Porcine Dentin Sialoprotein Is a Proteoglycan with Glycosaminoglycan Chains Containing Chondroitin 6-Sulfate*

Dentin sialoprotein (DSP) is a glycoprotein that is critical for proper tooth dentin formation, but little is known about the nature of its carbohydrate attachments and other post-translational modifications. We have isolated DSP from pig dentin and demonstrate that it is a proteoglycan. Polyclonal antibodies were raised in chicken against recombinant pig DSP, and used to identify native DSP in fractions of tooth dentin proteins extracted from developing pig molars. Amino acid analyses and characterization of lysylendopeptidase cleavage products confirmed that the purified protein was DSP, and that Arg391 is at the DSP C terminus. On SDS-PAGE and on urea gels, DSP appeared as a smear extending from 280 to 100 kDa, but in the presence of β-mercaptoethanol the top of the DSP smear disappeared. The high molecular weight material was likely comprised of covalent DSP dimers connected by a disulfide bridge at Cys205. Oligosaccharides were released from DSP following N- and O-linked glycosidase digestions, but these digestions had little effect on the apparent molecular weight of DSP on SDS-PAGE, when compared with the significant reduction following chondroitinase ABC digestion. Glycosaminoglycanases with assorted glycosaminoglycan (GAG) cleavage specificities coupled with Western analyses of the cleaved GAG “stubs” demonstrated that the DSP GAG attachments contain chondroitin 6-sulfate, but not keratan sulfate, heparan sulfate, chondroitin, or chondroitin 4-sulfate. DSP binds biotin-labeled hyaluronic acid, and such binding is inhibited by the addition of unlabeled hyaluronic acid. We conclude that DSP is a proteoglycan and that GAG attachments are the predominant structural feature of porcine DSP.

The dentin sialophosphoprotein gene (DSPP) 1 on human chromosome 4q21.3 encodes the two major noncollagenous pro-teins in tooth dentin: dentin sialoprotein (DSP) and dentin phosphoprotein (DPP) (1,2), and these proteins are critical for proper dentin formation. As of this writing, eight mutations have been identified in the human DSPP gene that cause hereditary dentin defects (3)(4)(5)(6)(7). In each case, the phenotype is inherited in an autosomal dominant pattern and is usually limited to the teeth. Occasionally the condition is associated with progressive neurosensory high frequency hearing loss (DFNA39/DGI1, MIM 605594). In addition, DSPP-null mice developed dentin defects in the absence of other symptoms (8).
Whereas it is known that DSPP gene products are essential for normal dentin biomineralization, DSP structural features are only partially characterized, and its functions are unknown.
Much of our current understanding of the structure and function of dentin sialoprotein comes from studies of the rat protein (9 -11). Rat DSP is the N-terminal portion of DSPP and is generated by proteolytic cleavages, primarily after Tyr 421 but also following His 406 (12). The molecular mass of the major rat DSP component, which contains 29.6% carbohydrate, was determined by sedimentation equilibrium analysis to be 52.57 kDa (13), although the glycoprotein has an apparent molecular mass of 95 kDa on SDS-PAGE (14). Most recently, higher molecular mass forms of rat DSP (designated HMW-DSP) have been detected. Rat HMW-DSP has 10.3 phosphates per molecule whereas DSP has 6.2. The majority of the carbohydrates attached to the protein backbone are by N-linked glycosylations (15). It has been proposed that the rat HMW-DSP may contain GAG chains and/or considerable amounts of sulfate, but these ideas have not been tested experimentally (16).
DSPP is expressed primarily during dentin formation (17,18), and in bone (19 -21). The extracellular matrices of dentin and bone share a similar inventory of noncollagenous proteins, but show differences in the relative quantities of specific protein components and in the way they are processed by proteases. DSP and DPP are the only extracellular constituents that appear to be more abundant in dentin than in bone (22). DSPP mRNA has also been detected in the inner ear. The finding of progressive neurosensory hearing loss in some DGI patients suggests that DSPP gene products may play an important role in processes not yet identified (5), and these processes may or may not involve biomineralization. Rat DSP is not an effective nucleator of hydroxyapatite (HA) formation in vitro, and has low affinity for seed crystals (23), suggesting that DSP is unlikely to regulate mineralization directly. Given the fact that the function of dentin sialophosphoprotein is currently unknown, isolating DSP from other organisms besides rat and characterizing its structural properties might highlight conserved structural features and provide insights into its potential functions.
