Báo cáo khoa học: Role of calcium phosphate nanoclusters in the control of calcification pot

Role of calcium phosphate nanoclusters in the control of
calcification
Carl Holt
1
, Esben S. Sørensen
2
and Roger A. Clegg
1
1 Hannah Research Institute, Ayr, UK
2 Protein Chemistry Laboratory, Department of Molecular Biology, University of A
˚
rhus, Denmark
Many biological fluids, including blood, milk, extracel-
lular fluid, saliva, urine, synovial fluid and cerebrospi-
nal fluid, are usually supersaturated with respect to
hydroxyapatite (HA) [1–5], but generally remain stable.
Nevertheless, dystrophic calcification does occur, and
vascular calcification or stone-forming biofluids, for
example, have serious consequences for human health.
Genetic ablation and other experiments on individual
serum proteins have demonstrated the importance of
serum fetuin A (FETUA), osteopontin (OPN) and
matrix Gla protein (MGP) for inhibiting the precipita-
tion of calcium phosphate (CaP) in serum and prevent-
ing ectopic calcification of soft tissues [6–8]. A
metastable, colloidal, complex of CaP with FETUA,
MGP and secretory phosphoprotein 24 (SPP-24) forms
when the serum is destabilized [9,10], but the physio-
logical mechanism is still unclear.
Milk provides an example of a biofluid that seldom
forms CaP precipitates or causes dystrophic calcifica-

tion of the mammary gland, even though it may con-
tain very much higher concentrations of calcium (Ca)
and inorganic phosphorus (P
i
) than does serum [11].
In milk, casein micelles sequester CaP through phos-
phate centre (PC) sequences, typically pSpSpSEE, in
Keywords
casein; dentin matrix acidic
phosphoprotein 1; fetuin; natively unfolded
protein; osteopontin
Correspondence
C. Holt, 47 Logan Drive, Troon KA10 6PN,
UK
Tel: +44 1292 317 615
E-mail:
(Received 21 November 2008, revised 17
January 2009, accepted 11 February 2009)
doi:10.1111/j.1742-4658.2009.06958.x
Calcium phosphate nanoclusters are equilibrium particles of defined chemi-
cal composition in which a core of amorphous calcium phosphate is
sequestered within a shell of casein phosphopeptides. Sequence analyses
and a structure prediction method were applied to secreted phosphopro-
teins of known importance in controlling calcification, and eight noncasein
phosphoproteins were identified as containing one or more subsequences
capable of forming nanoclusters. Small-angle X-ray scattering was used to
confirm that a plasmin phosphopeptide of one of the identified proteins,
osteopontin, formed a novel type of calcium phosphate nanocluster in
which the radius of the amorphous calcium phosphate core was four times
larger than is typical of casein nanoclusters. A thermodynamic treatment

of nanocluster formation identified the factors of importance in determin-
ing the equilibrium size of the core, and showed how a nanocluster solution
could be thermodynamically stable yet supersaturated with respect to the
mineral phase of bones and teeth. It is suggested that the ability of some
secreted phosphoproteins to form nanoclusters is physiologically important
for the control or inhibition of calcification in soft and mineralized tissues,
the extracellular matrix and a wide range of biofluids, including milk and
blood.
Abbreviations
ACP, amorphous calcium phosphate; CaP, calcium phosphate; CPN, calcium phosphate nanocluster; DCPD, di-calcium phosphate di-hydrate;
DMP1, dentin matrix acidic phosphoprotein 1; FETUA, fetuin A; HA, hydroxyapatite; MGP, matrix Gla protein; OCP, octacalcium phosphate;
OPN, osteopontin; PC, phosphate centre; pS, phosphoseryl residue; RBP, riboflavin-binding protein; SAXS, small-angle X-ray scattering;
SCPP, secretory calcium-binding phosphoprotein; SP, secreted phosphoprotein; SPP-24, secretory phosphoprotein 24.
2308 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
a
S1
-, a
S2
- and b-caseins. Understanding the sequestra-
tion process has been furthered through studies with
short casein phosphopeptides containing a PC. Thus,
the 25-residue N-terminal b-casein tryptic phosphopep-
tide (b-casein 1–25) sequestered CaP to form a calcium
phosphate nanocluster (CPN) [12–14] with a core of
amorphous, acidic and hydrated calcium phosphate
(ACP) of radius 2.4 nm surrounded by a shell of about
50 phosphopeptides with a thickness of 1.6 nm. Ini-
tially, it was thought that the CPNs were metastable
particles in a state of arrested precipitation, but it was
later shown that they were equilibrium particles with a

defined composition, size and structure. Most signifi-
cantly, they formed spontaneously when the phospho-
peptide was added to a pre-existing precipitate of
ACP. There is abundant evidence from infrared
spectroscopy, X-ray absorption spectroscopy, X-ray
and high-resolution electron diffraction and solid state
31
P-NMR spectroscopy that micellar CaP and the core
CaP of CPNs are amorphous. Thus, in terms of size,
structure, solubility and dynamics, the micellar CaP
and core CaP of CPNs appear to be very similar
[12–20].
The primary purposes of this investigation were to
provide a deeper understanding of the thermodynamics
of CaP sequestration and to define more closely the
structural characteristics of the phosphoproteins
responsible. A second aim was to identify a group of
proteins with the sequence and conformation predicted
to be needed for CaP sequestration and to undertake
an experimental test of the prediction for one of them.
For the experimental work, OPN was selected because,
unlike the caseins, it is expressed in a wide range of
species, tissues and biofluids [21,22]. A successful dem-
onstration would be a step towards establishing the
broader physiological importance of CPN formation.
OPN is a member of the same paralogous group as
the caseins, called the secretory calcium-binding phos-
phoproteins (SCPPs) [23,24]. Like the caseins, it has an
unfolded conformation [25] and clustered sites of phos-
phorylation [26], and among its many recognized func-

tions is an involvement in the control of mineralization
processes [21,22].
Results
Thermodynamics of CPN formation
Doc. S1 (see Supporting information) provides addi-
tional details of the treatment. The chemical formula
of an electroneutral CPN can be written as a multiple
of an empirical formula, or ‘monomer’ containing a
single PC:
Ca
R
Ca
H
R
H
ðP
i
Þ
R
P
ðH
2
OÞ
R
W
ðPep À PCÞ
1
hi

j

ð1Þ
The average molar ratios of water, Ca and P
i
to PC
are R
W
, R
Ca
and R
P
, respectively,

j is the average
number of PCs in the CPN and Pep is the chemical for-
mula of the peptide divided by the number of PCs it
contains (f). The formula of the monomer can be
further divided into an amorphous hydrated CaP and a
sequestering ligand of calcium phosphopeptide. The
empirical chemical formula of the electroneutral CaP is
CaðHPO
4
Þ
y
ðPO
4
Þ
2À2y
3
:xðH
2

OÞð2Þ
where 3y ⁄ (2 + y) is the mole fraction of P
i
in the
di-anionic form. The empirical chemical formula of
CaP can then be used to define a type of solubility
constant K
S
as an ion activity product. In a dilute
solution in which the activity of water is effectively
unity:
K
S
¼ a
1
Ca
2þ
a
y
HPO
2À
4
a
ð2À2yÞ=3
PO
3À
4
ð3Þ
K
S

can be used, just like the solubility product of a
pure bulk phase, to calculate the extent of formation
of CPNs.
The association of CaP monomers generates an
equilibrium distribution of core sizes, and it can be
shown by a simple adaptation of the capillary theory
of nucleation that an activity distribution results with
a modal core radius of:
r
Ã
core
%
2kDG
seq
3A
core
RTlnða
1
=a
s
Þ

