Enzymes Responsible for Synthesis of Corneal Keratan Sulfate Glycosaminoglycans*

Keratan sulfate glycosaminoglycans are among the most abundant carbohydrate components of the cornea and are suggested to play an important role in maintaining corneal extracellular matrix structure. Keratan sulfate carbohydrate chains consist of repeating N-acetyllactosamine disaccharides with sulfation on the 6-O positions of N-acetylglucosamine and galactose. Despite its importance for corneal function, the biosynthetic pathway of the carbohydrate chain and particularly the elongation steps are poorly understood. Here we analyzed enzymatic activity of two glycosyltransferases, β1,3-N-acetylglucosaminyltansferase-7 (β3GnT7) and β1,4-galactosyltransferase-4 (β4GalT4), in the production of keratan sulfate carbohydrate in vitro. These glycosyltransferases produced only short, elongated carbohydrates when they were reacted with substrate in the absence of a carbohydrate sulfotransferase; however, they produced extended GlcNAc-sulfated poly-N-acetyllactosamine structures with more than four repeats of the GlcNAc-sulfated N-acetyllactosamine unit in the presence of corneal N-acetylglucosamine 6-O sulfotransferase (CGn6ST). Moreover, we detected production of highly sulfated keratan sulfate by a two-step reaction in vitro with a mixture of β3GnT7/β4GalT4/CGn6ST followed by keratan sulfate galactose 6-O sulfotransferase treatment. We also observed that production of highly sulfated keratan sulfate in cultured human corneal epithelial cells was dramatically reduced when expression of β3GnT7 or β4GalT4 was suppressed by small interfering RNAs, indicating that these glycosyltransferases are responsible for elongation of the keratan sulfate carbohydrate backbone.

In higher eukaryotes, high numbers of cells interact with each other and form complicated but well organized tissue structures. During the tissue formation in the developmental stage, cells recognize the surrounding environment and interact with neighboring cells and the extracellular matrix. Proteo-glycans (PGs) 3 are glycoproteins that carry linearly elongated polysaccharides called glycosaminoglycans (GAGs) on core proteins and that are largely found in the extracellular matrix and play important roles for development, maintenance, and function of the tissues.
Production of the GAG chain takes place mostly in the Golgi apparatus. Several glycosyltransferases and carbohydrate-modifying enzymes such as sulfotransferases act on GAG elongation and modification. So far almost all of the enzymes involved in GAG production are cloned and analyzed for their enzymatic activities; however, the process of GAG production is still unknown in some GAGs because of the presence of multiple enzymes with redundant activities.
Keratan sulfate (KS) PGs are major components of the cornea and also found in cartilage and brain. Because of the high concentration of these molecules in the cornea, their biological function has been extensively studied and found to include maintenance of corneal extracellular matrix structure (1)(2)(3)(4)(5)(6)(7)(8)(9).
Corneal KS PGs consist of PG core proteins, such as lumican, keratocan, and mimecan, carrying KS GAGs in an N-linked manner (10 -12). KS GAG is a linear carbohydrate chain made of sulfated disaccharide repeats of -3Gal␤1-4GlcNAc␤1-with sulfate on the 6-O position of GlcNAc and galactose (10 -12) and exhibiting modifications such as fucose and sialic acid (13,14). Production of the KS GAG chain on PGs is processed by glycosyltransferases and sulfotransferases localized in the Golgi apparatus, and matured KS PGs are secreted into the extracellular matrix. Elongation of the carbohydrate backbone of the KS GAG chain is catalyzed by enzymes of two glycosyltransferase families, ␤1,3-N-acetylglucosaminyltransferase (␤3GnT) and ␤1,4-galactosyltransferase (␤4GalT), and sulfation of the chain is catalyzed by two carbohydrate sulfotransferases.
Recent studies of carbohydrate sulfotransferases have identified enzymes responsible for sulfation of the corneal KS GAG chain as KS galactose 6-O sulfotransferase (KSG6ST) and corneal N-acetylglucosamine 6-O sulfotransferase (CGn6ST, also known as GlcNAc6ST-5 and GST4␤) (15,16). Because CGn6ST only transfers sulfate on the nonreducing terminal GlcNAc but not onto internal GlcNAc, sulfation of GlcNAc * This work is supported by National Institutes of Health Grant EY014620 (to T. O. A.). 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 supplemental Tables S1-S5 and Fig. S1. 1  residues of KS GAG is coupled with KS GAG elongation (16). On the other hand, KSG6ST transfers sulfate on galactose located both internally and on the nonreducing terminal of the carbohydrate chain (15,17). KSG6ST also prefers a sulfated carbohydrate as substrate (15,17), suggesting that galactose sulfation occurs after production of the GlcNAc-sulfated poly-N-acetyllactosamine chain and that GlcNAc sulfation is necessary for sulfation of galactose residues by KSG6ST. Patients with macular corneal dystrophy type I, which is caused by deficiency of functional CGn6ST, have no detectable highly sulfated KS in the cornea (18), serum, and cartilage (19,20), suggesting that GlcNAc sulfation is required to produce highly sulfated KS GAG. Observations that macular corneal dystrophy patients with CGn6ST mutations develop corneal opacities (21) and that mice lacking the orthologous sulfotransferase display corneal thinning with abnormalities in corneal extracellular matrix structure (9) indicate important roles for KS GAG sulfation in the function and maintenance of the cornea. However, unlike KS sulfation, mechanisms underlying KS GAG chain elongation are poorly understood.