Previously we cloned and characterized a cDNA encoding porcine DSP (24). This provided us with the entire porcine DSP-derived amino acid sequence, and the mRNA code needed for the expression of recombinant protein. Here we report the expression of recombinant porcine DSP in bacteria and the generation of a polyclonal antibody (Ab) able to specifically detect DSP in dentin extracts. We have used this Ab to devise methods to isolate native DSP from developing pig teeth. Finally, we have performed a number of procedures to characterize the post-translational modifications of porcine DSP, and prove that porcine DSP is proteoglycan, with chondroitin 6-sulfate being the predominant glycan attachment. We provide evidence that DSP forms covalent dimers, connected by a disulfide bridge. We have identified the proteolytic cleavage site that releases DSP from the parent protein and show that porcine DSP can bind hyaluronic acid.

EXPERIMENTAL PROCEDURES
All experimental procedures involving the use of animals were reviewed and approved by the Institutional Animal Care and Use Program at the University of Michigan.
Expression of DSP in Bacteria Using the pGEX4T-1 Vector-The pig DSP cDNA (24) was amplified by PCR using the primers 5Ј-AGTC-GACTCATTCCAGTTCCTCAAATCAAG (forward) and 5Ј-AGCGGC-CGCTTCATAAACTAACCTTATCTT (reverse), which introduced unique SalI and NotI restriction sites at the 5Ј-and 3Ј-ends of the DSP coding region. The amplification product was purified from an agarose gel, ligated into the TA cloning vector (pCR2.1-TOPO, Invitrogen, Carlsbad, CA), and transfected into bacteria. The recombinant plasmid was isolated and confirmed by DNA sequencing. The pig DSP coding sequence was excised with SalI and NotI, and subcloned into the pGEX4T-1 vector (Amersham Biosciences) restricted with the same enzymes. The expression construct was transformed into E. coli strain BL21(DE3) (Stratagene, La Jolla, CA). The bacteria were cultured in NZCYM/Amp medium, and DSP expression was induced by the addition of isopropyl-␤-D-thiogalatopyranoside to a final concentration of 1.0 mM. The bacteria were cultured for 5 additional hours, and the cell pellet was collected by centrifugation. The cell pellet was resuspended in BugBuster reagent (Novagen, Madison, WI) and Protease Inhibitor Mixture Set II (Calbiochem, San Diego, CA) using 5 ml of reagent and 120 l of mixture per gram of wet cell paste at room temperature. The resupension was mixed with 25 units of benzonase per ml of BugBuster regent and was incubated for 20 min at the room temperature. Insoluble cell material was removed by centrifugation, and the supernatant was directly mixed into GST-Bind Resin, which was equilibrated with GST Bind/Wash buffer.
Purification of Recombinant DSP-The GST-Bind Resin was suspended and transferred to a column. The column was washed with 10 volumes of GST Bind/Wash buffer, and the bound protein (i.e. GST-DSP) was eluted with 3 volumes of GST Elute buffer. The eluate was dialyzed against the distilled water and lyophilized. The lyophilized sample (14.5 mg) was dissolved in 5 ml of 20 mM Tris-HCl, 150 mM NaCl, 2.5 mM CaCl 2 buffer (pH 8.4) and was incubated with 15 units of thrombin (Novagen) for 16 h at room temperature. The reaction mixture was fractionated on a TSK-gel phenyl-5PW RP-HPLC column (TOSOH, Tokyo, Jpn) using a flow rate of 1 ml/min of 0 -100% buffer B over 40 min (buffer A; 50 mM ammonia water, and buffer B; Aϩ80% acetonitrile). The recombinant DSP eluted in the third peak, was lyophilized, and stored at Ϫ80°C.
Production of DSP Antibody-Porcine recombinant DSP antibody was produced by Lampire Biological Laboratories (Pipersville, PA). The purified recombinant DSP (1.5 mg) was emulsified with Freund's complete adjuvant (first injection) or incomplete adjuvant (booster injections) and used to immunize a chicken. After two injections, eggs were collected and specific IgY antibodies were purified by the affinity chromatography. The titer of this purified Ab, as estimated by enzymelinked immunosorbent assay (ELISA), was 1:17,000. This polyclonal antibody was used in subsequent Western blot analyses to identify DSP-containing fractions from developing porcine molars.
Teeth Preparation-Tooth germs of permanent molars were surgically extracted with a hammer and chisel from the maxillae and mandibles of 16 six-month-old pigs, within minutes of each animal's termination at the Michigan State University Meat Laboratory (East Lansing, MI). The soft tissue was removed, and the hard tissue was quickly frozen in dry ice. All subsequent preparation steps were carried out at 4°C or on ice or on dry ice. Dental enamel was removed by scraping 55 molars with a curette. Each molar crown was separately embedded in paraffin with the dentin side facing up.