3V
core
4p

1=3
¼
2kDG
seq

À3A
core
DG
o
core

3V
core
4p

1=3
ð4Þ
where V
core
is the empirical formula volume of CaP,
k ¼ð36pV
2
core
Þ
1=3
, A
core
is the core surface area per PC,
DG
seq
is the free energy of sequestration of the core by
the shell of peptides, a
1
and a
s

are the activities of a
CaP molecule in the nanocluster solution and in a
solution saturated with respect to the bulk phase of
core material, respectively, and DG

core
is the free energy
of formation of the bulk core phase.
As r* must be a positive real number, two possible
solutions exist. In classical nucleation theory, the sur-
face energy and bulk free energy terms are positive;
precipitation occurs from a supersaturated solution in
which a
1
> a
s
. In the formation of CPNs, the effective
surface energy is negative, and hence the solution is
undersaturated with respect to the bulk phase of ACP
(a
1
< a
s
).
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2309
Stability and metastability in biofluids and the
extracellular matrix
Freshly formed ACP can be sequestered by phospho-
peptides but, if the rate of ACP formation and matu-

ration is faster than the rate of sequestration, the
nanoclusters cannot form and a metastable solution
results. Certain partial SP sequences have been identi-
fied as the starting point of controlled crystal growth
in the extracellular matrix of mineralized tissues. These
include long phosphorylated sequences in, for example,
phosphophoryn, the C-terminal sequence of OPN and
the N-terminal sequence of dentin matrix acidic phos-
phoprotein 1 (DMP1) [27,28] and long sequences of
Glu residues in, for example, integrin-binding sialo-
phosphoprotein II [29]. When a sequence that can
sequester ACP and a sequence that can accelerate the
maturation of ACP into HA are both present in a
given SP, the competing reactions of ACP maturation
and ACP sequestration may make the formation of
CPNs as the equilibrium product more difficult or even
impossible. The formation of the nanocluster solution
requires not only that maturation of the ACP should
be prevented, but also a stoichiometric excess of the
phosphopeptide over CaP. If [p] molÆL
)1
of P
i
can
precipitate as ACP from the initially supersaturated
solution, the condition for thermodynamic stability is
a ¼
½p
f ½PPR
P

1 ð5Þ
where [PP] is the phosphopeptide concentration. Under
these conditions, a is also the fraction of reacted PCs.
Although a nanocluster solution is stable with
respect to the formation of ACP, it remains supersatu-
rated with respect to HA (Fig. 7C). HA has never been
observed to nucleate directly from solution, but forms
by a solution-mediated maturation of ACP [30] and,
as the latter cannot form, the nanocluster solution is
stable with respect to this phase also.
Identification of sequestering phosphoproteins
Identification of PCs in secreted phosphoproteins (SPs)
The canonical PC used in the search was derived from
the known casein PCs, and comprised a sequence of 10
or fewer consecutive residues containing at least three
sites of phosphorylation, no Cys and fewer than three
hydrophobic residues. Example PC sequences found in
SPs with known involvement in mineralization are
shown in Table 1, and aligned sequences of their
orthologues are given in Doc. S2 (see Supporting infor-
mation). Most of the identified SPs and all of the proven
CPN-forming SPs are members of the SCPP paralogous
group. Most PCs contain a block of consecutive phos-
phorylation sites, followed by the primary recognition
site of the casein kinase 2 or Golgi kinase. The longest
block of consecutive sites of phosphorylation in a casein
PC is in rat a
S1
-casein with eight, with a ninth close by.
Longer sequences of phosphorylated residues, such as

those found in phosphophoryn and the C-terminal half
of OPN and N-terminal part of DMP1, have been
shown to promote the maturation of ACP into
more crystalline phases, and so were discounted as
CPN-forming sequences. A minor PC pattern involves
three or more repeats of a primary kinase recognition
triplet SXE (MGP) or SD[E,pS] (OPN). When the
aligned orthologue sequences were examined (Doc. S2),
it was found that not all PCs were conserved, particu-
larly when a protein contained more than one PC. For
example, the N-terminal half of bovine OPN contained
all three PCs coded by exons 3, 5 and 6. The last two
were not as highly conserved as the first, but none of the
orthologues had fewer than two PCs.
Table 1. Identified PC sequences formed by the action of the Golgi
kinase and casein kinase 2 on selected secreted phosphoproteins.
CSN1S1, a
S1
-casein; CSN1S2, a
S2
-casein; CSN2, b-casein; IBSP-II,
integrin-binding sialophosphoprotein II; MEPE, matrix extracellular
bone phosphoglycoprotein. Potential sites of phosphorylation are
shown in bold.
Protein Species Swiss-Prot No. PC
a
SCPPs
OPN Cow P31096 6- TSSGSSEEKQ -15
42- QNSVSSEETD -51
99– SDESHHSDES -108

DMP1 Mouse O55188 8- NTESESSEER -17
28- PTNSESSEES -37
49- HTHSSESGEE -58
120-SADTTQSSED -129
142-SDSKDQDSED -151
161-DSAQDSESEE -170
CSN1S1 Guinea pig P04656 19- SSSSSSSEER -28
54- IISESTEERE -63
65- SSISSSEEV -73
CSN1S2 Pig P39036 5- EHVSSSEESI -14
54- ASSSSSEESV -63
130- ELSTSEEPVS-139
CSN2 Human P05814 5- ESLSSSEESI -14
IBSP-II Human P21815 55- GDDSSEEEEE -64
MEPE Human Q9NQ76 498- DS
GSSSESDG -507
Non-SCPPs
FETUA Human P02765 307- SLGSPSGEVS -316
SPP-24 Human Q13103 108-SSSTSESYSS -117
MGP Human P08493 2- ESHESMESYE -11
PRB4 Human P10163 2- SSSEDVSQEE -11
RBP Chicken P02752 192-ESSSMSSSEE -201
a
Sequence numbers are for the mature peptide chain without the
signal sequence.
Calcium phosphate sequestration by osteopontin C. Holt et al.
2310 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
Conformation of secreted phosphoproteins containing
PCs
The PONDRÒ predictor is the oldest and most thor-

oughly tested of the predictors of partial or complete
disorder in proteins. It continues to perform well in
comparative tests with more recent methods [31], and
is one of the components in the most recent meta pre-
dictor, metaPrDOS [32]. According to PONDRÒ pre-
dictions, the positions of PC sequences in the SPs in
Table 1 were, with the exception of the globular pro-
tein riboflavin-binding protein (RBP), disordered, and
had disordered flanking sequences (Fig. 1A,B). The PC
motif of RBP was disordered and is undefined in the
crystal structure [33], but its N-terminal flanking
sequence was correctly predicted to be ordered. The
prediction for FETUA indicated a folded N-terminal
sequence containing the two cystatin-like domains, but
a flexible C-terminal half in which the PC lies. The
result for SPP-24 was the least clear-cut with only
short disordered sequences flanking the PC. Essentially
the same results were obtained by the top-idp predic-
tor [34], with the notable exception that SPP-24 was
borderline stable near the PC and stable in its flanking
sequences (Fig. 1C), but the metaPrDOS predictor [32]
agreed better with the PONDRÒ result for this protein
(Fig. 1D). All methods were in agreement in showing
that OPN has little or no stable conformation, and
hence can be described as a worm-like, or rheomorphic
[35], chain.
With the exception of proline-rich protein 1, all
other members of the SCPP paralogous group identi-
fied by Kawasaki and Weiss [24,36] were predicted by
PONDRÒ to be flexible over a substantial fraction of

their total sequence (results not shown).
Characterization of OPN and OPN 1–149 in free
solution
Small-angle X-ray scattering (SAXS) of OPN and
OPN 1–149
Both OPN and OPN 1–149 showed the scattering pat-
tern expected of a flexible but non-Gaussian chain with
short, rod-like segments (Fig. 2). The average of three
determinations of the radii of gyration of OPN and
OPN 1–149 in the concentration range 5–15 mgÆmL
)1
were 5.50 ± 0.17 and 2.17 ± 0.24 nm, respectively.
The worm-like chain model fitted to the OPN SAXS
gave b = 1.74 nm, which could correspond, for exam-
ple, to an average of five to six residues temporarily
arranged in a poly-l-proline II local helix. The lower
chain stiffness of OPN 1–149 (Fig. 2) is possibly a
result of the higher proportion of Pro residues in this
part of the sequence (eight of the total of 13), each of
which produces a sharp change in chain direction in
the cis configuration, and of Gly residues (four of
four), which allow markedly more chain flexibility than
other residues because of their short side-chain. Apart
from Asp, the other residues are present in similar pro-
portions in the two halves of OPN. It is possible,
therefore, that both OPN and OPN 1–149 contain sim-
ilarly sized runs of local poly-l-proline II structure
but, in the latter, the frequency of hinge residues is
greater.
Microcalorimetry of OPN 1–149

