Multiple non-reducing chain termini isolated from bovine corneal keratan sulfates.

Keratan sulfate-containing proteoglycans were isolated from bovine cornea (15-month-old to 3-year-old animals) and digested with the enzyme, keratanase II. The released oligosaccharides, which included non-reducing termini and repeat region oligosaccharides but not linkage regions, were reduced with alkaline borohydride and fractionated on a Spherisorb column. These oligosaccharides were examined by 600-MHz 1H NMR spectroscopy using one- and two-dimensional methods and, in addition to some oligosaccharide alditols previously recovered from skeletal keratan sulfate, the following new capping structures were identified: NeuAcalpha2-6Galbeta1-4GlcNAc(S)-ol, NeuAcalpha2-3Gal(S)beta1-4GlcNAc(S)beta1-3Galbeta1-4GlcNAc(S )-ol, NeuGcalpha2-6Galbeta1-4GlcNAc(S)beta1-3Galbeta1-4Gl cNA c(S)-ol, NeuGcalpha2-3Galbeta1-4GlcNAc(S)beta1-3Galbeta1-4Gl cNA c(S)-ol, NeuGcalpha2-3Gal(S)beta1-4GlcNAc(S)beta1-3Galbeta1-4GlcNAc(S )-ol, NeuGcalpha2-3Gal(S)beta1-4GlcNAc(S)beta1-3Gal(S)beta1-4GlcNAc(S)-o l, Galalpha1-3Galbeta1-4GlcNAc(S)beta1-3Galbeta1-4GlcNAc( S)-ol, Galalpha1-3Galbeta1-4GlcNAc(S)beta1-3Gal(S)beta1-4GlcNAc(S)- ol, GlcNAc(S)beta1-3Gal(S)beta1-4GlcNAc(S)-ol, and GalNAc(S)beta1-3Gal(S)beta1-4GlcNAc(S)-ol. These structures represent seven families of capping residues, whose relative molar proportions are given in parentheses: NeuAcalpha(2-3)- (12%), NeuAcalpha(2-6)- (41%), NeuGcalpha(2-3)- and NeuGcalpha(2-6)- families (12%), Galalpha(1-3)- (26%), GalNAc(S)beta(1-3)- (5%), and GlcNAc(S)beta(1-3)- (4%). It is not clear, at present, where each of these structures occurs on the bi-antennary N-linked corneal keratan sulfate chains, which themselves occur within three keratan sulfate proteoglycan species. However, examination of the relative proportions of the capping to the repeat structures and knowledge of the average molecular size suggests that the sum of these non-reducing termini represents the caps of two antennae.

The corneal stroma is mostly an extracellular matrix, which comprises collagens, proteoglycans, and matrix proteins. Its transparency is dependent upon its hydration and the orderly arrangement of the collagen fibrils. The fibrils in mammalian stroma measure 25-30 nm in diameter and have a mean interfibrillar distance of 66 nm within lamellae (1).
The proteoglycans of the corneal stroma are associated with the collagen fibrils, and they comprise the keratan sulfate (KS) 1 and the chondroitin/dermatan sulfate (CS/DS) families. Electron microscopic studies have shown that four separate proteoglycan binding sites lie within the D period of collagen fibrils in rabbit cornea. The KS proteoglycans were located at the a or c "step" bands and CS/DS proteoglycans were found at the d or e "gap zone." A similar pattern was obtained in bovine cornea (2). On the basis of such studies, Scott has proposed a model (3) for proteoglycan-collagen interactions in cornea in which duplexed glycosaminoglycan chains (both doublestranded DS and KS) may bridge collagen fibrils and thus ensure the precise interfibrillar distances vital for corneal transparency.
The major CS/DS proteoglycan of cornea is decorin (4). This is a member of a group of small interstitial proteoglycans, which includes biglycan and fibromodulin that all interact with fibrillar collagens via their protein cores.