In humans, ␤3GnT and ␤4GalT enzymes are encoded by eight and seven genes, respectively. Among these, ␤3GnT7 and ␤4GalT4 are thought to be responsible for elongation of the KS carbohydrate chain because these enzymes have higher activity for sulfated than nonsulfated substrates (22,23); however, direct evidence in support of this hypothesis has not been reported.
Here, using soluble forms of recombinant enzymes and glycosidase-assisted column chromatography, we demonstrate that ␤3GnT7 and ␤4GalT4 can produce KS GAG carbohydrate in vitro. We also observed that suppressing expression of either ␤3GnT7 or ␤4GalT4 reduced highly sulfated KS GAG production in cultured human corneal cells, indicating that these glycosyltransferases are responsible for KS GAG elongation.
Preparation of Soluble Enzymes-Expression vectors or an empty pcDNA3.1-HSH were transfected individually into Lec20 cells (26), kindly provided by Dr. Pamela Stanley, using Lipofectamine Plus reagent (Invitrogen) according to the man-ufacturer's instructions. The cells were then cultured for 24 h with ␣-minimum essential medium (Irvine Scientific, Santa Ana, CA) containing 10% fetal bovine serum, and the medium was replaced with Opti-MEM medium (Invitrogen) supplemented with 40 g/ml of L-proline and cultured cells for 24 h at 37°C. We then concentrated the culture medium up to 200fold using a Microcon YM-30 (Millipore Corp., Bedford, MA) and mixed that medium (1:1) with glycerol for storage at Ϫ20°C. Protein production was confirmed by Western analysis using alkaline phosphatase-conjugated anti-T7 tag antibody (Novagen, Madison, WI) and the LumiPhos WB chemiluminescence solution (Pierce).
Column Chromatography-To determine the size of reaction products, we performed gel filtration chromatography with a column (1.0-cm diameter ϫ 120-cm length) of Bio-Gel P-4 (Bio-Rad) equilibrated with 100 mM ammonium acetate buffer, pH 6.8. The reaction mixture samples were applied to the column, and the eluate was collected in 1-ml fractions. The elution pattern of carbohydrate products was monitored by counting 35 S radioactivity using a liquid scintillation counter. For further analyses, we pooled radioactive fractions, desalted them by Sephadex G25 gel filtration (Sigma) equilibrated with 7% (v/v) 1-propanol/water medium, and lyophilized the samples again.
To identify the degree of sulfation of products, we separated them by anion exchange HPLC using a Whatman Partisil SAX-10 column (4.6-mm diameter ϫ 25-cm length). The column was first equilibrated with 60% acetonitrile/water, and then after loading a sample, the products were eluted under the following conditions: 60% acetonitrile/water for 5 min; a linear gradient from 60% acetonitrile/water to either 45 mM (up to tetrasulfated materials); or 70 mM (up to heptasulfated materials) KH 2 PO 4 containing 60% acetonitrile/water for 75 min. The flow rate was 1 ml/min, 1-ml fractions were collected, and 35 S radioactivity was monitored as described. To evaluate sulfation status, we prepared multi-sulfated carbohydrate standards by enzymatic reactions as described previously (16).
Glycosidase-associated Carbohydrate Structure Determination-To determine nonreducing terminal structure of materials in P-4 gel filtration fractions, the purified materials were treated with glycosidases and analyzed as described above. For ␤-galactosidase and hexosaminidase A treatments, lyophilized samples were dissolved in water and incubated with either 25 milliunits of Jack bean ␤-galactosidase (Seikagaku Co., Tokyo, Japan) or 5 milliunits of human placental hexosaminidase A (Sigma) in a 30-l mixture containing 50 mM sodium citrate buffer pH 3.5, at 37°C for 5 min (␤-galactosidase) or for 16 h (hexosaminidase A). The samples were then boiled for 5 min, and the digests were separated by column chromatography. For keratanase treatment, we incubated purified samples with 75 milliunits of keratanase from Pseudomonas sp. (Seikagaku Co.) in a 20-l mixture containing 100 mM Tris-HCl, pH 7.5, at 37°C for an hour, stopped the enzymatic reaction by boiling, and subjected samples to column chromatography.
