Research article

Phosphorylation of Ser136 is critical for potent bone sialoprotein-mediated nucleation of hydroxyapatite crystals

Gurpreet S. Baht, Jason O'Young, Antonia Borovina, Hong Chen, Coralee E. Tye, Mikko Karttunen, Gilles A. Lajoie, Graeme K. Hunter, Harvey A. Goldberg


Acidic phosphoproteins of mineralized tissues such as bone and dentin are believed to play important roles in HA (hydroxyapatite) nucleation and growth. BSP (bone sialoprotein) is the most potent known nucleator of HA, an activity that is thought to be dependent on phosphorylation of the protein. The present study identifies the role phosphate groups play in mineral formation. Recombinant BSP and peptides corresponding to residues 1–100 and 133–205 of the rat sequence were phosphorylated with CK2 (protein kinase CK2). Phosphorylation increased the nucleating activity of BSP and BSP-(133–205), but not BSP-(1–100). MS analysis revealed that the major site phosphorylated within BSP-(133–205) was Ser136, a site adjacent to the series of contiguous glutamate residues previously implicated in HA nucleation. The critical role of phosphorylated Ser136 in HA nucleation was confirmed by site-directed mutagenesis and functional analyses. Furthermore, peptides corresponding to the 133–148 sequence of rat BSP were synthesized with or without a phosphate group on Ser136. As expected, the phosphopeptide was a more potent nucleator. The mechanism of nucleation was investigated using molecular-dynamics simulations analysing BSP-(133–148) interacting with the {100} crystal face of HA. Both phosphorylated and non-phosphorylated sequences adsorbed to HA in extended conformations with alternating residues in contact with and facing away from the crystal face. However, this alternating-residue pattern was more pronounced when Ser136 was phosphorylated. These studies demonstrate a critical role for Ser136 phosphorylation in BSP-mediated HA nucleation and identify a unique mode of interaction between the nucleating site of the protein and the {100} face of HA.

  • biomineralization
  • bone sialoprotein
  • hydroxyapatite
  • nucleation
  • phosphorylation
  • small integrin-binding ligand
  • N-linked glycoprotein (SIBLING)


Protein phosphorylation was initially detected by Levene and Alsberg in the early 1900s [1] and attributed to enzymatic activity by Burnett and Kennedy in 1954 [2]. Phosphorylation of proteins is now known to be a highly regulated event associated with cellular signalling, matrix assembly and protein processing, among other activities. In calcified tissues, protein phosphorylation has been postulated to be a critical factor in biomineralization [3].

Mineralization of the ECM (extracellular matrix) in bone is a complex and poorly understood process in which HA (hydroxyapatite) crystals are deposited on to a type I collagen scaffold. Postulated mechanisms involve non-collagenous phosphoproteins of the ECM that are capable of modulating crystal nucleation and growth, as well as binding to the collagenous network [4]. Studies on genetically modified mice have shown that members of the mineral-associated SIBLING (small integrin-binding ligand, N-linked glycoprotein) family [5] of proteins are involved in the biomineralization of bones and teeth.

In vitro studies have also implicated SIBLING proteins in biomineralization processes. Common to all of these SIBLING members is modulation of their activities by phosphorylation. For example, osteopontin [also known as BSP (bone sialoprotein) 1, SPP-1, uropontin, ETA-1] requires phosphorylation for its HA- and calcium oxalate monohydrate-inhibiting activities [6]; dentin matrix protein-1 has varied activity for mineral-crystal nucleation and growth dependent on its phosphorylation state [7]. In contrast, BSP (formerly known as BSP II), both the native form containing PTMs (post-translational modifications) [8] and the unmodified form [9], is a potent nucleator of HA.

BSP is an acidic phosphoprotein found at high levels in mineralized tissues [10]. Mammalian BSP has an average of 327 amino acid residues, including a 16-residue signal sequence, and an isoelectric point of 3.9 for the unmodified protein [10,11]. NMR [12], CD spectroscopy [9] and small angle X-ray scattering [9] studies have shown that BSP has a loose flexible structure, and thus could be classified as an intrinsically unstructured protein [13]. It is hypothesized that this lack of order allows BSP to simultaneously interact with multiple binding partners [12]. Different domains in mammalian BSP interact with type I collagen (residues 18–45) [14,15], HA crystals [16] and the cell-surface integrin αVβ3 (residues 274–276) [17]. Mice lacking BSP show decreased mineralization of cortical bones, shortened long bones, and altered activity of osteoblasts and osteoclasts in young animals, implicating involvement of BSP in early mineral formation in vivo and healthy bone biology [18].

Our previous work has shown BSP to be a potent nucleator of HA crystals [19], an activity that is increased upon binding to collagen [14]. The HANDs (HA-nucleating domains) of BSP have been identified as the two glutamic acid-rich regions within residues 42–100 and 133–205 [9]. This study also demonstrated that bone-extracted nBSP (native BSP) is a more potent nucleator than prokaryotic-expressed unmodified rBSP (recombinant BSP). Although BSP is heavily modified with a variety of PTMs, it is presumed that phosphorylation is critical in nucleation activity.

