Physicochemical characterization of the nucleational core of matrix vesicles.

While previous studies revealed that matrix vesicles (MV) contain a nucleational core (NC) that converts to apatite when incubated with synthetic cartilage lymph, the initial mineral phase present in MV is not well characterized. This study explored the physicochemical nature of this Ca2+ and Pi-rich NC. MV, isolated from growth plate cartilage, were analyzed directly by solid-state 31P NMR, or incubated with hydrazine or NaOCl to remove organic constituents. Other samples of MV were subjected to sequential treatments with enzymes, salt solutions, and detergents to expose the NC. We examined the NC using transmission electron microscopy, energy-dispersive analysis with x-rays, and electron and x-ray diffraction, Fourier transform-infrared spectroscopy, high performance thin-layer chromatographic analysis, and SDS-polyacrylamide gel electrophoresis. We found that most of the MV proteins and lipids could be removed without destroying the NC; however, NaOCl treatment annihilated its activity. SDS-polyacrylamide gel electrophoresis showed that annexin V, a phosphatidylserine (PS)-dependent Ca2+-binding protein, was the major protein in the NC; high performance thin-layer chromatographic analysis revealed that the detergents removed the majority of the polar lipids, but left significant free cholesterol and fatty acids, and small but critical amounts of PS. Transmission electron microscopy showed that the NC was composed of clusters of approximately 1.0 nm subunits, which energy-dispersive analysis with x-rays revealed contained Ca and Pi with a Ca/P ratio of 1.06 +/- 0. 01. Electron diffraction, x-ray diffraction, and Fourier transform-infrared analysis all indicated that the NC was noncrystalline. 1H-Cross-polarization 31P NMR indicated that the solid phase of MV was an HPO42--rich mixture of amorphous calcium phosphate and a complex of PS, Ca2+, and Pi. Taken together, our findings indicate that the NC of MV is composed of an acid-phosphate-rich amorphous calcium phosphate intermixed with PS-Ca2+-Pi, annexin V, and other proteins and lipids.

process. In growth plate cartilage, much evidence indicates that extracellular microstructures known as matrix vesicles (MV) 1 play a key role in the induction of the first solid phase mineral (1)(2)(3). MV are produced by budding from the plasma membrane into sites of intended calcification by cellularly controlled mechanisms (4 -6). Previous biochemical (7)(8)(9) and ultrastructural (10,11) studies have shown that MV contain large amounts of calcium and phosphate; however, the initial mineral phase has not yet been characterized. X-ray diffraction (12), and FT-IR (13) and FT-Raman (14) spectroscopic analyses all indicate that the initial mineral form in MV is noncrystalline.
In previous work, we showed that it was possible to strip away nonessential components from the MV using various chemical and detergent extractions leaving a residual core that retained the ability to induce mineral formation (15). The activity of this detergent-stable nucleational core (NC) was inhibited by Zn 2ϩ and was destroyed by treatment with pH 6 citrate buffer. This acid lability is similar to that of synthetic amorphous calcium phosphate (ACP) (15,16). Further, in our assay system, the calcification of both MV and ACP share the same pH-dependent kinetic profile when incubated in synthetic cartilage lymph (SCL) (17). Based on these observations, the NC of MV appears to contain ACP-like mineral; however, both FT-IR and FT-Raman spectroscopy failed to detect classical ACP in native MV (13,14).
Several other forms of calcium phosphate, i.e. dicalcium phosphate dihydrate (brushite) and octacalcium phosphate (OCP), have also been postulated to be precursors to hydroxyapatite (HAP) in vivo (18). By FT-IR and FT-Raman analyses (13,14), the first crystalline mineral phase formed by MV incubated in metastable SCL was OCP-like in character. After 24 h it appears to be poorly crystalline HAP, similar to that present in bone (19,20).
Lipids are present in the extracellular matrix at sites of calcification (21), and some are associated with the mineral phase (22). Varying proportions of the predominant lipid associated with the mineral, PS, are complexed with Ca 2ϩ and P i (22)(23)(24). These Ca 2ϩ -PS-P i complexes (CPLX) are present in bone, cartilage, and other mineralized tissues (25) and are especially enriched in MV (26). In vitro experiments have shown that both CPLX and ACP, when incubated in synthetic lymphs, can lead to formation of HAP (18,27,28). In these studies we sought to determine unequivocally whether CPLX and ACP are present in the NC of MV. Establishment of the physicochemical properties of the nucleational core (the first mineral phase) would provide an important clue to how MVs are formed and how they function as the primary nucleator in growth plate cartilage calcification.
