Biosynthesis of Chondroitin and Heparan Sulfate in Chinese Hamster Ovary Cells Depends on Xylosyltransferase II*

Xylosyltransferase (XylT) catalyzes the initial enzymatic reaction in the glycosaminoglycan assembly pathway for proteoglycan biosynthesis. Its activity is thought to be rate-limiting. Two xylosyltransferases have been found using genomic analyses, and one of these, XylT1, has been shown to have xylosyltransferase activity. On the other hand, the less studied XylT2 in recombinant form lacks xylosyltransferase activity and has no known function. Wild-type Chinese hamster ovary cells express abundant Xylt2 mRNA levels and lack detectable Xylt1 mRNA levels. Analysis of a previously described Chinese hamster ovary cell xylosyltransferase mutant (psgA-745) shows that it harbors an Xylt2 nonsense mutation and fails to assemble glycosaminoglycans onto recombinant biglycan. Transfection of this cell line with a murine Xylt2 minigene results in the production of recombinant chondroitin sulfate-modified biglycan core protein and restoration of fibroblast growth factor binding to cell surface-associated heparan sulfate. Expression analyses on 10 different human transformed cell lines detect exclusive XYLT2 expression in two and co-expression of XYLT1 and XYLT2 in the others but at disparate ratios where XYLT2 expression is greater than XYLT1 in most cell lines. These results indicate that XylT2 has a significant role in proteoglycan biosynthesis and that cell type may control which family member is utilized.

3). PGs are found on the cell surface and in the extracellular matrix. PGs are fundamental to cellular processes including enhancement of receptor binding of cytokines and growth factors and augmentation of enzyme-substrate interactions important in coagulation and lipid catabolism (4 -9). Extracellular matrix PGs maintain basement membrane integrity, augment growth factor and cytokine sequestration, and create morphogen gradients (4, 10 -12). Genomic analyses have identified two xylosyltransferase family members in mammals. XYLT1 encodes for xylosyltransferase I (XylT1), shown to have xylosyltransferase activity (13). XYLT2 encodes for xylosyltransferase II (XylT2), which when expressed in vitro failed to have xylosyltransferase activity (13,14), and therefore, its function is unknown. A highly useful tool for the study of PGs has been the xylosyltransferase-deficient Chinese hamster ovary cell (CHO) line pgsA-745 (15,16). This cell line was isolated from an ethylmethane sulfonate mutagenesis screen of CHO-K1 cells for sulfation incorporation mutants. This cell line produces neither chondroitin nor heparan-sulfated PGs (15,16). In our long term investigation of this enzymatic pathway and the function of XylT2, we found that wild-type CHO cells express abundant levels of Xylt2 mRNA and near undetectable levels of Xylt1 mRNA. Furthermore, we discovered that the pgsA-745 cell line harbors an inactivating genetic defect in Xylt2 and that overexpression of Xylt2 in these cells complements their inability to add chondroitin sulfate to a glycosaminoglycan-free biglycan core protein and to bind basic fibroblast growth factor to surface associated heparan sulfate proteoglycans.

Northern Blot, RT-PCR, and Real-time RT-PCR-Poly(A)
RNA was isolated as per the manufacturer's instructions (Invitrogen MicroFast Track). Northern blot filters were generated as described previously (17) using 2 g of poly(A) RNA. For human cell line analysis, a human cancer cell line poly(A) RNA Northern filter was used (Stratagene). Complementary mouse DNA probe templates were 1.5 kb for Xylt2 (according to nucleotides 753-2256 of GenBank TM accession number NM_145828) and 0.5 kb for Xylt1 (nucleotides 2584 -3102 according to GenBank accession number NM_175645), both encoding the final coding sequence of the cDNA sequence including the stop codon at the 3Ј end. The mouse Xylt1 probe template showed 91% homology with the hamster cDNA sequence. Human cDNA probe template for XYLT1 (nucleotides 2455-3009 according to GenBank accession number NM_022166) was a fragment that spanned the last 555 bp of the * This work was supported by American Heart Association Grant 0265270Z and National Institutes of Health Grant P20 RR018758. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains a supplemental figure and a supplemental Each sample was analyzed in triplicate with gene-specific primers (see supplemental Table 1) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primers on the same 96-well plate. The GAPDH product values were used for normalization. In addition, to ensure reproducibility of the reactions, a standard curve was performed with both gene-specific primers and GAPDH primers using control human kidney total RNA reverse transcription product.
