In Vitro Modeling of Matrix Vesicle Nucleation

Annexins A5, A2, and A6 (Anx-A5, -A2, and -A6) are quantitatively major proteins of the matrix vesicle nucleational core that is responsible for mineral formation. Anx-A5 significantly activated the induction and propagation of mineral formation when incorporated into synthetic nucleation complexes made of amorphous calcium phosphate (ACP) and Anx-A5 or of phosphatidylserine (PS) plus ACP (PS-CPLX) and Anx-A5. Incorporation of Anx-A5 markedly shortened the induction time, greatly increasing the rate and overall amount of mineral formed when incubated in synthetic cartilage lymph. Constructed by the addition of Ca2+ to PS, emulsions prepared in an intracellular phosphate buffer matched in ionic composition to the intracellular fluid of growth plate chondrocytes, these biomimetic PS-CPLX nucleators had little nucleational activity. However, incorporation of Anx-A5 transformed them into potent nucleators, with significantly greater activity than those made from ACP without PS. The ability of Anx-A5 to enhance the nucleation and growth of mineral appears to stem from its ability to form two-dimensional crystalline arrays on PS-containing monolayers. However, some stimulatory effect also may result from its ability to exclude Mg2+ and HCO–3 from nucleation sites. Comparing the various annexins for their ability to activate PS-CPLX nucleation yields the following: avian cartilage Anx-A5 > human placental Anx-A5 > avian liver Anx-A5 ≥ avian cartilage Anx-A6 >> cartilage Anx-A2. The stimulatory effect of human placental Anx-A5 and avian cartilage Anx-A6 depended on the presence of PS, since in its absence they either had no effect or actually inhibited the nucleation activity of ACP. Anx-A2 did not significantly enhance mineralization.

To elucidate the roles of these components of the nucleational core of MVs, we incorporated them into synthetic complexes, which were investigated for their ability to induce and propagate mineral formation when incubated in SCL. In this report, we focus on the contributions of the various MV annexins, comparing their mineral forming activities when incorporated into synthetic complexes. Although in previous studies, we used a simple KCl-KP i buffer to make CPLX to reconstruct the nucleational core (29), in vivo, where MVs form, a more complex ionic environment occurs. Therefore, in this study we used an intracellular phosphate buffer (ICP) (30) with an electrolyte content similar to that observed in ultrafiltrates of lysates of isolated growth plate chondrocytes (9) to construct ACP and CPLX, which were used to initiate mineral formation from SCL in vitro.

EXPERIMENTAL PROCEDURES
Synthesis of ACP, CPLX, and Annexin-containing Complexes-A 4ϫ stock emulsion of PS was prepared by drying 5 mg of PS in chloroform under N 2 to form a thin film in a test tube. Then 2 ml of the P i -rich ICP buffer was added. ICP buffer contained 106.7 mM K ϩ , 45.1 mM Na ϩ , 1.5 mM Mg 2ϩ , 115.7 mM Cl Ϫ , 23.0 mM P i , 10 mM HCO 3 Ϫ , 1.5 mM SO 4 2Ϫ , and 3.1 mM N 3 Ϫ as a preservative; its total molarity was 153.3 mM, and its pH was 7.2 (30). The tube was then sonicated for 1-2 min at 25°C in a water bath to form a uniform emulsion of small unilamellar vesicles. To make the PS-CPLX, 75 l of the 4ϫ PS stock emulsion were mixed with 225 l of ICP buffer, and then 6 l of 100 mM CaCl 2 was added dropwise with rapid stirring over a 5-10min period to form the insoluble complex, which was harvested by centrifugation for 5 min at ϳ15,000 ϫ g. The pellets were resuspended by sonication in 300 l of SCL.
To prepare the annexin-containing complexes, native proteins (chicken cartilage from ϳ20,000 metatarsals, Anx-A5, Anx-A2, and Anx-A6; chicken liver, Anx-A5; human placenta, Anx-A5) were purified as previously described (8,25,31) and dialyzed against the ICP buffer. Documentation of their purity is shown in Fig. 1. Protein levels were estimated by the method of Lowry et al. (32) and by Coomassie Blue staining of gels. Aliquots containing 75-225 g of the native Anx isolates were added to 75 l of the 4ϫ PS stock solutions, and the final volume was adjusted to 300 l before adding the CaCl 2 to precipitate the complex as above. These complexes were assayed for mineralization activity.