Extraction of Porcine DSP from Developing Pig Teeth-All extraction steps were carried out at 4°C or on ice and Protease Inhibitor Mixture Set III (1 mM AEBSF, 0.8 M aprotinin, 50 M bestatin, 15 M E-64, 20 M leupeptin, and 10 M pepstatin) (Calbiochem) and 1 mM 1,10phenanthroline (Sigma) were added into the buffer during the extraction. Dentin was first washed with Milli-Q water for 1 day. Dentin protein "G extract" was obtained by sequential extraction with 50 mM Tris-HCl/4 M guanidine buffer (pH 7.4) for each of 5 days and then combined. Then "E extract" was obtained by sequential extraction with 50 mM Tris-HCl/4 M guanidine/0.5 M EDTA buffer (pH 7.4) for 7 days, changing the buffer everyday, and then combined. The G and E extracts were centrifuged at 7,000 ϫ g, and then desalted and concentrated using the combination of a Spectra/Por 3 membrane (Spectrum Laboratories Inc., Rancho Dominguez, CA) with a YM-3 membrane (Amicon, Beverly, MA), and were lyophilized and stored at Ϫ80°C.
Isolation and Purification of Porcine DSP-The lyophilized E extract (100 mg) was dissolved in 10 ml of 50 mM Tris-HCl/6 M urea buffer (pH 7.4), and fractionated by an anion exchange chromatography with a Q-Sepharose Fast Flow column (1.6 cm ϫ 20 cm, Amersham Biosciences). The column was equilibrated with 50 mM Tris-HCl, 6 M urea buffer (pH 7.4) and eluted with a linear NaCl gradient (0 -0.5 M) containing 50 mM Tris-HCl, 6 M urea buffer (pH 7.4) at a flow rate of 0.2 ml/minute at 4°C. Protein was detected by monitoring the absorbance at 280 nm. Both DSP and DPP eluted in the third peak, which was concentrated with a YM-3 membrane, and further fractionated by hydroxyapatite affinity HPLC using a Beckman System Gold HPLC system equipped with a Bio-Scale CHT5 Ceramic Hydroxyapatite, Type I column (column volume, 5 ml) (Bio-Rad). The column was equilibrated with 10 mM sodium phosphate buffer (pH 6.8) and eluted with a linear gradient of 1.0 M sodium phosphate buffer (pH 6.8) at flow rate of 1.5 ml/minute at the room temperature. DSP eluted in the second peak, which was concentrated and separated by size exclusion chromatography with a Superdex 200 column (1.6 cm ϫ 60 cm, Amersham Biosciences). The column was equilibrated with 50 mM Tris-HCl/4 M guanidine buffer (pH 7.4) at a flow rate of 0.2 ml/minute at 4°C. DSP eluted in the first peak, was dialyzed and finally purified by reversed phase high performance liquid chromatography (RP-HPLC) using the Beckman System Gold HPLC system equipped with a TSK-gel phenyl 5PW-RP column (4.6 mm ID ϫ 7.5 cm, TOSOH). The column was equilibrated with 0.05% ammonia and eluted with linear acetonitrile gradient (5-95%) containing 0.05% ammonia at a flow rate of 1.0 ml/minute at room temperature. DSP eluted in the first and second peaks, was lyophilized and stored at Ϫ80°C.
SDS-PAGE-SDS-PAGE was performed using PAGEr® Gold Precast Gels (4 -20% gradient, 10% or 7.5% Tris-glycine gels) (Cambrex Bio Science Rockland Inc., Rockland, ME). SDS-PAGE-urea gels contained 10% polyacrylamide and 8 M urea. Samples were dissolved in Laemmli sample buffer (Bio-Rad) with or without 2% ␤-mercaptoethanol and electrophoresis was carried out using a current of 20 mA for 1.5 h. The gels were stained with Bio-Safe Coomassie (Bio-Rad), Silver Stain Plus (Bio-Rad), or Stains-all (Sigma). The apparent molecular masses of protein bands were estimated by comparison with SeeBlue® Plus2 Pre-Stained standards (Invitrogen).
Western Immunoblots-The proteins on electrophoresed gels were electrotransferred onto Hybond-P membrane (Amersham Biosciences). Polyclonal for porcine DSP and monoclonal for each of glycosaminoglycan antibodies were used at dilution of 1:50,000. In the Far-Western analysis of hyaluronic acid binding, biotin-labeled hyaluronic acid (1: 1000) was detected using a horseradish peroxidase-streptavidin conjugate (1:5000, Sigma). The membranes were immunostained by the chemiluminescent detection with ECL Advance TM Western blotting detection kit (Amersham Biosciences).  period, three volumes of ice-cold ethanol were added to the reaction mixture, which was then centrifuged for 10 min at 10,000 ϫ g. Both supernatant and pellet were evaporated to dryness, and the dried samples were stored at Ϫ80°C. An aliquot (20 g) of the pellet was electrophoresed on 4 -20% Tris-glycine gradient gel, and the degradation pattern was confirmed by Stains-all staining and Western immunoblots using the polyclonal anti-DSP antibody.