In this investigation we examined the physicochemical properties of the residual core material in MV remaining after sequential treatments with enzyme, buffered salt, and detergent solutions. We also used treatment with hydrazine or hypochlorite to remove organic matter. We assayed the kinetics of mineral phase formation after each treatment procedure to establish that nucleational activity had been retained. Focus was then placed on structural characterization of the residue (i.e. nucleational core, NC) using the following physicochemical methods: transmission electron microscopy (TEM) for structure, energy-dispersive analysis with x-rays (EDAX) and chemical analyses for elemental composition, electron and x-ray diffraction for crystallinity, and FT-IR and solid-state 31 P NMR spectroscopy for chemical structure. High performance thinlayer chromatography (HPTLC) was used to analyze lipids; SDS-PAGE was used to analyze proteins.

EXPERIMENTAL PROCEDURES
Isolation of Matrix Vesicles-Collagenase-released matrix vesicles (CRMV) were isolated from growth plate cartilage of 6 -8-week-old broiler strain chickens as described previously (15). Briefly, tissues were digested with 0.1% trypsin (type III, Sigma) at 37°C for 20 min in synthetic cartilage lymph (SCL). SCL contained 2 mM Ca 2ϩ and 1.42 mM P i , in addition to 104.5 mM Na ϩ , 133.5 mM Cl Ϫ , 63.5 mM sucrose, 16.5 mM TES, 12.70 mM K ϩ , 5.55 mM D-glucose, 1.83 mM HCO 3 Ϫ , 0.57 mM Mg 2ϩ , and 0.57 mM SO 4 2Ϫ . The trypsin solution was removed; tissue slices were rinsed twice with SCL, and then digested with collagenase (200 units/g of tissue, type IA, Sigma) at 37°C for 3-3.5 h. The partially digested tissue was vortexed, and the suspension was differentially centrifuged at 13,000 ϫ g for 20 min and at 100,000 ϫ g for 1 h. The MV pellet was resuspended as a stock suspension of 5.0 mg of vesicle protein/ml of SCL containing 1.0 mM Ca 2ϩ (one-half the normal level of Ca 2ϩ in SCL) to prevent dissolution of labile Ca 2ϩ and P i present (9), and also to minimize Ca 2ϩ uptake by the vesicles during storage. Protein levels in the vesicles were determined by the method of Lowry et al. (29).
Hyaluronidase-released matrix vesicles were isolated from the same tissues by incubating in SCL containing 0.25% hyaluronidase (Type IS, Sigma) at 37°C for 90 min, followed by incubation in SCL containing collagenase (500 units/g of tissue) at 37°C for 3 h (30). After vortexing, hyaluronidase-released matrix vesicles were isolated by differential centrifugation as before. The MV pellets were resuspended in SCL containing 1 mM Ca 2ϩ .
Sequential Extraction of CRMV-CRMV were extracted sequentially with hyaluronidase, water, pure collagenase, high salt, and then with detergent, to expose the acid-labile NC (Fig. 1). While some of these procedures had been explored individually before to determine their effects on MV mineralization (15), in this study these procedures were used sequentially to fully strip away the non-essential components. The final pellet was resuspended in SCL containing 1.0 mM Ca 2ϩ .
HPTLC Analysis of Lipids-CRMV (1.0 mg of Lowry protein), and the sequentially treated residual cores, were decalcified with 1 M EDTA, pH 7.5, for 30 min at 25°C, dialyzed against deionized water overnight at 4°C with constant stirring, and then lyophilized. The demineralized residue was extracted sequentially with chloroform-methanol (2:1, v/v) (Extract 1) and chloroform-methanol-concentrated HCl (200:100:1, v/v/v) (Extract 2) as described previously (22). The dried crude lipids from Extracts 1 and 2 were partitioned through a micro-Sephadex G-25 column to remove nonlipid contaminants (31). The pure lipids were then analyzed by HPTLC on Whatman LHP-K silica gel plates as described previously (15,32). Mixtures containing 10 g each of the various standard lipids were applied to separate lanes on the same plate as the lipid samples.
Mineral Formation-A multiwell microplate mineralization assay based on light scattering (33) was used for real-time measurement of mineral formation by MV, and their subfractions. Samples were prepared for mineralization assays as follows. Aliquots of the MV stock solution were diluted 100-fold to a final concentration of 50 g of protein/ml of SCL containing 2.0 mM Ca 2ϩ ; nucleational core pellets ( Fig. 1, Pellet F) were resuspended SCL to a final concentration of 50 g of protein/ml, based on the initial protein concentration (Fig. 1, Pellet  A). For kinetic assays of mineral formation, 160-l aliquots of the diluted vesicle or nucleational core suspensions were placed into 96well half-area tissue culture clusters (Costar, Cambridge, MA) and incubated at 37°C. An IEMS Reader MF instrument (Labsystems, Needham Heights, MA) was used to take absorbency readings at 340 nm at 15-min intervals.
