Glycolipids-In in vitro assays, GlcCer is a good acceptor for ␤4GalT-1 (11), and ␤4GalT-6 is a lactosylceramide synthase (15). To identify in vivo acceptors for these ␤4GalTs, we isolated glycolipids from Gat Ϫ 2 CHO, Gat Ϫ 2Lec20, Pro Ϫ 5 CHO, and Pro Ϫ 5Lec20 mutants and performed high-performance TLC with the standards GlcCer, Lac-Cer, and GM3. Gat Ϫ 2Lec8 CHO cells were used as a control because they have an inactive UDP-Gal Golgi translocase (35) and do not add Gal to glycolipids (29). CHO cells synthesize GM3 and a small amount of GlcCer and LacCer, but no complex gangliosides (29), as shown in Fig. 8. As expected, the Gat Ϫ 2Lec8 mutant possessed a very small amount of GM3 and LacCer and an increased amount of GlcCer compared with Gat Ϫ 2 CHO parental cells. Interestingly, the glycolipid expression pattern of CHO cells that lack functional ␤4GalT-1 (Gat Ϫ 2Lec20) or ␤4GalT-6 (Pro Ϫ 5 CHO) or both (Pro Ϫ 5Lec20) was very similar to that of parental Gat Ϫ 2 CHO cells, which express all six ␤4GalTs. The major glycolipid was GM3 in all CHO cell lines, and there was no significant increase in GlcCer levels in cells lacking ␤4GalT-1, ␤4GalT-6, or ␤4GalT-1 and ␤4GalT-6. These results show that although ␤4GalT-1 and ␤4GalT-6 have activity for GlcCer in vitro, in CHO cells, neither is required for glycolipid synthesis. ␤4GalT-5 and/or ␤4GalT-4 seem likely to be responsible for glycolipid synthesis in CHO cells.
In Vitro ␤4GalT Acceptor Specificities of CHO Cells Lacking ␤4GalT-1, ␤4GalT-6, or Both-To correlate the activities of ␤4GalT-2, -3, -4, and -5 in cell extracts with the glycans they produce in vivo, ␤4GalT assays were performed under a range of conditions. The mixture of ␤4GalTs present in Lec20 mutants had little activity for the transfer of Gal to GlcNAc when Lec20 cell extracts were assayed under various conditions of substrate concentration and pH or in the presence of different nonionic detergents (Fig. 9). When microsomal membranes from Gat Ϫ 2 CHO cells, which express the six ␤4GalT genes, were assayed under the optimized conditions determined in Fig. 9, a specific activity of ϳ24 nmol/h/mg of protein was obtained (Table IV). In Pro Ϫ 5 CHO cells, which are missing ␤4GalT-6, the specific activity was slightly but significantly reduced, suggesting that recombinant ␤4GalT-6 can use GlcNAc as a substrate, although not efficiently. In Gat Ϫ 2Lec20 cells lacking functional ␤4GalT-1, the activity with GlcNAc was reduced to ϳ34%. In Pro Ϫ 5Lec20 cells, with defective ␤4GalT-1 and ␤4GalT-6, activity with GlcNAc was only 28% of that in Gat Ϫ 2 CHO cells.
In the mammary gland, ␤4GalT-1 interacts with ␣-lactalbumin, resulting in a change of acceptor specificity from GlcNAc  Fig. 7. Among different spectra, species with the same peak number had a mass within 0.5 units of that given in the table.
b N-Glycans were predicted on the basis of the mass and composition of known sialylated N-glycan structures (S, sialic acid; G, galactose; Gn, N-acetylglucosamine; M, mannose; F, fucose).
c This mass was assigned the previously predicted structure (26). The origin of the ϳ10-mass unit difference between observed and predicted m/z is not known. d (ϩ), peaks were present, but in low amounts. e These are the only partially galactosylated N-glycans present in significant amounts in mutants and absent from Gat Ϫ 2 CHO.