Cloning the CHO Cell Xylt2 Locus-Long range PCR (Takara Bio Inc.) according to the manufacturer's recommended conditions was performed on genomic DNA from wild-type CHO cell line CHO-K1 (ATCC number CCL-61) and xylosyltransferase-deficient CHO cell psgA-745 (ATCC number CRL-2242) grown under recommended conditions. Primer sequences were designed based on human-mouse homologous sequence (see supplemental Table 1). Five overlapping genomic fragments spanning 13 kb in length were generated that included the sequence down to the provisional poly(A) addition site as identified by comparison with mouse cDNA sequence (according to GenBank accession number NM_145828). Some products were cloned into TOPO TA vector (Invitrogen) and sequenced, and others sequenced directly as PCR product. Both wild-type and mutant locus sequences have been deposited in the GenBank data base under GenBank accession numbers (EF051485 and EF051486, respectively).
Production of Cell Lines-The coding portion of the Bgn cDNA, encoding mouse biglycan, was cloned into the expression vector pcDNA3.1 (Invitrogen) that contains a neomycinselectable marker. A protein tag HPC4 (18) was engineered into the C terminus of the recombinant biglycan. This created the BigHPC4 expression construct. An Xylt2 minigene containing the coding portion of exon 1 spliced to the remaining portion of the locus spanning exon 2 to the translational stop codon in exon 11 was cloned into the expression vector pBudCE4.1 (Invitrogen) with the Zeocin cassette replaced with puromycin-and neomycin-selectable cassettes to create the expression vector XYIISCI-44. This put the Xylt2 minigene under the control of the EF-1␣ promoter. The parental cell line psgA-745 was purchased from ATCC. The psgA-745 CHO cells were grown as recommended and transfected with BigHPC4 using FuGENE (Roche Applied Science) to create reporter cell lines. Cells were selected using medium with 400 g/ml G418 (Invitrogen). Colonies were picked and transferred to 96-well plates, and conditioned media were analyzed by Western blot using a mouse monoclonal antibody to HPC4 (a gift from Dr. Charles Esmon, Oklahoma Medical Research Foundation) (18). Several clones had high expression levels of glycosaminoglycan-free mouse biglycan. One clone named A10 was transfected with XYI-ISCI-44 using FuGENE (Roche Applied Science) followed by 400 g/ml G418 and 0.02 mg/ml puromycin selection. Colonies were picked to 96-well plates, and conditioned media were analyzed by Western blot using a HPC4 monoclonal antibody as described above.
Western Blot Analyses-Conditioned media from individual clones were denatured, reduced, and separated on 4 -15% gradient acrylamide gels (Bio-Rad). Separated proteins were transferred to polyvinylidene difluoride membrane (Millipore) using semidry transfer (Bio-Rad). Blots were blocked in 5% nonfat milk in phosphate-buffered saline (PBS) containing 1 mM CaCl. The primary antibody, a mouse monoclonal to the HPC4 epitope tag (18), was used at 1:1000 for 1 h at room temperature. The blots were then washed in PBS containing 1 mM CaCl and incubated with a horseradish peroxidase-conjugated goat antimouse secondary antibody for 1 h. Blots were washed again and incubated with ECL reagent (Amersham Biosciences) followed by exposure to film. For chondroitinase ABC digestion, 10 l of conditioned medium was incubated 12 h at 37°C with 0.02 units of chondroitinase ABC (Seikagaku) in 0.1 M Tris-HCl, pH 8.0, 0.03 M sodium acetate buffer in a total volume of 20 l.
Sequence Analysis-On at least three independent products generated by long range PCR as described above, sequence was obtained on an Applied Biosystems 3730 capillary sequencer using the dideoxy chain termination method using Taq polymerase-and locus-specific primers. For each independent product, a consensus was generated from at least two overlapping sequence sets. A final master consensus was generated from the compilation of the independent product consensuses. Ambiguous areas were regenerated with low error pfu Turbo Taq (Strategene) DNA polymerase and reanalyzed. All sequence analyses were performed with standard software.