The thermogram shown in Fig. 3 shows an almost per-
fectly smooth increase in specific heat with temperature
in accord with the SAXS observations of a worm-like
chain and consistent with the low chemical shift
dispersion in
1
H-NMR spectra of OPN [25].
Binding of Ca ions to OPN 1–149
Three pK values and three Ca ion association con-
stants were allowed to vary during the fitting to the
experimental isotherms of the b-casein 1–25 peptide,
and the resulting fitted curves are shown in Fig. 4. The
three Ca ion association constants obtained were 3000,
400 and 30 m
)1
. The single phosphoseryl residue (pS)
had an effective pK value of 6.0 and the cluster of
three pS residues ionized with a pK value of 7.2. The
OPN 1–149 isotherm, also shown in Fig. 4, was fitted
by two Ca ion association constants of 3000 (dianionic
phosphate) and 30 m
)1
but, because it does not have
the triplet of pS residues, two pK values of 6.4 and 5.0
were required.
Formation of OPN 1–149 nanoclusters
OPNmix and OPN 1–149 were able to sequester CaP
to form nanoclusters, but OPN could not, suggesting
that the extended phosphorylated sequences in the
C-terminal half either were too large to form PCs or

the sequence catalysed the maturation of ACP into
more crystalline phases. Using the simple mixing
method at a peptide concentration of 30 mgÆmL
)1
of
OPNmix, there was no initial precipitation, even with
a single addition of the P
i
stock, provided that it was
added slowly with good stirring. The initial turbidity
slowly disappeared over about 1 week to give a slightly
opalescent solution, comparable to that of CPNs pre-
pared by the urea ⁄ urease method. When the peptide
concentration was reduced to below 10 mgÆmL
)1
,
an initial precipitate or turbid colloidal suspension
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2311
developed which did not fully redisperse on standing.
If, however, further peptide was added to a final con-
centration of 30 mgÆmL
)1
, soon after the development
of the initial precipitate, the solution clarified com-
pletely over about 1 week. However, if the addition of
the phosphopeptide was delayed, or if the initial pep-
tide concentration was below 5 mgÆmL
)1
, complete

redispersion was not achieved, even after 4 months.
These experiments demonstrated that, like the casein
CPNs, the OPN 1–149 CPNs can be formed by either
a forward reaction from a supersaturated solution or
by a back reaction from a two-phase system containing
a precipitate of ACP and sufficient sequestering pep-
tide to convert all the ACP to CPNs. Neither casein
nor OPN phosphopeptides could form the nanoclusters
from partially matured ACP.
Characterization of OPN nanoclusters
SAXS of OPN 1–149 nanoclusters prepared by the
urea ⁄ urease method
The results of the SAXS measurements on CPN subs-
amples, measured as a function of time after the addi-
tion of urease, are summarized in Fig. 5A,B. The first
AB
C
D
Fig. 1. Prediction of disorder as a function of residue position in SPs having known or potential PC sequences. The positions of known or
predicted PCs in the sequence are shown as full lines. (A) PONDRÒ predictions for SCPPs in Table 1. (B) PONDRÒ predictions for the other
secreted phosphoproteins in Table 1. (C) TOP-IDP predictions for h-OPN and h-SPP-24 plotted as the midpoint of a window of 51 residues.
(D) metaPrDOS predictions for h-OPN and h-SPP-24.
Calcium phosphate sequestration by osteopontin C. Holt et al.
2312 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
two subsamples were taken after 17 min, when the pH
was 6.82, and after 50 min, when the pH was 6.87,
but, by the third sample, the pH was essentially con-
stant and close to 7.0. Strongly scattering spherical
particles formed from an initial state dominated by the
scattering of a statistical polymer but, after about

2 days, the scattering profile showed hardly any fur-
ther change, as demonstrated by a measurement
5 months later.
SAXS of the matured system was modelled as a
mixture of free peptide and CPNs, as shown in Fig. 5C.
The worm-like chain representation of the free peptide
was used with the assumption that the PCs on the same
peptide all react together to give either fully bound or
fully free peptide, so that the fraction of free peptide
equals the fraction of unreacted PCs. The weighted
subtraction produced a scattering curve which is
characteristic of spherical, but polydisperse, particles
with a corona of statistical scattering elements. The
Gaussian copolymer micelle model of Pedersen and
Gerstenberg [37] with a log-normal size distribution
produced a reasonably close representation of the
scattering of the CPNs, although the OPN peptide
chains in free solution deviated from true Gaussian
behaviour.
Electrophoretic light scattering by nanoclusters
The maturation of a CPN solution prepared by the
rapid urea ⁄ urease method using the b-casein (f1-25)
phosphopeptide is shown in Fig. 6A. At pH 5.5, before
the urease was added and below the point at which
CPN formation begins, the intensity of scattered light
was low and the solution was apparently unchanged.
Nevertheless, inversion of the correlation function gave
an intensity-weighted size distribution of colloidal
particles, almost certainly CaP formed at the time of
mixing, as the solution is undersaturated with respect

to ACP at pH 5.5. All other results in Fig. 6A were
recorded after the final pH value of 7.0 was attained.
A progressive loss of colloidal particles at the expense
of the CPN component occurred as the solution
matured. The intensity distribution of a similar solu-
tion that had been stored at ambient temperature and
pH 7 for 1 day showed that the colloidal particles were
nearly absent. In another experiment, CPNs prepared
with a mixture of casein phosphopeptides [38] by the
simple mixing method were compared with those made
by the urea ⁄ urease method. The turbidity A
1 cm
600 nm
ÀÁ
of
Fig. 2. Kratky plots of the SAXS of OPN (in 20 mM P
i
buffer,
pH 7.0, ionic strength 80 m
M) and of OPN 1–149 (in the CaP dilu-
tion buffer used in the nanocluster experiments). Fitted curves are
from the worm-like chain model. Each set of results has been
scaled by the mean square radius of gyration determined by the
fitting procedure.
Fig. 3. Normalized differential scanning calorimetry thermogram of
OPN 1-149 at pH 7.0.
Fig. 4. Ca-binding isotherms of b-casein 1–25 as a function of pH
and of OPN 1–149 at pH 7.0.
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2313

the CPN solution made by the first method fell
from 0.017 to 0.003 over 5 days to equal that of the
CPN solution prepared by the urea ⁄ urease method,
which showed no change in absorbance over time.
The hydrodynamic radii of filtered solutions after
equilibration for 5 days were 6.05 and 6.75 nm for the
first and second methods, respectively (results not
shown).
The OPN 1–149 CPN had a hydrodynamic radius of
21.9 nm after 2 days of equilibration, whether pre-
pared by the urea ⁄ urease method or the simple mixing
method, although the mixing method produced an
initial slight precipitate which quickly dispersed,
confirming that an equilibrium size was attained. The
hydrodynamic radius is comparable with the radius of
gyration determined by SAXS. In the intensity-
weighted size distribution of the unfiltered OPN 1–149
CPN (Fig. 6B), there was a very small peak of much
larger particles which could be removed by filtration
through a 0.2 lm filter. The origin of these larger par-
ticles may have been the result of a very small amount
of cross-linking between nanoclusters produced by the
trifunctional peptides or of unequilibrated colloidal
CaP particles. Another peak, contributing 8.5% to the
total scattered intensity, on the low side of the main
CPN peak, corresponded to the hydrodynamic size of
the free peptide. The electrophoretic mobility of the
OPN 1–149 CPN was 1.4 lmÆs
)1
ÆV