KS Synthesis in Cultured Human Corneal Epithelial Cells-SV40-immortalized human corneal epithelial cells were maintained in Dulbecco's modified Eagle's medium/Ham's F-12 50/50 mix medium (Mediatech, Inc., Herndon, VA) supplemented with 15% fetal bovine serum, 4.2 g/ml of bovine insulin, 0.8 g/ml of cholera toxin (Invitrogen), 8.3 g/ml of mouse epidermal growth factor (Invitrogen) and 33 g/ml gentamycin (Sigma). To produce sulfated KS, we co-transfected CGn6ST and KSG6ST expression vectors into 1 ϫ 10 6 cells using Nucleofector electroporator (Amaxa Inc. Gaithersburg, MD) according to the manufacturer's instruction and cultured cells in the above medium at 37°C for 48 h. To analyze effects of glycosyltransferase expression on KS production, we co-transfected CGn6ST and KSG6ST expression vectors together with specific glycosyltransferase siRNAs (sequence information is listed in supplemental Table S1), which were obtained from Ambion (Austin, TX). As a negative control, we used a commercially available negative control siRNA (NC#2; Ambion).
Western Blot Analysis of Highly Sulfated KS-Transfected cells were washed with cold PBS five times, and lysates were prepared using 120 l of lysis buffer containing 50 mM Tris-HCl, pH 7.4, 1% Nonidet P-40, 150 mM NaCl, a 1ϫ proteinase inhibitor mixture (Sigma), and 1 mM phenylmethylsulfonyl fluoride. After SDS-PAGE, the proteins were electroblotted to an Immobilon-P transfer membrane (Millipore). The membranes were treated with blocking buffer (10% nonfat milk in PBS) at room temperature for 1 h and then probed with 5D4 anti-KS antibody (1:4000) (Seikagaku Co.) in blocking buffer for 1 h. The membranes were washed in 0.05% Tween 20 in PBS three times and reacted with horseradish peroxidase-conjugated goat anti-mouse IgG antibody (1:4000) in blocking buffer for 1 h. After washing membranes in 0.05% Tween 20 in PBS three times, signals were detected using the ECL Plus reagent (GE Healthcare, Piscataway, NJ). Signal intensity was calculated using NIH Image 1.62 software.
Quantitative Reverse Transcription-PCR Analysis-Total RNA was isolated from transfected cells by RNeasy Mini kit (Qiagen) according to the manufacturer's instruction. After DNase I treatment, we subjected samples to reverse transcription using Superscript II reverse transcriptase (Invitrogen). Quantitative PCR analysis was carried out using Power SYBR Green PCR master mix (Applied Biosystems, Foster City, CA) with a specific primer set for each glycosyltransferase gene (see supplemental Table S2). Amplification of DNA products was monitored by the Mx 3000 QPCR system (Stratagene, La Jolla, CA) using the following reaction conditions: initial denaturation at 95°C for 10 min followed by 40 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 1 min, and extension at 72°C for 30 s. Cyclophilin A mRNA served as the internal control.

RESULTS
GlcNAc-sulfated Poly-N-acetyllactosamine Production by ␤3GnT7, ␤4GalT4, and CGn6ST in Vitro-To analyze enzymatic activity of ␤3GnT7 and ␤4GalT4 for KS carbohydrate synthesis in vitro, we constructed expression vectors encoding soluble glycosyltransferases and the sulfotransferases, KSG6ST and CGn6ST, and transfected Lec20 cells, a cell line derived from Chinese hamster ovary cells, with vectors individually to secrete enzymes into the medium (Fig. 1). Because Lec20 cells do not produce endogenous ␤4GalT1 (26), most glycosyltransferase activity in the medium originates from the transfected expression vector (Fig. 1B). Enzymes prepared from the culture Ϫ -6GlcNAc␤1-6Man␣1-6Man␣-octyl, was treated with concentrated culture medium of mock-transfected Chinese hamster ovary (CHO) cells (open circles), mock-transfected Lec20 cells (closed circles), or Lec20 cells transfected with an expression vector encoding soluble ␤4GalT4 (closed triangles), and analyzed by P-4 gel filtration. The retention position of the monosulfated trisaccharide substrate is marked with S. Note that addition of galactose onto the substrate molecule was observed in the reaction with culture medium of mock-transfected Chinese hamster ovary cells but not in that of mock-transfected Lec20 cells. OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41 medium were mixed in several combinations and incubated with a monosulfated trisaccharide substrate, 35 SO 3 Ϫ -6Glc-NAc␤1-6Man␣1-6Man␣-octyl, and carbohydrate donor substrates, UDP-GlcNAc, UDP-Gal, and PAPS. After 16 h of incubation at 37°C, the reaction products were analyzed by P-4 gel filtration column chromatography (Fig. 2). When the substrate was reacted with a mixture of ␤3GnT7 and ␤4GalT4 glycosyltransferases, two major products were detected as molecules larger than the substrate ( Fig. 2A), indicating that the two enzymes can add carbohydrate onto the carbohydrate sub- strate. When we incubated the substrate with a mixture of ␤3GnT7, ␤4GalT4, and CGn6ST, we detected several products of much larger size (Fig. 2B) than products from a mixture lacking CGn6ST ( Fig. 2A).

