Role of hrCEMP1 in Human Biomineralization: Characterization, Notas de estudo de Engenharia Elétrica

Baixe Role of hrCEMP1 in Human Biomineralization: Characterization e outras Notas de estudo em PDF para Engenharia Elétrica, somente na Docsity! Biochemical and Biophysical Research Communications 384 (2009) 49–54Contents lists available at ScienceDirect Biochemical and Biophysical Research Communications journal homepage: www.elsevier .com/locate /ybbrcCharacterization of recombinant human cementum protein 1 (hrCEMP1): Primary role in biomineralization Eduardo Villarreal-Ramírez a, Abel Moreno b, Jaime Mas-Oliva c, Juan Luis Chávez-Pacheco a, A. Sampath Narayanan d, Ivet Gil-Chavarría a, Margarita Zeichner-David e, Higinio Arzate a,* a Laboratorio de Biología Periodontal y Tejidos Mineralizados, Facultad de Odontología, UNAM, México D.F. 04510, Mexico b Instituto de Química, UNAM, México D.F. 04510, Mexico c Instituto de Fisiología Celular, UNAM, México D.F. 04510, Mexico d Department of Pathology, School of Medicine, University of Washington, Seattle 98195, USA e Center for Craniofacial Molecular Biology, University of Southern California, USA a r t i c l e i n f o a b s t r a c tArticle history: Received 2 April 2009 Available online 23 April 2009 Keywords: Cementum Biomineralization Cementum protein 1 Mineralized tissues Hydroxyapatite Octacalcium phosphate Periodontal regeneration0006-291X/$ - see front matter  2009 Elsevier Inc. A doi:10.1016/j.bbrc.2009.04.072 * Corresponding author. Fax: +52 5556225563. E-mail address: harzate@servidor.unam.mx (H. ArCementum protein 1 (CEMP1) has been recently cloned, and in vitro experiments have shown functions as regulator of cementoblast behavior and inducer of differentiation of non-osteogenic cells toward a cementoblastic/osteoblastic phenotype. In this study, we have produced a full-length human recombi- nant CEMP1 protein in a human gingival fibroblast cell line. The purified protein (hrCEMP1) has a Mr 50,000. Characterization of hrCEMP1 indicates that its secondary structure is mainly composed of b-sheet (55%), where random coil and alpha helix conformations correspond to 35% and 10%, respectively. It was found that hrCEMP1 is N-glycosylated, phosphorylated and possesses strong affinity for hydroxyapatite. Even more important, our results show that hrCEMP1 plays a role during the biomineralization process by promoting octacalcium phosphate (OCP) crystal nucleation. These features make CEMP1 a very good candidate for biotechnological applications in order to achieve cementum and/or bone regeneration.  2009 Elsevier Inc. All rights reserved.Introduction Cementum is a unique avascular mineralized connective tissue that covers the root surface of teeth and provides the interface through which the root surface is anchored to collagen Sharpey’s fibers of the periodontal ligament. Nevertheless, the complex pro- cesses that regulate cementogenesis and normal cementum metab- olism remain unclear to date. Recent evidence indicates that cementum formation is critical for appropriate maturation of the periodontium [1]. Recently we have isolated and characterized a human cementum protein which we named Cementum Protein 1 (CEMP1), (GenBank Accession No. NP_001041677; HGNC: ID 32553) [2]. Antibodies against this protein recognize the cementoid layer and adjacent cementoblastic cell layer, cementocytes, progen- itor cells located near the blood vessels in the periodontal ligament, cells located in the endosteal spaces of human alveolar bone, dental follicle-derived cells and human periodontal ligament cells [2–4]. CEMP1 mRNA is highly expressed in cementoblasts, subpopulations and progenitor cells of the human periodontal ligament [5]. In vitro experiments showed that CEMP1 promotes cell attachment, differ-ll rights reserved. zate).entiation [6,7], and deposition rate, composition, and morphology of hydroxyapatite crystals formed by human cementoblast cells [7]. Since CEMP1 is synthesized by cementoblast cells and, a re- stricted periodontal