)1
Æcm. According to
the Henry equation [39], it corresponds to a f potential
of )15.4 mV.
A
B
C
D
Fig. 5. Study by SAXS of the maturation of nanoclusters prepared with OPN 1–149 by the urea ⁄ urease method. (A) Effect of time on the
radius of gyration determined by the Guinier method. (B) Normalized, q
2
-weighted SAXS of the nanoclusters diluted to 5 mg Æ mL
)1
after the
given times. (C) Model of the scattering of the matured nanocluster solution as a mixture of scattering from copolymer micelle-like nanoclus-
ters and free peptide. The scattering of the nanoclusters was obtained by subtracting the scattering of the free peptide from the total scat-
tering. Model calculations used the parameters b = 0.07 nm, A
core
= 0.25 nm
2
, r
o
= 12.5 nm, b = 0.35. (D) Representation of an OPN 1–149
nanocluster. An eighth section of the spherical core of ACP is shown. Surrounding the core is a shell of OPN 1–149 molecules, each
anchored to the core through its three PCs. For clarity, only one phosphopeptide molecule is shown. The mesh illustrates the position of the
surface of shear, which determines the hydrodynamic radius of the nanocluster. The diagram is scaled to give approximately correct impres-
sions of the relative magnitudes of A
core
, r
g,peptide

and r
h
for a core radius of 12.5 nm.
Calcium phosphate sequestration by osteopontin C. Holt et al.
2314 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
Calculation of the ionic equilibria and partition of salts
in OPN 1–149 and OPNmix nanocluster solutions
An invariant ion activity product in the ultrafiltrates
was found for a TCP stoichiometry (y =0,
K
S
= 7.6 · 10
)10
m
1.66
, results not shown). This is a
more basic ACP than was found in the casein CPNs,
which have y = 0.4 [13]. Below pH 5.97, no CPNs
could form because the ion activity product was below
K
S
. Above pH 5.97, the extent of reaction of PCs with
ACP was found which allowed the ion activity product
in the CPN solution to equal K
S
. The casein CPN val-
ues for R
Ca
and R
P

were used, and peptide binding
was calculated on the assumption that all the peptides
in the OPNmix sample had the same binding isotherm
as OPN 1–149. The complete model of ionic equilibria
was then used to calculate the composition of an equi-
librium diffusate, so that it could be compared with
the composition of the experimental ultrafiltrate
A
B
Fig. 6. Intensity distribution curves derived from the dynamic light
scattering measurements. (A) Unfiltered nanocluster solution
prepared with b-casein (f1–25) by the urea ⁄ urease method from an
initial pH of 5.5 to a final pH of 7.0. The larger particles observed at
pH 5.5 are probably colloidal ACP formed during mixing, which
gradually dissolve at the expense of the nanoclusters formed above
pH 6. (B) Mature OPN 1 149 nanoclusters.
A
B
C
Fig. 7. Calculated properties of OPNmix nanocluster solutions. (A)
Comparison of calculated ultrafiltrate concentrations of P
i
, Ca and
free Ca
2+
with experimental values shown as symbols. (B) Calcu-
lated fraction of reacted PCs. (C) Log of the saturation index versus
pH for DCPD, OCP and HA.
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2315

(Fig. 7A). The general agreement of the model with
experiment is evident. Figure 7B shows how the calcu-
lated extent of reaction of the PCs varied with pH.
The saturation indices, defined as the ratio of the ion
activity product to the solubility product, for di-cal-
cium phosphate di-hydrate (DCPD), octacalcium phos-
phate (OCP) and HA are shown in Fig. 7C. Above
pH 5.97, the CPN solutions were undersaturated or
only slightly supersaturated with respect to DCPD and
OCP, but over the entire pH range, the nanocluster
solution was highly supersaturated with respect to HA.
In addition to the work with the OPNmix sample, a
single determination was made of the partition of salts
in a CPN solution at pH 7.0 prepared with the pure
OPN 1–149 peptide. The experimental and, in paren-
theses, model, ultrafiltrate concentrations of P
i
, Ca and
Ca
2+
were 12.1 (10.5), 1.4 (1.1) and 1.1 (0.65) mm,
respectively, which compare quite closely with the
values obtained with the OPNmix nanoclusters.
Discussion
Structure of CPN-forming phosphopeptides and
phosphoproteins
Detailed structural studies on CPNs have been made
using purified short peptides of lengths between 21 and
42 residues, namely a
S1

-casein 59–79 and b-cas-
eins 1–25 and 1–42. The results from the present work
utilized a peptide of 149 residues, and it is most likely
that the individual micellar CaP particles comprise the
core of equilibrium complexes formed from proteins of
more than 200 residues. It may be concluded that the
length of the peptide or protein is not an important
consideration. The OPN plasmin peptide has no signif-
icant sequence similarity to any casein sequence out-
side of the PCs. Flanking sequences of all the SPs in
Table 1 are deficient in hydrophobic residues and Cys,
and so they tend to have a low degree of sequence
complexity and favour an unfolded conformation.
On the larger scale, all PC-containing SCPPs and
the non-SCPPs proline-rich basic phosphoprotein 4
and MGP are known or predicted to be unfolded over
most or all of their length. The absence of a globular
structure close to the surface of the core allows a
higher density of PCs to bind to the surface, and so
clearly a fully globular protein is at a disadvantage.
The unfolded conformation may also allow a faster
rate of CaP sequestration, which may be of importance
when the rate of maturation of ACP nuclei is compa-
rable with the rate of sequestration. Nevertheless, it
can be envisaged that a globular domain, if it has an
extended, flexible, linker sequence connecting it to a
PC, could be just as effective as a natively unfolded
protein or short peptide. FETUA, with two cystatin-
like domains in the N-terminal half, and SPP-24, with
one, are predicted to have part of their sequence

remote from the PC in a more stable globular confor-
mation. If it can be demonstrated that these proteins
are also able to sequester CaP through their PCs, the
requirement for an unfolded flexible conformation
could be limited to a more restricted region adjacent
to the PC.
Thermodynamic stability of the OPN 1–149
nanocluster solution
CPNs could be prepared with OPN 1–149 by either
the urea ⁄ urease method or simple mixing and, after a
few days of maturation, during which the turbidity
decreased to a constant, low value, achieved an equi-
librium size which did not change in the following
5 months of storage. The results shown in Fig. 6A and
the changes in turbidity with time show that particles
larger than the equilibrium size were produced during
mixing and, to a lesser extent, by the urea ⁄ urease
method, but during maturation, the larger particles
disappeared at the expense of CPNs (Fig. 6A); the
same equilibrium size was achieved whichever method
was employed to make CPNs. Although the nanoclus-
ter solution was stable, the ion equilibria calculations
showed that it was highly supersaturated with respect
to HA; however, as this phase can only form via solu-
tion-mediated transformation of ACP, there is no
means by which it can be generated when there is a
sufficient excess of the sequestering peptide.
Core shell structure of the OPN 1–149
nanocluster
The radius of gyration of the peptide on the core sur-

face was about one-third of its value in free solution,
and this can be understood qualitatively if it is
assumed that the peptide is attached to the surface
through the three PCs (Fig. 5D). Compared with the
casein CPNs, the core CaP is more basic, correspond-
ing to the empirical chemical formula of TCP, and
nearly four times larger, but the molar ratios of Ca or
P
i
to PC were calculated to be the same. Most proba-
bly, the core is simply more hydrated. According to
Eqn (4), the size is determined mainly by the ratio of
the free energy of sequestration to the free energy of
formation of the bulk core phase and the core surface
area per PC. The latter was found to be 0.25 nm
2
,
which is about one-quarter of that for the
b-casein 1–25 CPN, and so this alone could account
Calcium phosphate sequestration by osteopontin C. Holt et al.
2316 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
for the difference. It is more probable that the differ-
ence in chemical composition and hydration in the
core affects the two free energy terms equally, so that
their ratio is unchanged.
Notwithstanding the difference in hydration in the
core, it is most probable that the core is amorphous,
similar to CaP in casein micelles and the core CaP of
casein CPNs, otherwise the particles would not have
equilibrated to a path-independent constant size.