Release of O-linked Oligosaccharide Chains by O-Glycosidase Digestion-
The pellet (ϳ0.2 mg) after the glycopeptidase A digestion in 250 mM sodium phosphate buffer (pH 5.0) was incubated with 5 milliunits of sialidase Au, 0.5 milliunits of ␣-(1,3,4)-fucosidase and 2.5 milliunits of O-glycosidase (QA-bio, San Mateo, CA) at 37°C for 20 h. At the end of the incubation period the reaction was treated as described for the glycopeptidase A digestion.
The fluorescence labeling for both N-and O-linked oligosaccharides after glycopeptidase A and O-glycanase digestions was previously described (25). The sample was applied to a column of Sephadex G-15 (1.5 ϫ 35 cm, Amersham Biosciences) equilibrated with 10 mM ammonium acetate buffer (pH 6.0) and was eluted with the same buffer. The effluent was continuously monitored by a fluorescence monitor (FP-2020, JASCO, Tokyo, Jpn) using an excitation wavelength of 315 nm and an emission wavelength of 400 nm.
Glycosaminoglycanase Digestion of DSP-Four glycosaminoglycanases (protease-free chondroitinase ABC, heparitinase I, keratanase, and keratanase II) and six monoclonal antibodies raised against dermatan sulfate (DS), keratan sulfate (KS), and the stubs formed by partial digestion of chondroitin (⌬Ch), chondroitin 4-sulfate (⌬Ch-4S), chondroitin 6-sulfate (⌬Ch-6S), and heparan sulfate (⌬HS) glycosaminoglycan chains were purchased from Seikagaku America. Each glycosaminoglycanase digestion of glycopeptidase A and O-glycanase-treated DSP (ϳ50 g) was separately carried out with 0.2 unit of protease-free chondroitinase ABC in 0.1 ml of 40 mM Tris-HCl/40 mM sodium acetate buffer (pH 8.0), 0.005 units of the heparitinase I in 50 mM sodium acetate buffer (pH 7.0) containing 5 mM CaCl 2 , 0.05 units of the keratanase in 10 mM Tris-HCl buffer (pH 7.4) or 0.001 units of the keratanase II in 10 mM sodium acetate buffer (pH 6.0) at 37°C for 24 h, respectively. At the end of each glycosaminoglycanase digestion, an aliquot (10 g) of the digest was electrophoresed on a 4 -20% Trisglycine gradient gel, and the degradation patterns were confirmed by Stains-all staining and/or Western immunoblots using the polyclonal anti-DSP Ab and each of monoclonal anti-proteoglycan antibodies.
Digestion of Chondroitinase ABC-treated DSP by Endoproteinase LysC-Digestion of the chondroitinase ABC-treated DSP (0.2 mg) was carried out in 30 mM Tris-HCl buffer (pH 8.8) with an enzyme/substrate ratio of 1:100 (w/w) at 37°C for 18 h. After the reaction, the digests were applied to a Discovery C18 column (10 mm ID ϫ 25 cm, SUPELCO, Bellefonte, PA) and fractionated. The column was equilibrated with 0.05% trifluoroacetic acid and eluted with a linear acetonitrile gradient (5-95%) containing 0.05% trifluoroacetic acid at a flow rate of 1.0 ml/minute.
Amino Acid Analysis-The purified DSP samples (0.2 mg) were hydrolyzed with 6 N HCl at 115°C for 16 h. The amino acid analyses were performed using a Beckman Model 7300 automatic amino acid analyzer.