A number of studies have examined sites of phosphorylation for human [20] and bovine [21] BSP. It is apparent that most of the serine and threonine residues phosphorylated in bone-extracted BSP are highly conserved sites among mammalian species (summarized in Figure 1). Additional studies have investigated native and recombinant bovine BSP treated in vitro with protein kinase CK2, providing further information on the potential sites of phosphorylation [21,22].

Figure 1 Map of potential phosphorylation sites located within BSP

Potential sites of phosphorylation for BSP are shown based on a compilation of the literature investigating the PTMs of BSP from various species (rat, GenBank® accession number BAA19245; mouse, GenBank® accession number NP_032344; bovine, GenBank® accession number Q28862; human, GenBank® accession number NP_004958; and porcine, GenBank® accession number AAA19822). Conserved sites of phosphorylation are shown in black highlighted with a symbol appearing above (◆, ▲, or ■) corresponding to the citation: ◆, [20] (in vivo phosphorylation of human nBSP); ▲, [22] (in vitro phosphorylation of bovine rBSP by CK2); and ■, [21] (in vivo phosphorylation of bovine nBSP, in vitro phosphorylation of bovine nBSP by CK2 and in vitro phosphorylation of bovine rBSP by CK2). Residues identical in all species are identified with an * and conserved substitutions by a :.

CK2, Golgi casein kinase, mammary casein kinase and microsomal casein kinase have all been proposed as enzymes responsible for BSP phosphorylation in vivo. These enzymes share similar consensus sequences for protein modification. CK2 has a consensus sequence of serine/threonine in the ‘n’ position followed by an acidic residue in the n+3 position (S/T-X-X-E/D/P~S) [2325]. This consensus sequence is rather promiscuous in that an acidic residue at n+1 (and/or n+2, n+4) can also result in a phosphorylation event [26]. Golgi, mammary and microsomal casein kinases all have the same strict consensus sequence: serine/threonine followed by an acidic residue in the n+2 position (S/T-X-E/D/P~S) [27,28], which suggests that these three kinases are one and the same.

A number of studies have focused on CK2-mediated phosphorylation of BSP [21,29]. CK2 is a ubiquitously distributed enzyme required for life in eukaryotic organisms [30]. Most of the confirmed phosphorylation sites of BSP reside within highly conserved CK2 consensus sequences. Recently, Boskey et al. [31] used limb-bud mesenchymal cell micromass cultures to show that CK2 is critical for proper bone matrix assembly, function and mineralization. Upon inhibition of CK2, matrix mineralization was delayed and total mineral content was diminished.

The present study investigates the phosphorylation-dependent HA-nucleation activity of BSP. Through MS, mutagenesis and functional analysis, a specific region and a specific site of phosphorylation important for BSP function are identified. Furthermore, MD (molecular dynamics) simulations are used to generate a mechanism of interaction between BSP and HA that may explain the role of this protein in mammalian biomineralization.


Expression and purification of proteins and peptides

Native bone-extracted BSP was purified from rat bones as described previously [32]. Approved animal experiments were conducted according to guidelines established by the Animal Use Subcommittee at The University of Western Ontario. Rat rBSP and two rBSP-derived peptides containing residues 1–100 [rBSP-(1–100)] and 133–205 [rBSP-(133–205)] were prepared as described previously [9,15].

A pET-28a expression plasmid (Novagen) subcloned with the rat BSP sequence was used as a template to induce mutation of serine and threonine residues to alanine. Three full-length mutants of BSP were generated in this way: rBSP(S136A) (Ser136 is converted into alanine); rBSP(SXA) (Ser171,181,193 and Thr172 are converted into alanine); and rBSP(SALLA) (Ser136,171,181,193 and Thr172 are converted into alanine) as summarized in Figure 2. Proteins were expressed in Escherichia coli strain BL21(DE3) cells and purified by fast protein liquid chromatography following established protocols [9]. The purity and content were determined by SDS/PAGE, amino acid analysis and MS.

Figure 2 Summary of mutants generated

Full-length mutants of rBSP were produced by converting serine residues at possible sites of phosphorylation into alanine within residues 133–205, as shown.

A pGEX expression plasmid (Novagen) subcloned with the CK2α′ sequence and a GST (glutathione transferase) tag was used for expression of CK2. The resulting holoenzyme was expressed in E. coli strain BL21(DE3) and purified by GST-affinity chromatography. The enzyme was exhaustively dialysed [25 mM Tris/HCl, 100 mM NaCl, 0.5 mM EDTA and 0.5 mM DTT (dithiothreitol) (pH 7.4)] and stored at −80 °C in storage buffer [25 mM Tris/HCl, 100 mM NaCl, 0.5 mM EDTA and 0.5 mM DTT (pH 7.4) containing 50% glycerol]. Purity was assessed by MS and enzyme activity was assessed using an established protocol [33].