Moreover, highly phosphorylated OPN, like casein, is
a very powerful inhibitor of ACP maturation, even at
much lower concentrations than those employed here
[40].
Nonequilibrium, pathway and time-dependent phe-
nomena are commonly observed in CaP precipitation
from solution at near-neutral or alkaline pH, and the
usual product is a poorly crystalline HA or OCP phase
(Fig. 8A). Numerous reports exist of the effects of
phosphoproteins or phosphopeptides on the lag time
before precipitation, the rate and extent of precipita-
tion and rate of conversion of ACP into more crystal-
line phases (recently reviewed by George and Veis
[41]). Invariably, the studies have been made under
conditions in which there is a large molar excess of
CaP over the peptide [in Eqn (5), a ) 1], so that the
results involve metastable phases or metastable colloi-
dal solutions, some with very long lifetimes (Fig. 8C).
When much higher concentrations of phosphopeptide
are employed, such that 0 < a £ 1, the maturation of
ACP may be completely inhibited and, provided that
the free energy of sequestration by the phosphopeptide
is sufficiently high, it can form the equilibrium com-
plexes called CPNs (Fig. 8B).
Is CaP sequestration to form equilibrium nanocl-
usters of broad physiological importance?
The properties of nanocluster solutions that can be
exploited in biofluids are, firstly, that they are ther-
modynamically stable, so that mineralization of soft
tissues should not occur. Second, when a fresh ecto-

pic deposit of ACP does form, it can be removed by
an excess of the sequestering protein or peptide.
Third, in contact with hard tissue, the nanocluster
solution cannot cause demineralization and could
indeed act as a reservoir of CaP for crystal growth
or tissue remineralization. Fourth, Eqn (5) places no
upper limit on the concentrations of Ca and P
i
in
the fluid. For example, the free Ca ion concentra-
tions and supersaturation with respect to HA in milk
A
BC
Fig. 8. Schematic drawing of the alternative fates of ACP nuclei formed from a supersaturated solution. (A) In the absence of a competent
sequestering peptide [i.e. a in Eqn (5) is infinite], ACP nuclei grow and mature into a crystalline or poorly crystalline calcium phosphate;
under physiological conditions, the final state is usually poorly crystalline OCP or HA or, in the case of tooth enamel, highly crystalline HA.
(B) In the presence of a stoichiometric excess or equivalence of PCs (0 < a £ 1), a thermodynamically stable solution of CPNs may form if
all the CaP is sequestered by the competent SPs. The CPNs have a defined composition and size at equilibrium. If some of the nuclei
escape sequestration to grow and mature to a poorly crystalline state, they cannot subsequently form the equilibrium nanoclusters. (C) In
the presence of a substoichiometric concentration of competent SPs (1 < a < ¥), the growth and maturation of the ACP nuclei may be slo-
wed to give a metastable colloidal suspension or precipitate of complexes of variable stoichiometry, size and degree of crystallinity.
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2317
remain comparable with those in blood, even though
the total Ca concentration in milk may be two
orders of magnitude higher. Fifth, there is scope for
an exquisite degree of control of mineralization
through the degree of phosphorylation of the compe-
tent SPs, particularly when, as in OPN and DMP1,
they have opposing functional subsequences that can

be separated by proteolytic cleavage.
Price and coworkers [6,9,42,43] have described the
composition and properties of a complex of CaP
with FETUA, MGP and SPP-24 in the serum of rats
treated with the bisphosphonate etidronate or in
serum treated with 10 mm additional Ca and P
i
to
induce the formation of the complex. The molar
concentration ratio of P
i
to PC was at least two
orders of magnitude higher than needed to form
CPN particles (i.e. a ) 1). According to Heiss et al.
[10], inhibition of CaP precipitation by FETUA is
caused by the transient formation of colloidal
spheres of 30–150 nm in diameter, which are initially
amorphous, but become progressively more crystal-
line with time and eventually form a precipitate.
Although Heiss et al. [10] could not isolate the
complex from normal serum, an artificial serum con-
taining a similar concentration of FETUA (10 lm)
and buffered Ca and P
i
solutions remained in a
metastable state for about 6 h before precipitation,
which they considered long enough for the complex
to be removed from the circulation. Although
these studies produced metastable products, it is
nevertheless possible that, under conditions in

which a < 1, the product of interaction between
FETUA and CaP may be a thermodynamically
stable serum.
In conclusion, the present findings provide the first
description of equilibrium CPN formation by a phos-
phopeptide that is widely distributed among extant
species, tissues and biofluids, and was one of the earli-
est members of the paralogous group of SCPPs to
emerge in the late Cambrian together with CaP tissue
mineralization [23,24]. Other members of the paralo-
gous group and some non-SCPPs (Table 1) have been
identified as having the potential PCs and predicted
flexible conformations that are the hallmark of CPN-
forming phosphopeptides. One or more of these pro-
teins is physiologically important in the tissues of
bone, dentine, cementum and osteoid, or is secreted
into biofluids, such as blood, milk, saliva and urine.
Our contention is that among the physiologically
important functions of the non-casein SPs of Table 1
are presently unrecognised ones that involve the
formation of thermodynamically stable complexes such
as CPNs.
Experimental procedures
Sequence analyses
Identification of PCs in secreted phosphoproteins
Searches for PCs were made by manual and automated
methods in the sequences of SPs, known to the authors to
be involved in CaP mineralization processes, using the Uni-
Prot (=Swiss-Prot + TrEMBL) database on the ExPASy
server () of the Swiss Institute of

Bioinformatics. Alignment of orthologous sequences and
the generation of general motifs used the clustalW2
method [44] and the pattern searching routine pratt [45],
respectively, both implemented on the European Bioinfor-
matics Institute server (). Because of
their remarkable sequence variability, low complexity and
the rarity of high scoring residues, standard scoring matri-
ces were not found to work well without the use of supple-
mentary constraints [35]. Alignments were made of
sequences coded by individual exons or neighbouring exons,
making use of the tendency for the splice junctions to be
conserved. In addition, known and predicted sites of phos-
phorylation were edited to become Cys residues in order to
give them a higher alignment score than Ser or Thr. Pre-
dicted sites of phosphorylation were identified according to
the consensus rules for phosphorylation by the Golgi kinase
and casein kinase 2. For comparative purposes, chicken
RBP was included, although it is not presently known to be
involved in mineralization, and is globular rather than
unfolded.
Prediction of flexible sequences
Predictions were made for the unphosphorylated SPs of
Table 1 using the PONDR
Ò
VL-XT predictor (http://
www.pondr.com/), which integrates three feed-forward
neural networks: the VL1 predictor, the N-terminus pre-
dictor (XN) and the C-terminus predictor (XC). Predic-
tions of long flexible regions of 40 or more residues with
a score of 0.5 are considered to be more reliable than

