Purification and partial characterization of small proteoglycans I and II, bone sialoproteins I and II, and osteonectin from the mineral compartment of developing human bone.

Using nondegradative isolation procedures we purified and characterized five major noncollagenous proteins from developing human bone. Small bone proteoglycan I, Mr approximately 350,000 on sodium dodecyl sulfate (SDS), 4-20% gradient polyacrylamide gels has a different amino-terminal sequence of NH2-Asp-Glu-Glu-()-Gly-Ala-Asp-Thr and is not cross-reactive with the small bone proteoglycan II, Mr approximately 200,000 on SDS-gradient polyacrylamide gels. Bone proteoglycan II is 95% N terminally blocked and the small amount that can be sequenced has an amino-terminal sequence (NH2-Asp-Glu-Ala-()-Gly-Ile. . .) that is apparently similar but not identical to a small proteoglycan isolated by Brennan, M.J., Oldberg, A., Pierschbacher, M.D., and Ruoslahti, E. (1984) J. Biol. Chem. 259, 13742-13750 from human fetal placenta membrane. Two bone sialoproteins, each of which migrates at a Mr approximately 80,000 on SDS gels, have also been isolated. Bone sialoprotein I has an amino-terminal sequence of NH2-Ile-Pro-Val-Lys-Gln-Ala. . . which is different from that of bone sialoprotein II with an amino-terminal sequence of NH2-Phe-Ser-Met-Lys-Asn-Leu. . . The two bone sialoproteins do not cross-react on Western blot analysis. Human bone osteonectin contains a large number of cysteines, more than 90% of which appear to be in disulfide bonds. The N-terminal amino acid sequence of human bone osteonectin was nearly identical to bovine bone osteonectin and had many similarities to a protein found in mouse parietal endoderm (Mason, I.J., Taylor, A., Williams, J.G., Sage, H., and Hogan, B.L.M. (1986) EMBO J. 5, 1831-1837.

Noncollagenous proteins constitute about 10% of the organic matrix of mammalian bone. While no noncollagenous protein has had its biological role unambiguously assigned, it is generally agreed that the secretion, assembly, maturation, mineralization, and maintenance of the bone collagen matrix may be aided or directed by one or more of these proteins (for review see Refs. [1][2][3]. Several proteins in the mineral compartment, including serum albumin (4) and the a2HS glycoprotein (5) are serum components that bind to the bone hydroxyapatite crystals. Other bone proteins are synthesized by bone-derived cells in culture, including bone proteoglycans I and I1 (6,7), bone sialoprotein I1 (6), osteonectin (7), and osteocalcin, also known as bone gla protein (8).
The mineral compartment of developing bovine bone has * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
" been shown to contain two small chondroitin sulfate proteoglycans as distinguished by SDS-polyacrylamide gel electrophoresis (9). The two bovine bone proteoglycans differed in amino acid composition (proteoglycan I being more rich in leucine) and in apparent lack of cross-reactivity of proteoglycan I with antisera made to proteoglycan I1 (10). It was suggested that proteoglycan I contains two chondroitin sulfate chains and proteoglycan 11, one chain (10). Recent work with limited V-8 protease digestion of proteoglycan 11-like molecules from bovine bone, cartilage, and tendon suggests that while these molecules are closely related, the bone proteoglycan I1 core protein is subtly different from those in the other connective tissues (11). During the mid 1960s Herring and co-workers (12) purified and characterized a sialic acid-rich protein of approximately 25,000 daltons from bovine bone. This 25,000 M , bovine bone sialoprotein has recently been shown to be a degradation product of a larger molecule, bone sialoprotein I1 (13,14). This intact glycoprotein is 70,000-80,000 M , on SDS gels and contains approximately 50% protein, 12% sialic acid, 7% glucosamine, and 6% galactosamine. Bovine bone sialoprotein I1 does not stain with Coomassie unless pretreated with neuraminidase (13) but can be easily stained with Alcian blue or Stains All. A second bone glycoprotein, sialoprotein I, has been identified in bovine bone (2,14).