Recent studies of corneal KS have revealed the existence of several discrete proteoglycans. Originally two core proteins (37 and 25 kDa) were identified (5). Subsequently, two forms of the 37-kDa protein (37A and 37B) were resolved (6). Then, a cDNA clone was isolated and sequenced that coded for a chicken corneal KSPG. This protein, lumican, had a deduced molecular mass of 37 kDa and a sequence related to fibromodulin, decorin, and biglycan (7). This protein was shown to have a high homology with the bovine 37B core protein, which was thus identified as bovine lumican (8). The bovine 37A core protein has recently been sequenced and named keratocan (9).
Keratan sulfate (KS) was first isolated from the bovine corneal stroma by Meyer (10), who subsequently classified KS types (11). Corneal KS with an alkali-stable bond between N-acetylglucosamine and asparagine was called KS-I, and skeletal KS with the alkali-labile bond between N-acetylgalactosamine and serine or threonine was designated KS-II. This skeletal type has been further subclassified into articular, KS-II-A, and non-articular, KS-II-B (12), on the basis that the former contains ␣(1-3)-fucose and ␣(2-6)-N-acetylneuraminic acid, which are absent in the latter. Finally, a third type of KS obtained from rat brain was shown to be O-linked between mannose and serine or threonine (13).
The characteristic structure of KS involves a disaccharide repeat unit, -4GlcNAc␤1-3Gal␤1-, which may be sulfated on C-6 of both the N-acetylglucosamine and galactose residues. However, the detailed structures of specific KS types are probably best considered to be composed of three regions: linkage, repeat, and chain-capping.
In corneal KS these structures are not so well understood. The linkage region is, apparently, well characterized and is of the complex type (25,26), as shown in Structure 1.
The repeat region (or regions) have been examined in detail (26,27). However, little is known about the chain caps in corneal KS. Sialic acid was found as a minor component in chain termini (28) and further studies showed that this residue was linked to the 3-position of galactose residues, in a ratio of sialic acid to mannose of 0.41 (29). Another study (26) concluded that sialic acid was located close to the linkage region.
The ultimate goal is to identify the structures and functions of the KS chains in each of the corneal KSPGs. However, in this study the structures of the keratanase II fragments derived from the non-linkage regions of the entire corneal KS chain population are examined.
Sepharose CL-6B, Q-Sepharose Fast Flow gel, and a Mono-Q HR 10/10 column were purchased from Pharmacia (Uppsala, Sweden  toir and immediately cooled to 4°C. They were processed within 2 days. The corneal stroma (with endothelium) was excised after the epithelium was scraped off. In a typical preparation, 110 g (wet weight) of corneas were obtained from 176 eyes, an average of 0.63 g/eye.
All of the extraction procedures were performed at 4°C. The corneas (110 g) were extracted with 500 ml of 4 M guanidine HCl containing 0.05 M sodium acetate, pH 5.8, 0.1 M 6-aminohexanoic acid, 0.01 M EDTA, and 0.005 M benzamidine HCl for 24 h. The extract was recovered by filtering through a nylon sheet, and the residual material was extracted twice more with 200 ml for 24 h. The three extracts were combined and dialyzed against 0.15 M NaCl, 7 M urea, 0.05 M Tris-HCl, pH 7.0.
The dialysate was chromatographed on a Q-Sepharose column (1.5 cm ϫ 15 cm) in five runs. The unbound material was eluted with loading buffer: 0.15 M NaCl, 7 M urea, 0.05 M Tris-HCl, pH 7.0 (data not shown). The column was connected to a Bio-Rad HPLC system, and the bound material was eluted with a linear gradient of 0.15 M to 1.0 M NaCl containing 7 M urea and 0.05 M Tris-HCl, pH 7.0, within 40 min, and finally to 1.6 M within 10 min, at a flow rate of 2 ml/min. Fractions of 2 ml (1 min) were collected and analyzed with 1,9-dimethylmethylene blue (1-l aliquots) for glycosaminoglycan (30) and by direct enzymelinked immunosorbent assay (ELISA) using monoclonal antibody 5D4 for KSPGs and antibody LF95 for CS/DSPGs.