shorter regions. Predictions were checked against the more
recent predictor from the same group, TOP-IDP [34], and
human OPN and SPP24 sequences were submitted to a
recent meta predictor, metaPrDOS, which, on this occa-
sion, employed an optimized combination of five predic-
tors [32].
Fractionation of the OPNmix sample by gel filtration
chromatography
An OPN fraction (OPNmix) was isolated from bovine
milk by the method of Sorensen et al. [26] It comprised
better than 95% OPN or OPN peptides, phosphorylated
to some degree at all sites. It was fractionated further by
Superdex 75 gel filtration chromatography using a Pharmacia
Calcium phosphate sequestration by osteopontin C. Holt et al.
2318 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
XK 16 column (Pharmacia Ltd., Sandwich, UK) with
a bed length of 64 cm at a flow rate of 0.3 mLÆmin
)1
.
The sample of 10 mg was dissolved in 1 mL of elution
buffer (50 mm phosphate, 300 mm NaCl, 0.02% NaN
3
,
pH 7.0) and dialysed overnight against 120 mL of elution
buffer before loading on to the column. Fractions of
1 mL were collected and examined by SDS–Mops gel elec-
trophoresis before pooling into four fractions (F1, F2, F3a
and F3b) which were dialysed exhaustively against deion-
ized water, freeze dried and the recovered masses recorded.
The procedure was repeated as required to collect enough

of each fraction for the physicochemical studies. The F2
and F3a fractions formed single bands on the SDS–Mops
gel, but F3b contained a mixture of smaller peptides. The
mass percentages of F2, F3a and F3b from the recovered
masses were 10.2, 56.7 and 33.1, respectively. MALDI-MS
measurements and calculations showed that F2 was the
full-length form (OPN 1–262) and the F3a fraction con-
tained N-terminal fragments probably formed by proteo-
lytic cleavage of K149 (or possibly also R147 or K150) by
the principal milk proteinase plasmin. For convenience,
this fraction is identified simply as OPN 1–149. Full-length
native OPN had an experimentally determined molecular
mass of 33.9 kDa, 1.7 kDa of which was phosphate
groups and approximately 2.9 kDa was O-bound glycans.
For the N-terminal fragment in F3a, the corresponding
values were 19.8 kDa with contributions of 0.9 kDa from
phosphate groups and 2.9 kDa from glycans. Thus, the
average degree of phosphorylation of the 28 potential
phosphorylation sites in the whole protein was 79%. The
N-terminal fragment analysis was consistent with an aver-
age of about 60–65% phosphorylation of the 16 sites in
the N-terminal plasmin peptides. All the glycans (three to
four sites) were present in both.
Preparation of CPNs
The object is to produce a precise number of moles of CaP
in the presence of an excess concentration of sequestering
phosphopeptide, as required by Eqn (5). Although this can
be performed by simply mixing together stock solutions
containing high concentrations of Ca, P
i

and the phospho-
peptide, the concentration gradients generated and their
persistence during inefficient mixing can allow the inital
ACP to mature into a more stable state which can no
longer form equilibrium CPNs. Even when this does not
happen, there can be an overshoot past the equilibrium
state to generate colloidal metastable complexes of ACP
and the phosphopeptide, and the subsequent re-equilibra-
tion can take days or weeks to complete. To circumvent
these problems, we developed the urea ⁄ urease method [12].
In this method, the reagents are mixed together to give an
initial undersaturated solution with a low pH, typically
5–5.5. The pH is then increased homogeneously by hydroly-
sis of a precise number of moles of urea using urease to
catalyse the reaction, producing the strong base ammonia
and weak carbonic acid. The amount of urea determines
the final pH (typically 6–8) and the amount of enzyme can
vary the time taken to approach the target pH from 2 min
to 2 h. Rapid attainment of the target pH is followed by
several hours during which the nanoclusters grow to their
equilibrium size [13].
Nanoclusters were prepared by the urea ⁄ urease method
[12] using either the OPNmix or OPN 1–149 (fraction F3a)
and magnesium-free Buffer A [13]. The standard concentra-
tion of peptide was either 25 or 30 mgÆmL
)1
, which corre-
sponds to a similar molar concentration to that used in
previous work with caseins. It did not prove possible to
find suitable conditions for CPN formation with OPN or

the F3b fraction. Neither was it possible to sequester CaP
with the highly phosphorylated chicken protein phosvitin
or a tryptic digest of phosvitin (results not shown). Experi-
ments were also performed using the simple mixing method,
whereby the pH was increased by sequential small addi-
tions, with good stirring, of phosphate salts of different
basicities.
For SAXS, electrophoretic light scattering and differen-
tial scanning calorimetry measurements, the OPN 1–149 or
CPNs were diluted to a suitable concentration with a buffer
that preserved their integrity. The dilution buffer was
designed to match as closely as possible the salt composi-
tion and pH of an ultrafiltrate prepared from the CPN
solution, but with the addition of sodium azide (1.5 mm)as
a preservative and 0.01% of the whole casein tryptic phos-
phopeptide mixture to inhibit CaP precipitation in the
buffer during storage.
Calculation of the chemical species in
nanocluster solutions
At a given fraction of reacted PCs, the concentrations of
nondiffusible (complexed) Ca and P
i
are given by:
½P
i

c
¼ f ½PPR
P
a

½Ca
c
¼ f ½PP R
Ca
a þð1 À aÞ

t
Ca
ð½Ca
2þ
; pHÞ
ÀÁ
ð6Þ
where [PP] is the phosphopeptide concentration and the
function

t
Ca
ð½Ca
2þ
; pHÞ is the phosphopeptide binding iso-
therm. From these values, the diffusible concentrations
were calculated, and hence the composition of the ultrafil-
trate was obtained from the Donnan equilibrium across the
membrane [46].
Equation (6) requires the calculation of Ca ion binding
to the free peptide at any pH in the range 5.0–8.0. A
semi-empirical model was used to describe the binding
isotherms obtained previously for the b-casein 1–25 phos-
phopeptide in this pH range, and the same model was

adapted to fit the binding isotherm of OPN 1–149 mea-
sured at pH 7.0. The rescaled model was then used to
predict binding at any other value of the pH. Further
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2319
details of the model are provided in Doc. S3 (see Sup-
porting information).
The Ca
2+
binding isotherm of OPN 1–149 was deter-
mined by measuring the concentration of free Ca ions with
a Ca ion selective electrode after successive small additions
of a stock solution of 100 mm Ca(NO
3
)
2
to a solution con-
taining 25 mgÆmL
)1
of the peptide, buffered to pH 7.0 with
20 mm P
i
and 50 mm KNO
3
. The total Ca concentration
ranged from 0 to 19.3 mm and the corresponding ionic
strengths varied from 80 to 93 mm.
Partition of salts by ultrafiltration
Ten solutions of the Ca-free OPNmix in magnesium-free
Buffer A were prepared by the urea ⁄ urease method with

pH values between 5.0 and 7.5. They were allowed to equil-
ibrate before ultrafiltration through a 0.5 mL centrifugal
concentrator with a molecular mass cut-off of 10 000 Da,
(Vivascience AG, Hanover, Germany) using a centripetal
field of 5000 g for 15 min. The concentrations of Ca, free
Ca
2+
and P
i
in the ultrafiltrate and starting solution were
determined as described previously [13]. The peptide-free
ultrafiltrate composition was then used to calculate the
ionic equilibria [5] and ion activity product for CaP,
according to Eqn (3), for y in the range 0–1.
Small-angle X-ray scattering
Measurements were made on station 2.1 at the CCLRC
Daresbury Laboratory (Warrington, Cheshire, UK), as
described previously [13]. The radii of gyration and the
intercept at q = 0 were determined by a Guinier plot of
ln(I) versus q
2
. Scattering curves were normalized by divid-
ing by the Guinier intercept and weighted by q
2
to empha-
size the low-intensity features (Kratky plot). The scattering
of OPN and OPN 1–149 in free solution was fitted to a
worm-like chain model [47,48]. Chain stiffness was mea-
sured by means of the Kuhn segment length b, which is
the peptide bond length in an equivalent freely jointed