Enzymes Responsible for Corneal KS GAG Synthesis
To identify the carbohydrate structure of the products, we collected products separately and analyzed them by glycosidase treatment followed by column chromatography. Fraction I, containing the most abundant product, was slightly larger than the starting substrate by P-4 gel filtration column chromatography, and its size was not altered by hexosaminidase A treatment (Fig. 2C). However, a component of fraction I was converted to a digested fraction with the same retention position as the starting substrate by ␤-galactosidase treatment (Fig. 2C). This result indicates that the product in fraction I was a tetrasaccharide carbohydrate resulting from addition of one galactose on the nonreducing terminal of the starting trisaccharide substrate by ␤4GalT4. This tetrasaccharide product was next analyzed by SAX-10 anion exchange chromatography, and the product was found to be a monosulfated carbohydrate (Fig.  3A). From these results, we conclude that the product was monosulfated tetrasaccharide Gal␤1- We next analyzed the carbohydrate structure found in fraction II (Fig. 2B) using the same strategy. A chromatogram of fraction II on a P-4 column showed a broader peak (Fig. 2D), but the pattern was converted to a sharper peak by ␤-galactosidase treatment (fraction IIЈ in Fig. 2D). Meanwhile, hexosaminidase A treatment, which hydrolyzes both nonsulfated GlcNAc and 6-Osulfated GlcNAc located on nonreducing terminal of a carbohydrate substrate (27), converted constituents of fraction II into two components (II-1 and II-2 in Fig. 2D). Because the retention positions of fractions IIЈ and II-1 were very close but not identical and because both peaks overlapped with the original broad peak of fraction II (Fig. 2D), we conclude that fraction II contains a mixture of fractions IIЈ and II-1 and that the II-1 component is hydrolyzed by ␤-galactosidase and converted to component IIЈ (Fig.  2D). We also performed sequential digestion of fraction II components with ␤-galactosidase followed by hexosaminidase A and found that all products were converted to a single fraction having the same elution position as fraction II-2 and also fraction I (Fig. 2D). These results indicate that the carbohydrate backbone structure of fraction II is a mixture of a pentasaccharide, GlcNAc␤1-3Gal␤1-4GlcNAc␤1-6Man␣1-6Man␣1octyl (fraction IIЈ), and a hexasaccharide, Gal␤1-4GlcNAc␤1-3Gal␤1-4GlcNAc␤1-6Man␣1-6Man␣1-octyl (fraction II-1), and that hexosaminidase A treatment converted the pentasaccharide product (fraction IIЈ) into a tetrasaccharide carbohydrate, Gal␤1-4GlcNAc␤1-6Man␣1-6Man␣1-octyl (fraction II-2).
The sulfation status of components of II, II-1, and II-2 was analyzed by a SAX-10 column (Fig. 3, B-D). Fraction II was separated into two components, monosulfated material and disulfated material (Fig. 3B). Fraction II-1 was also separated into two components, both of which eluted at the same positions as fraction II, but the proportion of mono-to disulfated materials in fraction II-1 was nearly 1:1 (Fig. 3C). Because increased sulfation was apparently due to the presence of CGn6ST, disulfated components found in fractions II and II-1  OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41
Similar to fraction II, components of fractions III and IV were separated into two fractions by hexosaminidase A treatment (Fig. 2, E and F), indicating that each fraction contains carbohydrates of two different lengths, with or without exposed Glc-NAc at the nonreducing terminal. Similar to what was seen in fraction II, sequential treatment of the carbohydrate components in fraction III with ␤-galactosidase and hexosaminidase A converted the two carbohydrates into a single component exhibiting a sharp peak with a retention position identical to fraction II-1 on the P-4 chromatogram (Fig. 2E). This result indicates that carbohydrate components of fraction III are hepta-and octasaccharides, both of which are elongated products of fraction II-1 carbohydrates, to which either Gal␤1-4Glc-NAc␤1or GlcNAc␤1-structures have been added onto the fraction II-1 carbohydrate backbone. These results suggest that all observed carbohydrate products were made of an incomplete poly-N-acetyllactosamine backbone consisting of repeating N-acetyllactosamine units, with or without galactose on the nonreducing terminal on the starting substrate. We therefore estimated that fraction IV consisted of nona-and decasaccharide carbohydrates made of poly-N-acetyllactosamine repeats attached to the starting substrate.