Fractions containing KSPGs were pooled, dialyzed against water, and lyophilized (yield 482 mg). This sample was then dissolved in 15 ml of eluent: 0.15 M NaCl, 7 M urea, 0.05 M Tris-HCl, pH 7.0, and chromatographed on a Sepharose CL-6B column (2.5 cm ϫ 155 cm) eluted at a flow rate of 25.6 ml/h. Fractions of 12.8 ml (30 min) were collected and analyzed by absorbance at 276 nm, 1,9-dimethylmethylene blue assay (2-l aliquots), and direct ELISA using antibody 5D4 and LF95 (2 l aliquots). Fractions 42-65, positive to the 1,9-dimethylmethylene blue assay, contained both KSPGs and CS/DSPGs as recognized by 5D4 and LF95. They were pooled and recovered after dialysis and lyophilization (yield 374 mg). The CS/DSPGs were digested with the enzyme chondroitin ABC lyase. Briefly, 150 mg of proteoglycans was dissolved in 15 ml of 0.1 M Tris acetate, pH 7.3, and incubated with chondroitin ABC lyase (0.6 units) at 37°C for 10 h. The digestion was terminated by the addition of 60 ml of 0.15 M NaCl, 7 M urea, 0.05 M Tris-HCl, pH 7.0. The resulting CS/DS-oligosaccharides and core proteins (decorin) were removed by chromatography on a Q-Sepharose column (1.5 cm ϫ 15 cm) as described above, except that the gradient started after 10 min of isocratic elution. The KSPGs were recovered after dialysis against water and lyophilization (yield 100 mg, 0.23% of the wet weight).
SDS-PAGE-Corneal samples were prepared as described below. Enzymatic deglycosylation with recombinant peptide N-glycosidase F was performed following the manufacturer's instructions. Aliquots were removed after 1 h and 20 h of incubation time. Enzymatic treatment with enzymes specific to keratan sulfate was carried out under the following conditions; 1 mg of KSPGs was dissolved in 100 l of digestion buffer and incubated at 37°C with various enzymes. For endo-␤-galactosidase (0.02 units), digestion was performed for 22 h in 0.05 M sodium acetate, pH 5.8, containing 0.2 mg of bovine serum albumin/ml. For keratanase (1.5 units), digestion was for 5 h in 0.05 M Tris acetate, pH 7.3, containing 5 mM sialidase inhibitor. For keratanase II (0.01 units), digestion was for 5 h in 10 mM sodium acetate, pH 6.5.
SDS-PAGE was performed using a discontinuous SDS system (31) with a 3% stacking gel and 10% running gel. The dithiothreitol-reduced samples were subjected to electrophoresis at 6 mA/1.5-mm gel for 20 h at 15°C. The gel was stained with Coomassie Blue (Fig. 4).
Preparation of Corneal Keratan Sulfate Chains-Corneal KSPGs (40 mg) were dissolved in 0.8 ml of 1% SDS and boiled for 5 min. To this solution was added 7.2 ml of 0.1 M phosphate buffer, pH7.3, containing 20 mM EDTA and 7% CHAPS, prior to the addition of 40 units of peptide N-glycosidase F. The enzyme was inactivated by heating at 100°C for 5 min after incubation at 37°C for 20 h. The digest was centrifuged at 100,000 ϫ g for 1 h and the supernatant dialyzed against water and lyophilized. The released KS chains were further purified by gel-per-meation chromatography on a Sepharose CL-6B (83 cm ϫ 1.5 cm) column and ion-exchange chromatography on a Mono-Q HR 10/10 column (data not shown).
The resulting KS was examined by 1 H NMR spectroscopy (see below). Sugar compositional analysis using a CarboPac PA1 column after acid hydrolysis with 4 M trifluoroacetic acid (19) showed that the ratio of Fuc:GlcNAc:Gal:Man was 1.45:35:32:3. The weight-average molecular weight (M w ) of the KS chains was estimated (32) on a pre-calibrated Bio-Gel TSK-30 XL column (300 ϫ 7.8 mm).
Isolation of Oligosaccharides Derived by Keratanase II Digestion of KSPGs-Corneal KSPGs (100 mg) were dissolved in 5 ml of 10 mM sodium acetate, pH 6.5 and incubated with 0.25 units of keratanase II at 37°C for 30 h. The enzyme was inactivated by heating at 100°C for 5 min. The core proteins were removed by centrifugation at 100,000 ϫ g for 1 h. The digest was then reduced with 4 M NaBH 4 for 3 h and recovered after desalting on a Bio-Gel P2 column (1 cm ϫ 10 cm).