As a control, PS was omitted, and the 100 mM CaCl 2 stock was added dropwise into the ICP buffer with rapid stirring over a 5-10-min period to form ACP. As another control, purified native annexins were prepared in the ICP buffer (without PS); CaCl 2 was then added to form Anx-A5-ACP complexes, which were harvested by centrifugation. As a further control to ensure that the observed turbidity was due to mineral formation and not to simple aggregation of the seeded complexes, P i was omitted from the calcifying medium to preclude mineral formation.
Mineralization Assay-Mineral formation was monitored by light scattering (33) using the multiwell microplate assay system described previously by Wu et al. (23). Using this method, an increase of 0.1 A at 340 nm is roughly equivalent to precipitation of ϳ10% of the total Ca 2ϩ in the 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, 12.7 mM K ϩ , 5.55 mM glucose, 1.83 mM HCO 3 Ϫ , 0.57 mM Mg 2ϩ (9). After centrifugation of the CPLX-(and ACP)-forming reaction mixtures and resuspensions, 60 -80 l of these nucleators were added to 1 ml of SCL followed by brief sonication to yield uniform suspensions. Quadruplicate samples (140 l) from each 1-ml suspension were distributed into four wells of a 96-well half-area Costar microplate; turbidity measurements were made and recorded automatically at 15-min intervals for 12-16 h using a Labsystems iEMS MF microplate reader (Needham Heights, MA). Prior to each absorbence reading, the plate reader automatically agitated the sample to ensure that all particulate matter was resuspended (34). When P i was omitted from SCL, there was no significant increase in turbidity at 340 nm over the entire incubation period, regardless of which nucleator was introduced into the medium. Thus, the observed increase in turbidity is due to mineral formation and not to simple aggregation of the nucleation components.
Analysis of the Kinetics of Mineral Formation-The kinetics of mineral formation was analyzed using the method of Genge et al. (34). This method involves first derivate analysis of mineral formation curve (the plot of absorbence at 340 nm versus incubation time) to precisely measure 1) the time needed to induce mineral formation (T I ), 2) the average rate of mineral formation during the rapid formation period (RMF R ), and 3) the nucleation potential (NP) ((RMF R /T I ) ϫ 100), a sensitive measure of the ability of each factor to induce and support mineral formation. This method, illustrated in Fig. 2, shows both a mineral formation curve and its first derivative (dA/dT) to illustrate how T I , RMF R , and NP were calculated. In addition, as described in the legend to the figure, a five-parameter logistic curve-fitting algorithm was used to quantify the maximal amount of mineral formed (AMF Max ). The effects of the different factors (e.g. PS, the different annexins, ACP, etc.) on these four nucleational parameters enabled an accurate evaluation of how each contributed to the mineral formation process.

Effect of Annexin Level Used in Forming PS-CPLX Nucleators-Although
Anx-A5 is known to be a component of the nucleational core (23) and to stimulate MV mineralization (34), it was not clear how much of the protein was required. To answer this question, the mineralizing activity of annexin-free PS-containing complexes (PS-CPLX) and those containing graded amounts of avian liver Anx-A5 (ALAnx-A5) were compared. In the absence of ALAnx-A5, the addition of 60 or 80 l of PS-CPLX to SCL formed much less mineral than was seen in the presence of the Anx (Fig. 3, A and B). Induction times were slow (13-15 h), and the rate of rapid mineral formation, the Shown is an SDS, 7.5-15% acrylamide gradient gel. Approximately 3-5 g of each purified native protein were loaded onto each lane. The gel reveals that each protein was greater than 98% pure; their identity (see top of each lane) was previously established by type-specific antibodies, amino acid sequence analysis, and molecular biological methods (31). Crystallography of purified Anx-A5 also shows that the 35 kDa band is authentic Anx-A5 (61).
nucleation potential, and the total amount of mineral formation were less than one-third of that seen with ALAnx-A5-containing complexes ( Table 1).