to Glc and the production of lactose. ␤4GalT-2 is also able to produce lactose efficiently (9). However, ␤4GalT-2 in CHO cells does not appear to associate with ␣-lactalbumin in vitro since Gat Ϫ 2Lec20 cells, which lack ␤4GalT-1 activity but have a normal complement of ␤4GalT-2 transcripts (Fig. 5A), had only 3% of the parental Gat Ϫ 2 CHO activity for transfer of Gal to Glc in the presence of ␣-lactalbumin (Table IV). When more complex N-glycans were assayed as acceptors, Gat Ϫ 2 CHO extracts always had a higher specific activity than Pro Ϫ 5 CHO extracts, suggesting that ␤4GalT-6 is acting on complex acceptors in vitro (Table IV). For Gat Ϫ 2Lec20 cells lacking functional ␤4GalT-1, the most severe reduction in activity (ϳ97%) was observed for the biantennary N-glycan acceptor GnGn. Therefore, under the assay conditions used, none of the five other ␤4GalTs efficiently galactosylated a biantennary complex glycopeptide in vitro. The tetraantennary Nglycan acceptor GnGnGnGn was more effectively galactosylated in the Gat Ϫ 2Lec20 ␤4GalT-1 Ϫ mutant (ϳ27% compared with Gat Ϫ 2 CHO cells). However, the triantennary acceptor was a significantly better acceptor for the mixture of ␤4GalT-2, -3, -4, -5, and -6 in Gat Ϫ 2Lec20 cell extract (ϳ54% compared with wild-type Gat Ϫ 2 CHO cells). Therefore, whereas ␤4GalT-1 appears to be the major activity transferring Gal to complex N-glycans in CHO cell microsomes, other ␤4GalTs efficiently transfer Gal to the triantennary complex acceptor GnGn␤4Gn. ␤4GalT-6 is also able to use GnGn␤4Gn efficiently as an acceptor since the absence of ␤4GalT-6 in Pro Ϫ 5 CHO resulted in a reduction of 43% activity (Table IV).
Lactosylceramide synthase activity was equivalent in wildtype Gat Ϫ 2 CHO cells and each mutant line (Table IV), as would be predicted from the glycolipid analysis in Fig. 8. Since ␤4GalT-6 is known to be a lactosylceramide synthase (15), it was surprising that Pro Ϫ 5 CHO cells, which lack ␤4GalT-6, had equivalent in vitro activity for GlcCer. Similarly, Gat Ϫ 2Lec20 and Pro Ϫ 5Lec20, which lack functional ␤4GalT-1 or ␤4GalT-1 and ␤4GalT-6, respectively, showed no decrease in transfer to GlcCer. Thus, neither ␤4GalT-1 nor ␤4GalT-6 appears to contribute to glycolipid synthesis in CHO cells.
Finally, both ␤4GalT-1 and ␤4GalT-6 contributed to the transfer of Gal to the mucin core 2 acceptor (Table IV). Compared with Gat Ϫ 2 CHO cells, Pro Ϫ 5 CHO cells were reduced ϳ25%, suggesting that ␤4GalT-6 adds Gal to core 2; Gat Ϫ 2Lec20 cells were reduced ϳ40%, suggesting that ␤4GalT-1 also transfers Gal to core 2. Clearly, other ␤4GalTs such as ␤4GalT-4, which has been identified as having a high degree of specificity for a core 2 acceptor (12), contribute in CHO extracts to the transfer of Gal to the core 2 oligosaccharide. DISCUSSION The knowledge that mammals have six ␤4GalTs with overlapping in vitro acceptor specificities (8,36,37) presents the challenge of sorting out their unique biological functions. An important question is the degree of redundancy between different members of the ␤4GalT family in transferring Gal to complex glycoprotein and glycolipid acceptors. To begin to address this question, we have identified N-glycan and glycolipid structures synthesized in CHO cells that express the six ␤4GalTs compared with mutant cells that lack ␤4GalT-1, ␤4GalT-6, or both. We have shown that mutants of the Lec20 complementation group (21) lack ␤4GalT-1 activity. In Gat Ϫ 2Lec20 CHO cells, it is due to a deletion mutation that removes exons III and IV of the ␤4GalT-1 gene so that only the N-terminal 214 amino acids of ␤4GalT-1 can be synthesized. Pro Ϫ 5Lec20 CHO cells have a mutation that results in extremely low steady-state levels of ␤4GalT-1 transcripts. The overall phenotype of both Lec20 mutants is essentially identical (21); and thus, it was of interest to discover that Pro Ϫ 5Lec20 cells have no detectable ␤4GalT-6 transcripts. The ␤4GalT-6 deficiency in Pro Ϫ 5Lec20 cells originated from the parental Pro Ϫ 5 CHO cells, which are also completely devoid of ␤4GalT-6 gene transcripts. Therefore, we used the four cell lines to investigate the relative contributions of ␤4GalT-1, ␤4GalT-6, and the four remaining ␤4GalTs to in vivo Gal transfer to glycoproteins and glycolipids and to in vitro galactosyltransferase acceptor specificity for various acceptors.