Flow Cytometry Analyses with FGF-2-This assay was performed as described previously (19). Biotinylated fibroblast growth factor 2 (FGF-2) was graciously provided by Dr. Jeff Esko of the University of California at San Diego. Flow cytometry was performed on a BD FACSCalibur (BD Biosciences). In some experiments, the cells were incubated with 50 milliunits/ml heparitinase (EC 4.2.2.8) in PBS for 20 min at 37°C prior to staining with FGF-2.
Immunocytochemistry for Heparan Sulfate in CHO Cells-Wild-type CHO-K1, clone 3, and psgA-745-expressing biglycan (parental clone) cells were applied to slides by cytospin. Slides were fixed 10 min with 4% paraformaldehyde and rinsed in heparitinase digestion buffer followed by digestion with hep-aritinase I (Seikagaku, 100703) in the same buffer. After several rinses in PBS and saponin, a 5-min incubation in hydrogen peroxide in methanol, and rinsing again in PBS and saponin, the slides were blocked for 1 h at room temperature in 3% bovine serum albumin, 0.01% saponin, and PBS followed by incubation with 10E4 (Seikagaku, 370255) at 1:50. After three washes in PBS with saponin, the slides were incubated with peroxidaseconjugated goat antimouse IgG/IgM (Jackson ImmunoResearch Laboratories) antibody for 1 h at room temperature followed by NovaRED (Vector Laboratories) incubation, counterstaining, dehydration, and coverslip mounting.

Wild-type CHO Cells Express Abundant Xylt2, and Mutant psgA-745 Cells Lack Detectable Levels of Xylt2 and Xylt1 mRNA-
To identify the candidate protein responsible for the xylosyltransferase deficiency in the psgA-745 cells, the expression levels of Xylt1 and Xylt2 in both wild-type cell line CHO-K1 and psgA-745 mutant cell lines were determined. In addition, since little is known about the regulation of the PG biosynthetic pathway, it was of interest to determine the impact of xylosyltransferase deficiency on the expression of downstream enzymes in this pathway. Consequently, the mRNA level for ␤4Galt7 was measured. For controls, Extl2 and Gapdh expression was measured. The Extl2 protein product can catalyze the fifth biosynthetic step of heparan sulfate glycosaminoglycan assembly onto core proteins.
Although XylT1 has been shown to be a xylosyltransferase capable of initiating glycosaminoglycan assembly onto PG core proteins, we did not detect Xylt1 mRNA levels in the wild-type CHO cells by poly(A) Northern blot analyses using a mouse probe that we have determined to be 91% homologous with the hamster cDNA (Fig. 1A). To confirm this result, we performed RT-PCR with hamster-specific primers and generated a faint product similar in size to the Xylt1 RT-PCR product from hamster kidney (Fig. 1C), but this product was nonspecific as determined by Southern blot analysis and sequence analysis. An additional primer set that spanned from exon 10 to 11 also failed to detect any Xylt1 mRNA in the wild-type CHO cells. On the other hand, the wild-type cells expressed abundant levels of Xylt2 mRNA (Fig. 1B). In contrast, the mutant psgA-745 did not express either (Fig. 1, A-C). In both cell lines, steady state levels of ␤4Galt7 and Extl2 mRNA were similar (Fig. 1B). These results suggest that XylT2 is a potential candidate for PG biosynthesis in wild-type CHO cells and that the psgA-745 CHO cell mutants are xylosyltransferase-deficient because of a defect in Xylt2 expression.