Hyaluronic Acid (HyA) Binding Assay-HyA binding assay determined the amount of biotin-labeled hyaluronic acid (bHyA) binding to DSP protein immobilized on a microtiter plate in the presence of competing unlabeled HyA. All dilutions, incubations, and washes for the HyA binding assay were performed in 10 mM sodium acetate, 150 mM NaCl, 0.05% (v/v) Tween 20 (pH 5.8) at room temperature unless otherwise stated. Wells of plastic microtiter plates (Linbro/Titertek, ICN FLOW, Bucks, UK) were coated overnight either with 200 l per well of protein solution in 20 mM carbonate buffer (pH 9.6) or with buffer alone. The coating solution was removed and the plates were washed three times. The plates were blocked by incubation for 90 min at 37°C with 5% (w/v) nonfat dry milk (Bio-Rad). The plates were washed three times and 200 l of 1:1000 diluted bHyA (Sigma), which either contained no HyA or various amounts of human umbilical cord HyA (39 ng to 10 g) (Sigma), was added to the wells and incubated overnight at room temperature. The plates were washed three times and 200 l of 1:1000 diluted alkaline phosphatase conjugated ExtrAvidin (Sigma) was added and incubated for 1 h at room temperature, followed by three washes. A 10 mg/ml solution of disodium p-nitrophenylphosphate in 100 mM Tris-HCl, 100 mM NaCl, 5 mM MgCl 2 (pH 9.3)(200 l) was added to all wells, incubated for 15 min, and the reaction terminated by the addition of 50 l of 0.2 M NaOH. The absorbance at 405 nm was determined, and all absorbances were corrected against blank wells, which contained substrate solution alone.
Optimum amounts of RP2, rpDSP or recombinant pig amelogenin (rP172) used in this assay were determined by coating with a range of protein concentrations (5-800 pmol per well) and binding bHyA in the absence of competitor. On this basis, 600 pmol per well of RP2, 250 pmol per well of rpDSP or 800 pmol per well of rP172 were used in the competition assay.
Automated Edman Degradation-Automated Edman degradation was performed with an Applied Biosystems Procise 494 cLC protein sequencer. The purified peptide samples were suspended in 20 l of 100% trifluoroacetic acid and loaded 100% on the precycled cartridge filter.

RESULTS
Expression of Recombinant Pig DSP-The first goal in the characterization of porcine dentin sialoprotein was to isolate the protein from developing teeth. Polyclonal antibodies were raised against recombinant pig DSP and used to identify DSP in the various dentin fractions. Recombinant porcine DSP was expressed as a GST-fusion protein in bacteria (Fig. 1). The intact GST-DSP fusion protein was purified by HPLC and affinity chromatography (Fig. 1A). Recombinant DSP was excised from the GST domain by thrombin and purified (Fig. 1, B and C). The final recombinant porcine DSP product had an apparent molecular mass of 65 kDa (Fig. 1B, lane 2). The identity of the purified protein was confirmed by N-terminal sequencing, which gave the sequence GSPEFPGRLIPVPQI (the DSP N-terminal sequence is underlined), as expected. 2 mg of recombinant pig DSP were used to generate specific DSP polyclonal antibodies in chicken.
Isolation of Porcine DSP from Developing Molars-Dentin proteins were extracted from developing porcine molars ( Fig.  2A) using sequential guanidine (G extract) and guanidine/ EDTA (E extract) extractions (26). DSP was identified as a wide band above 250 kDa by Western blot analyses in the E extract (Fig. 2B, lane E). The E extract was first fractionated by anion exchange chromatography using a Q-Sepharose column, which sorted the E extract into 3 well-defined chromatographic peaks (Fig. 2C). DSP was identified in the third peak (Fig. 2D,  lane Q3), the contents of which were further fractionated over an HA column. The HA column divided the Q3 component into 3 parts (Fig. 3B, left). After passing the HA2 fraction over with a protein G column to remove immunoglobulins, the three HA fractions were analyzed by SDS-PAGE and immunostained with the DSP Ab (Fig. 3A). Two major DSP components were detected on the Western blot at ϳ280 and 140 kDa. We associated the appearance of the lower DSP band at 140 kDa with the removal of immunoglobulins, which may have blocked immunodetection of this band in previous fractions. This band was not prominent in the previous fractions (compare lanes E and Q3 in Fig. 2, B and C with lane HA2 in Fig. 3A). The smaller immunostained bands, having apparent molecular masses in the 25-30 kDa range (Fig. 3A, HA2), also immuno-stained in negative controls in which only the secondary antibody was included (Fig. 3D, S2-S3). Fraction HA2 was further separated by size exclusion chromatography (Fig. 3, C and D) into four parts. The DSP components were in the first fraction (S1). Several Stains-all-positive proteins, which were not detected by the DSP Ab, were now sufficiently concentrated by the fractionation procedures to be detected in each of the four size exclusion fractions. Stains-all is a cationic carbocyanine dye that stains most proteins red or pink, while Ca 2ϩ -binding proteins stain dark blue or purple (27).