Analysis of Ca, Inorganic P (P i ), and Total P-MV and their subfractions were digested in 0.1 N HCl at 25°C overnight. The digests were clarified by centrifugation using a Beckman Airfuge (22 p.s.i., 8 min) and then samples were removed for mineral ion analysis. Ca 2ϩ was analyzed by the method of Baginsky et al. (34); P i was determined by the method of Ames (35). For total P analysis, a small volume of 70% HClO 4 containing 0.01% ammonium molybdate (36) was added to the dry samples, which were digested at 140°C for 15 min or longer until the samples were completely clear. Then aliquots of the digestate were taken for P i analysis as above.
Gel Electrophoresis-SDS-PAGE was used to analyze the protein composition of the various MV fractions. A 3.0% stacking gel and a 7.5-15% gradient resolving gel were used, following the method of Laemmli (37). Gels were stained with Coomassie Brilliant Blue R-250.
FT-IR and Solid-state 31 P NMR Spectroscopic Analysis-For FT-IR analyses, lyophilized samples of the NC (ϳ1 mg) were incorporated into KBr (ϳ300 mg) and pressed into a pellet under vacuum at 12,000 psi pressure, and then examined over a range from 4000 to 400 cm Ϫ1 using a Perkin Elmer model 1600 series spectrometer (13). (Some samples were rapidly frozen in liquid N 2 and not dried; they were examined at Ϫ196°C by solid-state NMR.) Solid-state 31 P NMR spectra of the CRMV and the synthetic PS-Ca 2ϩ -P i complex (33) were obtained using a Varian Unityϩ spectrometer operating at 121.42 mHz for 31 P. Pulse widths were 4.4 s, and a contact time of 2.5 ms was used. Samples (10 -20 mg) were packed into 7-mm Doty Scientific SiN 3 rotors and spun at 2 kHz, stabilized to a precision of Ϯ1 Hz. Chemical shifts were referenced to external H 3 PO 4 at 0 ppm (38).
TEM, EDAX, and Electron Diffraction-MV subfractions were lyophilized and samples (Ͻ1 mg) dispersed in 100% ethanol by sonication. One drop of the suspension was added to one drop of deionized H 2 O, and transferred to carbon-coated grids, blotted, and dried. The grids were viewed using an Hitachi H-8000 TEM at 200 keV accelerating voltage. Grids were not stained to increase contrast. For EDAX analyses the following parameters were used: accelerating voltage, 25 keV; angle of incidence beam, 90°; x-ray emergence angle, 41.8°; x-ray window incidence angle, 6.8°; counting time, 200 s. At least five representative areas were analyzed for elemental composition. Values were calculated in atomic percent normalized to 100 after eight iterations. Areas shown to be rich in Ca and P by EDAX analysis were analyzed by electron diffraction for the presence of crystalline mineral.
Micro X-ray Diffraction Analyses-Each MV subfraction was analyzed by x-ray diffraction for the presence of crystalline solids before and after incubation in SCL using a Phillips Norelco x-ray diffractometer using CuK␣ radiation, a nickel filter, 40 kV and 20 mA, 1°receiving slit, and a scanning speed of 1/8°2/min (39).
Hydrazine Treatment of CRMV-Lyophilized CRMV were treated with hydrazine monohydrate (Aldrich, 98%, 166 l/mg of protein) at 60°C for 24 h (40). The mixture was centrifuged using a Beckman Airfuge at 22 p.s.i. for 8 min and the insoluble residue washed twice with 100% ethanol. The residue was 1) dried under vacuum before FT-IR analysis, 2) dispersed in 100% ethanol for TEM analysis, or 3) dispersed in SCL containing 1 mM Ca 2ϩ before measuring Ca 2ϩ uptake.
Sodium Hypochlorite and Sodium Hydroxide Treatment of CRMV-Vesicle samples containing 50 g of protein were treated with a solution of either NaOCl, pH 12 (50 or 100 mM final concentration), or NaOH, pH 12 and 13 (10 or 100 mM final concentration) in a total volume of 20 l of SCL at 23°C for 15 min. NaOCl treatment has been shown to remove organic matter (41). After treatments, the reaction mixtures were diluted to 1.0 ml with SCL and 160-l aliquots were removed and used for mineral formation assays.

Fractionation-
The fractionation scheme for the sequential treatment of the CRMV to remove nonessential materials and reveal the functionally active NC is outlined in Fig. 1. Hyaluronidase and water treatments were used to remove proteoglycan-related link protein and hyaluronic acid-binding protein.
Highly purified collagenase and 1 M NaCl in SCL were then used to remove residual collagen and its fragments. To remove further nonessential components, the vesicle remnants from the preceding steps were then incubated with detergent, either CHAPS or Triton X-100, to yield the NC.