The reduced keratanase II digest was chromatographed on a Spherisorb S5 SAX column (1 cm ϫ 25 cm) in three identical runs. The sample was first eluted isocratically with 2 mM LiClO 4 , pH 5.0, for 10 min, followed by a 2-250 mM LiClO 4 gradient over 240 min, and finally a 250 -500 mM LiClO 4 gradient over 30 min at a flow rate of 2 ml/min. The a The weight of each component was determined from refractive index data obtained from the chromatogram on a Bio-Gel P2 column. The molar ratio is calculated from the weight and molecular weight of each characterized oligosaccharide. eluate was monitored for the absorbance at 206 nm (Fig. 5). The fragments were recovered after desalting on a Bio-Gel P2 column. 1  H NMR spectra at 600 and 500 MHz were acquired using Bruker AMX600 and AM500 spectrometers, respectively. The 400-MHz 1 H NMR spectra were obtained on a Jeol GSX400 spectrometer.

RESULTS
KS Proteoglycans-The entire population of KSPGs from the bovine corneal stroma (plus endothelium) were prepared after dissociative extraction and three stages of chromatography. The first step (Fig. 1) involved ion-exchange chromatography to separate the proteoglycans from protein contaminants. The material recovered (pooling bar) contained nearly all the KS, but substantial amounts of decorin (LF95-positive fractions) were present. The gel permeation chromatography step (Fig. 2) permitted further purification on the basis of size. Finally, after chondroitin ABC lyase treatment to digest the glycosaminoglycans in the CS/DSPGs, the final ion-exchange step (Fig. 3) was used to separate the KSPGs from the core proteins of the CS/DSPGs. To establish the apparent number of proteoglycan species present in the preparation, the protein core sizes after several different enzymatic treatments were determined by SDS-PAGE (Fig. 4). Using the enzyme endo ␤-galactosidase,    37(B) kDa. The comparative core protein sizes after enzyme treatment increase in the order: recombinant peptide N-glycosidase F, endo ␤-galactosidase, keratanase, and then keratanase II, reflecting that the relative proximity of the enzyme cleavage sites to the core protein is in this sequence. The pattern of core proteins released by recombinant peptide Nglycosidase F (EC 3.2.2.18) in this study is slightly different from that observed by Funderburgh et al. (5), who used Nglycanase (EC 3.5.1.52), and this, perhaps, is due to different fine specificities in the enzymes used.
KS Chains-The KS chains prepared from these KSPGs by peptide N-glycosidase F digestion were assessed for their molecular sizes on a calibrated TSK-30 column (32). From these data (not shown), it is clear that the total KS population exhibits considerable chain size polydispersity (M n 5000 -30,000).
The mean weight-average M w was found to be 14000 (calculated mean M n ϭ 10,000). These are believed to be the first data reporting the sizes of isolated KS chains, rather than those of peptido-KS fragments. The NMR spectrum of this KS chain preparation (see Fig. 14) is discussed below.
Keratanase II-derived Fragments-The oligosaccharides were derived from a limit digest of the KS proteoglycan population and thus lack the linkage regions, which remained attached to the protein cores. After borohydride reduction, these fragments were chromatographed on a Spherisorb S5 SAX column (Fig. 5) and recovered. The resulting oligosaccharides listed in Table I can be categorized into those deriving from either the repeat regions or the capping regions.