Incorporation of increasing levels of ALAnx-A5 into the PS-CPLXs produced progressively more active nucleators. For example, incorporating 15, 30, and 45 g of ALAnx-A5 into the complex added at 60 l/ml of SCL, reduced the induction times by 2.7-, 4.0-, and 4.8-fold, and increased the nucleation potential by 9.7-, 15.6-, and 18.5-fold, respectively, compared with the annexin-free PS-CPLX controls (Fig. 4, A and B). Although ALAnx-A5 increased the rate of rapid mineral formation 3.5-4.0-fold and the maximal amount of mineral formation 2.5-3.0-fold, they appeared to have nearly reached their maximal values at these levels ( Fig. 4, C and D). Increasing the volume of nucleator from 60 to 80 l/ml of SCL led to small further increases in mineralizing activity. Half-maximal effects were obtained at only ϳ5 g of ALAnx-A5 in PS-CPLX/ml of SCL (ϳ140 nM). Thus, only low levels of Anx-A5 were sufficient to significantly stimulate mineralization of the PS-CPLX.
Effect of Different Classes of Annexins-MVs have been shown to contain three classes of annexins: Anx-A5, Anx-A6, and Anx-A2; however, their direct effects on mineral formation had not been previously examined. Nor had annexin A5s from different sources been compared. Therefore, these were incorporated into both ACP and PS-CPLX nucleation complexes prepared from ICP buffer, and their effects on the rate and amount of overall mineral formation as well as the various nucleation parameters were compared.
As is evident from Fig. 5, the effects of the various annexins on mineral formation induced by ACP (left panels A, C, E, G, and I) as opposed to PS-CPLX (right panels B, D, F, H, and J) were quite different. As previously noted, annexinfree PS-CPLX had low mineralforming activity, inducing mineral formation only after ϳ10 h. Incorporation of all three Anx-A5s and Anx-A6 into PS-CPLX caused much quicker induction and more rapid mineral formation (Fig. 5, B, D, F, and H); however, Anx-A2 ( Fig.  5J) had minimal effect. Of the three MV annexins, avian cartilage Anx-A5 (ACAnx-A5) was the most potent activator of mineralization, shortening the induction time by 5.8-fold and increasing the nucleation potential by 13.5-fold as well as doubling the rate and amount of maximal mineral formation ( Table 2, top). The stimulatory effects of the other Anx-A5s and avian cartilage Anx-A6 (ACAnx-A6) on PS-CPLX were less, but still highly significant.
With ACP complexes, the effects of the annexins were quite different. Although both ACAnx-A5 and ALAnx-A5 accelerated the onset of mineral formation, human placental annexin-A5 (HPAnx-A5) did not (cf. Fig. 5, A, C, and E), and both ACAnx-A6 and avian cartilage annexin-A2 (ACAnx-A2) delayed the onset of mineral formation (Fig. 5, G and I). ACAnx-A5 had the highest nucleation potential but had little effect on the rate of rapid mineral formation or the amount of maximal mineral formation when compared with the no-annexin control ( Table 2, bottom). ALAnx-A5 caused slower induction of mineral formation than ACAnx-A5 but a some- To analyze raw data from the microplate reader, base-line absorbence (A) at 340 nm was established by averaging readings before any statistical increase was observed. This base line was subtracted from the observed absorbence values to obtain base line-corrected mineral formation, smoothed by calculating a running average of three successive measurements. The resulting mineral formation curve (blue diamonds) and dA/dT (the change in absorbency in each 15-min period; rose triangles) depict the changing rate of mineral formation during the incubation. Illustrated in red is the equation for the linear portion of the ascending slope of the first derivative curve; when y ϭ 0, the value of x ϭ T I ϭ 3.12 h. Illustrated in blue is the equation for the linear portion of the descending slope of the first derivative curve; when y ϭ 0, x ϭ 5.44 h, the end of the rapid formation period (EFP R ). The average rate of mineral formation during the rapid formation period (RMF R ) is calculated by dividing the amount of mineral formation (A at 340 nm at EFP R , ϭ 0.339) by the length of the rapid formation period (EFP R Ϫ T I , 5.44 Ϫ 3.12 ϭ 2.32 h); thus, RMF R ϭ 0.339/2.32 ϭ 0.146 dA/h. The NP, a measure of the ability of a nucleator to induce and sustain mineral formation, is