Nucleational Activity and General Composition- Fig. 2 shows the time course of mineral formation by the NC after CHAPS treatment. It is evident that mineralization of the NC was actually increased compared to the untreated CRMV. Lowry analysis indicated that only about one-third of the original protein remained in the NC after the CHAPS treatment. However, chemical analyses revealed that the Ca and P i content of the MV was relatively unchanged; ϳ98% of the Ca 2ϩ and ϳ86% of the P i were recovered in the NC.
Protein Composition- Fig. 3 shows the SDS-PAGE pattern of MV proteins in fractions from the sequential treatments ending with CHAPS detergent, as depicted in Fig. 1. The protein profile of the NC (Pellet F) was much simpler than that of the original MV. Annexin V (ϳ33 kDa) was most abundant, but significant amounts of unidentified high (Ͼ300 kDa), and low (ϳ10 kDa) molecular mass proteins, as well as several lesser bands: annexin VI (ϳ66 kDa), annexin II (ϳ36 kDa), alkaline phosphatase (ϳ80 kDa), and three unidentified proteins of ϳ50, 45, and 38 kDa were seen. Because of their tight association with the NC, it is possible that these proteins may regulate (influence) the rate of growth, shape, and size of crystals during the nucleational process.
Lipid Composition-Prior to detergent extraction (four-step extracted CRMV), the residual pellet had a lipid profile similar to that of the original CRMV when analyzed by HPTLC (Fig.  4A). A variety of phospholipids, phosphatidylethanolamine, phosphatidylinositol, PS, lysophosphatidylethanolamine, phosphatidylcholine, and sphingomyelin, as well as significant amounts of free cholesterol, fatty acids, and triacylglycerols were present. However, upon extraction with detergent (either CHAPS or Triton X-100), a large amount of lipid was removed. The lipid profile of the NC after Triton X-100 extraction ( Fig.   FIG. 1. Sequential extraction scheme for isolating the nucleational core from MV.

FIG. 2.
Mineral phase induction by CRMV and the CHAPSnucleational core. CRMV were sequentially extracted as shown in Fig. 1. Samples of the original CRMV and the final CHAPS-NC were incubated in SCL, and their ability to induce mineral formation was measured by light scattering at 340 nm as described under "Experimental Procedures." Note that the NC actually formed more mineral than did the original CRMV. 4B) contained significant amounts of free cholesterol and fatty acids, and an unidentified nonpolar lipid migrating near to monoacylglycerols, as well as small amounts of phosphatidylethanolamine, PS, and phosphatidylcholine. A similar lipid profile was seen in the NC after CHAPS extraction, but the amount of residual phospholipid was further reduced.
Hydrazine, Sodium Hypochlorite, and Sodium Hydroxide Treatment of CRMV-To more rigorously remove organic constituents, CRMV were treated with either hydrazine monohy-drate (40), or with sodium hypochlorite (41); sodium hydroxide treatment was used as a non-oxidizing, highly alkaline control (Fig. 5). It is important to note here that hydrazine treatment has been shown not to damage either ACP or HAP (40); hypochlorite treatment has been shown not to damage HAP crystals (41). A marked reduction in Ca 2ϩ uptake was noted after hydrazine treatment, whereas after treatment with NaOCl, nearly all Ca 2ϩ uptake ability of the CRMV was destroyed. In contrast, NaOH treatment had no effect (10 mM) or stimulated (100 mM) mineral formation. Chemical analysis of the hydrazine-treated vesicles revealed that essentially all of the Ca 2ϩ and P i were retained, but about 22% of the total P was removed, presumably from loss of phospholipids and phosphoproteins. Hydrazine removed ϳ95% of the proteoglycan associated with the MV and ϳ85% of the Lowry protein. Elemental composition of the hydrazine core by EDAX analysis revealed Ca and P in nearly equimolar levels (see below).
X-ray Diffraction Analysis-X-ray diffraction was used to determine the crystallinity of the NC and to follow the pattern of mineral accumulation when the NC was incubated in SCL. At 0 h, with both the Triton X-100 and the CHAPS NC, a broad, diffuse diffraction spectrum unlike that of synthetic ACP was observed; it appeared to be derived primarily from organic material present in the NC (data not shown). After incubation in SCL for 1-3 h, small diffraction bands appeared in the region of 32°and 25.8°2 indicative of the formation of poorly crystalline apatite. After 16 h incubation, there were increased amounts of diffractive material (Fig. 6). The mineral phase may be a poorly crystalline carbonate hydroxyapatite, similar to that seen in other biological minerals (18,20).
