18
19
20
21
22
II-D. Transformation of Octacalcium Phosphate to Hydroxyapatite: A Study of the Molecular Mechanism by SEM, TEM, XRD and Solid-State NMR Spectroscopy (submitted)
Biomineralization is a biological process describing the formation of minerals in living organisms.1 Calcium phosphates are the major inorganic constituents of biological hard tissues in vertebrates, existing in the form with close resemblance to hydroxyapatite (HAp, Ca10(PO4)6(OH)2).2 The so-called biological apatite or dahllite refers to poorly crystallized nonstoichiometric carbonate-containing HAp. Whether biological apatite is formed by direct precipitation or through an intermediate phase remains an unsettled issue in the field of biomineralization.2,3 Because of the structural similarity between HAp and octacalcium phosphate (OCP, Ca8H2(PO4)6⋅5H2O), OCP has been hypothesized as the precursor phase of biological apatite.4 The most compelling evidence for this hypothesis is the observation of an OCP “central dark line” in many biological apatites and in some synthetically prepared HAp.5-7 OCP is thermodynamically less stable than hydroxyapatite and it is often found as an intermediate phase during the precipitation of HAp.8,9 Many in-vitro studies of calcium phosphate precipitation have been carried out to elucidate the dependence of the thermodynamic events and the crystal morphology on the degree of supersaturation,10 temperature,11,12 pH conditions13,14 and the rate of precipitation.15-22 The most informative results hitherto reported were obtained from X-ray powder diffraction (XRD), scanning electron microscopy (SEM) and transmission electron microscopy (TEM).3
OCP can be described as an alternating layer structure of apatite layer and hydrated layer,23 where the apatite layer is structurally very similar to HAp.24,25 According to the OCP precursor model proposed by Brown,4 the first calcium phosphate crystals formed in a supersaturated solution under physiological condition are OCP like. The subsequent hydrolysis step leads to the formation of HAp, where the c axes of the OCP and HAp unit cells are along the same direction during the structural transition. While this model is consistent with the results of some TEM studies,26,27 a recent computational study shows that the c axes of the OCP and HAp should be at opposite directions in order to minimize the free energy at the interface between OCP and HAp.28 It is by no means trivial to verify this computational prediction experimentally because TEM results cannot distinguish the alignment of two crystallographic axes from parallel to anti-parallel fashion. Furthermore, OCP is meta-stable and it will appear only if the pH in the crystallization system is below 6. In our previous work, we developed an in-vitro system to realize a single-crystal-to-single-crystal transformation from OCP to HAp in the presence of gelatin and urea.29 The transformation of OCP to HAp is initiated by raising the pH condition from acidic to alkaline. Since a uniform pH increase of the reaction mixture is achieved by slow decomposition of urea at 100°C, our in-vitro system is an ideal model system for the study of the molecular mechanism of OCP to HAp transition. While all the phosphorus atoms in a unit cell of HAp are equivalent, there are six crystallographically non-equivalent phosphorous sites in OCP. As such, solid-state 31P NMR is well suited to characterize the OCP to HAp transformation at the molecular level.
23
The high-resolution 31P NMR studies of OCP can be dated back to the 1980s30,31 and the complete spectral assignment has been made very recently.32,33 Throughout the years, a large variety of advanced NMR techniques such as 31P{1H} cross-polarization (CP) at variable contact delays,30 dipolar dephasing technique,31 heteronuclear correlation spectroscopy (HETCOR),34-36 differential cross polarization,37,38 and multi-nuclear double-resonance techniques39,40 had been successfully used to characterize the structures of synthetic hydroxyapatite, calcified tissues and apatite formation. Therefore, in the present study we chose to use a series of solid-state 31P NMR techniques including 31P{1H} Lee-Goldburg spectroscopy41,42 and 31P homonuclear double-quantum (DQ) NMR43 to monitor the OCP to HAp transition. The in-vitro system we developed earlier, in the absence of gelatin, is used to prepare calcium phosphate precipitated at different pH conditions. From the DQ NMR measurements we are able to show that the c axes of the OCP and HAp unit cells are at opposite directions during the transformation. Furthermore, the data of the 31P{1H} cross-polarization NMR suggest that water molecules enter the hydration layers of OCP crystals via the hydrolysis reaction HPO42- + OH- = PO43- + H2O, which also accounts for the deprotonation of the HPO42- ions during the transformation. Overall, our NMR data provide hitherto the most detailed description of the OCP to HAp transformation mechanism at the molecular level.