calculated by dividing the RMF R by the T I ϫ 100; thus, NP ϭ (0.146/3.12) ϫ 100 ϭ 4.70. Calculation of the AMF Max was made using a five-parameter logistic curve-fitting algorithm for analysis of sigmoidal data whose equation is as follows: (75). For the above mineral formation curve, the values of the five parameters (shown in parentheses) are as follows: a ϭ 0, the observed base line absorbance at 340 nm corrected for background; b ϭ 37.157, a value related to the rate of mineral formation at the inflection point; c ϭ 2.451, a value related to the induction time; d ϭ 0.5587, the asymptotic value where absorbence at 340 nm reaches its maximum, AMF Max ; e ϭ 0.0313, an asymmetry factor related to the skewness from classic Gaussian distribution of the first derivative curve. To obtain these five parameters, the background-corrected absorbance values for the data set were linked to a customized Excel spreadsheet, generating the graph of mineral formation (blue diamonds) shown above. A first approximation of the five parameters was made by assigning values to each from visual inspection of the curve, observing the goodness of fit of the calculated curve generated from embedded formulas and macros within the customized Excelா spreadsheet. A customized Visual Basic code was then run to activate the Solver Tool within Excel, refining the estimated five parameters after 20,000 iterations to optimized values that minimize the sum of the squares of the errors of the differences between the experimental and the calculated curve. Typically, the parameters converge to sum of the squares of the errors values of less than 10 Ϫ5 . Using this five-parameter equation, the most reproducible values for AMF Max (typically with an S.D. value of Ͻ5%) were obtained by using data points from the end of the rapid formation period to the end of the incubation.
what higher amount of maximal mineral formation. In contrast, HPAnx-A5 not only did not accelerate onset of mineral formation by ACP but caused significant reductions in the rate of mineral formation, the nucleation potential, and the maximal amount of mineral formation. ACAnx-A2 also significantly slowed the onset of mineral formation, reducing the rate and amount of mineral formation when compared with the no-annexin control. On the other hand, ACAnx-A6 caused an even greater inhibition of ACP-induced mineral formation. It slowed the induction time by ϳ45% and reduced the rate of rapid mineral formation by 24%, the nucleation potential by 47%, and the maximal amount of mineral formation by 18%. The inhibitory effect of ACAnx-A6 on ACP-nucleated mineral formation was opposite its stimulatory effect on PS-CPLX mineral formation.
The distinct effect that the various annexins have on the nucleation potential and the time needed for induction of mineral formation is illustrated in Fig. 6. Here the marked enhancement of mineral formation (nucleation potential) by the Anx-A5s, and especially by cartilage Anx-A5, is contrasted with the lack of effect of Anx-A2. Also note the inverse relationship between the nucleation potential and induction time. Also note how the combination of PS with ACP to form the ICP-based PS-CPLX (Fig. 6A) clearly reduced the nucleation potential and lengthened the induction time when compared with the ACP control (Fig. 6B). Comparison of the two panels shows that with both PS-CPLX and ACP, Anx-A5 markedly increased the rate of mineral formation and shortened the time needed to trigger its onset.

DISCUSSION
The primary purpose of this study was to elucidate the specific role that Anx-A5 plays in the induction and propagation of MV mineral formation. This quantitatively major protein is a known component of the nucleational core of MVs (29). However, since there are two other annexins present (Anx-A2 and Anx-A6), another objective was to define the contributions that each made to the nucleational core. We determined their effects on mineralization when incorporated into a variety of synthetic biomimetic nucleational complexes (simple binary annexin-ACP complexes as well as ternary annexin-PS-CPLXs). Our studies reveal that Anx-A5 plays two major roles in regulating mineral formation from PS-CPLX; it strongly accelerates the onset of mineral formation, and it significantly increases its rate.