The sulfation status of carbohydrates in fractions III and IV was again analyzed by SAX-10 HPLC before and after hexosaminidase A treatment (Fig. 3, E-J). By this analysis, we observed multi-sulfated carbohydrate products in the fractions up to tetrasulfated status (Fig. 3, H and I). Because we conclude that fraction III contains hepta-and octasaccharides, both of which carry 3 GlcNAc in their carbohydrate backbones, and because we included CGn6ST in the reaction mixture, the product should be trisulfated carbohydrate if the molecules were fully sulfated. By SAX-10 HPLC analysis, however, we found that more than 50% of the products were mono-and disulfated products in fraction III (Fig. 3E), indicating that the products were not fully sulfated on their GlcNAc residues. To understand the sulfation pattern of undersulfated carbohydrate products, we treated components of fraction III with keratanase, which recognizes GlcNAc-sulfated disaccharide SO 3 Ϫ -6GlcNAc␤1-3Gal linked to GlcNAc (with or without sulfation) via a ␤1-4 linkage and hydrolyzes that linkage (Fig. 4A). Thus, when a nonradioactive sulfate is attached to GlcNAc located close to 35 SO 3 Ϫ -GlcNAc, the molecule will be hydrolyzed and converted to radiolabeled trisaccharide. If nonradioactive sulfate is attached to GlcNAc located close to nonreducing terminal, the molecule will be converted to pentasaccharide (Fig. 4A). Following keratanase treatment, almost all fraction III components were converted to a trisaccharide component (Fig. 4B). The sulfation status of components of fraction III, which was

TABLE 1
Determined carbohydrate structures and their molar ratio in GlcNAc-sulfated poly-N-acetyllactosamine product synthesized by a mixture of ␤3GnT7/␤4GalT4/CGn6ST in vitro F, Galactose; ■, GlcNAc; E, Mannose. *, actual components found in these fractions are the illustrated structures without nonreducing terminal GlcNAc, because of hexosaminidase A treatment. **, these structures are estimated from the results of keratanase treated disulfated hepta/octasaccharide structures. OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41 originally identified as a mixture of mono-, di-, and trisulfated carbohydrates, was also identified as a monosulfated product following keratanase treatment, indicating that carbohydrates in fraction III were sulfated on the internal GlcNAc located closest to 35 SO 3 Ϫ -GlcNAc. This result suggests that GlcNAc sulfation of carbohydrate products by a mixture of ␤3GnT7/ ␤4GalT4/CGn6ST was consecutively but not randomly occurring during elongation of carbohydrate backbone. Thus, we determined the carbohydrate structure of the products synthesized by the three enzymes ␤3GnT7, ␤4GalT4, and CGn6ST in Table 1. Based on the result, we illustrated synthetic flow of carbohydrate products generated by the three enzymes and calculated efficiencies of each enzymatic reaction step (supplemental Fig. S1). We also calculated average reaction efficiency for each enzymatic reaction that repeatedly takes place by the same enzyme for production of the same nonreducing terminal structure during GlcNAc-sulfated poly-N-acetyllactosamine chain production (Fig. 5).

Enzymes Responsible for Corneal KS GAG Synthesis
Highly Sulfated KS Production by ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST in Vitro-Corneal KS GAG is highly sulfated on both GlcNAc and galactose residues (10 -12). Sulfation of galactose residues is catalyzed by KSG6ST (15). From its substrate specificity, sulfation of galactose by KSG6ST likely occurs after production of GlcNAc-sulfated poly-N-acetyllactosamine GAG by CGn6ST and glycosyltransferases (11,16,23). Using soluble enzymes including KSG6ST, we tested pro-duction of highly sulfated KS GAG in vitro. The reaction products of a mixture of ␤3GnT7, ␤4GalT4, and CGn6ST included elongated carbohydrates of several lengths (Fig. 6A), and one fraction containing a mixture of nona/decasaccharides was identified to have mostly tri-and tetrasulfated structures (Fig.  6E). When a mixture of ␤3GnT7/␤4GalT4/CGn6ST products was further treated by KSG6ST, the degree of sulfation of the nona/decasaccharide fraction was increased up to heptasulfated products (Fig. 6F), indicating that GlcNAc-sulfated poly-N-acetyllactosamine carbohydrates were utilized and converted to highly sulfated KS carbohydrate following KSG6ST treatment. When a trisaccharide carbohydrate substrate was treated with all four enzymes (␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST), elongation efficiency was equivalent to that found in a reaction of three enzymes (Fig. 6, A and C). The products were also sulfated up to hexasulfated components (Fig. 6G), which was a higher degree of sulfation than that seen in the ␤3GnT7/␤4GalT4/CGn6ST mixture (Fig. 6E) but lower than that seen in a two-step reaction with ␤3GnT7/␤4GalT4/ CGn6ST followed by KSG6ST (Fig. 6F). Interestingly, treatment of a trisaccharide substrate with a mixture of KSG6ST and ␤3GnT7 and ␤4GalT4 glycosyltransferases elongated the products up to nona/decasaccharides, and these products were also sulfated up to tetrasulfated components (Fig. 6H). Because treatment of the substrate with two glycosyltransferases without any sulfotransferase only produced up to penta/hexasaccharide carbohydrates ( Fig. 2A), this result indicates that KSG6ST enhanced elongation of the poly-N-acetyllactosamine backbone by the glycosyltransferases. From these results, we conclude that four enzymes: ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST, are sufficient to synthesize highly sulfated KS carbohydrate in vitro.
␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST Are Required for Sulfated KS Production in Cultured Corneal Cells-We next analyzed production of KS GAG in cultured corneal cells. Under normal culture conditions, SV40-transformed human corneal epithelial cells do not produce detectable levels of highly sulfated KS GAG, which is recognized by the 5D4 monoclonal antibody (28,29). However, the corneal cells began producing 5D4-positive, highly sulfated KS carbohydrate when CGn6ST and KSG6ST were overexpressed (Fig. 7, A and D), and this was a consistent result with our previous observation of KS production in HeLa cells (25). These findings indicate that both sulfotransferases are required to produce highly sulfated KS carbohydrate. Human genes B3GNT7 and B4GALT4 encode ␤3GnT7 and ␤4GalT4, respectively, and corneal epithelial cells endogenously express both genes (Fig. 7, G and H). 4 When we suppressed B3GNT7 expression in the cultured corneal epithelial cells by transfection of specific siRNAs, 5D4positive KS production was dramatically reduced to less than 10% of levels seen in cells transfected with negative control siRNA; even both CGn6ST and KSG6ST were overexpressed (Fig. 7, D and G). Similarly, we observed a 50% reduction of 5D4-positive product in the cells using B4GALT4 siRNAs (Fig.  7, E and H). Because siRNAs for other glycosyltransferases,

Enzymes Responsible for Corneal KS GAG Synthesis
such as ␤1,3-N-acetylglucosaminyltransferease-2 (␤3GnT2) and ␤1,4-galactosyltransferase-1 (␤4GalT1), both of which are likely major contributors to poly-N-acetyllactosamine elongation, did not have a significant effect on production of highly sulfated KS production in cells (Fig. 7, F and I), we conclude that both ␤3GnT7 and ␤4GalT4 are required for elongation of the KS carbohydrate backbone in human corneal cells.

DISCUSSION
In this study, we demonstrate that ␤3GnT7 and ␤4GalT4 are required for KS GAG production in vitro and in cultured human corneal cells. ␤3GnT7 was originally identified as a gene involved in tumor invasiveness and later as a member of the ␤3GnT family (30). ␤4GalT4 was identified as a member of ␤4GalT family by expressed sequence tag data base searches (31) and suggested to function in neolacto-series glycosphingolipid biosynthesis (32). Seko et al. (22) reported ␤4GalT activity that acts preferentially on sulfated substrates in human colorectal mucosa and later found that the enzyme responsible for that activity was ␤4GalT4. Using in vitro analysis, they suggested that among seven enzymes of the ␤4GalT family, ␤4GalT4 functions in biosynthesis of sulfo sialyl Lewis X structure as well as in production of KS GAG (22). They also analyzed enzymatic activity of a COS-7 cell membrane fraction containing overexpressed ␤3GnTs using several carbohydrate substrates and found that ␤3GnT7 has higher enzymatic activity for sulfated N-acetyllactosamine substrates than for nonsulfated substrates, suggesting that ␤3GnT7 functions in KS biosynthesis (23). Their studies suggested that four enzymes, ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST, could produce sulfated KS GAG in vitro (23), but such an observation has not been reported until this study.