Synthetic peptides BSP-(133–148) and phosphorylated BSP-(133–148) were manually synthesized with free N- and C-termini using Fmoc (fluoren-9-ylmethoxycarbonyl) chemistry as described previously [34], and were purified by HPLC on a C18 column. Final peptide preparations were >98% pure. From ESI (electrospray ionization)-MS, the following masses were obtained (theoretical values in parentheses): BSP-(133–148), 2013.91 (2013.91 Da); phosphorylated BSP-(133–148), 2093.91 (2093.91 Da).

Phosphorylation by CK2α′

Purified proteins and peptides (1 mg/ml) were incubated with either 5 or 10 m-units of CK2α′ in reaction buffer [150 mM NaCl, 10 mM MgCl2, 1 mM DTT, 0.1 mM ATP and 50 mM Tris/HCl (pH 7.4)]. The reaction was allowed to proceed for 2 h at 30 °C and the reaction was then quenched by the addition of 0.7 vol. of water at 4 °C, and dialysed exhaustively against 2 mM ammonium bicarbonate. The samples were freeze-dried and purity, protein content and moles of phosphate added were determined by SDS/PAGE, amino acid analysis and MS respectively.

Phosphatase treatment of reagents

Purified protein or peptide (1 mg/ml) was incubated with CIP (calf intestinal phosphatase; 10 units) in reaction buffer [100 mM NaCl, 10 mM MgCl2, 1 mM DTT and 50 mM Tris/HCl (pH 7.9)]. The reaction was allowed to proceed for 1 h at 37 °C then quenched by the addition of 0.7 vol. of water at 4 °C, and exhaustively dialysed against 2 mM ammonium bicarbonate. The sample was freeze-dried and purity, protein content and moles of phosphate removed were determined by SDS/PAGE, amino acid analysis and MS respectively.

MS analysis

All proteins and peptides generated were investigated with linear MALDI–TOF-MS (matrix-assisted laser-desorption ionization–time-of-flight; Bruker Reflex IV MALDI TOF MS) to determine the average molecular mass. These masses were used to calculate the number of phosphate groups added or removed per molecule. Specific sites of phosphorylation were determined by partial digestion of CK2-treated full-length rBSP and rBSP-(133–205) using Asp-N, thermolysin, trypsin or proteinase K. For example, unmodified and phosphorylated rBSP-(133–205) (1 mg/ml) were incubated with enzyme (0.25 unit Asp-N, 0.30 unit proteinase K or 0.50 unit thermolysin) in 2 mM ammonium bicarbonate for 30 min at 25 °C (Asp-N and proteinase K) or at 30 °C (thermolysin). Resultant digestion products were analysed using MALDI–TOF-MS in linear and reflectron modes (Applied Biosystems; 4700 Proteomics Discovery System). Additionally, phosphorylated rBSP-(133–205) was analysed with LC-ESI-MS/MS (Micromass; Q-TOF Global Ultima) to further identify phosphorylated residues within this sequence.

Steady-state HA-nucleation assay

HA-nucleation activity was studied using a method described previously [19]. Briefly, the protein or peptide of interest was suspended in freshly prepared 1% agarose in blank steady-state buffer [150 mM NaCl, 0.01% sodium azide and 20 mM Hepes (pH 7.4)]. The mixtures were poured into the central chambers of modified equilibrium dialysis cells. The cells were then perfused at 37 °C for 5 days with steady-state buffers containing either 7.10 mM Ca(NO3)2 or 4.30 mM Na2HPO4 at a flow rate of 1 ml/h per cell. Proteins were studied in triplicate over a range of concentrations, and the lowest concentration capable of nucleating HA was used as a comparative indicator of nucleation activity or potency. Total mineral formation was determined by measurement of total calcium and phosphate content within the gels after ashing, as described previously [14]. The statistical significance of experimental data was determined by comparing with the negative control value (no protein) by one-way ANOVA followed by a Tukey test.

Structural analysis by CD

Full-length proteins and modified protein were dissolved at a concentration of 0.2 mg/ml, whereas peptides and peptide derivatives were dissolved to 0.75 mg/ml in 2 mM Tris/HCl and 150 mM NaCl (pH 7.4). Far-UV spectra were collected in a 1-mm pathlength quartz cell using a Jasco J-810 CD spectropolarimeter from 190 nm to 250 nm at 25 °C using 0.5-nm steps. Buffer lacking protein or peptide was used as a blank and spectra reported as molar ellipticity (θ) expressed in degrees cm2·dmol−1 as calculated on the basis of the mean residue molecular mass. Estimates of protein secondary structure from CD data were made using the Dicroprot package with analysis from the CDSTRR, Contin and Selcon methods.


Atomic-scale MD simulations were performed with the GROMACS suite [35] using a protocol similar to our previously published work [6]. Extended conformations were used as the initial peptide structure and peptides were oriented parallel to the crystal surface and placed approx. 3 nm from the {100} crystal face of HA, a surface previously implicated in binding BSP [16]. The co-ordinates for the HA {100} face were taken from previously obtained experimental results [36]. The crystal surface was placed at the centre of a periodic cell with dimensions of 8.4 nm×6.2 nm and the atoms were constrained to their equilibrium positions. The system was solvated with SPC (simple point charge) water [37]. Chloride counter ions were added to neutralize the net charge of the system. GROMACS force field parameters were used for chloride ions. Energy minimization was performed prior to simulation using the steepest descent method without constraints. Simulations were then performed in the NVT (constant particle number, volume and temperature) ensemble at 300 K with the bond lengths constrained using the SHAKE algorithm [38]. Temperature was kept constant using the weak-coupling thermostat with a time constant of 0.1 ps. The PME (particle-mesh Ewald) method was used for electrostatic interactions [39,40].