chain.
The CPNs were prepared either on the SAXS station and
measured immediately, or were prepared at least 2 days
previously to allow the system to come to equilibrium. The
sample was then stored at room temperature and re-mea-
sured 5 months later during a later allocation of beamtime.
Preliminary experiments with CPNs prepared with the
OPNmix sample established that the scattering was inde-
pendent of concentration below a peptide concentration of
7mgÆ mL
)1
. Accordingly, all measurements were made after
dilution to 5 mgÆmL
)1
. Scattering of the CPN solutions was
treated as a mixture of free chains and nanoclusters, with
the latter modelled as copolymer micelles with a homoge-
neous core and corona of Gaussian chains [49]. The size
distribution of copolymers was represented by a log-normal
number distribution with a modal core radius r
o
and width
parameter b, and the number fraction of free chains was
assumed to equal the fraction of unreacted PCs. Further
details of the scattering models are given in Doc. S4 (see
Supporting information).
Electrophoretic light scattering
The intensity-averaged diffusion coefficient was determined
with a Malvern Zetasizer Nano instrument (Malvern,
Worcestershire, UK). Inversion of the intensity autocorrela-

tion function by means of the Multiple Narrow Modes
algorithm in the instrument’s software gave an intensity-
weighted size distribution. The mean hydrodynamic radius
(

r
h
) was calculated from the diffusion coefficient using the
Stokes–Einstein equation. Electrophoretic mobilities were
measured by the phase analysis method in the disposable
single-use cells supplied with the instrument. The zeta
potential (f) was calculated from the electrophoretic mobil-
ity in a unit field (u
e
) using the Henry equation [39], as the
particle size is between the Hu
¨
ckel and Smoluchowski limit-
ing formulae.
An OPN 1–149 CPN sample was prepared at pH 7 by
the rapid urea ⁄ urease method at a peptide concentration of
25 mgÆmL
)1
and matured for 1 day. It was diluted with
dilution buffer to a concentration of 5 mgÆmL
)1
, filtered
and the electrophoretic mobility and diffusion coefficient
were determined. The urea ⁄ urease method was also used to
make CPNs with b-casein (f1-25), and a study was made of

maturation in unfiltered solutions. A comparison was also
made of the hydrodynamic radii of CPNs prepared with
the mixture of tryptic phosphopeptides from whole casein
[38] by the urea ⁄ urease reaction and the simple mixing
method.
Differential scanning calorimetry
A differential scanning calorimetry scan of OPN 1–149 at a
concentration of 5 mgÆmL
)1
in 10 mm P
i
buffer, pH 7.0,
was recorded using a MicroCal MCS calorimeter (North-
ampton, MA, USA), as described previously [50].
Acknowledgements
This work was sponsored by the Scottish Executive
Rural Affairs Department, the Biotechnology and Bio-
logical Sciences Research Council and the Science and
Technology Facilities Council. We thank Elaine Little,
Jim Dunsmuir and Irene Stewart for their careful tech-
nical assistance. Gu
¨
nter Grossmann and Kalotina
Geraki provided invaluable help with the SAXS work.
Jason Dalby and Malvern Instruments Ltd. are
thanked for carrying out the electrophoretic light
scattering measurements. Access to PONDR
Ò
was
provided by Molecular Kinetics (6201 La Pas Trail –

Calcium phosphate sequestration by osteopontin C. Holt et al.
2320 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
Ste 160, Indianapolis, IN 46268, USA; Tel: 00 1 317
280 8737; E-mail: ). VL-
XT is copyright ª1999 by the WSU Research Founda-
tion (Pullman, WA, USA), all rights reserved.
PONDRÒ is copyright ª2004 by Molecular Kinetics,
all rights reserved.
References
1 Robertson WG & Marshall RW (1979) Calcium mea-
surements in serum and plasma – total and ionized.
CRC Crit Rev Clin Lab Sci 11, 271–304.
2 Glinkina IV, Durov VA & Mel’nitchenko GA (2004)
Modelling of electrolyte mixtures with application to
chemical equilibria in mixtures – prototypes of blood’s
plasma and calcification of soft tissues. J Mol Liquids
110, 63–67.
3 Silwood CJL, Grootveld M & Lynch E (2002) H-1
NMR investigations of the molecular nature of low-
molecular-mass calcium ions in biofluids. J Biol Inorg
Chem 7, 46–57.
4 Gyory AZ & Ashby R (1999) Calcium salt urolithiasis –
review of theory for diagnosis and management. Clin
Nephrol 51, 197–208.
5 Holt C, Dalgleish DG & Jenness R (1981) Inorganic
constituents of milk. 2. Calculation of the ion equilibria
in milk diffusate and comparison with experiment. Anal
Biochem 113, 154–163.
6 Price PA & Lim JE (2003) The inhibition of calcium
phosphate precipitation by fetuin is accompanied by the

formation of a fetuin–mineral complex. J Biol Chem
278, 22144–22152.
7 Schafer C, Heiss A, Schwarz A, Westenfeld R, Ketteler
M, Floege J, Muller-Esterl W, Schinke T & Jahnen-
Dechent W (2003) The serum protein alpha(2)-Here-
mans–Schmid glycoprotein ⁄ fetuin-A is a systemically
acting inhibitor of ectopic calcification. J Clin Invest
112, 357–366.
8 Moe SM & Chen NX (2005) Inflammation and vascular
calcification. Blood Purif 23, 64–71.
9 Price PA, Thomas GR, Pardini AW, Figueira WF,
Caputo JM & Williamson MK (2002) Discovery of a
high molecular weight complex of calcium, phosphate,
fetuin, and matrix gamma-carboxyglutamic acid protein
in the serum of etidronate-treated rats. J Biol Chem
277, 3926–3934.
10 Heiss A, DuChesne A, Denecke B, Grotzinger J,
Yamamoto K, Renne T & Jahnen-Dechent W
(2003) Structural basis of calcification inhibition by
alpha(2)-HS glycoprotein ⁄ fetuin-A – formation of
colloidal calciprotein particles. J Biol Chem 278,
13333–13341.
11 Holt C (1983) Swelling of Golgi vesicles in mammary
secretory-cells and its relation to the yield and
quantitative composition of milk. J Theor Biol 101,
247–261.
12 Holt C, Wahlgren NM & Drakenberg T (1996) Ability
of a beta-casein phosphopeptide to modulate the precip-
itation of calcium phosphate by forming amorphous
dicalcium phosphate nanoclusters. Biochem J 314,

1035–1039.
13 Little EM & Holt C (2004) An equilibrium thermody-
namic model of the sequestration of calcium phosphate
by casein phosphopeptides. Eur Biophys J Biophys Lett
33, 435–447.
14 Holt C, Timmins PA, Errington N & Leaver J (1998) A
core-shell model of calcium phosphate nanoclusters sta-
bilized by beta-casein phosphopeptides, derived from
sedimentation equilibrium and small-angle X-ray and
neutron-scattering measurements. Eur J Biochem 252,
73–78.
15 Holt C, Vankemenade M, Nelson LS, Sawyer L, Har-
ries JE, Bailey RT & Hukins DWL (1989) Composition
and structure of micellar calcium-phosphate. J Dairy
Res 56, 411–416.
16 Holt C (2004) An equilibrium thermodynamic model of
the sequestration of calcium phosphate by casein
micelles and its application to the calculation of the
partition of salts in milk. Eur Biophys J Biophys Lett
33, 421–434.
17 Holt C, de Kruif CG, Tuinier R & Timmins PA (2003)
Substructure of bovine casein micelles by small-angle
X-ray and neutron scattering. Colloids Surf A
Physicochem Eng Asp 213, 275–284.
18 Holt C, Hasnain SS & Hukins DWL (1982) Structure
of bovine-milk calcium-phosphate determined by x-ray
absorption-spectroscopy. Biochim Biophys Acta 719 ,
299–303.
19 Kruif CGd & Holt C (2003) Casein micelle structure,
functions and interactions. In Advanced Dairy