ligament cell subpopulations (cementoblast precursors), it is suggested that this molecule is a cementum-specific biological marker and it might play a role as regulator of cell differ- entiation. Furthermore, CEMP1 transfection into non-osteogenic cells such as adult human gingival fibroblasts results in differentia- tion of these cells into a ‘‘mineralizing” cell phenotype [8]. Although the physiological function of CEMP1 is not completely understood, it is our hypothesis that this molecule plays an important role during the cementogenesis process and also as an inducer of the formation of mineralizing nodules and calcium deposition during hydroxyapa- tite formation. Therefore, the aim of the present study was to charac- terize the physic-chemical characteristics of hrCEMP1 expressed in a human gingival fibroblast cell line and determine post-translational modifications and their influence on CEMP1’s functional properties during the mineralization process.Materials and methods Expression and purification of CEMP1. The open reading frame of CEMP1 (GenBank Accession No. NP_001041677), was subcloned 50 E. Villarreal-Ramírez et al. / Biochemical and Biophysical Research Communications 384 (2009) 49–54into the pENTR/SD/D vector (Invitrogen, Carlsbad, CA) and the resultant pENTR/SD/D-CEMP1 cDNA construct ligated into a pcDNA40(+) vector pcDNA40-CEMP1(+). Human gingival fibro- blasts (HGF) were isolated and grown as previously described [3]. The plasmid pcDNA40-CEMP1(+) was transfected into human gin- gival fibroblasts cells as described elsewhere [8]. Recombinant human CEMP1 protein collected from conditioned media of HGF expressing CEMP1 was purified by Ni2+ affinity chro- matography (HiTrap Chelating HP column, Invitrogen, Carlsbad, CA). Determination of protein purity was performed by 12% SDS– PAGE. Western blot analysis. Recombinant human CEMP1 (10 lg) was separated by 12% SDS–PAGE and electroblotted onto Inmobilon-P (PVDF) nitrocellulose membrane (Millipore Corp., Bedford, MA). Anti-hrCEMP1 and anti-6XHis (C-term) polyclonal antibodies were used to specifically identify the CEMP1 gene product and the fused histidines. Peroxidase-conjugated goat anti-rabbit IgG was used and secondary antibody and detection was performed as previ- ously described [2]. Hydroxyapatite affinity chromatography. To determine if hrCEMP1 has affinity to hydroxyapatite, an Econo-Pac CHT-II car- tridge (1 mL) (Bio Rad, Hercules, CA) was used. The column was equilibrated with 10 mM sodium phosphate, pH 7.2. Fifty micro- grams of purified hrCEMP1 was loaded and unbound proteins removed with a solution containing 10 mM sodium phosphate, pH 7.2. Bound proteins were eluted with a solution containing 0.05, 0.1, 0.2, 0.3, 0.4 and 0.5 M sodium phosphate, pH 7.2. Frac- tions were subjected to 12% SDS–PAGE and Western blotting. Circular dichroism spectroscopy. hrCEMP1 protein was dissolved in PBS, pH 7.4, at 200 lg/mL. The concentration was calculated from the absorption at 280 nm using an extinction coefficient of 28,125 M1 cm1 and deduced from the amino acid sequence [9]. CD spectra were recorded in thermostatted (25 C) quartz cells of 1-mm optical path length within a wavelength range of 190– 260 nm using a AVIV62DS spectropolarimeter. The molar ellipticity (h) expressed in degrees.cm2 dmol1 was calculated on the basis of a mean residue of Mr 50,000. Five spectra were accumulated to im- prove the signal to noise ratio. A baseline with buffer (PBS, pH 7.4) was recorded separately and subtracted from each spectrum. The program CONTIN was used to calculate secondary structure con- tent [10]. Dynamic light scattering (DLS). Light scattering experiments were performed using a Zetasizer Nano S (Malvern Instruments, Ltd., UK) molecular sizing instrument which employs a 4 mw, 633 nm semiconductor laser as light source and NIBS technology (Malvern Instruments, Ltd., UK) [11]. During experiments the tem- perature was held at 25 (0.1 C) via a Peltier unit. Data analysis was performed using the Zetasizer Nano S DTS