TEM Analysis-The NCs were examined by TEM to reveal their ultrastructure. At 90,000 ϫ magnification, the CHAPS-NC appeared to be composed of clusters of particles in aggregates ranging in size from ϳ10 to 110 nm in diameter (Fig. 7A). At 450,000 ϫ magnification, these aggregates appeared to be made of branched chains of particles 6.5-7.0 nm in  5. Effect of hydrazine, sodium hypochlorite, and sodium hydroxide treatment on mineral formation by CRMV. CRMV were treated with hydrazine monohydrate, sodium hypochlorite, or sodium hydroxide to dissolve organic matter, as described under "Experimental Procedures," and incubated in SCL for measuring mineral formation. Note after hydrazine treatment, that while mineral formation occurred more quickly than normal after introduction into the SCL, the period of rapid mineral formation was less extended and at a slower rate than that of native CMRV. After 100 mM sodium hypochlorite treatment (pH 12), there was no appreciable mineral formation for the duration of the assay, whereas 10 mM NaOH, (pH 12) had little effect on the vesicle's activity. Also note that 100 mM NaOH (pH 13) stimulated both the rate and extent of MV-mediated mineralization. This indicates that there was no crystalline mineral present in the CRMV, otherwise mineral formation would have occurred rapidly and extensively. diameter (arrows) in arm-like extensions from the main body of the aggregate (Fig. 7B). At 900,000 ϫ magnification, these ϳ7-nm particles appeared to be composed of clusters of individual subunits ϳ1.0 nm in diameter (Fig. 7C, arrows).
When the Triton X-100 NC was examined at 90,000 ϫ magnification, aggregates ranging in diameter from ϳ10 to 100 nm were again seen (Fig. 7D). At 360,000 ϫ magnification, these aggregates appeared to be composed of small hollow spheres (ϳ7 nm outer, ϳ3 nm inner diameter, arrows) (Fig. 7E). At 600,000 ϫ magnification, these hollow spheres appear to be composed of subunits ϳ1.0 nm in diameter (Fig. 7F). Thus, although there were differences in general appearance, the basic structures of both the CHAPS and the Triton X-100 cores were similar. The ϳ1.0-nm dot-like subunits appear to be the smallest functional component of nucleational core of MV. Their size, and the overall Ca 2ϩ and P i composition of the NC (see below), indicate that they may be clusters of Ca 2ϩ and P i (HPO 4 2Ϫ ) ions stabilized by lipids (detergent) and proteins. EDAX-Well defined clusters of the detergent-extracted NC were also analyzed by EDAX to determine their elemental composition. For the CHAPS-treated NC, the relative abundance in atomic percent of the major elements detected was: Ca, 51.4 Ϯ 0.1%; P, 48.6 Ϯ 0.1% (mean Ϯ S.E.). This gave a Ca:P ratio of 1.058 Ϯ 0.005. For the Triton X-100-treated NC, the atomic percentages were: Ca, 51.6 Ϯ 0.1%; P, 48.4 Ϯ 0.1%, giving a Ca:P ratio of 1.066 Ϯ 0.004. EDAX analysis of the hydrazine core gave the following atomic percentages: Ca, 51.8 Ϯ 0.5%; P, 48.2 Ϯ 0.5%, giving a Ca:P ratio of 1.076 Ϯ 0.022. Thus, the NC obtained after extraction with each detergent, or with hydrazine, yielded material with very low Ca/P molar ratios close to that of brushite. However in each case, electron diffraction showed the material to be amorphous (data not shown).
FT-IR Analyses-FT-IR analyses of the NC after CHAPS extraction (Fig. 8A) revealed a strong P-O stretch band at 1054 cm Ϫ1 with shoulders at 1096 and 1000 cm Ϫ1 and a broad unsplit P-O absorption band (558 cm Ϫ1 ) characteristic of noncrystalline calcium phosphate minerals. There were the strong O-H stretch bands at 3332-3507 cm Ϫ1 of hydrates, strong C-H stretch bands of lipids at 2921.5 and 2851.7 cm Ϫ1 , as well as the strong amide bands of proteins (1654 and 1544 cm Ϫ1 ). The NC after Triton X-100 extraction (Fig. 8B) had an FT-IR spectrum similar to that seen with CHAPS. In both NCs there was a well defined peak at 1238 cm Ϫ1 characteristic of Ca 2ϩ bound to the O-P-O phosphodiester of the polar head group of PS (43).
When the material remaining after treatment of CRMV with hydrazine was examined by FT-IR, a pattern with greatly attenuated organic constituents was seen. There was a broad O-H stretch band at 3400 cm Ϫ1 ; however, the C-H stretch region of the lipid acyl chains and the two amide bands of proteins were much reduced. The P-O stretch bands of mineral phase phosphates were strong (data not shown). NaOCltreated MV were not examined by FT-IR because previous studies by Tomazic et al. (44) had shown that hypochlorite treatment altered all mineral phases except HAP.