KS GAG biosynthesis apparently occurs in two stages: the production of GlcNAc-sulfated poly-N-acetyllactosamine chain by ␤3GnT7, ␤4GalT4, and CGn6ST and the production of highly sulfated KS by sulfation of galactose by KSG6ST (11,16,22,23). CGn6ST transfers sulfate only on the nonreducing terminal GlcNAc (16); therefore, sulfation of GlcNAc residues must be coupled to elongation of the carbohydrate backbone processed by glycosyltransferases. On the other hand, KSG6ST has much higher sulfation activity on substrates containing N-acetyllactosamine disaccharide with sulfate on GlcNAc (15,17), suggesting that sulfation of galactose occurs after production of the GlcNAc-sulfated poly-N-acetyllactosamine chain. Furthermore, because ␤3GnT7 has very weak glycosyltransferase activity on a sulfated substrate when the nonreducing terminal galactose is sulfated (23), sulfation of the nonreducing terminal galactose by KSG6ST may inhibit elongation of the KS GAG backbone. However, when we reacted a sulfated carbohydrate substrate with all four enzymes including KSG6ST, elongation of carbohydrate products was not significantly inhibited (Fig. 6C), although the degree of product sulfation was not as high as that seen with ␤3GnT7/␤4GalT4/CGn6ST treatment followed by KSG6ST reaction (Fig. 6, F and G). This result indicates that production of a GlcNAc-sulfated poly-N-acetyllactosamine chain is not inhibited by the presence of KSG6ST. Surprisingly, when we reacted the substrate with three enzymes (␤3GnT7, ␤4GalT4, and KSG6ST) without CGn6ST, we observed elongated carbohydrates, which were not seen in a reaction of two glycosyltransferases with no sulfotransferase ( Figs. 2A and 6D). This unanticipated result can be explained as follows. Firstly, elongation of the carbohydrate chain by ␤3GnT7 and ␤4GalT4 is not coupled to sulfation of galactose by KSG6ST because the nonreducing terminal galactose is a less favored substrate for KSG6ST (17). Second, nonsulfated carbohydrate products made of repeating disaccharides of Gal␤1-4GlcNAc are less hydrophilic and therefore less efficient substrates for continuous chain elongation by glycosyltransferases because of poor solubility. Indeed, production of nonsulfated or less sulfated carbohydrate may be a primary cause of corneal deposit formation of macular corneal dystrophy patients (33). Thus, when KSG6ST is present during carbohydrate synthesis by ␤3GnT7 and ␤4GalT4, elongated carbohydrates are occasionally but not cooperatively sulfated on galactose residues by sulfotransferase, such that sulfated products become more hydrophilic and are utilized for further elongation by glycosyltransferases. Taken together, we conclude that highly sulfated KS is mainly produced by ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST and that biosynthesis of carbohydrate occurs in two steps: production of GlcNAc-sulfated poly-N-acetyllactosamine structure followed by sulfation of galactose residues. Because natural corneal KS GAG chains are efficiently elongated more than 10 sulfated N-acetyllactosamine repeats (11,12,34), we hypothesize that in KS GAGproducing cells, ␤3GnT7, ␤4GalT4, and CGn6ST are tightly associated or closely co-localized with each other to achieve efficient production of GlcNAc-sulfated poly-N-acetyllactosamine chains in the Golgi, whereas KSG6ST may be localized elsewhere or not tightly associated with the other enzymes. Further analysis of cellular localization or biochemical examination such as immunoprecipitation is required to test this hypothesis.
A mixture of ␤3GnT7, ␤4GalT4, and CGn6ST produced GlcNAc-sulfated poly-N-acetyllactosamine carbohydrate in vitro. We confirmed production of up to a tetrasulfated decasaccharide (Table 1); however, the products clearly contained longer and more sulfated carbohydrates, because P-4 column chromatography detected peaks that likely contain larger molecules (Fig. 2B). Although long carbohydrate products were made by a simple reaction with a three-enzyme mixture, the length of products was much less than that of corneal KS GAG (11,12), indicating that reaction conditions of this study were not as high as that in vivo. Further optimization, such as altering the ratio of the three enzymes or the concentration of donor molecules may improve the efficiency of KS carbohydrate production in vitro. In our reaction conditions, although galactosylation of the nonreducing terminal GlcNAc by ␤4GalT4 was more than 90%, the efficiencies of the other two reactions were about 50% (Fig. 5), indicating that reaction conditions of these steps must be improved to generate longer carbohydrate products. However, the most inefficient step of KS GAG production in vitro was the first step of ␤3GnT7 reaction, which transfers GlcNAc to the nonreducing terminal galactose of monosulfated tetrasaccharide carbohydrate. Because 82.2% of products were remained as Gal␤1-4(SO 3 Ϫ -6)GlcNAc␤1-6Man␣1-6Man␣1-octyl (Table 1), the effi-