Porcine DSP Is a Covalent Dimer-The two large DSP components in S1 were partially separated using a TSK-gel phenyl FIG. 2. Sequential extraction of pig proteins from dentin and anion exchange chromatography. Panel A, hard tissue of a developing pig molar embedded in paraffin. Dentin proteins were extracted by filling the pulp chamber with deionized water (W), guanidine (G), and guanidine/EDTA (E). Following dialysis with water and centrifugation, the guanidine extract was divided into supernatant (GS) and pellet (GP). Panel B, analysis of the W, GS, GP, and E extracts by 4 -18% gradient SDS-PAGE stained with CBB, Stains-all, or immunostained using the DSP polyclonal antibody (DSP Ab). The EDTA extract contained DSP and was separated into 3 fractions (Q1-Q3) by HPLC using a Q-Sepharose column. Panel C shows the chromatogram from this column. Following dialysis with water and centrifugation, the Q1 fraction was divided into supernatant (Q1S) and pellet (Q1P). Panel D, analysis of the Q1S, Q1P, Q2 and Q3 fractions by 4 -18% gradient SDS-PAGE stained with CBB, Stains-all, and the DSP Ab. DSP was detected as an immunopositive band at 280 kDa in fraction Q3. 5PW RP-HPLC column (Fig. 4A). As DSP was concentrated in the final fractionation, it became detectable by Stains-all. The first peak to elute (RP1, 2.8 mg), when analyzed under nonreducing conditions (minus ␤-mercaptoethanol); contained predominantly a smear between 100 and 150 kDa and a band at 280 kDa (Fig. 4B, Ϫ lanes RP1), while RP2 (1.6 mg) contained predominantly a smear between 130 and 180 kDa, and a band at 280 kDa (Fig. 4B, Ϫ lanes RP2). These two fractions (RP1 and RP2) were reduced with ␤-mercaptoethanol and analyzed by SDS-PAGE and by Western blotting (Fig. 4B, plus lanes). The 280 kDa DSP could not be detected in either of the reduced fractions. The amino acid compositions of RP1 and RP2 were similar to each other and to the predicted composition of DSP from pig and other mammals, suggesting that the fractions were comprised mostly of DSP protein (Table I). When DSP was denatured with urea, DSP appeared as a smear extending from 280 to 100 kDa. When both ␤-mercaptoethanol and urea were added, and the samples were run on a urea gel, the top of the smear disappeared (Fig. 4B, right). We interpret these findings to indicate that the higher molecular mass material contains DSP dimers covalently linked by a disulfide bridge. As porcine DSP has only one cysteine residue per molecule, the two DSP monomers are necessarily joined at Cys 205 .
Porcine DSP Is a Proteoglycan-The nature of DSP carbohydrate attachments was investigated. The RP1 and RP2 frac-tions were separately digested with N-glycosidase. This digestion did not produce a noticeable difference in the DSP profile on SDS-PAGE and on Western blots (Fig. 5A, lanes 1 versus 2); however, carbohydrate chains were released from both DSP fractions and could be detected following fluorescent labeling (Fig. 5B, left). The N-linked glycosylations released from RP1 differed from those released by RP2 (compare Fig. 5B, left top to left bottom). We interpret these results to indicate that porcine DSP has N-linked glycosylations, but that these glycosylations do not explain the substantial discrepancy between the DSP apparent molecular mass (100 -280 kDa), and the apparent molecular mass of the recombinant protein (65 kDa), or the predicted molecular mass of the unmodified protein (39 kDa).
Following N-linked glycosidase digestion, the RP1 and RP2 DSP fractions were separately digested with enzymes to remove fucose and sialic acid (which can interfere with cleavage by O-glycosidase) and with O-glycosidase, to remove O-linked oligosaccharides. As in the case of the N-linked glycosylations, no differences were observed in the apparent mobility of DSP on SDS-PAGE and on Western blots (Fig. 5A, lanes 2 versus 3). Carbohydrate chains released by O-glycosidase digestion from both DSP fractions could be detected following fluorescent labeling, but the amount was less than the amount released by N-glycosidase digestion (Fig. 5B, right). We interpret these results to indicate that porcine DSP contains O-linked carbo- hydrate, but less than the amount of N-linked carbohydrate.
Following the N-and O-linked glycosidase digestions, the RP1 and RP2 DSP fractions were separately digested with chondroitinase ABC. This enzyme partially removes glycosaminoglycan chains and had a profound effect on the apparent mobility of DSP on SDS-PAGE and Western blots (Fig. 5A,  lanes 3 versus 4). When visualized by immunostaining with an antibody specific for the cut ends (or "stubs") left behind by cleavage of chondroitin 6-sulfate, the apparent molecular masses of the two DSP smears from the RP1 fraction centered around 30 and 50 kDa, which was smaller than the corresponding smears, centered around 40 and 65 kDa, from the RP2 fraction digest (compare Fig. 5A, lanes 4 for RP1 and RP2).