The one-dimensional spectrum for 19 (Fig. 6) resembles those from oligosaccharides C1-C5. The spectrum may be fully assigned with the aid of COSY-45 data (Fig. 7). Complete sets of resonances corresponding to five sugar residues are found (see Table II). Signals at 1.815 and 2.755 ppm connect through to a series of protons with positions characteristic for ␣(2-3)linked N-acetylneuraminic acid, residue E. The presence of an internal 6-sulfated N-acetylglucosamine (residue C) is indicated by the anomeric proton signal at 4.730 ppm through to the non-equivalent methylene resonances at 4.293 and 4.445 ppm (Fig. 7). A reducing terminal 6-sulfated N-acetylglucosaminitol, residue A, shows connections almost identical to those found for oligosaccharide C2 (see above), in which Glc-NAc(S)-ol is attached to an unsulfated galactose (33). The two remaining sets of signals can be assigned to galactose residues B and D by reference to previously presented data (33, 35). The Oligosaccharides Containing ␣(2-6)-linked N-Acetylneuraminic Acid-The oligosaccharides 11 and 21 have 1 H NMR spectra (data not shown) identical to those for the previously characterized C1 and C3 from KS type II (33,34). Therefore, the structures of 11 and 21 are as follows: 11, NeuAc␣2-6Gal␤1-4GlcNAc(S)␤1-3Gal␤1-4GlcNAc(S)-ol; 21, NeuAc␣2-6Gal␤1-4GlcNAc(S)␤1-3Gal(S)␤1-4GlcNAc(S)-ol.
The 1 H NMR spectrum of 6 (data not shown) also displays signals characteristic of an ␣(2-6)-linked N-acetylneuraminic acid residue with H-3 ax at 1.693 ppm. However, a signal at 2.747 ppm has not, hitherto, been observed from an ␣(2-6)-linkage environment. The spectrum is simple, and the complete assignment of the chemical shifts (see Table III) for 6 is consistent with the following trisaccharide structure: NeuAc␣2-6Gal␤1-4GlcNAc(S)-ol.
The spectrum for fraction 10 (Fig. 8) shows two groups of signals with intensity ratio 2:1, representing two components: 10i and 10ii. The spectral profile of 10ii is similar to those of 18 and 23; this includes signals typical for an ␣(2-3)-linked Nglycolylneuraminic acid residue. It may, therefore, possess the same core structure as 18 and 23.
Oligosaccharides Capped with N-Acetylated Residues-Oligosaccharide 20 elutes in the region between trisulfated and tetrasulfated repeat-unit tetrasaccharides. In this region only sialylated, trisulfated pentasaccharides have been observed in studies of cartilage KS (33,34). However, the 1 H NMR spectrum for 20 shows no sialic acid resonances (Fig. 12a). It therefore must be a new oligosaccharide derived from corneal KS.
Oligosaccharide 17 elutes just after the trisulfated repeatunit tetrasaccharides. The one-dimensional (Fig. 12b) and twodimensional 1 H NMR (Fig. 13) spectra demonstrate that it is a trisaccharide. Responses corresponding to GlcNAc(S)-ol, residue A, and Gal(S), residue B, are located at positions almost identical with those for the corresponding signals from 20. The third set of connected resonances is, however, quite distinct. The anomeric proton signal at 4.680 ppm has a novel shift position. It connects to H2 at 3.973 ppm with a coupling constant of 8.34 Hz, suggesting that it has a ␤-configuration.  (33,34), except that H-1 and H-2 are shifted significantly downfield. The one-dimensional spectrum also shows an extra N-acetyl methyl proton signal at 2.052 ppm, together with a signal at 2.071 ppm arising from the N-acetylglucosaminitol. These data (summarized in Table III) suggest that this sugar must be a ␤(1-3)-linked 6-sulfated N-acetylgalactosamine (41). The structure of 17 is, therefore, GalNAc(S)␤1-3Gal(S)␤1-4Gl-cNAc(S)-ol.
NMR Spectrum of Peptide N-Glycosidase F Released and Reduced KS Chains-A partial 600-MHz 1 H NMR spectrum for the peptide N-glycosidase F released and borohydride reduced KS chains is shown in Fig. 14 The major methyl resonance of fucose is seen at 1.235 ppm corresponding to ␣(1-6)-linked fucose in the linkage region (42), but a smaller signal at 1.164 ppm can be assigned to ␣(1-3)-linked fucose (43) from the repeat regions (see below).