software package (Mal- vern Instruments, Ltd., UK). Presence of cysteine disulfide bonds. Human recombinant CEMP1 at a 2.5 mM concentration was dissolved with 6 M guanidine–HCl containing 200 mM DTT. The protein was reduced at 37 C over- night, and boiled for 5 min before the protein was loaded into a gel filtration column (1.5  10 cm Sephadex G-10, Pharmacia, Upp- sala, SW) equilibrated with 300 mM acetic acid. The reduction state of hrCEMP1 was assessed by quantitation of thiols using an assay for dithiodipyridine. Briefly, hrCEMP1 was incubated with 6 M guanidine–HCl, 10 mM EDTA, 120 mM Na2HPO4, pH 6.6, and DTPD (4,40-dithiodipyridine) added to a final concentration of 500 nM. Samples were incubated for 30 min at 25 C and the A324 was monitored to estimate the number of cysteine residues present. Glycosylation analysis. Carbohydrates contained in hrCEMP1 were determined using the ECL glycoprotein detection system (Amersham Biosciences, UK). Briefly, samples were separated by 12% SDS–PAGE and electro-transferred as described above. Oxida-tion was carried out in the dark with 10 mM sodium metaperio- date dissolved in 100 mM acetate buffer, pH 5.5. Samples were treated with biotin hydrazide to incorporate biotin into the oxi- dized carbohydrate and biotin was detected by the horseradish peroxidase-conjugated streptavidin system using enhanced chemi- luminescence (ECL, Millipore Corp., Bedford, MA). Release of N-linked oligosaccharides. N-Glycans were released from 200 lg of hrCEMP1 by enzymatic cleavage using peptide N- glycosidase F (Calbiochem Glycoprotein Deglycosylation kit, Merck Biosciences Ltd., Nottingham, UK). The protein was resus- pended with 10 lL of 250 mM sodium phosphate buffer, pH 7.0 and 2.5 lL of denaturation solution (2% w/v SDS, 1 M b-mercap- toethanol). The mixture was heated at 100 C for 5 min. One unit of PNGasa F was added and incubated for 24 h at 37 C. The N- linked glycosylation pattern of hrCEMP1 was resolved by 12% SDS–PAGE. hrCEMP1 phosphorylation. According to ‘in silico’ analysis, hrCEMP1 possesses multiple potential phosphorylation sites. Eigh- teen phosphorylation sites (10 serine, 8 threonine) were predicted by the NetPhos 2.0 program [2]. Tyrosine is not present in the hrCEMP1 amino acid sequence. To determine if serine and threo- nine phosphorylation is present in hrCEMP1, hrCEMP1 secreted to the media and purified by Ni2+ affinity chromatography was used. Human recombinant CEMP1 was subjected to 12% SDS–PAGE and electrotransfered as described above. Membranes were blocked as described, and incubated with primary polyclonal anti- bodies against phosphothreonine and phosphoserine (Zymed, San Francisco, CA, USA). After washing, membranes were incubated with the horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody. Membranes were washed and detection of secondary antibody performed as previously described. hrCEMP1 effect on apatite formation. To determine if hrCEMP1 promotes apatite formation, a capillary counterdiffusion system was used [13–15]. Briefly, 1% (w/v) agarose gel containing 20 lg/ mL of hrCEMP1 was poured into the capillaries (0.5 mm diameter and 30 mm long). The ends of the capillaries were injected with 100 mM CaCl2 and 100 mM NaH2PO4. All experiments were carried out at 37 C. After 7 days, the crystals were recovered by dissolving the gel in hot milli Q water and air-dried. Energy-dispersive X-ray micro-analysis (EDX). The composition of crystals formed by induction of hrCEMP1 into the capillaries was analyzed using a Jeol 5600 scanning electron microscope fitted with a detector of energy dispersive X-ray microanalysis micro- probe. All analyses were performed at 20 kV for 300 s [16]. Crystals were analyzed in low vacuum and the calcium/phosphate (Ca/P) ratio was calculated from the intensity of the peaks present in the EDX pattern. After determining the composition of the crystals, they were covered with a thin gold film, 100 nm thick, to avoid electron disturbances that could interfere with the SEM images. Electron