A summary of mutants and their properties is given in Table  V. The loss of ␤4GalT-1 had the most profound effect on in vitro lactose synthase activity and on the transfer of Gal to the biantennary GnGn glycopeptide (Table IV). One or more of the remaining five ␤4GalTs transferred Gal quite efficiently to GlcNAc, tri-and tetraantennary glycopeptides, and the core 2 oligosaccharide, although ␤4GalT-1 provided Ն50% of the activity with these acceptors. Interestingly, ␤4GalT-1 did not contribute to the transfer of Gal to GlcCer in CHO cell extracts, even though recombinant ␤4GalT-1 uses GlcCer as an acceptor (9). Thin-layer chromatography of glycolipids from both the Lec20 mutant lines confirmed that ␤4GalT-1 does not contribute to the synthesis of LacCer or GM3 in CHO cells (Fig. 8).
The in vitro results with microsomal membranes suggest that ␤4GalT-1 is the most important ␤4GalT in galactosylating biantennary N-glycans (Table IV). This conclusion is supported by MALDI-TOF-MS analysis of neutral and sialylated N-glycans (Tables II and III). Gat Ϫ 2 CHO cells, which express all six ␤4GalTs, synthesize fully galactosylated tetra-or triantennary neutral N-glycans, but have appreciable amounts of undergalactosylated biantennary N-glycans. In Lec20 mutants that lack functional ␤4GalT-1, almost every possible partially galactosylated N-glycan is synthesized, but the predominant species is undergalactosylated biantennary N-glycans. Thus, ␤4GalT-1 is required for efficiently generating fully galactosylated bi-, tri-, and tetraantennary neutral N-glycans. Most interestingly, only a very small proportion of the partially galactosylated structures that predominate in Lec20 mutants acquire sialic acid (Table III). In fact, no biantennary N-glycans containing 1 Gal residue capped with sialic acid were detected. This strongly suggests that the first sialic acid is not transferred by the ␣2,3 sialyltransferase in CHO cells until both Gal residues are present on a biantennary structure. In addition, it is apparent FIG. 8. Thin-layer chromatography of glycolipids. Purified glycolipids extracted from the four cell lines and Gat Ϫ 2Lec8 cells were spotted on a Silica Gel 60 high-performance TLC plate with standards (Std) of GlcCer (20 g), LacCer (27 g), and GM3 (4 g). The plate was developed by ascending chromatography in chloroform, methanol, and 0.02% CaCl 2 (60:40:9) and stained with resorcinol/H 2 SO 4 reagent. Glycolipids are marked with arrowheads. The band marked with an asterisk in the Gat Ϫ 2Lec8 lane is not GM3. that only one partially galactosylated triantennary structure (SG 2 Gn 3 M 3 Gn 2 ) and one tetraantennary structure (SG 1 Gn 4 M 3 Gn 2 ) were present in Lec20 mutants. None of the other neutral N-glycans with 1, 2, or 3 Gal residues (see Table  II) were sialylated (see Table III). Also of interest are the several N-glycans that appear to have polylactosamine se-quences among the sialylated species.