Enzymes Responsible for Corneal KS GAG Synthesis
ciency of GlcNAc transfer at this step was only 17.8%, which was one-third the average efficiency of ␤3GnT7 enzyme observed throughout sulfation-coupled poly-N-acetyllactosamine elongation (Fig.  5), suggesting that the tetrasaccharide, Gal␤1-4(SO 3 Ϫ -6)GlcNAc␤1-6Man␣1-6Man␣1-octyl, may not be a suitable ␤3GnT7 substrate. A commonly recognized carbohydrate structure of N-linked corneal KS consists of an extended linear KS GAG chain attached to a complextype N-glycan core structure via one or two repeats of nonsulfated N-acetyllactosamine disaccharides (11,12,34). ␤3GnT7 may have less activity on galactose located close to the N-glycan core structure, and formation of N-acetyllactosamine repeats immediately adjacent to the N-glycan core may be catalyzed by other glycosyltransferases. Indeed, among ␤3GnT family proteins, ␤3GnT7 shows the lowest activity on a nonsulfated N-acetyllactosamine connected to a complex-type N-glycan core (23), supporting this idea. During continuous sulfation-coupled carbohydrate elongation, we observed that both ␤3GnT7 and ␤4GalT4 showed at least 4-fold higher activity toward an N-acetyllactosamine structure having sulfated GlcNAc at the position closest to the nonreducing terminal than toward a terminal structure without sulfate (Fig.  5). This finding is consistent with substrate specificity of these enzymes observed in vitro (22,23) and strongly suggests that ␤3GnT7 and ␤4GalT4 are responsible for KS GAG production. We also observed that there was no carbohydrate product exhibiting more than a tetrasaccharide structure without sulfation at the nonreducing terminal during the glycosyltransferase reaction with CGn6ST (Table 1) or FIGURE 6. Enzymatic synthesis of highly sulfated KS chains in the presence of KSG6ST in vitro. The starting substrate was treated with a mixture of three enzymes, ␤3GnT7, ␤4GalT4, and CGn6ST (A) followed by treatment of KSG6ST (B), and analyzed by P-4 gel filtration chromatography. The substrate was also treated with either a mixture of four enzymes, ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST (C), or a mixture of three enzymes, ␤3GnT7, ␤4GalT4, and   A, D, and G), ␤4GalT4 (B, E, and H), ␤3GnT2 or ␤4GalT1 (C, F, and I). Forty-eight hours later, the cell lysates were analyzed by SDS-PAGE followed by Coomassie Brilliant Blue staining (A-C) or by Western blot analysis using the 5D4 antibody for highly sulfated KS production (D-F). Transfected cells were also analyzed by quantitative reverse transcription-PCR to confirm knockdown effects of each siRNA (G-I). NC, negative control siRNA. Calculated values in G-I represent percentages of target gene expression in a specific siRNA-transfected cells over that of negative control siRNA-transfected cells. OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41 JOURNAL OF BIOLOGICAL CHEMISTRY 30095 without sulfotransferase ( Fig. 2A), suggesting that a carbohydrate molecule with a nonsulfated tetrasaccharide at the nonreducing terminal is not utilized as a glycosyltransferase substrate. Presently, however, it is unclear whether this is due to lower hydrophilicity of the carbohydrate molecule or less preferable substrate structure.

Enzymes Responsible for Corneal KS GAG Synthesis
Decreasing ␤3GnT7 mRNA levels dramatically reduced sulfated KS production to less than one-tenth of controls in cultured human corneal epithelial cells (Fig. 7D). Reduced ␤4GalT4 mRNA also inhibited sulfated KS production but less effectively than suppression of ␤3GnT7 mRNA, although in both cases siRNAs suppressed mRNA expression more than 70% (Fig. 7, G and H). siRNAs specific for ␤3GnT2 and for ␤4GalT1 did not significantly suppress sulfated KS production, indicating that reduction of sulfated KS production in cultured corneal cells was due to specific suppression of ␤3GnT7 or ␤4GalT4 by corresponding siRNAs. The weaker effect of ␤4GalT4 siRNAs may be because ␤4GalT4 has much higher enzymatic activity for elongation of KS GAG chains than ␤3GnT7; thus, the remaining ␤4GalT4 mRNA may account for ␤4GalT4 activity in producing sulfated KS. Further investigation is required to eliminate the possibility that other ␤4GalT enzymes compensate for ␤4GalT4 in sulfated KS production.
In summary, we have detected production of sulfated KS by four enzymes, ␤3GnT7, ␤4GalT4, CGn6ST, and KSG6ST, in vitro and in cultured cells. Because specific suppression of either ␤3GnT7 or ␤4GalT4 reduced sulfated KS production in cultured corneal cells, we conclude that ␤3GnT7 and ␤4GalT4 play major roles in elongating sulfated KS in the cornea. Further characterization of the glycosyltransferases and the sulfotransferases in producing sulfated KS may establish roles for the carbohydrate chain for corneal extracellular matrix organization.