diffraction pattern by transmission electron microscopy. Crystals were mounted on carbon-coated 150-mesh gold grids and examined for diffraction techniques. D-spacings of diffraction patterns were calibrated against those used as gold standard with identical diffraction conditions. The mineral phase was analyzed by means of a JEOL 100 CX analytical transmission microscope employing 100 kV. Results and discussion Isolation of human CEMP1 by recombinant expression Previously, we have expressed hrCEMP1 in a prokaryotic expression system; however, this system is not able to express the full-length recombinant CEMP1 [2]. In this study, human recombinant CEMP1 was expressed in human-derived gingival fibroblasts as a secreted 6XHis fusion protein and latter purified E. Villarreal-Ramírez et al. / Biochemical and Biophysical Research Communications 384 (2009) 49–54 53related to the mineralization process elute at lower concentrations of sodium phosphate [26], whereas proteins implicated with the biomineralization process elute at concentrations between 200 and 300 mM sodium phosphate. Accordingly to this result, hrCEMP expressed in a prokaryote system present the same affinity to hydroxyapatite (data not shown), as the protein expressed in eukaryotic cells. These results indicate that hrCEMP1 possesses intrinsic properties to bind to hydroxyapatite even without post- translational modifications. hrCEMP1 effect on apatite formation The morphology of apatite crystals grown in absence of hrCEMP1 had raft-plaque-like shape and microscopic crystals with a globular-spherulite-like shape (Fig. 3E). However, crystals grown in the presence of hrCEMP1 showed drusa-like shape and a com- bined drusa-like and needle-like morphology (Fig. 3F). Human re- combinant CEMP1 induces the formation of polymorphous crystals as confirmed by X-ray diffraction. Elemental analysis performed with EDX determined the Ca/P ratio to be 1.67 for control crystals, equal to the theoretical Ca/P ratio of 1.67 for hydroxyapatite (Ca10(PO4)6(OH)2) according to ICDD file: PDF#24-0033 (Fig. 3G). Whereas, experimental conditions using hrCEMP1 determined that the crystals are OCP, (Ca8H2(PO4)65H2O) and EDX elemental analy- sis determined a Ca/P ratio of 1.33 according to ICDD file: PDF#26- 1056 (Fig. 3H). Furthermore, diffraction patterns for the crystals formed under control and experimental conditions were analyzed and index assigned. Diffraction patterns for the control conditions represent hydroxyapatite according to the Miller index (hkl: 1 0 2; hkl: 3 0 0 and hkl: 2 1 1) with interplanar distances of 3.17 Å, 2.72 Å and 2.81 Å, respectively (Fig. 3I). Crystals formed in presence of hrCEMP1 represent OCP according to the Miller index (hkl: 2 2 1; hkl: 1 2 2 and hkl: 4 4 1) with interplanar distances of 3.74 Å, 2.94 Å and 2.15 Å, respectively (Fig. 3J). Taken all together, these results demonstrated that biologically active hrCEMP1 plays a role during the biomineralization process, that it is required for the synthesis of needle-shaped OCP crystals and responsible for OCP crystal nucleation activity. OCP is found to be a transient phase during the growth of biological crystals. In small crystals, OCP is completely transformed into HA by hydro- lysis and can only be detected in larger crystals because of its slow kinetics of transformation. OCP has also been presumed a neces- sary precursor of biological apatites in both normal (enamel, den- tine, cementum, and bone) and pathological (e.g., phosphatic renal stones) calcifications. The initial basis of CEMP1’s as a possible therapeutic agent rests on evidence showing that CEMP1 has a role during the biological mineralization process of cementum-like tissue, and that induces in vitro phenotypic changes from non-osteogenic cells to an osteo- blastic/cementoblastic phenotype [8]. Accordingly, other mole- cules related to the mineralization process such as BMP-2, 4 and 7 have been established to induce reparative/regenerative mineral- ized tissue formation [27]. Sustained