The observation of a peak at 1238 cm Ϫ1 (Fig. 8, A and B) suggested the presence of Ca 2ϩ complexed with PS in both the CHAPS and Triton X-100 NCs, a finding consistent with earlier data indicating the presence of CPLXs in MV (26). Thus, FT-IR spectra of synthetic CPLXs were also obtained. All samples of CPLX had a distinct peak at ϳ1238 cm Ϫ1 (Fig. 8C). Other features were: strong C-H stretch bands at 2852 and 2922 cm Ϫ1 typical of lipid acyl chains, strong peaks at 1740, 1637, and 1618 cm Ϫ1 from the CϭO groups of acyl esters and free carboxylate groups of PS, strong P-O stretch peaks at 1035-1101 cm Ϫ1 , and two sharp peaks in the P-O absorption region at 561 and 600 cm Ϫ1 . CPLX isolated after brief (Ͻ10 min) reaction of Ca 2ϩ with PS and P i , showed only a broad band in the 1035-1101 cm Ϫ1 region.
To directly verify the presence of CPLX in the NC, without stripping the MV with detergents, solid-state 31 P NMR analyses of lyophilized CRMV were performed (Fig. 9). Bloch decay spectra of the MV (Fig. 9A) had a somewhat asymmetric central peak at 3.2 ppm, surrounded by an asymmetric pattern of six spinning sidebands of varying intensity. The central peak of this Bloch decay spectra is characteristic of PO 4 3Ϫ groups (e.g. in ACP), but the six asymmetric sidebands indicated that this was not the only solid phase. Proton cross-polarization 31 P NMR spectra of the MV (Fig. 9B) greatly enhanced the intensity of the sidebands, indicating an abundance of acid phosphate (i.e. HPO 4 2Ϫ ) groups in the solid phase. The complex pattern of spinning sidebands seen in MV was unlike that of any known crystalline calcium phosphate (45); however, it was similar to that observed for synthetic CPLX (Fig. 9C).

DISCUSSION
In this study, we focused on identification and characterization of the NC present in MV. Previous FT-IR and FT-Raman spectroscopic studies had failed to detect the presence of well defined P-O stretch or bending bands in freshly isolated MV (13,14), even though the vesicles were rich in Ca 2ϩ and P i (7). Apparently, the P i initially present in MV was masked by large amounts of organic constituents. Therefore, in this paper we used a sequence of extraction steps, digestions with purified enzymes, and treatments with detergents, to remove interfering organic material from the MV. The residual mineral-rich NC was then examined by a variety of physicochemical methods to characterize its structure and composition.
The majority of the original Ca 2ϩ and P i were recovered in the NC, but only about 30% of the original protein. SDS-PAGE analyses of the NC revealed that annexin V (33 kDa) was the principal protein, although several others, including annexins FIG. 6. X-ray diffraction analysis of the mineral formed by the nucleational core after incubation in SCL. CRMV were sequentially treated as shown in Fig. 1, and then extracted with either CHAPS or Triton X-100 to yield the residual cores, which were incubated in SCL. At the specified times, the incubation mixture was centrifuged at 100,000 ϫ g for 1 h and the pellet lyophilized and analyzed by micro x-ray diffraction. A, CHAPS-NC incubated for 1 h in SCL; B, CHAPS-NC incubated for 16 h; C, Triton X-100-NC incubated for 1 h in SCL; D, Triton X-100-NC incubated for 16 h. Note the buildup of the diffracting mineral phase after the 16-h incubation, although there is already detectable crystalline mineral present after 1 h of incubation. Before incubation with SCL, the diffraction patterns of the NC were very broad and no peaks characteristic of crystalline phosphates were evident (data not shown).
II and VI (36 and 67 kDa) and alkaline phosphatase (ϳ80 kDa), were also present (Fig. 3, and Ref. 15). Distinct bands at 45, 50, and Ͼ300 kDa were also seen (Fig. 3). Recent studies indicate that a 50-kDa protein binds to annexin V in the presence of Ն800 nM Ca 2ϩ , preventing its removal from membranes (46). In view of the abundance of annexin V and Ca 2ϩ , the 50-kDa protein in the NC may be related to this protein. The Ͼ300-kDa protein seen in the SDS gels of the NC remains to be identified.
The role of the annexins present in the NC will now be considered. By virtue of their well known acidic phospholipiddependent Ca 2ϩ -binding properties (47), it is not surprising that they were intimately associated with the mineral-rich NC (15). In fact, the majority of the annexins in MV can only be released by extraction with EGTA (48); however, some can be removed by extraction with acidic chloroform-methanol, which also removes mineral-associated acidic phospholipids (49). Thus in MV, annexins are tightly associated with the mineral and the acidic phospholipid, PS, and exhibit proteolipid-like properties. Recent studies indicate that annexin V may be important for the orientation of Ca 2ϩ and P i in the NC, stimulating crystal growth (33).