DISCUSSION
In this study seven families of chain caps have been isolated from the entire corneal KS population and their proportions are shown in Fig. 15. The most surprising aspect of this research is the diversity of these chain-capping structures. Thus, the terminal residues and their linkages, NeuAc␣2-3, NeuAc␣2-6, NeuGc␣2-3, NeuGc␣2-6, Gal␣1-3, GlcNAc(S)␤1-3, and Gal-NAc(S)␤1-3, have all been identified in this bovine corneal KSPG preparation. In principle, Gal␤1-4 is a potential chaincapping residue, but separate studies (data not shown) using the enzyme keratanase indicate that Gal␤1-4 is unlikely to be a chain terminator.
highly antigenic in humans. The Gal␣1-3Gal␤1-4GlcNAc structure is also not observed in humans because of a nonfunctional ␣(1-3)-galactosyltransferase (45). It seems possible that the diversity of chain caps is related to the presence of three or more KSPGs in cornea, but an assessment of probable cap function will require identification of the antenna and PG distribution of individual caps.
The fucose content of the corneal KS was addressed in three ways. First, the carbohydrate analysis of the released KS chains showed 1.45 fucose per triple-mannose. As discussed below, this is an overestimate of fucose (perhaps by 30%) because of the underestimation of mannose. Second, the 1 H NMR spectrum (Fig. 14) of the released KS chains shows a 10:1 ratio of ␣(1-6)-to ␣(1-3)-fucose, and as a KS chain has maximally one ␣(1-6)-fucose in its linkage region (such oligosaccharides have not been studied here), it is clear that the ␣(1-3)-fucose content cannot be higher than one/10 chains. Third, in this study two fucose-containing oligosaccharides, Gal␤1-4Glc-NAc(S)␤1-3Gal␤1-4(Fuc␣1-3)GlcNAc(S)-ol and Gal␤1-4-(Fuc␣1-3)GlcNAc(S)␤1-3Gal␤1-4GlcNAc(S)-ol (33,34), have been identified (data not shown) as contaminants in oligosaccharides 6 and 7, respectively. The small amount of ␣(1-3)fucose is distributed among several fragments (33,34) and is, therefore, difficult to study by keratanase II analysis, but it is clear from the above structures that it occurs in the repeat regions. In comparison with bovine articular cartilage, KS-IIA, the ␣(1-3)-fucose content of this bovine corneal KS preparation is, perhaps 20 -30 times lower on a weight basis.
Several pieces of evidence contribute to our understanding of corneal KS molecular size and antennary structure. The molar ratio of the recovered keratanase II-derived capping residues to the repeating disaccharides may be calculated from the data presented in Table I to be 1:11. Thus, there is one cap for every 11 disaccharide repeating sequences, equivalent to a molecular weight of 5,500 (an average molecular weight of mono-and disulfated disaccharides is about 500). The molecular size of KS was estimated by two methods. Gel-permeation chromatography showed that the KS chains had a weight average, M w , of 14,000, corresponding to a number average, M n , of about 10,000. These values are probably an underestimate because the column was calibrated with linear keratanase-derived oligosaccharides and not bi-antennary structures. Sugar compositional analysis showed that there are 32 galactose residues for each triple mannose. This permits an estimated molecular size of 17,000 (assuming 32 disaccharides, a linkage region, plus capping residues). This is probably an overestimate because the recovery of mannose is much lower than that of other sugars in the acid hydrolysis (47), i.e. the ratio of galactose to mannose is overestimated as is the molecular size based upon the number of galactoses. Therefore, the number-average molecular weight of the entire KS chain population lies between 10,000 and 17,000.
The smallest possible linkage region, which is not cleavable by keratanase II, contains one fucose, one chitobiose, three mannoses, and two lactosamine disaccharides, and has a molecular weight of about 1,800. If the two estimated M r values of 10,000 and 17,000 are used for the KS chains, then the remaining repeat and capping regions would be expected to have masses of between 8200 and 15200. Dividing these by 5500 yields between 1.5 and 2.7 caps (cleavable by keratanase II) per average KS chain. This is broadly consistent with a bi-antennary model (25,26). However, it must be emphasized that KS chains from different corneal KSPGs may have discrete structures, and thus no single generalized model of corneal KS may be appropriate.
This study has characterized a large number of, hitherto, unrecognized capping structures in corneal KS. However, further studies will be required to identify how the caps are distributed between the three known KSPGs and their antennae, and whether their relative contents vary with age and animal species. It is clearly a priority to purify the individual KSPGs and characterize their keratan sulfates either by spectroscopic or fingerprinting methods (48).