Antibodies specific for dermatan sulfate and keratan sulfate, and for the stubs left behind by cleavage of chondroitin, chondroitin 4-sulfate, and heparan sulfate were negative, but gave positive results for controls (Fig. 5C). We interpret these results to mean that DSP is a proteoglycan with a large and variable amount of chondroitin 6-sulfate attachments, and that such attachments constitute a major structural feature of porcine DSP.
Identification of the DSP C Terminus-Endoproteinase Lys-C digestion of pig RP2 generated DSP cleavage products that could be resolved into 10 chromatographic peaks by C-18 RP-HPLC (Fig. 7A). These cleavage products were characterized by N-terminal sequencing (Fig. 7B). The N terminus of seven different digestion products were characterized. Only DSP sequences were observed. The locations of the digestion products in the porcine DSP primary sequence are underlined in Fig. 7C. The protein sequence obtained from peak 10 (GIEIAAPR) did not end in lysine and the sequence chromatograms showed an abrupt loss of signal, indicating that the end of the peptide was reached. This suggests that Arg 391 is at the DSP C terminus. This conclusion was further supported by the isolation and characterization from pig dentin of a 19 kDa DSPP cleavage product that has Ser 392 at its N terminus (data not shown). After eliminating the 15 amino acid signal peptide, the porcine DSP protein, in the absence of any post-translational modifications, has 376 amino acids, a predicted molecular mass of 39 kDa, and an isoelectric point of 4.4. We were not able to obtain protein sequence for the DSP N terminus or for any DSP peptide containing the predicted N terminus. The DSP sequence for the peptide in peak 9 (NEDVGXAXQ) gave blank cycles for Asn 170 and Ser 172 , suggesting that these residues are likely to be modified, by glycosylation and phosphorylation, respectively.

DISCUSSION
Proteoglycans have been proposed to play important roles in the mineralization of tooth dentin. It is believed that the predominant proteoglycans in predentin and dentin are chondroitin sulfate-rich decorin and biglycan and keratan sulfate-rich lumican and fibromodulin, but other proteoglycans, such as versican are also present (29). In a situation analogous to the relationship between osteoblasts, osteoid and bone, the surface of the dentin forming cell (odontoblast) is separated from the mineralized extracellular matrix by an unmineralized layer of predentin. The profile of proteoglycans changes during the transition of predentin to dentin, supporting the idea that some proteoglycans inhibit, whereas others promote and guide the mineralization process (30). Based upon the calcium-and hydroxyapatite-binding properties of glucuronic acid-rich and iduronic acid-rich glycosaminoglycans and proteoglycans (31), it has been proposed that proteoglycans containing dermatan sulfate within predentin act as inhibitors of mineralization and that highly sulfated ChS proteoglycans function in the control of mineralization (30).
Among the noncollagenous proteins, DPP is considered to be the most abundant, and DSP the second most abundant macromolecules in the dentin matrix (16). DSP and DPP, however, are synthesized as a single protein (DSPP), which is proteolytically processed to form the molecular species resident in the dentin matrix. DSP-only transcripts have been reported for rat (32) and pig (24), but no full-length mRNA transcripts encoding only the DPP portion of the protein have ever been identified. In the absence of selective proteolytic degradation of DSP, one might expect that DSP should be the most abundant noncollagenous protein and DPP should be expressed in almost equal amounts. In the current study we saw no evidence opposing this expectation, although we also did not confirm it. DPP is easier to isolate and concentrates into discrete bands on SDS-PAGE that are readily detected by Stains-all. DSP is difficult to isolate and forms a smear on SDS-PAGE that does not readily stain with CBB or Stains-all. In the partial characterization of DSP lysylendopeptidase cleavage products, we did not detect the specific C-terminal cleavage product (VSL) that would have been present on DSP protein expressed from the DSP-only transcript.
Porcine DSP is a heterogeneous proteoglycan that was separable into two fractions, that appeared as wide bands on Western analyses migrating at roughly 280 and 140 kDa, respectively, and on urea gels as a smear between 280 and 100 kDa. Reduction with ␤-mercaptoethanol caused the material at 280 kDa to disappear, suggesting that the 280-kDa band was comprised of DSP covalent dimers connected at the lone cysteine, at position 205. The DSP dimers appear to be in low abundance compared with the DSP monomers, and reduction of the disulfide bond did not contribute perceptibly to the intensity of monomer. Cys 205 is conserved in human DSP, but not in rodents. Mouse and rat DSP also have one cysteine per molecule, but this cysteine is located closer to the DSP C terminus.