While the sequential enzyme and salt treatments removed little of the original lipid, the detergents extracted most of the polar lipid from the MV. In view of their amphipathic properties, it seems probable that these detergents displaced and/or substituted for the polar lipid, stabilizing the NC. However, they left behind much of the free and esterified cholesterol, free fatty acid, and an unidentified lipid migrating near monoacylglycerol, and critical small amounts of PS (Fig. 4). The persistence of sizable amounts of nonpolar lipids in the NC indicates that they are difficult to extract with detergents. That they may have a function is indicated by recent studies by Sarig et al. (50), who suggest that cholesterol plays a role in pathological mineralization. Previous reports (23)(24)(25)(26)(27) had noted that the acidic phospholipids interact and form complexes with the initial mineral phase, yet data presented here indicate that these lipids can be removed by detergent extraction. Since the NC was shown to be rich in calcium and phosphate, we assessed its nature by direct analysis of various native MV preparations (i.e. no treatment after isolation), using solid-state 31 P NMR. We compared these spectra with those of synthetic Ca 2ϩ -PS-P i complexes and calcium phosphate mineral standards (38). These studies (Fig. 8) provided important insight into the nature of the P-rich NC. All FIG. 7. TEM analysis of the nucleational core obtained by sequential extractions and finally detergent treatment. CRMV were sequentially treated as described in Fig. 1 to yield the CHAPS-NC (A-C) or the Triton-NC (D-F). After lyophilization, small samples were dispersed in 100% ethanol, and one drop of the alcoholic suspension was mixed with one drop of SCL and immediately transferred to carbon-coated grids and dried. Grids were not stained. The structures of the NC were examined by TEM at successively higher magnifications (see "Experimental Procedures"). CHAPS-NC: A, 90,000 ϫ (bar ϭ 200 nm); B, 450,000 ϫ (bar ϭ 40 nm); C, 900,000 ϫ (bar ϭ 20 nm). Triton-NC: D, 90,000 ϫ (bar ϭ 200 nm); E, 360,000 ϫ (bar ϭ 50 nm); F, 600,000 ϫ (bar ϭ 30 nm). In A, the arrow points to an aggregate of particles ϳ110 nm in diameter. In B, the arrows point to particles ϳ6 nm in diameter at the end of chains. In C, the arrows point to areas where clusters of individual subunits ϳ1.0 nm in diameter can be seen to make up the particles. (Some isolated subunits can be seen between some of the clusters of particles in B). In D, the arrow points to an aggregate of particles ϳ100 nm in diameter, which in E can be seen to be hollow spheres (arrows) ϳ7 nm outer, ϳ3 nm inner diameter. In F, careful inspection reveals that walls of the hollow spheres are composed of subunits (arrows) ϳ1.0 nm in diameter. 0-h MV samples had spectra similar to those of mixtures of ACP and synthetic CPLX, when examined by proton crosspolarized 31 P NMR. The distinct central peak at 3.2 ppm is characteristic of ACP or HAP, but diffraction studies showed no evidence of crystallinity. The proton cross-polarization spinning sideband pattern of the CRMV was highly distinctive and matched that of the CPLX; it did not resemble that of any of the known crystalline calcium phosphates (45). By measurement of the relative intensities of the central peak, and the spinning sidebands of the CRMV and known mixtures of ACP and CPLX, the relative abundance of the two phases could be assessed. These estimates gave a value of 8.2% of the total vesicle P as CPLX and 91.8% as ACP. This proportion of the total P as CPLX is in very close agreement with the 7% value obtained from solvent partition of MV by Wuthier and Gore (26).
FT-IR, x-ray and electron diffraction analyses, and TEM of the NC all revealed that the initial mineral phase was noncrystalline. The broad unsplit P-O stretch pattern of the FT-IR spectra of both the CHAPS and Triton-NC at 560 cm Ϫ1 is characteristic of noncrystalline calcium phosphates. X-ray diffraction and electron diffraction analyses similarly failed to detect significant crystallinity. TEM examination revealed that both the CHAPS-and the Triton X-100 cores were composed of small 6 -8-nm diameter aggregates composed of small dot-like subunits ϳ1.0 nm in diameter, similar to those previously described by Höhling et al. (51).
The CHAPS-NC had branched chains of aggregates in struc-tures resembling the high molecular weight sulfated proteoglycans (52). The formation of these aggregates may be related to the high molecular weight band seen in SDS-PAGE analysis of the CHAPS core (Fig. 3), but EDAX analyses did not show any S in the NC. TEM of the Triton X-100 NC revealed the presence of numerous hollow spheres ϳ7 nm in outside diameter. These are similar in structure and size to those seen with synthetic PS-Ca 2ϩ -P i complexes (53) or with ACP prepared by precipitation from aqueous media (54). At higher magnification, the walls of these hollow spheres appeared to contain the ϳ1.0-nm subunit particles. We interpret these findings to indicate that the hollow spheres are composed of a residual lipid (and/or detergent) in which are absorbed the ϳ1.0-nm clusters of Ca 2ϩ and HPO 4 2Ϫ /PO 4 3Ϫ ions. The ϳ1.0-nm diameter of the dot-like subunits is similar in size to the core structure of ACP measured by radial distribution function analysis (42,55).