The results of in vitro galactosyltransferase assays and glycan analyses reveal subtle but significant effects of the absence of ␤4GalT-6 in CHO cells. The Pro Ϫ 5 CHO cell extract, which lacks ␤4GalT-6, had significantly, although slightly, reduced activity with all acceptors except GlcCer (Table IV). By far the FIG. 9. ␤4GalT activities of Pro ؊ 5 CHO and Pro ؊ 5Lec20 under different assay conditions. ␤4GalT activities of Pro Ϫ 5 CHO and Pro Ϫ 5Lec20 cell extracts were assayed using GlcNAc as acceptor as described under "Experimental Procedures." The reaction contained 5 mol of MES, pH 6.5 (C) or pH 5.7 (B and D), 3 mol of MnCl 2 , 1.2% Triton X-100, 25 nmol of UDP-[6-3 H]Gal (ϳ10,000 cpm/ nmol) (A-C), 0.5 mol of GlcNAc (A, C, and D), and ϳ100 g of protein. In A, MES was used for pH 5ϳ6.5, and MOPS was used for pH 7-8. In C, all detergents were used at a final concentration of 1%. TX-100, Triton X-100; NP-40, Nonidet P-40; DOC, sodium deoxycholate.  biggest effect of the loss of ␤4GalT-6 in Pro Ϫ 5 CHO was in the ϳ44% reduced transfer of Gal to the triantennary acceptor from fetuin (GnGn␤4Gn). This was not reflected in an abundance of undergalactosylated triantennary N-glycans in Pro Ϫ 5 CHO glycoproteins, however. In fact, there were only minor peaks of bi-and triantennary neutral N-glycans lacking Gal residues in Pro Ϫ 5 CHO glycoproteins. By contrast, for the sialylated N-glycans, the absence of ␤4GalT-6 gave rise to the same species of undergalactosylated tri-and tetraantennary structures as did the absence of ␤4GalT-1. Thus, Pro Ϫ 5 CHO cells lacking only ␤4GalT-6 contained only a subset of the fully galactosylated, sialylated N-glycans synthesized by the full complement of six ␤4GalTs in Gat Ϫ 2 CHO cells. Finally, it can be seen from the spectrum of complex N-glycans synthesized in the double mutant Pro Ϫ 5Lec20 that the effects of missing ␤4GalT-1 and ␤4GalT-6 are essentially additive.
Perhaps the most unexpected result with cells lacking ␤4GalT-6 was the fact that this was not reflected in an increased amount of GlcCer due to reduced synthesis of LacCer and GM3 (Fig. 8). ␤4GalT-6 was called LacCer synthase when first cloned (15) and has been proposed to be a major ␤4GalT responsible for LacCer synthesis. Although it is true that all the ␤4GalTs can synthesize LacCer in vitro, ␤4GalT-6 and ␤4GalT-5 are more closely related to each other than to the other ␤4GalTs at the amino acid level. Thus, it may be that ␤4GalT-5 is the ␤4GalT that synthesizes LacCer in CHO cells because it is clear that neither ␤4GalT-1 nor ␤4GalT-6 is responsible. In summary, the Gal transfer properties of Gat Ϫ 2 CHO cells, which express all six ␤4GalTs, compared with those of glycosylation mutants lacking functional ␤4GalT-1 (Gat Ϫ 2Lec20), ␤4GalT-6 (Pro Ϫ 5 CHO), or both (Pro Ϫ 5Lec20) show that ␤4GalT-1 is a key enzyme for the galactosylation of complex N-glycans and that neither ␤4GalT-1 nor ␤4GalT-6 is involved in glycolipid synthesis in CHO cells (Table V).