All simulations were run on the Shared Hierarchical Academic Research Computing Network (SHARCNET). A time step of 2 fs was used and all simulations were run for 10 ns. The system was evaluated in terms of centre-of-mass distance for the whole peptide and amino-acid side-chains from the top layer crystal surface atoms, averaged from 3 to 10 ns and sampled every 20 ps.


Endogenous phosphorylations contribute to the HA-nucleation activity of bone-extracted nBSP

To quantify the effect of phosphorylation on the HA-nucleating activity of BSP, nBSP and rBSP were compared using a steady-state assay for HA formation. HA nucleation was observed at concentrations as low as 0.001 nmol of nBSP, whereas 0.100 nmol of rBSP was required to initiate mineral formation (Figure 3 and Table 1). Dephosphorylation of nBSP by CIP decreases the nucleation potency, confirming that phosphorylation of BSP is a requirement for potent nucleation activity.

Figure 3 Phosphorylation-derived nucleation potency of BSP

Nucleating activities of nBSP (grey bars), CIP-treated nBSP (hatched bars) and rBSP (solid bars) within 1% agarose gels at subthreshold concentrations of calcium (7.10 mM) and phosphate (4.30 mM) were compared. Water was used as a no-protein negative control (−), demonstrating that the level of calcium and phosphate reached within the gel in the absence of any mineral crystal. Values are means±S.E.M. *P<0.05, compared with the no-protein control value as measured by one-way ANOVA followed by a Tukey test (n=3).

View this table:
Table 1 Phosphorylation of BSP and BSP peptides increases HA-nucleation potency

HA-nucleation potencies of various BSP reagents (as shown, in part, in Figures 3 and 4) with levels of incorporated phosphates after treatment with CK2 are summarized. C, minimum concentration of protein resulting in a significant increase in total calcium and phosphate accumulation.

Phosphorylation increases HA-nucleation potency of rBSP

rBSP was treated with CK2 and the level and sites of phosphorylation were determined. CK2 treatments resulted in the addition of an average of 4.6, 6.8 or 8.5 phosphates per BSP molecule (Table 1). Phosphorylation sites that were confirmed by two different enzymatic treatments followed by MS analysis are listed in Table 2. Analysis of tryptic and Asp-N-derived peptides confirmed the sites of phosphorylation as Ser15, Ser50, Ser51 and Ser136. In addition, other sites located within peptides corresponding to residues 57–78 (Table 2), 157–182 and 242–273 also contain phosphorylated residues, although the specific sites were not confirmed (results not shown). The CK2-treated rBSP was investigated for HA-nucleation potency and compared with the unmodified form. As before, 0.100 nmol of unmodified rBSP was required to induce HA nucleation; however, 0.010 nmol of phosphorylated rBSP initiated mineral formation (Figure 4B and Table 1). This 10-fold gain in apparent potency observed for CK2-treated rBSP remained unchanged, regardless of the number of phosphates added to the protein (Table 1). CIP treatment of the phosphorylated rBSP completely removed the attached phosphates (results not shown) and decreased potency to that of the unmodified form of rBSP (Figure 4B), confirming that phosphorylation of BSP increases its mineral-formation potency.

View this table:
Table 2 Summary of phosphorylated peptides after Asp-N and/or trypsin digestion of CK2-treated rBSP

*Indicates digested by Asp-N or trypsin

Figure 4 Phosphorylation-dependent nucleation activity of the BSP HANDs

(A) rBSP and its two nucleation-active peptides (highlighted in grey), rBSP-(1–100) and rBSP-(133–205), are shown. (BD) Untreated, CK2-treated (with an average of 6.8 phosphates) and CK2-treated followed by CIP-treated full-length rBSP (B), rBSP-(1–100) and CK2-treated rBSP-(1–100) (5.5 phosphates) (C), and rBSP-(133–205), CK2-treated (1.0 phosphate) and CK2-treated followed by CIP-treated (D) were assayed for HA-nucleation activities in 1% agarose gels at subthreshold concentrations of calcium (7.10 mM) and phosphate (4.30 mM). Open bars indicate the no-protein negative control (−); solid bars indicate unmodified protein or peptide; grey bars indicate CK2 treatment; and hatched bars indicate CK2 treatment followed by CIP treatment. Values are means±S.E.M. *P<0.05 compared with the no-protein control value as measured by one-way ANOVA followed by a Tukey test (n=3). P ~rBSP, phosphorylated rBSP.