The first issue that needed to be addressed was how much Anx-A5 was needed to activate the PS-CPLX. We found that incorporation of 15-45 g of avian liver Anx-A5 into PS-CPLX seeded into 1 ml of SCL was sufficient to cause marked shortening of the induction time as well as major increases in the rate and amount of mineral formation. Impressively, the nucleation potential was increased 10 -20-fold. Half-maximal increases were obtained with only 5 Ϯ 1 g of the annexin/ml of SCL. This would amount to only one Anx-A5 for ϳ5000 PS-CPLX units. Considering this low stoichiometric relationship, it is evident that a catalytic mechanism must be involved.
How did the annexins do this? To answer this question, it is necessary to understand the basic structure of PS-CPLX. Electron microscopic examination reveals that it forms extended planar two-dimensional bilayers (40 Ϯ 2 Å thick) rather than vesicular structures (10,35). The packing arrangement of the PS-CPLX head groups at the surface of the membrane probably is hexagonal, but the area occupied by each head group is currently unknown. Based on the condensation that is known to occur in the presence of Ca 2ϩ , it would be expected to be in the range of 40 -55 Å 2 (36). Such structures would facilitate nucleation of crystalline mineral simply because their planar surface matches that of the crystals themselves. Evidence for this is seen in the structure of MVs undergoing mineral formation; the plane of the developing crystal often flattens the edge of the vesicle, distorting its shape (37). Further, radial distribution function-extended x-ray absorption fine structure analysis reveals that the polar head group of PS-CPLX has interatomic Ca-P distances (ϳ3.2 Å) similar to those in crystalline calcium phosphate minerals (38). The Ca-P distances indicate Ca 2ϩ coordinates with oxygen atoms of the carboxylate and the phos-phodiester bridge of PS, as well as with P i of the ACP component of the CPLX. Thus, PS-CPLX appears to create an interface between the lipid bilayer and the developing mineral crystal.
How then could AnxA5 contribute to this interfacial structure? First, Anx-A5 forms highly ordered two-dimensional crystalline arrays upon Ca 2ϩ -dependent binding to PS-rich monolayers (39 -42). Two main types of arrays have been found, with either p6 or p3 symmetry, both of which derive from building units composed of a trimer of Anx-A5 monomers (40,42). The more open p6 form occurs when the PS content of the planar membrane is low (5-20%); the p3 form occurs when the PS level is high (Ն40%). Comparison of the lattice parameters of the p3 crystal form of Anx-A5 with those of hydroxyapatite (43), octacalcium phosphate (44,45), and the core structure of ACP (46) indicates that the pattern formed by the clusters of PO 4 3Ϫ and Ca 2ϩ ions of the unit cell (ϳ9.5-Å radius) match the trilobate space at the center of each Anx-A5 trimer. Thus, the PS-CPLX-Anx-A5 arrays appear to form a two-dimensional crystalline pattern that facilitates nucleation of crystalline CaP i mineral formation.
In addition, there is a second way that AnxA5 may significantly stimulate mineral formation. Our initial studies revealed that PS-CPLXs formed from simple K 2 HPO 4 / KCl buffer were potent nucleators of hydroxyapatite mineral formation (29,47). However, those prepared from ICP buffer FIGURE 4. Dose response of nucleation parameters to avian liver annexin-A5 level. Samples were prepared as described in the legend to Fig. 3; responses to seeding with 60 and 80 l/ml of SCL are shown. Dose responses to level of ALAnx-A5 are shown as follows: T I (A); RMF R (B); NP (C); AMF Max (D). Note that maximal effects were nearly attained with 60 g of ALAnx-A5 in the complex. Using data from the four panels, halfmaximal effects of ALAnx-A5 were calculated to be 5 Ϯ 1 g/ml SCL.

TABLE 1 Influence of seeding level and level of incorporated avian liver annexin-5 on key parameters in the induction of mineral formation by PS-CPLX
Values are means Ϯ S.E. of the indicated number of samples. Differences between the mean of control (NO annexin) and ALAnx-5-containing samples were compared using the two-tailed homoscedastic Student's t test. Superscript values to the right of the S.E. are the exponential values of the probability that the differences were due to chance.