Identification of Mutation Responsible for Undetectable Xylt2 mRNA Levels in psgA-745 Cells-To determine the reason for the undetectable levels of Xylt2 in the psgA-745 cells, the hamster Xylt2 genomic locus from both the wild-type and the mutant cell lines was cloned and sequenced. Overlapping fragments spanning the start of translation down to the provisional poly(A) addition site in the last exon were generated and sequenced. The total locus spanned 13 kb, containing 11 exons as in human (accession number NT_010783) and mouse (accession number NT_039521). This analysis revealed the locus to have two point mutations consisting of a nonsense mutation in exon 1 (Fig. 1D) and a G to A transition mutation in intron 7 (position 9031 with respect to the ATG of exon 1). The nonsense mutation is predicted to encode a truncated protein of 38 amino acids. However, this premature stop codon likely induces nonsense-mediated mRNA degradation (18), explaining the undetectable levels of Xylt2 mRNA in the psgA-745 mutants. Therefore, these results further suggest that XylT2 is responsible for xylosyltransferase activity in the parental wildtype CHO cells and that xylosyltransferase deficiency in the psgA-745 CHO cells results from a translational stop codon in Xylt2.
XylT2 Can Restore Chondroitin Sulfate Modification of Biglycan-To determine whether the Xylt2 mutation is responsible for xylosyltransferase deficiency in the psgA-745 CHO cell mutant line, we attempted to complement the deficient cells by overexpression of an Xylt2 minigene. Several xylosyltransferase-deficient reporter CHO cell lines were made by expressing recombinant tagged biglycan in psgA-745 CHO cells. These cell lines produce only unmodified biglycan core protein as expected ( Fig. 2A). A representative clone called A10 was transfected with a mouse Xylt2 minigene. Subsequent clones were isolated and analyzed, showing that they now produced glycosaminoglycan-containing biglycan ( Fig. 2A) that was sensitive to chondroitinase digestion (Fig. 2B), confirming that XylT2 can restore the chondroitin sulfate modification of a core protein in the xylosyltransferase-deficient pgsA-745 cell line.
Xylt2 Can Restore Functional Heparan Sulfate Modification-To determine whether heparan sulfate biosynthesis has been restored in the mutant cells stably transfected with Xylt2, we used a standard functional binding assay. This assay has been used by others to assess rescue of the psgA-745 cells transfected with the Caenorhabditis elegans xylosyltransferase sqv-6 (20). Biotinylated FGF-2 binds to cell surface heparan sulfate proteoglycan, and this binding can be measured using flow cytometry. As compared with the parental A10 cell line from Fig. 2, A and  B, 34.5% of clone 3 cells transfected with Xylt2 bound FGF-2 (Fig. 2C). Clone 3 cells that bound FGF-2 had a geometric mean of 88.63 as compared with the geometric mean of 6.5 for the parental and 242.4 for the wild-type CHO cells, respectively. These results are similar to the biglycan Western blot analysis where not all the biglycan was converted to the glycosaminoglycan-containing form, suggesting a heterogeneous population of cells despite being maintained on double selection growth conditions. Nevertheless, Xylt2 can restore FGF-2 cell surface binding mediated by heparan sulfate. To confirm that heparan sulfate mediates this binding, we digested the cells with heparitinase prior to staining with FGF-2 and found that with digestion, there is loss of FGF-2 binding (Fig. 2D). Furthermore, we also screened the rescued cells with heparan sulfate-specific antibodies and found heparan sulfate on the rescued cells (Supplemental Fig. 1). Therefore, transfected Xylt2 can restore heparan sulfate biosynthesis in xylosyltransferase-deficient cells.
XylT2 Is Expressed in Multiple Cell Types-In humans (13), there is a wide tissue distribution of Xylt2 mRNA as opposed to Xylt1 mRNA. This suggests that preferential expression of Xylt2 is not unique to CHO cells. To determine this, we examined Xylt2 mRNA expression in 10 transformed human cell lines using Northern blot analyses and real-time RT-PCR. The Northern blot analyses showed that most cell lines express XYLT2, and fewer express only XYLT1 (Fig. 3). In addition, these analyses illustrate that when either XYLT2 or XYLT1 is expressed, there are single transcripts for each gene, indicating that there are no post-transcriptional modifications such as alternative splicing or alternative use of multiple polyadenylation sites. The real-time RT-PCR analyses likewise showed that all of the cell lines had detectable levels of XYLT2 mRNA (Fig. 4A), and XYLT1 expression when detected was considerably lower than XYLT2 expression in most cell types. Two exceptions are HL-60 and MG-63 cell lines that had a predominance of XYLT1 expression (Fig. 4B). Fewer cell types appear to express XYLT1, but on closer inspection, some cell types express very low mRNA levels (Fig. 4B, inset). Only HeLa and  K562 cells had exclusive XYLT2 expression (Fig. 4, A and B,  inset). These results imply that although some cells may differentially utilize XylT1 and XylT2 for xylosyltransferase activity, many cell types, similar to CHO cells, may rely primarily on XylT2 for PG biosynthesis.