Osteonectin, a phosphorylated glycoprotein with an apparent molecular weight of 38,000 on SDS gels (15,16), is a dominant (10-15%) noncollagenous protein in the mineral compartment of developing bovine bone (17). Other fetal bovine tissues have been shown by radioimmune assay to contain small quantities of a cross-reactive species, although the levels are only 0.1-0.2% of that found in bone (17). mRNA for osteonectin has also been shown to be present in tendon tissue and in a variety of cultured cells (18). Osteonectin-like products have been shown to be present in serum (17), at least some of which appears to be associated with platelets (19). The function of osteonectin in bone is unknown but it has been shown to initiate the formation of hydroxyapatite onto collagen in vitro (20). Osteonectin binds to collagen fibrils, calcium, and hydroxyapatite, the last with high affinity (16,20). Further, the levels of bone osteonectin in one bovine model of osteogenesis imperfecta (Texas variant) was severely depressed compared to the bone of unaffected siblings (21). Affected individuals of a second bovine model of osteogenesis imperfecta (Australia variant) have clinically identical symptoms to the first model but have normal levels of bone osteonectin (22), suggesting that bone osteonectin content is a biochemical marker that may be useful in distinguishing The abbreviations used are: SDS, sodium dodecyl sulfate; HPLC, high performance liquid chromatography. genetic differences in clinically similar patients with this disease.
In this paper we present conclusive evidence that the two bone proteoglycans are distinctly different from each other as are the two bone sialoproteins. Human bone osteonectin is also partially characterized.

EXPERIMENTAL PROCEDURES
Extraction and Chromatography-Prenatal and neonatal human calvaria were a generous gift of Drs. G. Tschank and H. M. Hanauske-Abel of the Prenatal Diagnosis Laboratory, Gutenberg University Mainz, West Germany and Dr. Stuart Weinstein, University of Iowa. The bone was trimmed free of sutures and soft tissue, cut into small pieces, and processed to a fine powder under liquid nitrogen in a Spex impact mill. For each preparation, 30 g of milled bone was extracted in 4 liters of 4 M guanidine HCI, (Bethesda Research Laboratories (BRL)) 0.05 M Tris (BRL), 0.1 M 6-aminocaproic acid (Sigma), 5 mM benzamidine HCI (Behring Diagnostics), and 1 mM phenylmethylsulfonyl fluoride (BRL), pH 7.4, for 48-72 h at 4 "C as described previously (23). The supernatant was poured off, the insoluble material rinsed two times with 100 ml of fresh buffer, and then stirred for 72 h at 4 "C in a demineralizing buffer that contained all of the above plus 0.5 M tetrasodium EDTA (Sigma). The supernatant was clarified by filtering through Whatman No. 4 paper and then concentrated to a volume of 60 ml by ultrafiltration (Amicon YM-10 membranes). One-third of each preparation was filtered (0.45 pm filter) and chromatographed on tandem Sepharose CL-GB (Pharmacia) columns (2.6 X 190 cm) in 4 M guanidine HCI, 0.05 M Tris, pH 7.4, as described previously (15), except the flow rate was 18 ml/h. The elution profile was monitored at 234 nm, and 19-min fractions were collected. Appropriate fractions from the Sepharose CL-GB columns were pooled, concentrated by ultrafiltration, and the buffer exchanged to fresh 7 M urea, 0.05 M Tris, pH 6.0, by repeated ccrncentration and dilution in the Amicon-stirred cell at 4 'C. The sample was then chromatographed on a freshly prepared DEAE-Sephacel (Pharmacia) column (1.6 X 6 cm) equilibrated in the urea buffer. The sample was loaded and the column washed at 30 ml/h until the base line (226-240 nm) was re-established. A linear 16-h gradient from 0 to 1 M NaCl in the same buffer was used to elute the proteins, and 10-min fractions were collected. All samples in urea were either concentrated by ultrafiltration and rapidly desalted (see below) or stored at -20 "C. Individual fractions from either the Sepharose CL-GB or the DEAE-Sephacel column were often monitored by SDS-polyacrylamide gel electrophoresis (see below) by centrifuging 0.2-1.0 ml of a fraction in a Centricon 10 or 30 microconcentrator (Amicon) and flushing with excess water. The sample was recovered from the membrane using electrophoresis sample buffer containing an excess of sucrose and electrophoresed as described below.