Parameter
Level of avian liver AnxA5 in nucleator/ml of SCL (seeding volume 60 l)

Influence of annexins on key parameters in the induction of mineral formation by PS-CPLX and ACP nucleational complexes
Values are means Ϯ S.E. of the indicated number of samples. Differences between the mean of NO annexin and annexin-containing samples were compared using the two-tailed homoscedastic Student's t test. Superscript values to the right of the S.E. are the exponential values of the probability that the differences were due to chance. (30), an Mg 2ϩ -and HCO 3 Ϫ -containing biomimetic of the intracellular fluid of growth plate chondrocytes (9), had markedly lower nucleational activity (34). It is most likely that Mg 2ϩ and HCO 3 Ϫ in ICP buffer are responsible for their impaired nucleational activity. Both Mg 2ϩ and HCO 3 Ϫ are known to stabilize ACP and prevent its conversion to hydroxyapatite (48,49). Analysis of MVs shows that for every four Ca 2ϩ present in their lipid-calcium-phosphate complexes, there are about three Mg 2ϩ (19). However, from the work now reported, it is clear that Anx-A5 enabled these ICP-based PS-CPLXs to become highly effective nucleators. A plausible explanation is that incorporation of Anx-A5 enables the exclusion of Mg 2ϩ and HCO 3 Ϫ from key nucleation sites in the ICP-based PS-CPLX. Supporting this possibility, we have found that Anx-A5 does not form complexes between Mg 2ϩ , P i , and PS (8). Further, P i does not interfere with the binding between Ca 2ϩ , PS, and Anx-A5 needed for the formation of the PS-CPLX (8). Thus, exclusion of Mg 2ϩ in addition to the two-dimensional arrangement of the Anx-A5 on the membrane surface helps explain why Anx-A5 so markedly stimulates crystalline mineral formation from ICP-based PS-CPLX.

PS-CPLX nucleation parameter
Comparing the various classes of annexins, our findings show that Anx-A5s and Anx-A6 were significantly more effective stimulators of mineral formation using PS-CPLX nucleators than was Anx-A2. Further, among the Anx-A5s tested, avian cartilage Anx-A5 was significantly more potent than either avian liver Anx-A5 or human placenta Anx-A5. The selective stimulation of avian cartilage Anx-A5 cannot be due to the presence of mineral contaminants from the cartilage. The extensive methods, including use of the calcium chelator glycol-bis (2-amino-ethylether)-N,N,NЈ,NЈ-tetraacetic acid required for isolation and purification of native annexin A5 would dissolve and remove any of the labile ACP mineral present in this tissue (25). Further, the lack of tissue-specific amino acid sequence variants in Anx-A5 points to a possible post-translational modification, perhaps phosphorylation of specific amino acid residues (50 -52). One commonly considered effect of Anx-A5 (i.e. formation of Ca 2ϩ channels) (27,(53)(54)(55) can be ruled out a priori, since the CPLXs are planar, not vesicular, structures.
Of the different nucleation parameters identified, those most dramatically altered by incorporation of Anx-A5 were the induction time and the rate of rapid mineral formation, as well as their derivative, the nucleation potential. The marked shortening of the time for induction of mineral formation suggests that Anx-A5 organizes the array of charges on the PS-CPLX lamellar surface to accelerate the kinetics of mineral nuclei formation. However, in addition, the tightly associated ACP moiety would provide a readily available source of Ca 2ϩ and P i to support crystal growth once nucleation occurs. The marked increase in rate during the rapid formation period suggests that Anx-A5 also enhances the growth of the initial calcium phosphate crystals by preventing them from becoming "poisoned" by surface adsorbents, such as Mg 2ϩ and HCO 3 Ϫ . Anx-A5 appears to protect kink sites where mineral growth is known to be propagated (56). In vivo, the assembly of Anx-A5 on the MV membrane may be facilitated by cell surface proteoglycans (57).