DISCUSSION
The main finding of this study is that XylT2 has an important role in chondroitin and heparan sulfate proteoglycan biosynthesis as shown by expression and complementation analyses. We have demonstrated a correlation between expression of Xylt2 in wild-type CHO cells and chondroitin sulfate assembly on biglycan and heparan sulfate on cell surface proteoglycans and between loss of Xylt2 mRNA expression and absence of chondroitin sulfate modification and cell surface binding of FGF-2 in the xylosyltransferase-deficient CHO cells. Furthermore, we have shown that XylT2 can complement the loss of chondroitin sulfate and heparan sulfate modification of core proteins in these cells. We have also demonstrated that many cell lines have abundant Xylt2 mRNA and near undetectable Xylt1 mRNA levels similar to CHO cells. Furthermore, two cell lines that exclusively express XYLT2 (21)(22)(23)(24)(25)(26) can produce heparan sulfate proteoglycans, clearly supporting a role for XylT2 in PG biosynthesis. Thus the role of Xylt2 in PG biosynthesis may not be exclusive to CHO cells.
Since the XylT2 sequence is similar to the sequence of XylT1, which has been shown to have xylosyltransferase activity in heterologous expression systems (13,14,27), the simplest interpretation is that XylT2 also functions as a xylosyltransferase. However, this activity has yet to be demonstrated directly (13,14). Since recombinant XylT2 fails to have activity under conditions where XylT1 does (13,14), then another protein may be important for XylT2 activity and is present in cell lysate. This protein is unlikely to be XylT1 since its message was not detected in the CHO cells and in two other cell lines that produce proteoglycans. Other glycosyltransferases are known to function in multiprotein complexes that can include other glycosyltransferases in the pathway (20). This may also occur for XylT2, which in addition to its areas of homology with XylT1 may contain other protein domains of unknown function that might be involved in critical protein-protein interactions important for catalysis or compartmentalization. Indeed, such a complex with xylosylprotein 4-␤-galactosyltransferase I has been proposed (28).
We found XYLT2 expressed in all cell types examined. In 8 out of 10 cell lines, XYLT2 expression was higher than XYLT1. These data suggest that many cell types, similar to CHO cells, may rely primarily on XylT2 for PG biosynthesis because of tightly regulated XYLT1 expression. Alternatively, XylT1 may be a more efficient enzyme, thus requiring lower expression levels than XylT2 for PG biosynthesis. In our studies, although most cell lines expressed high levels of XYLT2, some cell lines expressed both enzymes. This co-expression suggests that these enzymes must act on different substrates or differ in kinetics toward the same substrates or are differentially compartmentalized in the cell. The latter appears to not be the case for HEK-293 cells since both enzymes appear localized to the cis-Golgi or medial Golgi compartments (29). Differences in substrate specificity or enzyme kinetics may help explain the observed expression patterns.
These enzymes are similar in many ways. They have an overall amino acid sequence identity of 55%, with the proposed catalytic domain reaching over 80% sequence identity (13). Schon Real-time RT-PCR of total RNA from the indicated cell types is shown. A, XYLT2 expression in all cell types is found. B, XYLT1 expression comparable with that of XYLT2 is restricted to a limited number of cell types. The inset illustrates that some of those cell types appearing to lack XYLT1 expression actually have low levels of mRNA. However, HeLa and K562 cells completely lack XYLT1 mRNA by this analysis. FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 JOURNAL OF BIOLOGICAL CHEMISTRY 5199 et al. (29) found that XylT1 is secreted, and Gotting et al. (13) showed that CHO-K1 cells secrete XylT activity, which we now know to be XylT2. In addition, both enzymes appear localized to the same cellular compartment (29). However, the inability of XylT2 to have activity in vitro and the contrasting expression levels between XylT1 and XylT2 suggests that the biology of these enzymes is complex.