Appropriate samples were chromatographed on a MONO Q HR5/ 5 fast protein liquid chromatography column (MONO Q) (Pharmacia) equilibrated in fresh 7 M urea, 0.05 M Tris, pH 7.4. The sample (up to 5 mg) was loaded onto the column in the same buffer and washed for 5 min at 1 ml/min. A 40-min linear gradient to 1 M NaCl in the same buffer was used to elute the proteins. The elution was monitored at 280 nm, and 0.5-ml fractions were collected. Fractions were pooled, rapidly desalted, and lyophilized.
Desalting-Due to the presence of endogenous proteases that appear to be active during dialysis (24) all samples (up to 50 ml) were rapidly desalted by chromatography on Trisacryl GF 05 (LXB). The column (2.6 x 34 cm) was run at 90 ml/h in 0.1 M ammonium acetate and monitored at either 226 or 280 nm. Proteins (>2500 M,) eluted in the excluded volume were immediately frozen and lyophilized. Because we noted that most commercial sources of ammonium acetate contain significant levels of glycine, samples used for amino acid analysis were desalted in 0.1 M acetic acid brought to a pH of 7.0 with ammonium hydroxide.
Enzymatic Digestion-50-pgdry weight of either bone proteoglycan I or I1 were digested with 10 milliunits of chondroitinase ABC (Miles) in 25 pl of 0.05 M Tris, 0.063 M NaCI, 0.25 mg/ml bovine serum albumin for 1 h at 37 "C (9). The digestion was stopped by boiling for 2 min in an equal volume of SDS sample buffer. Polyacrylamide Gel Electrophoresis and Electrohlotting-Polyacrylamide gradient (4-2096) SDS slab gels (160 X 140 X 1.5 mm) were formed and topped with a 3% stacking gel as described previously (9). In addition to the Coomassie and Alcian blue staining (9), Stains All was used (23). Electrotransfer of proteins out of the SDS gels and onto nitrocellulose sheets (Schleicher & Schuell) was accomplished in 15-60 min at 100 V in a Bio-Rad Transblot apparatus with cooling to 4 "C according to the method of Towbin et al. (25).
Analytical Procedures-Amino acid compositions were determined on samples hydrolyzed in 6 N HCI in vials sealed under nitrogen for 20 h at 110 "C. Analyses were obtained on a single cation-exchange column as described elsewhere (26). Detection of the amino acids was by ninhydrin complex monitored at 570 and 440 nm. Automated Protein Sequencing-Lyophilized samples were dissolved in 1% trifluoroacetic acid and subjected to automated Edman degradations using an Applied Biosystems model 470A gas-phase sequencer employing the standard "NoVac" program supplied by the manufacturer (30). Phenylthiohydantoin derivatives were identified by HPLC on an IBM cyano column (31). The HPLC system used with this column consisted of a Perkin-Elmer series 4 liquid chromatograph, an LC-85B spectrophotometric detector equipped with a 1.4-pl flow cell, an LC1 100 computing integrator, and a model 7500 computer employing Chrom I11 software. These analyses were supplied under contract by the University of California at San Diego.
Immunodetection-Nitrocellulose electrotransfen were processed for immunodetection as described previously (13)  . Bars indicate the range over which individual proteins can be found in significant amounts. Elution conditions are detailed under "Experimental Procedures." One ml each of representative pooled fractions was concentrated and desalted by ultrafiltration (Centricon 30 for fractions 1-7 and Centricon 10 for fraction 8). recovered from the membrane in SDS sample buffer, electrophoresed on a 4-20% gradient polyacrylamide gel (reduced), and stained with Coomassie Blue and Alcian blue. Human serum albumin ( h a ) , procollagen (procol), and osteocalcin or bone gla protein (oc). Other abbreviations as in Fig. 1. injection) or incomplete adjuvant (2nd and 3rd injections) at 2-week intervals. The rabbits were bled from the ear vein a t 2-week intervals after titer was detected by enzyme-linked immunoassays on microtiter plates using the peroxidase-conjugated second antibody technique (13). Initial injections for bone sialoproteins I and I1 and proteoglycan I1 were 200, 320, and 760 pg, respectively, with the two boosts consisting of half as much by dry weight each. Antiserum specific for human osteonectin was made by cutting lightly stained human osteonectin bands out of a 2-mercaptoethanol-reduced gradient SDS gel and milling the bands in a Spex freezing mill. The powder was emulsified and injected as above. 1.2 mg of human bone osteonectin was divided among 10 lanes for electrophoresis. All antisera were stored a t -70 "C until use.