Although Anx-A2 is next in abundance to Anx-A5 in MV, it did not enhance nucleation of PS-CPLX. Its unique N-terminal 14 amino acids comprise a high affinity binding site for p11, a member of the S100 family of Ca 2ϩ -binding proteins (58). The heterotetrameric complex formed by interaction of two Anx-A2 and two p11 subunits appears to be involved in exocytosis. Thus, its function may be associated with MV formation rather than with mineralization. Thus, although Anx-A2 shares a common morphological core structure with Anx-A5 and exhibits Ca 2ϩ -dependent association with PS-enriched membranes, recent studies have shown that Anx-A2 is incapable of forming trimers upon binding to phospholipid bilayers (59). Anx-A2 also had much lower Ca 2ϩ stoichiometry for membrane binding (ϳ4 mol of Ca 2ϩ /mol of protein) than Anx-A5 (ϳ12 mol of Ca 2ϩ /mol of protein), which readily forms trimers upon interaction of PS-enriched membranes. It is of FIGURE 6. Effect of various annexin types on the relationship between the induction time and nucleation potential. Samples were prepared as described in Fig. 5. Paired values of T I and NP for each individual sample were plotted against each other. Note the striking inverse relationship between induction time and nucleation potential as well as the marked differences between annexin in their effect on this relationship. Also notice the lesser response of the ACP-based nucleators to the annexins when compared with the PS-CPLXbased nucleators. In A, with PS-CPLX nucleators, note that ACAnx-A5 clearly had the highest NP and the shortest T I of any of the annexins, whereas ACAnx-A2 had the longest T I and lowest NP of any of the annexins. In B, with ACP nucleators, note that ACAnx-A5 again had the highest NP and shortest T I but that ACAnx-A6 had the lowest NP and longest T I . Also note that the annexin-free ACP nucleator had higher NP and shorter T I values than ACAnx-A6 and some ACAnx-A2 values. A, PS-CPLX⅐annexin complexes; B, ACP⅐annexin complexes. significance that Zn 2ϩ , which strongly inhibits mineral formation by MV (35,60) also, has been shown by protein crystallography to disrupt the planar array of Anx-A5 trimers (61). Taken together, these data indicate that planar Anx-A5 trimers assembled on the surface of the laminar PS-CPLX substrate serve as nucleation sites on the inner MV membrane.
Anx-A6, the least abundant of the MV annexins (26), is a flexible 67-kDa bilobular protein that folds into two compact core structures, each essentially equivalent to that of a ϳ35-kDa annexin (62). Like Anx-A5, Anx-A6 also can form two-dimensional crystals when bound to membranes (63); however, the overall pattern is less ordered. Thus, it was not entirely surprising that avian cartilage Anx-A6 was nearly as effective as avian liver Anx-A5 in inducing mineral formation by the PS-CPLX. On the other hand, it was the most potent of the annexin classes in inhibiting mineral formation by ACP nucleational complexes, significantly lengthening their induction time and reducing the rate of rapid mineral formation. Thus, Anx-A6 may help preserve ACP structure in MV prior to induction of crystalline mineral formation. Anx-A6 also has been implicated in the organization of membrane domains (rafts), in particular their Ca 2ϩ -dependent association with the cytoskeleton (64). Membrane rafts are lateral assemblies rich in sphingolipids and cholesterol (65), two lipids that are enriched in MV (10,11). Thus, Anx-A6 may participate in organizing membrane domains that contribute to the nucleational activity of CPLX in MV. The unique ability of Anx-A6 to associate with sphingomyelin and cholesterol, both of which reduce the nucleational activity of PS-CPLXs, indicates that Anx-A6 may have a special role in enhancing MV mineralization.
Despite the obvious stimulation of mineral formation by Anx-A5, knock-out of the Anx-A5 gene does not interfere with skeletal development in mice (66). Such a finding is not unprecedented. Deletion of type X collagen, a unique protein abundant in the extracellular matrix of growth plate hypertrophic chondrocytes (67) and thought to participate in calcification (68), was found not to interfere with skeletal development (69). However, expression of a defective form of type X caused distinct skeletal abnormalities (70). Thus, it is possible that the lack of Anx-A5 was compensated for by the presence of Anx-A6, which had significant but less potent activation of the PS-CPLX mineralization.