Experimental
Sample Preparation and Characterization. Urea ( 99.5%) , sodium phosphate monobasic dehydrate (H2NaPO4･2H2O) (99%) and calcium nitrate tetrahydrate (Ca(NO3)2･4H2O) (99%) were used as received (Acros). A mixture of 10 mmol Ca(NO3)2 ･4H2O, 10 mmol H2NaPO4･ 2H2O and 20 mmol urea were dissolved in 400 mL doubly distilled water and then sealed in a polypropylene container. The aqueous solution was kept at 100°C for different periods. The precipitates thus obtained were filtered, washed and then dried at 60°C for one day. A series of samples were obtained at different reaction times, viz. 1.5 h, 3 h, 4 h, 5 h, 6 h, 7 h, 9 h, 12 h and 96 h. All the samples will henceforth be labeled based on their reaction times. X-Ray diffraction analysis was performed on a Philips X’Pert diffractometer, using Cu-Kα radiation (λ = 1.5418 Å).
The Rietvald analyses were done by the software EXPGUI.44,45 The field emission SEM were taken on a JEOL-JSM-6700F field emission scanning electron microscope operated at 10 kV. The TEM and electron diffraction (ED) patterns were taken on Hitachi S-7100 and Philips FEI Tecnai 20 G2 instruments operating at 75 kV and 200 kV, respectively.
Solid-State NMR. All NMR experiments were carried out at 31P and 1H frequencies of 121.5 and 300.1 MHz, respectively, on a Bruker DSX300 NMR spectrometer equipped with a commercial 4-mm probe. All spectra were measured at room temperature. The sample was confined to the middle 1/3 of the rotor volume using Teflon spacers. The variation of magic-angle spinning (MAS) frequency was limited to ± 3 Hz using a commercial pneumatic control unit.
Chemical shifts were externally referenced to 85% phosphoric acid and TMS for 31P and 1H, respectively. The 31P MAS spectra were measured at a spinrate of 10 kHz and with 70 kHz proton
24
decoupling. An exponential window function of 20 Hz line broadening was applied to each FID before the Fourier transformation.
The 31P{1H} CP heteronuclear correlation (HETCOR) spectra were measured at a spinrate of 10 kHz. During the contact time (2.5 ms) the 1H nutation frequency was set equal to 50 kHz and that of 31P was ramped through the Hartmann-Hahn matching sideband.46 Quadrature detection in the F1 dimension was achieved by the hypercomplex approach. Typically, for each t1 increment 32 transients were accumulated, and a total of 50 increments were done at steps of 100 µs.
The 31P{1H} Lee-Goldburg CP (LG-CP) spectra were measured at a spinrate of 10 kHz. The flip angle of the pulse after the t1 evolution is adjusted so that the spin-temperature inversion can be realized by phase alternating the first π/2 pulse. During the contact time the 1H nutation frequency and the resonance offset were set equal to 50 and 35.35 kHz, respectively, to fulfill the Lee-Goldburg irradiation condition. The 31P DQ experiments were carried out under MAS spinning frequency of 10 kHz based on the so-called HSMAS-DQ technique.33,47 To prepare the initial spin system identically for each transient, a saturation comb was applied prior to the recovery delay (8 s). During the DQ excitation and reconversion periods, the 31P π/2 and π pulses were set to 5 and 30 µs long, respectively. The π pulse trains were phase cycled according to the XY-8 scheme.48 The DQ reconversion period was set equal to the excitation period. Proton decoupling was set to 85 kHz during the DQ excitation/reconversion periods. A more detailed description of the experiment was given elsewhere.33