delivery of BMP-2 using gene therapy has shown to induce bone formation in vivo [28]. Recom- binant human amelogenin protein (rHAM(+)), has also been shown to be effective to induce in vivo regeneration of all tooth-support- ing tissues after creation of an experimental periodontitis model in dogs, where the recruitment of mesenchymal progenitor cells is a key factor during the formation of regenerated periodontal tissues [29]. Amelogenin and other proteins such as OCN, BSP and OPN have the ability to bind and to regulate hydroxyapatite crystal growth and nucleation as well as to promote mineralization [30,31]. New therapeutic strategies such as the use of PLGA scaffolds for tissue engineering applications could be applied to demonstrate the ability of a single protein to bring about regener- ation of mineralized tissues. These strategies could pave the wayfor development of new therapeutic devices for treatment of periodontal and bone diseases based on recombinant human CEMP1. Conclusions Studies on the function and structure of CEMP1 have been ham- pered by the difficulties in the isolation of protein from human and/or bovine cementum, due mainly to its low metabolic turnover rate, the amount of tissue being deposited and, its entrapment in hydroxyapatite. In this study, we showed the production, purifica- tion and characterization of a full-length hrCEMP1 employing for its synthesis an eukaryotic system. In addition, we demonstrated that CEMP1 is post-translational modified, N-glycosylated and phosphorylated. We also showed that hrCEMP1 plays a role during the mineralization process by promoting OCP crystal growth. These results provide the basis for future studies analyzing the po- tential characteristics of this novel protein towards new therapeu- tic strategies in order to treat individuals affected by disorders affecting mineralized tissues and to promote regeneration of cementum and/or bone in adult tissues.Acknowledgments The authors thank Dr. Pedro Bosch and Dr. José Reyes. This study was supported by funds from DGAPA-UNAM (IN200908), CONACyT 48638 to H.A. and 82888 to A.M. References [1] D.D. Bosshardt, S. Zalzal, M.D. McKee, A. Nanci, Developmental appearance and distribution of bone sialoprotein and osteopontin in human and rat cementum, Anat. Rec. 250 (1998) 13–33. [2] M.A. Alvarez-Perez, S. Narayanan, M. Zeichner-David, B. Carmona-Rodríguez, H. Arzate, Molecular cloning, expression and immunolocalization of a novel human cementum-derived protein (CP-23), Bone 38 (2006) 409–419. [3] A.S. Narayanan, R.C. Page, Biochemical characterization of collagens synthesized by fibroblasts derived from normal and diseased human gingiva, J. Biol. Chem. 251 (1976) 5464–5471. [4] M. Zeichner-David, K. Oishi, E. González, Z. Su, V. Zakartchenko, L.S. Chen, H. Arzate, P. Bringas, Role of Hertwig’s epithelial root sheath cells in tooth root development, Dev. Dyn. 228 (2003) 651–663. [5] H. Arzate, L.F. Jiménez-García, M.A. Alvarez-Pérez, A. Landa, I. Bar-Kana, S. Pitaru, Immunolocalization of a human cementoblastoma conditioned medium-derived protein, J. Dent. Res. 81 (2002) 541–546. [6] H. Arzate, J. Chimal-Monroy, L. Hernández-Lagunas, L. Díaz de León, Human cementum protein extract promotes chondrogenesis and mineralization in mesenchymal cells, J. Periodont. Res. 31 (1996) 144–148. [7] M.A. Alvarez Pérez, S. Pitaru, O. AlvarezFregoso, J. Reyes Gasga, H. Arzate, Anti- cementoblastoma-derived protein antibody partially inhibits mineralization on a cementoblastic cell line, J. Struct. Biol. 143 (2003) 1–13. [8] B. Carmona-Rodríguez, M.A. Alvarez-Pérez, A.S. Narayanan, M. Zeichner-David, J. Reyes-Gasga, J. Molina-Guarneros, A. García-Hernández, J. Suárez-Franco, I. Gil Chavarría, E. Villarreal-Ramírez, H. Arzate, Human cementum protein 1 induces expression of bone and cementum proteins by human gingival fibroblasts, Biochem. Biophys. Res. Commun. 358 (2007) 763–769. [9] P. Mendoza-Espinosa, A. Moreno, R. Castillo, J. Mas-Oliva, Lipid dependant disorder-to-order conformational transitions in apolipoprotein CI derived peptides, Biochem. Biophys. Res. Commun. 