Phosphorylation-dependent increase in activity is located in the second HAND

In order to determine the domain(s) responsible for the phosphorylation-derived increase in HA-nucleation potency of rBSP, two recombinant peptides, rBSP-(1–100) and rBSP-(133–205), each containing one of the two glutamate-rich HAND regions of BSP, were analysed for phosphorylation-dependent nucleation activity (as shown in Figure 4A). Both peptides were treated with CK2 and the average number of incorporated phosphates was determined by MALDI–TOF-MS. CK2 treatment of rBSP-(133–205) reproducibly resulted in an average of one phosphate incorporated, whereas rBSP-(1–100) had an average of 2.5 and 5.5 phosphates/peptide (Table 1). Unmodified rBSP-(1–100) required a minimum of 2.50 nmol to nucleate HA, a potency that was unaltered after CK2 treatment (Figure 4C and Table 1). In contrast, while unmodified rBSP-(133–205) required 0.250 nmol to nucleate HA, upon treatment with CK2 only 0.025 nmol was required (Figure 4D and Table 1). CIP treatment of the phosphorylated rBSP-(133–205) removed the phosphates from the peptide and returned its nucleation potency to that of unmodified rBSP-(133–205) (Figure 4D). Taken together these data demonstrate that the phosphorylation site(s) involved in enhanced HA-nucleation potency are located within the second, more potent, HAND of BSP.

Phosphorylation within the second HAND is located at Ser136

MALDI–TOF-MS analysis of untreated and CK-treated rBSP-(133–205) (Figures 5A and 5B) consistently revealed an average of one phosphate being added to the peptide, with a major monophosphorylated and a minor diphosphorylated species being evident (Figure 5B). The derived data and sequence information are shown in Table 2. Of interest, the low level of phosphate incorporation is in spite of having up to five potential sites of modification that are in the consensus sequence for CK2. This observation is in agreement with other investigations examining in vitro CK2 phosphorylation of BSP [21,22]. Enzymatic digestion and MALDI–TOF-MS were used to determine the phosphorylation site(s) within rBSP-(133–205). Phosphorylated rBSP-(133–205) was digested with Asp-N and the resultant fragments characterized by MS (Figure 5C). The two sets of peaks each show an isotopic distribution. The set of peaks centred at 2911 in Figure 5(C) indicate the presence of phosphorylated rBSP-(133–155) peptide [M(ox)KES(P)DEEEEEEEEEENENEEAEV], within which only Ser136 is a possible phosphorylation site for CK2 (Table 3). Examination by MS/MS (tandem MS) confirmed that residues other than Ser136 are probably not phosphorylated. The set of peaks centred at 2933 in Figure 5(C) correspond to the mass of a phosphorylated peptide containing a sodium adduct and oxidized methionine residue [(+Na)M(ox)KES(P)DEEEEEEEEEENENEEAEV]. Modified sodium adducts were also observed in the analysis of intact rBSP-(133–205) (Figures 5A and 5B). MALDI–TOF-MS analysis of the peptides generated by thermolysin and proteinase K digestion of CK2-treated rBSP-(133–205) also provides additional confirmation that Ser136 is the major site of phosphorylation within the second HAND peptide of BSP (results not shown).

View this table:
Table 3 Summary of CK2 treatment of rBSP-(133–205)

rBSP-(133–205) was treated with CK2 and digested with Asp-N. The products were determined by MS analysis.

Figure 5 MS analysis of phosphorylated BSP-(133–205)

MALDI–TOF-MS was used to determine the average mass of (A) unmodified rBSP-(133–205) and (B) CK2-treated rBSP-(133–205). Peaks observed corresponded to the analyte and its oxidized sodium adduct. (C) The phosphorylated form was digested with Asp-N and analysed using MALDI–TOF-MS in reflectron mode to identify the site(s) of phosphorylation. Masses of major peaks are shown in Table 3.

Phosphorylation of Ser136 enhances HA-nucleation potency

To determine whether phosphorylated Ser136 contributes to nucleation activity of full-length BSP, three full-length rBSP substitution mutations were developed and functionally characterized using the steady-state HA-formation assay: rBSP(S136A), rBSP(SXA) and rBSP(SALLA) (as defined in Figure 2). Note that the rBSP(SXA) mutant contains Ser136, whereas the other two mutants contain an Ala136 residue. These mutants were treated with CK2 and phosphorylation levels were verified by MALDI–TOF-MS analysis (Table 1). Each of the unmodified full-length mutants nucleated HA at concentrations as low as 0.100 nmol, similar to wild-type rBSP. However, upon treatment with CK2, the nucleation potency of rBSP(SXA) was increased (0.010 nmol), whereas those of rBSP(S136A) and rBSP(SALLA) remained unchanged (Table 1), confirming that phosphorylated Ser136 is critical for the enhanced HA-nucleation activity mediated by phosphorylated BSP. Since CK2 treatment produces variable levels of phosphorylation, and to ensure maximal phosphorylation, the CK2-treated full-length rBSP mutant rBSP(S136A) was incubated with fresh CK2 and ATP and functionally analysed in the steady-state HA-formation assay. No differences in nucleation potency were observed compared with the potencies for untreated rBSP (shown in Table 1).