The two final DSP fractions (RP1 and RP2) show porcine DSP to be a concentrated smear of proteoglycan on SDS-PAGE that stains weakly with Stains-all. Both fractions have amino acid compositions that match what is predicted for DSP (Table  I). The distribution of the DSP smears on SDS-PAGE is not altered perceptibly following successive cleavages by N-and O-glycosidases, although DSP has eight potential N-linked glycosylation sites (Asn 52 ,Asn 82 , Asn 92 , Asn 151 , Asn 170 , Asn 176 , Asn 191 , Asn 288 , Asn 330 ), and three potential O-linked glycosylation sites (Thr 199 , Thr 213 , and Thr 264 ). Some carbohydrate was released by both digestions. A major reduction was achieved, however by chondroitinase ABC. Furthermore, antibody analyses demonstrated that all of the detectable glycosaminoglycans were chondroitin 6-sulfate. This demonstrates for the first time that DSP is a proteoglycan. Xylosyltransferase, the enzyme that catalyzes the initial step of GAG glycosylation, targets serine followed by glycine residues (33). There are four such sites in porcine DSP (Ser 217 , Ser 220 , Ser 238 , and Ser 250 ), which are shown in Fig. 8A. Ser 238 and Ser 250 , but not Ser 217 -Ser 220 , are in a highly conserved region of the DSP protein and are most likely to be the glycan attachment sites (Fig. 8B).
The finding that porcine DSP has major chondroitin 6-sulfate attachments supports a re-evaluation of its potential roles in dentin biomineralization. The functions of proteoglycans are often closely coupled to the roles of their glycosaminoglycan chains. Glycosaminoglycans (GAGs) are highly extended, unbranched, negatively charged repeating disaccharide units that increase the viscosity and structural integrity of extracellular matrices. Some proteoglycans, such as aggrecan, interact with hyaluronic acid and a link protein to form large aggregates. The glycosaminoglycan chains of such aggregates generate a large osmotic swelling pressure and increase resistance to compressive forces.
The DSP chondroitin 6-sulfate chains might function to bind and organize water molecules and contribute to the hydration of dentin, as aggrecan does for cartilage (34). The hydration of dentin is important clinically. Devital or endodontically treated teeth dry out over time and become brittle, requiring the placement of crown-type restorations to prevent subsequent tooth fractures. Another potential function of the DSP chondroitin 6-sulfate chains could be to mediate specific binding to other macromolecules. The negative charges of the DSP chondroitin sulfate chains would be expected to repel negatively charged proteins, such as DPP. Interactions, both positive and negative, with other proteins could play a role in matrix assembly. In addition, the binding of signaling molecules such as growth factors to DSP GAG chains might give it a role in tooth development. The binding of heparan sulfate chains to growth factors is well characterized (35). Small leucine rich proteoglycans bind collagen, TGF␤, and other macromolecules (36). Similar properties are likely to be true of DSP. Recently it has been shown that DSP accumulates, presumably during mantle dentin formation, in the border zone between root dentin and cellular cementum and may participate in the induction of periodontal ligament cells (37). In the developing teeth of the DSPP knockout mouse, the levels of biglycan and decorin were increased in the widened predentin zone (8).
We have found that DSP isolated from the forming dentin of developing pig teeth binds hyaluronic acid in vitro, even in the absence of a link protein. This finding suggests that DSP might contribute to large molecular assemblies built upon a backbone of hyaluronic acid. The molecular basis for the in vitro binding of DSP to hyaluronic acid cannot be predicted from the DSP primary sequence. Some extracellular hyaluronan-binding proteins contain a link module (ϳ100 amino acids), which is involved in hyaluronic acid binding (38). Others contain a BX7B cluster of amino acids (two basic residues B separated by 7 non-acidic residues X) (39). Porcine DSP contains neither a link module nor a BX7B domain. A protein-protein BLAST search of the nonredundant protein databases hits DSP sequences from other organisms without matching any known proteoglycans, let alone proteoglycans known to bind hyaluronic acid. Despite this, there can be no doubt that porcine DSP is a proteoglycan, although its binding to hyaluronic acid in vivo remains to be demonstrated. We suspect that the lack of amino acid sequence homology between DSP and other proteoglycans discouraged the pursuit of biochemical analyses that would have previously demonstrated the presence of glycosaminoglycan chains in DSP isolated from other organisms.
The isolation of porcine DSP and the finding of major structural features not previously characterized will facilitate future studies to more fully characterize the DSP structure, determine its relative abundance to DPP in dentin, and to explore the function of its proteoglycan attachments.