365 (2008) 8–15. [10] S.W. Provencher, J. Glöckner, Estimation of globular protein secondary structure from circular dichroism, Biochemistry 20 (1981) 33–37. [11] A. Moreno, J. Mas-Oliva, M. Soriano-García, C. Salvador-Oliver, V.M. Bolaños-García, Turbidity as a useful optical parameter to predict protein crystallization by dynamic light scattering, J. Mol. Struct. 519 (2000) 243–256. [12] N. Blom, T. Sicheritz-Ponten, R. Gupta, S. Gammeltoft, S. Brunak, Prediction of post-translational glycosylation and phosphorylation of proteins from the amino acid sequence, Proteomics 4 (2004) 1633–1649. [13] L. Silverman, A.L. Boskey, Diffusion systems for evaluation of biomineralization, Calcif. Tissue Int. 75 (2004) 494–501. [14] C.E. Tye, K.R. Rattray, K.J. Warner, J.A.R. Gordon, J. Sodek, G.K. Hunter, H.A. Goldberg, Nucleation of hydroxyapatite by bone sialoprotein, J. Biol. Chem. 278 (2003) 7949–7955. [15] J.M. Garcia-Ruiz, Counterdiffusion methods for macromolecular crystallization, Methods Enzymol. 368 (2003) 130–154. 54 E. Villarreal-Ramírez et al. / Biochemical and Biophysical Research Communications 384 (2009) 49–54[16] F.J. Cuisinier, R.W. Glaisher, J.C. Voegel, J.L. Hutchison, E.F. Brès, R.M. Frank, Compositional variations in apatites with respect to preferential ionic extraction, Ultramicroscopy 36 (1991) 297–305. [17] A.K. Dunker, Z. Obradovic, The protein trinity-linking function and disorder, Nat. Biotechnol. 19 (2001) 805–806. [18] Q.Q. Hoang, F. Sicheri, A.J. Howard, D.S.C. Yang, Bone recognition mechanism of porcine osteocalcin from crystal structure, Nature 425 (2003) 977–980. [19] E. Hohenester, P. Maurer, C. Hohenadl, R. Timpl, J.N. Jansonius, J. Engel, Structure of a novel extracellular Ca(2+)-binding module in BM-40, Nat. Struct. Biol. 3 (1996) 67–73. [20] B. Kaufmann, S. Müller, F. Hanisch, U. Hartmann, M. Paulsson, P. Maurer, F. Zaucke, Structural variability of BM-40/SPARC/osteonectin glycosylation: implications for collagen affinity, Glycobiology 14 (2004) 609–619. [21] C.C. Chen, A.L. Boskey, Mechanisms of proteoglycan inhibition of hydroxyapatite growth, Calcif. Tissue Int. 37 (1985) 395–400. [22] W.J. Grzesik, C.R. Frazier, J.R. Shapiro, P.D. Sponseller, P.G. Robey, N.S. Fedarko, Age-related changes in human bone proteoglycan structure: impact of osteogenesis imperfecta, J. Biol. Chem. 277 (2002) 43638–43647. [23] M.A. Torres-Quintana, S. Lécolle, M. Goldberg, Effects of inositol hexasulphate, a casein kinase inhibitor, on dentine phosphorylated proteins in organ culture of mouse tooth germs, Arch. Oral. Biol. 43 (1998) 597–610.[24] S. Jono, C. Peinado, C.M. Giachelli, Phosphorylation of osteopontin is required for inhibition of vascular smooth muscle cell calcification, J. Biol. Chem. 275 (2000) 20197–20203. [25] M. Zeichner-David, F. Hall, R. Williams, F. Thiemann, S. Yen, M. MacDougall, H.C. Slavkin, Characterization of protein kinases involved in dentinogenesis, Connect. Tissue Res. 33 (1995) 87–95. [26] S. Hjerten, O. Levin, A. Tiselius, Protein chromatography on calcium phosphate columns, Arch. Biochem. Biophys. 65 (1956) 132–155. [27] M. Nakashima, Induction of dentin formation of canine amputated pulp by recombinant human bone morphogenetic protein (BMP)-2 and 4, J. Dent. Res. 73 (1994) 15515–15522. [28] B. Yue, B. Lu, K.R. Dai, X.L. Zhang, C.F. Yu, J.R. Lou, T.T. Tang, BMP2 gene therapy on the repair of bone defects of aged rats, Calcif. Tissue Int. 77 (2005) 395–403. [29] A. Haze, A.L. Taylor, S. Haegewald, Y. Leiser, et al., Regeneration of bone and periodontal ligament induced by recombinant amelogenin after periodontitis, J. Cell. Mol. Med. (2009), doi:10.1111/j.1582-4934.2009.00700.x. [30] Q.Q. Hoang, F. Sicheri, A.J. Howard, D.S. Yang, Bone recognition mechanism of porcine osteocalcin from crystal structure, Nature 425 (2003) 977–980. [31] G.K. Hunter, H.A. Goldberg, Nucleation of hydroxyapatite by bone sialoprotein, Proc. Natl. Acad. Sci. USA 90 (1993) 8562–8565.