Phosphorylation increases nucleation activity of the glutamate-rich region within the second HAND of BSP

To provide additional confirmation that phosphorylation of Ser136 increases the HA-nucleating activity of the adjacent glutamate-rich region, phosphorylated and non-phosphorylated peptides corresponding to the sequence predicted to be required for nucleation activity, BSP-(133–148), were synthesized. Non-phosphorylated BSP-(133–148) required 2.5 nmol to nucleate HA, whereas 1.0 nmol of phosphorylated BSP-(133–148) was sufficient (Table 1), confirming that phosphorylation of Ser136 is a key factor in nucleation potency.

BSP maintains a flexible structure

CD spectroscopy was used to investigate the secondary structures of nBSP, rBSP, derived peptides and their CK2-treated forms. A representative plot is shown in Supplementary Figure S1 (at All proteins and peptides studied showed the characteristic profile of an intrinsically unstructured protein [13]. Similar profiles were observed for CK2-treated rBSP and rBSP peptides, indicating that phosphorylation does not affect the secondary structure of BSP or BSP peptides.

MD of BSP-(133–148) and BSP-(59–74)

To provide a better understanding of the role phosphates play in matrix-mediated mineral formation, MD simulations were used to investigate the interaction between regions of BSP and the {100} face of HA. The {100} face of HA, which is calcium-rich, has previously been shown to be the crystal face that binds BSP and other mineral-associated proteins [16,41]. Both of the contiguous glutamate-containing sequences in rat BSP, BSP-(59–74) with a phosphate at Ser59 or Ser60, and BSP-(133–148) with and without a phosphorylated Ser136, were analysed. All peptides rapidly adsorbed to the crystal surface (Supplementary Figures S2A and S2B at The distance measurements of residue to crystal surface are shown in Figures 6(A) and 6(B). BSP-(133–148) shows an alternating pattern of residue interaction with the crystal surface that is more pronounced upon phosphorylation of Ser136. BSP-(59–74), based on the HA-nucleation inactive rat rBSP-(42–100) peptide [9], shows a less-pronounced pattern that is different from the nucleation-active BSP-(133–148). MD simulations of BSP-(133–148) adsorption are shown in Figures 6(C) and 6(D). While more evident for the phosphorylated peptide, both peptides adsorb in such a manner that alternate residues are close to and distant from the {100} face. Furthermore, the crystal-interacting residues show uniformity in distance of side chains to crystal face of approx. 0.5 nm, which was not observed for the non-nucleating BSP-(59–74) peptide.

Figure 6 MD simulations of BSP-(59–74) and BSP-(133–148) adsorption to the {100} crystal face of HA

(A) The vertical distance between the centre-of-mass of each side chain and the surface layer of crystal atoms was measured for BSP-(133–148) at 10 ns [grey line, MKESDEEEEEEEEEEN; black line, MKES(P)DEEEEEEEEEEN]. (B) The vertical distance between the centre-of-mass of each side chain and the surface layer was measured for BSP-(59–74) [grey line, S(P)SEEEGEEEETSNEEE; black line, SS(P)EEEGEEEETSNEEE] at 10 ns of the simulation. (C) Unmodified BSP-(133–148) and (D) phosphorylated BSP-(133–148) at 10 ns of the simulation. The atomic assignments are as follows: green, calcium; red, oxygen; orange, phosphate; grey, carbon; blue, nitrogen; white, hydrogen; and yellow, sulfur.


Our studies are directed toward understanding the mechanisms underlying bone biomineralization, and the roles ECM proteins play in mediating mineral crystal nucleation, inhibition and growth. Phosphorylation of these proteins has been shown to affect their ability to modulate mineral formation (reviewed in [42]). It is postulated that this altered activity results from changes in local protein structure and/or the addition of negative charges. Either of these systemic changes could alter protein–mineral interactions and in turn either destabilize mineral precipitates, allowing dissolution, or stabilize them allowing for nucleation. In the present study, we determined the functional significance of BSP phosphorylation and elucidated a role for region- and site-specific phosphorylation(s) in matrix-mediated mineral formation.

BSP is a known nucleator of HA. Whereas our previous studies have shown that PTMs are not an absolute pre-requisite for nucleation activity [19], phosphorylation contributes substantially to the nucleation potency of the protein. Both the number and location of phosphorylations in BSP have been studied using a variety of techniques [2022]. The consensus from these studies is that the average BSP molecule contains approx. six phosphates that are distributed among 11 serine and threonine residues. CK2 treatment of BSP results in a number of these consensus sites being phosphorylated; however, most of these appear to be in the N-terminal half of the protein [21]. Our studies agree with these findings. Salih and Fluckiger [21] found that, both in vivo and in vitro, different sites in BSP are phosphorylated at various levels. This heterogeneity probably exists in our CK2-treated species as well.