Alkylation reacted first with 20-fold excess of dithiothreitol (over protein), then with the radioactive iodoacetamide. All tubes were incubated in the dark for 4 h on ice. Low molecular weight activity was removed by centrifugation in a Centricon 30 in 7 M guanidine HCI buffer containing an excess of cold iodoacetamide (two washes) followed by repeated water washes. The filters were removed and counted in Hydrofluor (National Diagnostics) using a Beckman model LS8100 liquid scintillation counter.

RESULTS AND DISCUSSION
We have shown earlier that developing bone is a good source of mineral compartment proteins (most of the noncollagenous proteins are degraded in older bone (24)). The first denaturing extract is used to remove the proteins associated with the soft tissues of the bone, including cells, blood, and adhering connective tissues (32). The second, demineralizing extract contains virtually all of the mineral compartment noncollagenous   15 pg of intact proteoglycan each were electrophoresed on SDS 4-20% gradient polyacrylamide gels and stained with Alcian blue and Coomassie Blue (a). 50 pg of each proteoglycan was digested with chondroitinase ABC to release the core products, electrophoresed, and stained with Coomassie Blue ( b ) . Similar amounts of the core proteins were electrophoresed, electroblotted onto nitrocellulose (Western blot), and detected with antiserum raised against bone proteoglycan I1 and a peroxidase-conjugated second antiserum (c). See Fig. 1 for molecular weight standards.

. Identification of (a) intact bone proteoglycans I and I1 and ( b and c) their corresponding core proteins by SDS-gel electrophoresis and Western blot/immunodetection.
proteins of the bone, including serum-derived products (such as serum albumin and asHS glycoprotein) that are absorbed to the crystal surfaces as well as proteins thought to be an integral part of the matrix. Fig. 1 shows a SDS gel of the total ASP (Dl" Thr (T) Ser (S) veloping human bone. Several proteins are high in relative abundance and are described in this report. Approximately 95% of the glycosaminoglycan contained in the mineral compartment of bone is in the two small proteoglycans I and I1 (9). Proteoglycan I is thought to contain two chondroitin sulfate chains (M, -40,000 each) attached to a core protein of M , -45,000 (10). Proteoglycan 11, a faster migrating species on SDS gels, is thought to contain only one chondroitin sulfate chain on a similar-sized protein core (10). Differences in the amino acid compositions of these two proteoglycans from bovine bone and a potential difference in immunogenicity and cross-reactivity had been noted (10).
To purify proteoglycans I and 11, we relied on the separation of the two proteoglycans on Sepharose CL-GB under denaturing conditions, using SDS gels to analyze individual fractions from the column. We pooled high molecular weight fractions (the leading half of the bar labeled proteoglycan I in the top panel of Fig. 2, typically fractions 67-75) in the isolation of proteoglycan I, and low molecular weight fractions (the trailing half of the bar labeled proteoglycan I1 on Fig. 2, typically fractions 82-89) for proteoglycan 11. The bottom panel of Fig.   2 shows representative fractions of each major peak areas electrophoresed on SDS-polyacrylamide gels. The peak fractions also correspond to the peak numbers first described for the bovine bone noncollagenous proteins described earlier (15). DEAE-Sephacel chromatography, while not useful in separating bone proteoglycan I from bone proteoglycan I1 was used in each case to remove nonproteoglycan proteins from the pooled fractions (see top panel, Fig. 3 ) . The bottom panel of Fig. 3 shows representative fractions of the DEAE column, desalted, and electrophoresed on an SDS-polyacrylamide gel. Final purification was accomplished with the MONO Q anionexchange column (Fig. 4).