It should be noted that although numerous studies have been made on the nucleational properties of various lipid⅐Ca 2ϩ ⅐P i complexes, these have led to variable findings, sometimes inhibiting and sometimes stimulating mineral formation. Some of these differences relate to the methods by which these lipidmineral complexes were formed. For example, if PS is added to preformed ACP, it markedly delayed its conversion to hydroxyapatite; however, if PS is present during the formation of ACP, its inhibitory effect is less pronounced and greatly alters the structure of the complex formed (35). Using a different approach, Boskey et al. (18) induced formation of "Ca⅐acidic phospholipid⅐P i " complexes by incubating the lipids with a Ca 2ϩ -and P i -containing synthetic lymph for several hours, followed by washing with diethylether/ethanol to remove noncomplexed lipids and finally by washing with HCl, pH 5, to remove any noncomplexed mineral. Such complexes are generally stimulatory to mineral formation. However, as had been noted earlier (16), the addition of P i prior to or in combination with Ca 2ϩ was essential for formation of nucleationally active CPLX. Further, nonacidic phospholipids by themselves were incapable of forming CPLXs.
The method chosen here for forming the PS-CPLX was predicated on what appears to occur in growth plate chondrocytes when they form MV. Thus, ICP buffer was designed to mimic the electrolyte composition of the chondrocytes, which was shown to contain Mg 2ϩ and HCO 3 Ϫ and to be rich in K ϩ and P i (9). The consistent finding of high levels of noncrystalline calcium phosphate in nascent MV indicates that MV must form under conditions conducive for ACP formation (i.e. solutions rich in either Ca 2ϩ or P i or both), although vesicular proteins typically mask this amorphous mineral (71)(72)(73). This condition is found in the intracellular fluid of growth plate chondrocytes, where P i levels are typically 20 -25 mM (9). Confocal imaging studies of living growth plate cartilage slices have revealed that the cells have increasingly active Ca 2ϩ metabolism as they approach the site of MV formation. Ca 2ϩ levels become markedly elevated near the plasma membrane, which becomes incorporated during MV formation (74). Thus, in our view, ACP and CPLX formation occurs at the time of MV formation. It is evident that the MV annexins, acidic phospholipiddependent Ca 2ϩ -binding proteins that our current studies reveal enhance mineral formation of CPLX, also become incorporated into the MV at this time.
With regard to the turbidometric assay, we find that it has many advantages for assessing mineral formation. It is highly sensitive and enables continual monitoring of mineral formation; it does not disturb the assay system; and it yields precise, detailed data on mineral formation. Nevertheless, it does have limitations. It cannot distinguish between the number and size of particles in suspension and hence does not distinguish between crystal nucleation and growth. Nor can it assess the composition of the crystals formed. However, phenomena such as flocculation and aggregation of added materials can be ruled out as factors in the observed turbidity. Using Ca 2ϩ -containing P i -free SCL, which precludes mineral formation, there was no significant increase in turbidity over the full incubation period. Since both Ca 2ϩ and P i were required together, the observed increases in turbidity must be due to mineral formation.
Further, when applying the findings of this in vitro system to the in vivo situation, other facts need to be kept in mind. For example, unlike living systems, the total amount of Ca 2ϩ and P i present in each well is fixed and does not change during the course of the experiment. Thus, as mineralization begins and the amount of mineral increases, the amount of solution phase Ca 2ϩ and P i in the SCL must decrease in direct proportion. (This would not necessarily be true in vivo, where as an open system these ions can be delivered continuously by the blood). Nevertheless, in either case, the final amount of mineral formed is dictated by the solubility product (Ksp) of the solid phase, hydroxyapatite. Thus, in vitro, as more and more solid phase forms, the driving force for mineral formation progressively decreases (i.e. the activity of Ca 2ϩ and P i in the solution phase decreases). The amount of mineral that can form therefore reaches an asymptotic maximum (AMF Max ) that depends on the availability of ions as well as the presence of surface-adsorbed entities that influence its Ksp. The fact that AnxA5 shortens the time of onset, as well as markedly increasing the rate and final amount of mineral formed by PS-CPLX, indicates that it must not only enhance nucleation of mineral formation but also must protect the growing crystals from the adsorption of inhibitors. Thus, Anx-A5 enables more extensive and perfect crystal growth. Applied to the open in vivo condition, it is reasonable to expect that, given a continuous supply of Ca 2ϩ and P i , Anx-A5 could cause a substantial increase in the overall amount of mineral formed until constrained by other physiological factors that restrict bone growth.