Post-translationally modified nBSP is 100-fold more active an HA nucleator than unmodified rBSP, confirming that PTMs play an important role in HA-nucleation activity. Decreased nucleation potency after phosphatase (CIP) treatment of nBSP confirms that phosphorylation enhances mineral formation capability. In agreement, CK2 treatment of unmodified rBSP increases the HA-nucleation potency 10-fold. This gain-in-function observed with CK2-treated rBSP is reversed by dephosphorylation with CIP.

Our data coincide with the findings of Boskey et al. [31], who found that inhibiting CK2 activity in limb-bud mesenchymal cell cultures significantly decreases mineral formation, attributed to decreased phosphorylation of ECM proteins. The lower mineral content is postulated to be due to a reduced number of mineral foci. Similar inferences were made by Salih et al. [43], who investigated the expression levels of BSP and CK2 in response to bone injury. These authors concluded that, during bone-wound repair, BSP and CK2 expression levels are increased leading to higher phosphorylation levels of BSP and consequently more rapid mineralization. Both studies confirm ECM protein phosphorylation as being critical in proper biomineralization of bone; specifically, BSP phosphorylation probably plays a central role in de novo mineral formation.

Two nucleation-active peptides have been generated by truncation of BSP, rBSP-(1–100) and rBSP-(133–205). rBSP-(1–100) is shown in the present study to contain nucleation activity, whereas rBSP-(133–205) has previously been described as a nucleator [9]. These peptides each contain one of the HAND regions of BSP and potential sites of phosphorylation that may alter the nucleation activity of BSP.

CK2 treatment of the first HAND peptide results in the addition of several phosphates (2.5 and 5.5); however, it does not result in an altered HA-nucleation potency. This indicates that either the phosphorylated residues responsible for the increased nucleation potency of full-length BSP do not reside within residues 1–100, or that CK2 cannot modify a critical site in HAND1. The consensus sites known to be phosphorylated in vivo appear to also be phosphorylated in vitro by CK2. Other studies have also shown that the N-terminal half of BSP contains the majority of the phosphates, but the functional implications of this have yet to be elucidated. It should be noted that a truncated variant of this peptide, rBSP-(42–100), is not nucleation-active, despite containing the first HAND of BSP [9]. Upon including the N-terminal collagen-binding domain, the first HAND becomes nucleation-active. Binding to collagen also alters the nucleation activity of BSP, increasing it 10-fold [14]. Thus it follows that the significance of the first HAND may reside in its increased nucleation potency upon binding to collagen rather than PTM-derived HA-nucleation activity.

Phosphorylation of the second HAND of BSP increases its nucleation potency 10-fold. rBSP-(133–205) is the more potent of the two BSP HANDs and contains the longer stretch of contiguous glutamic acid residues. While multiple treatments of rBSP and rBSP-(1–100) under differing kinase reaction conditions resulted in variable phosphorylation states, CK2-treatment of rBSP-(133–205) consistently resulted in an average of one phosphate per molecule. Partial digestion of CK2-treated rBSP-(133–205) with Asp-N followed by MALDI–TOF-MS confirms that Ser136 is the major site of phosphorylation on rBSP-(133–205), a finding that is consistent with those of others [21,22].

As described above, human and bovine BSPs show multiple sites of phosphorylation. Within the present study, mass spectrometric analysis of tryptic and Asp-N-derived peptides confirmed sites of phosphorylation as Ser15, Ser50, Ser51 and Ser136, in addition to other sites that were not confirmed, but are located within the peptides corresponding to residues 57–78, 157–182 and 242–273. These findings are in general consistent with other studies [2022]. Among these, Ser136 is a highly conserved site that is phosphorylated in native bone-extracted protein and by treatment with CK2 in vitro [2022]. This site is upstream of a highly conserved contiguous glutamate sequence. Acidic stretches such as these play regulatory roles in protein phosphorylation as they are often required for modification by both CK2 and Golgi casein kinase.

Ser136 was mutated to alanine within rBSP to determine the functional relevance of this phosphorylation site within the context of the entire molecule. Other serine and threonine residues found within sequences similar to the CK2 consensus sequence in HAND2 and consistent with the location of potentially phosphorylated residues within the human BSP sequence [20] were also mutated to ensure integrity of functional data. Analysis of the HA-nucleation activity confirms that phosphorylation of Ser136 is critical for increased nucleation potency of CK2-treated full-length BSP. Mutation of all potential sites, except Ser136, still resulted in equivalent nucleation potency to the wild-type CK2treated BSP. Mutation of only Ser136 and analysis of the CK2-treated rBSP(S136A) resulted in a loss of the phosphorylation-dependent increase in nucleation potency. Of relevance, mutation of Ser136 does not affect the overall phosphorylation level of the whole molecule as the mutated molecule takes up a similar number of phosphates as wild-type rBSP (Table 1).

Synthetic peptides BSP-(133–148) and phosphorylated BSP-(133–148) were also studied in the steady-state HA-nucleation assay system. As expected, HA nucleation required higher concentrations of these short peptides, however, nucleation potency was higher with the phosphorylated synthetic peptide. These results thus provide strong evidence to support the concept that Ser136 acts as a switch in modulating HA-nucleation potency in BSP.