The marked enhancement in the 31 P NMR spectrum by proton cross-polarization, and the coupling constants observed (38), can only be accounted for by the presence of HPO 4 2Ϫ , not by adjacent H 2 O molecules. These findings reveal that the NC of native MV must contain substantial levels of protonated (HPO 4 2Ϫ ) phosphate. This is in agreement with EDAX analyses FIG. 8. FT-IR analysis of the nucleational core and of a synthetic PS-Ca 2؉ -P i complex. CRMV were sequentially extracted to yield the CHAPS-NC or the Triton X-100-NC as described in Fig. 1. The PS-Ca 2ϩ -P i complex was synthesized as described previously (33). All samples were lyophilized, and ϳ1 mg of each was incorporated into ϳ300 mg of KBr, compressed into a transparent pellet, and examined by FT-IR. A, CHAPS-NC; B, Triton X-100-NC; C, synthetic CPLX. The band at ϳ1235 cm Ϫ1 seen in the CHAPS and Triton X-100 NCs and the CPLX is characteristic of Ca 2ϩ -PS complexes (43).
FIG. 9. Solid-state 31 P NMR analysis of native CRMV. Lyophilized CRMV (10 -20 mg) were packed into 7-mm Doty Scientific SiN 3 rotors, spun, and recorded at room temperature as described under "Experimental Procedures." Bloch decay and cross-polarization pulse sequences were employed with high proton decoupling; 10-s recycle time was used for the Bloch decay, and 3-s times were used for the cross-polarization experiments. A, CRMV, Bloch decay; B, CRMV, proton cross-polarization; C, CPLX, proton cross-polarization. Note the marked increase in intensity of the spinning sidebands in B compared to A, when subjected to proton cross-polarization. Note the close similarity in the position and relative intensity of the spinning sidebands of the CPLX and CRMV upon proton cross-polarization. The large central band in A and B is due to the presence of ACP. Based on the relative intensity of the central peak and the sidebands, compared to known mixtures, ϳ8% of the P in the CRMV is composed of the CPLX. of the NC. Ca 2ϩ /P i ratios from ten separate areas of the CHAPS and Triton X-100 NC were 1.062 Ϯ 0.010 (mean Ϯ S.D.); those of the hydrazine core from which the most organic phosphate was removed were 1.076 Ϯ 0.039. Since EDAX showed no significant other elements (e.g. no sodium, magnesium, or potassium), these values indicate that the NC contains calcium phosphates with predominantly HPO 4 2Ϫ (see NMR data). Assuming that ACP was composed of mixtures of both Ca 9 (PO 4 ) 6 (deprotonated phosphate, Ca/P ratio ϭ 1.50) and CaHPO 4 (Ca/ P ϭ 1.00) the relative abundance of these two forms can be estimated. Taking into account the presence of the CPLX (ϳ8% of the P), the HPO 4 2Ϫ form appears to represent about 71% of the total P in MV.
FT-IR spectra of the synthetic CPLX revealed the presence of P-O stretch peaks characteristic of crystalline calcium phosphates, but it was not crystalline to electron diffraction. EDAX analyses of pure CPLX reveal a Ca/P stoichiometry of 1:2 (i.e. Ca 2ϩ :1 PS:1 P i ). Taken together these data suggest that the NC contains CPLX with Ca 2ϩ bound to the phosphodiester group of PS with bond orientations and energies that mimic the short range order of crystalline calcium phosphates. However, diffraction data indicate a lack of long range order seen in the crystalline state.
Recent solid-state 31 P NMR studies of bone mineral (56) have revealed the presence of a unique type of protonated phosphate (HPO 4 2Ϫ ) group, which has an isotropic chemical shift precisely like that of OCP, and not like that of brushite. In contrast, its anisotropic chemical shift is similar to that of brushite, but significantly different from that of OCP. It is possible that this protonated phosphate may be related to the CPLX complex we have identified in MV. This idea is supported by our previous finding that the early mineral induced by MV has characteristics, when examined by FT-IR and FT-Raman, like that of OCP (13,14).
To conclude, major advancement has been made in the structural and physicochemical characterization of the NC of MV. It is now clear that the initial mineral phase is primarily composed of a complex of the acidic phospholipid, PS, and a noncrystalline calcium acid phosphate. However, adsorbed to this complex are additional lipids such as free cholesterol, the lipiddependent Ca 2ϩ -binding proteins (the annexins), and some other proteins still to be characterized. Given the complexity of this core material, much additional work will be required to fully characterize its remaining constituents and how they contribute to NC properties.