Chemistry (Fox PF & McSweeney PLH, eds),
pp. 675–698. Kluwer Academic ⁄ Plenum, New York,
NY.
20 Lyster RLJ, Mann S, Parker SB & Williams RJP (1984)
Nature of micellar calcium-phosphate in cows milk as
studied by high-resolution electron-microscopy. Biochim
Biophys Acta 801, 315–317.
21 Mazzali M, Kipari T, Ophascharoensuk V, Wesson
JA, Johnson R & Hughes J (2002) Osteopontin –
a molecule for all seasons. QJM Int J Med 95,
3–13.
22 Giachelli CM & Steitz S (2000) Osteopontin: a versatile
regulator of inflammation and biomineralization.
Matrix Biol 19, 615–622.
23 Kawasaki K & Weiss KM (2003) Mineralized tissue
and vertebrate evolution: the secretory calcium-binding
phosphoprotein gene cluster. Proc Natl Acad Sci USA
100, 4060–4065.
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2321
24 Kawasaki K, Buchanan AV & Weiss KM (2007) Gene
duplication and the evolution of vertebrate skeletal min-
eralization. Cells Tissues Organs 186, 7–24.
25 Fisher LW, Torchia DA, Fohr B, Young MF &
Fedarko NS (2001) Flexible structures of SIBLING
proteins, bone sialoprotein, and osteopontin. Biochem
Biophys Res Commun 280, 460–465.
26 Sorensen ES, Hojrup P & Petersen TE (1995) Posttrans-
lational modifications of bovine osteopontin – identifi-
cation of 28 phosphorylation and 3 O-glycosylation

sites. Protein Sci 4, 2040–2049.
27 Hunter GK, Hauschka PV, Poole AR, Rosenberg LC
& Goldberg HA (1996) Nucleation and inhibition of
hydroxyapatite formation by mineralized tissue pro-
teins. Biochem J 317, 59–64.
28 He G, Ramachandran A, Dahl T, George S, Schultz D,
Cookson D, Veis A & George A (2005) Phosphoryla-
tion of phosphophoryn is crucial for its function as a
mediator of biomineralization. J Biol Chem 280, 33109–
33114.
29 Hunter GK & Goldberg HA (1994) Modulation of
crystal-formation by bone phosphoproteins – role of
glutamic acid-rich sequences in the nucleation of
hydroxyapatite by bone sialoprotein. Biochem J 302,
175–179.
30 Nancollas GH (1982) Phase transformation during pre-
cipitation of calcium salts. In Biological Mineralization
and Demineralization (Nancollas GH, ed.), pp. 79–99.
Springer-Verlag, Berlin.
31 Ferron F, Longhi S, Canard B & Karlin D (2006) A
practical overview of protein disorder prediction meth-
ods. Proteins Struct Funct Bioinform 65, 1–14.
32 Ishida T & Kinoshita K (2008) Prediction of disordered
regions in proteins based on the meta approach. Bioin-
formatics 24, 1344–1348.
33 Monaco HL (1997) Crystal structure of chicken ribofla-
vin-binding protein. EMBO J 16, 1475–1483.
34 Campen A, Williams RM, Brown CJ, Meng JW, Uver-
sky VN & Dunker AK (2008) TOP-IDP-scale: a new
amino acid scale measuring propensity for intrinsic dis-

order. Protein Pept Lett 15, 956–963.
35 Holt C & Sawyer L (1993) Caseins as rheomorphic
proteins – interpretation of primary and secondary struc-
tures of the alpha-S1-caseins, beta-caseins and kappa-
caseins. J Chem Soc Faraday Trans 89, 2683–2692.
36 Kawasaki K & Weiss KM (2006) Evolutionary genetics
of vertebrate tissue mineralization: the origin of secre-
tory calcium-binding and evolution of the phosphopro-
tein family. J Exp Zool B Mol Dev Evol 306B, 295–316.
37 Pedersen JS & Gerstenberg MC (1996) Scattering form
factor of block copolymer micelles. Macromolecules 29,
1363–1365.
38 Ellegard KH, Gammelgard-Larsen C, Sorensen ES &
Fedosov S (1999) Process scale chromatographic isola-
tion, characterization and identification of tryptic
bioactive casein phosphopeptides. Int Dairy J 9, 639–
652.
39 Henry DC (1931) The cataphoresis of suspended parti-
cles. Part 1. – The equation of cataphoresis. Proc R Soc
Lond A 133, 106–129.
40 Gericke A, Qin C, Spevak L, Fujimoto Y, Butler WT,
Sorensen ES & Boskey AL (2005) Importance of phos-
phorylation for osteopontin regulation of biomineraliza-
tion. Calcif Tissue Int 77, 45–54.
41 George A & Veis A (2008) Phosphorylated proteins and
control over apatite nucleation, crystal growth, and
inhibition. Chem Rev
108, 4670–4693.
42 Price PA, Caputo JM & Williamson MK (2002) Bone
origin of the serum complex of calcium, phosphate, fet-

uin, and matrix Gla protein: biochemical evidence for
the cancellous bone-remodeling compartment. J Bone
Miner Res 17, 1171–1179.
43 Price PA, Nguyen TMT & Williamson MK (2003) Bio-
chemical characterization of the serum fetuin–mineral
complex. J Biol Chem 278, 22153–22160.
44 Larkin MA, Blackshields G, Brown NP, Chenna R,
McGettigan PA, McWilliam H, Valentin F, Wallace
IM, Wilm A, Lopez R et al. (2007) Clustal W and clus-
tal X version 2.0. Bioinformatics 23, 2947–2948.
45 Jonassen I (1997) Efficient discovery of conserved pat-
terns using a pattern graph. Comput Appl Biosci 13,
509–522.
46 Holt C (1997) The milk salts and their interaction with
casein. In Advanced Dairy Chemistry (Fox PF, ed.), pp.
233–256. Chapman & Hall, London.
47 Kholodenko AL (1993) Scattering function for semiflex-
ible polymers – Dirac versus Kratky–Porod. Phys Lett
A 178, 180–185.
48 Potschke D, Hickl P, Ballauff M, Astrand PO &
Pedersen JS (2000) Analysis of the conformation of
worm-like chains by small-angle scattering: Monte-
Carlo simulations in comparison to analytical theory.
Macromol Theory Simul 9, 345–353.
49 Svaneborg C & Pedersen JS (2002) Form factors of
block copolymer micelles with excluded-volume interac-
tions of the corona chains determined by Monte Carlo
simulations. Macromolecules 35, 1028–1037.
50 Qi XL, Holt C, McNulty D, Clarke DT, Brownlow S &
Jones GR (1997) Effect of temperature on the second-

ary structure of beta-lactoglobulin at pH 6.7, as deter-
mined by CD and IR spectroscopy: a test of the molten
globule hypothesis. Biochem J 324, 341–346.
Supporting information
The following supplementary material is available:
Doc. S1. Thermodynamics of calcium phosphate nano-
cluster formation.
Doc. S2. Phosphate centre sequences.
Calcium phosphate sequestration by osteopontin C. Holt et al.
2322 FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS
Doc. S3. Binding of Ca
2+
to b-casein 1–25 and
OPN 1–149.
Doc. S4. Scattering models for calcium phosphate
nanoclusters.
This supplementary material can be found in the
online version of this article.
Please note: Wiley-Blackwell is not responsible for
the content or functionality of any supplementary
materials supplied by the authors. Any queries (other
than missing material) should be directed to the corre-
sponding author for the article.
C. Holt et al. Calcium phosphate sequestration by osteopontin
FEBS Journal 276 (2009) 2308–2323 ª 2009 The Authors Journal compilation ª 2009 FEBS 2323