Identification of Ser136 as a phosphorylation site critical to nucleation potency implicates the importance of local charge in crystal nucleation and/or a structural change upon phosphorylation responsible for the increased activity. Analysis by CD spectroscopy showed that there were no alterations in secondary structure after phosphorylation of BSP or BSP-derived peptides. Interestingly, CK2-mediated phosphorylation and the CK2 consensus sequence itself have been determined to be ‘helix breakers’ [44], inducing and stabilizing an extended conformation. However, it is possible that there is a change in the local structure within BSP upon phosphorylation that would not be detected with CD analysis.

Atomic-scale MD is a powerful simulation technique for predicting the interactions between small molecules [45]. Because crystals are repeating three-dimensional lattices, MD can also be used to study interactions between peptides and crystal faces. To perform such an analysis, simulations of peptide and crystal face incorporating all known characteristics (bond lengths, bond angles, etc.) are placed in a virtual simulation ‘box’ and acted upon by a force-field representing all known interactions acting within the system at the atomic level (van der Waals, electrostatics, various elastic modes, etc.) until a stable state is achieved. The work performed in the present study on BSP is similar to our previous studies in which 16-mer virtual peptides were used to investigate sequence-specific requirements of osteopontin binding to calcium oxalate crystals [6,46].

Regions of BSP from both of the HANDs, BSP-(59–74) and BSP-(133–148), were examined by MD. Crystal-bound BSP-(133–148) is in a pseudo-extended conformation. Strikingly, although less apparent for the unphosphorylated BSP-(133–148), our simulations of phosphorylated BSP-(133–148) show that every second amino acid [(P)Ser4, Glu6, Glu8, Glu10, Glu12 and Glu14] is close enough to the crystal face to form an electrostatic bond. Also, the side-chains of these amino acids are at a very uniform distance (~0.5 nm) from the crystal surface. The remaining amino acids are on the other ‘side’ of the strand, facing the solvent. Simulations of BSP-(59–74) show a quite different spatial distribution of side chains, characterized by a lack of alternation and much greater heterogeneity in distance from the crystal surface despite containing phosphates at sites known to be modified in vivo [21,22].

To our knowledge, the alternating residue pattern of adsorption predicted for BSP-(133–148) has neither been observed nor hypothesized previously. We believe that this unique mode of interaction may be critical for the HA-nucleating activity of BSP. HA-nucleation analysis of the variably phosphorylated rBSP, rBSP-(1–100), rBSP-(133–205) and rBSP(S136A) indicate that the total level of incorporated phosphate in BSP does not correlate with nucleation potency, but rather that site-specific phosphorylation results in the observed increase in activity. While studies have previously investigated the importance of phosphorylation by CK2 in cell-culture biomineralization [31,47] and the effect global protein phosphorylation has on the in vitro activity of mineral-modulating proteins [7,48], this is the first time that site- or even region-specific effects have been established related to mineral-formation activity.

In the present study we have identified that phosphorylation enhances the ability of BSP to nucleate HA in vitro. This enhancement has been mapped to Ser136, residing in the second HAND of BSP. Although there is no global change in secondary structure upon phosphorylation, MD analysis reveals an extended alternating residue pattern with equivalent side-chain distances upon binding, indicating a possible mechanism for HA nucleation by the acidic amino-acid-rich proteins found in bones and teeth.


Gurpreet Baht performed the experiments and with Graeme Hunter and Harvey Goldberg designed the study, analysed the results and wrote the manuscript. Jason O'Young and Mikko Karttunen performed and analysed the MD simulations. Antonia Borovina prepared some of the BSP mutant constructs. Hong Chen provided reagents and assisted in MS analysis. Coralee Tye provided some of the reagents and nucleation potencies. Gilles Lajoie provided guidance for peptide synthesis and mass spectrometric analysis.


This work was supported by the Canadian Institutes of Health Research [grant number MOP 93598]; and the Natural Sciences and Engineering Research Council of Canada.


We would like to thank Yinyin Liao for technical support, Dr D. Litchfield (Department of Biochemistry, University of Western Ontario, London, ON, Canada) for generously supplying pGEX-CK2α′, Jacob Turowec for his help with CK2 analysis, and Dr K.S. Gill for advice. Mass spectrometric analysis was performed in the University of Western Ontario MALDI MS Facility, peptide synthesis was conducted in the University of Western Ontario Biological MS Laboratory, and CD analysis was performed in the Biomolecular Interactions and Conformations Facility, Department of Biochemistry, University of Western Ontario. SHARCNET ( provided the computational resources.

Abbreviations: BSP, bone sialoprotein; CIP, calf intestinal phosphatase; DTT, dithiothreitol; ECM, extracellular matrix; ESI, electrospray ionization; GST, glutathione transferase; HA, hydroxyapatite; HAND, HA-nucleating domain; MALDI–TOF-MS, matrix-assisted laser-desorption ionization–time-of-flight MS; MD, molecular dynamics; nBSP, native BSP; PTM, post-translational modification; rBSP, recombinant BSP; SIBLING, small integrin-binding ligand, N-linked glycoprotein


View Abstract