Both the final intact products and their core proteins (generated by digestion with chondroitinase ABC to remove the attached chondroitin sulfate chains) are shown in Fig. 5. The selection of the fractions for the larger of the proteoglycan I and the smaller proteoglgcan I1 molecules to overcome the complement of proteins in the mineral compartment of de-propensity of thetwo heterogenous populations to co-elute to " TABLE I1 Amino acid sequences X, no assigned amino acid; ( ) tentative amino acid assignment; -, deletion of one amino acid.   I1 ( c ) . Immunodetection by peroxidase-conjugated second antibody and 4-chloro-1-naphthol is described under "Experimental Procedures." some degree did, however, skew the apparent molecular weight of the purified proteoglycans (thus purified proteoglycan I was apparently larger than the average proteoglycan I and the purified proteoglycan I1 population apparently smaller). For this reason we have chosen to assign the M , of proteoglycans I and I1 using the original crude extract as seen in Fig. 1 to more accurately describe the average size of the original proteoglycans. Therefore, the estimated M , of bone proteoglycans I and I1 are 350,000 and 200,000 respectively, although the M , of any small proteoglycan is, as always, a function of the SDS gel system used. The lunes in Fig. 5a are overloaded to show the relative purity of proteoglycan I and proteoglycan I1 preparations. The minor components seen at lower molecular weight in the proteoglycan I1 lane may be some free chondroitin sulfate chains or some minor amounts of proteoglycan I1 breakdown products, but they did not interfere with the N-terminal amino acid sequence described below. Bone proteoglycan I generated a single band core ( M , -46,000) while the bone proteoglycan I1 core was a closely spaced doublet ( M , -45,000 and 47,000), Fig. 5b. These results are similar to those found in bovine bone (data not shown). Antiserum raised against bone proteoglycan I1 in rabbit did not cross-react with the core protein of proteoglycan I on Western blot analysis (Fig. 5). While the amino acid compositions of proteoglycans I and I1 (Table I) (Table 11) is similar to that of fetal placental membrane proteoglycan (33) that is thought to have a glycosaminoglycan chain attached to the number 4 position, the serine of a serine-glycine sequence commonly accepted to be the attachment site of certain glycosaminoglycans. The human bone proteoglycan I1 protein sequenced at only about 5% yield so the reliability and significance of the differences at positions 7 and 15 between the bone and fetal membrane forms of this small proteoglycan is not known. Interestingly, the bone form of proteoglycan I1 is cleaved differently by V-8 protease than are the tendon, articular cartilage, and skin proteoglycan I1 molecules (11).
The first few N-terminal amino acids of the bone proteo-  glycan I molecule (Table 11) are similar to that of proteoglycan I1 (with an extra glutamic acid at position 3), suggesting that position 5 Q€ proteoglycan I may serve as a potential site for the additim of a glycosaminoglycan chain. No direct evidence of a substituted serine is yet available, however. Interestingly, a second serine-glycine sequence is seen at positions 10 and 11. In sequence analysis, the signal for the position 10 serine was relatively weak, indicating that a glycosamsitrkoglycan may be present at this location also. The two bone sialoproteins in the mineral compartment of human bone are very similar to those found in bovine bone (2, 13,14). Bone sialoprotein I, so named because it elutes at a lower salt concentration from ion-exchange columns, is similar in size to bone sialoprotein I1 on SDS-polyacrylamide electrophoretic gels but chromatographs as a slightly smaller protein on Sepharose CL-GB under denaturing conditions (Fig. 2). To purify these two proteins, pooled fractions from the molecular sieve column (bars, Fig. 2) were concentrated and exchanged into urea buffer by ultrafiltxation and chromatographed first on DEAE-Sephacel, then on MONO Q (see "Experimental Procedures" and Figs. 3 and 4). Unlike bovine bone sialoprotein I which stained poorly with Coomassie Blue and not at all with Alcian blue, human bone sialoprotein I did not stain well with either dye. Stains All, however, gave a strong blue-purple reaction product (typical of highly acidic proteins) for both human bone sialoprotein I (Fig. 6a) and its bovine counterpart (data not shown). The N-terminal sequence of human bone sialoprotein I shown in Table I1 had several cycles with no identifiable amino acid residues. Both oligosaccharide attachment sites and cysteine residues lead to a lack of an identifiable amino acid in sequence analysis.
The purified human bone sialoprotein 11, like its bovine counterpart (13), did not stain with Coomassie Blue under our routine conditions but did stain with Alcian blue (lane 4, bottom panel of Fig. 3). Fig. 6a shows this protein on SDSpolyacrylamide gel electrophoresis, stained with Stains All. Antisera made against bone sialoprotein I did not cross-react with bone sialoprotein I1 or vice versa on Western blots (Fig.  6, b and e) indicating that these proteins are not closely related. This is in contrast to partial cross-reactivity reported by Franzen and Heinegard (14) for the two bovine sialoproteins. Further evidence that the two sialoproteins are different are their different amino acid compositions (Table I) and different N-terminal sequences (Table 11). As was the case for bone sialoprotein I, the sequence of bone sialoprotein I1 has several analytical cycles with no identifiable amino acids, indicating either post-translational modification sites or cysteine residues.
Purified human osteonectin, like its bovine counterpart (15,16) is a phosphorylated glycoprotein of M , -46,000 on reduced SDS gels (Fig. 7). Unreduced, the protein is considerably smaller on SDS gels (MI -38,000, Fig. 7) suggesting that it contains several intramolecular disulfide bonds. This large change in M , upon reduction was reported earlier for fetal porcine osteonectin (34). Both human and bovine bone osteonectin incorporated much more of the radiolabeled alkylating agent, ['4C]iodoacetamide, after reduction with dithiothreitol (13.9-and 15.4-fold increases, respectively) compared with the same protein unreduced. This suggests that most (>go%) of the cysteines in the bone osteonectin are incorporated into disulfide bonds. Table I1 shows that the N-terminal sequences of bovine and human bone osteonectin are highly conserved, indeed there are only two conserved changes (positions 22 and 26). In contrast, a sequence for a related protein from mouse parietal endoderm (derived from cDNA, Ref. 35), contains many changes in the N terminus.
It is an open question at this time whether the differences between the human/ bovine bone-derived protein and the soft tissue mouse protein arise from species differences (in a nonconserved region) or from a tissue-specific difference within a given species. The latter possibly has at least two precedents in fibronectin (36) and ab-crystallin (37) in which a single gene results in two closely related but different primary structure proteins. Table I11 shows the cross-reactivity of the various antisera produced against €our of the five human bone-derived pcoitehs described in this report to a wide variety of species. Fix each species, the bones were cleaned, milled under liquid n.i%scgen, extracted first with the 4 M guanidine buffer, then with the demineralizing buffer. Lyophylized mineral compartment extract (200 pg)' was electrophoresed on our standard SDS, &,, electrotransferred for 20 min, and detected using the appropriate antiserum and peroxidase-conjugated second antiserum.
In this report we have presented evidence that the two small proteoglycans of developing human bone (I and 11) are different proteins. We also have shown that the two human bone sialoproteins, I and 11, are different proteins. Human bone osteonectin contains a substantial number of disulfide bonds and has an N terminus that is highly conserved with respect to bovine bone osteonectin. Using the standard nomenclature recommended by the American Society of Bone and Mineral Research (in which each protein, or core protein in the case of proteoglycans, is named by its fist three aminoterminal amino acids and its apparent molecular weight on a 4-20% gradient polyacrylamide SDS, reduced gel), the bone proteins described in this report are identified as: DEE-46,000 (proteoglycan I), DEA-47,000 (proteoglycan 11), IPV-80,000 (bone sialoprotein I), FSM-80,000 (bone sialoprotein 11), and APQ-46,000 (osteonectin).