Formation of Nanoparticles by Fetuin-A and Albumin in Metastable Versus Supersaturated Calcium and Phosphate Solutions
With respect to nanoparticle formation and calcification, our experiments clearly indicate that both inhibitory and seeding influences are seen to be associated with serum, and, furthermore, that at least some of these inhibitory influences can be removed through protease and heat treatment of the serum tested. These results would also suggest a critical role for proteins in this same process of mineral precipitation, perhaps by their initiating nanoparticle formation, itself probably representing an early step of the biomineralization process. This possibility prompted us to investigate more carefully the role of purified serum proteins in mediating nanoparticle formation.
For our protein studies, we chose to use albumin and fetuin-A not only because they had been shown to be involved in calcium nanoparticle formation [1]–[3], [46] but also because of extensive studies pointing to an active role for both proteins in regulating calcium and apatite homeostasis in the blood [56], [57], [59]–[68] as well as in mediating calcification reactions elsewhere in the body [69]–[77]. Fetuin-A, for instance, is a potent calcification inhibitor, known to bind strongly to both calcium and nascent apatite crystals. By itself, fetuin-A accounts for about half to one-third of the total serum inhibitory capacity directed against spontaneous apatite precipitation associated with supersaturated solutions of calcium and phosphate [57], [66]. As for albumin, according to Garnett and Dieppe, it “not only accounts for half of the protein concentration in serum, but also contributes half of the inhibitory activity [directed against apatite crystal growth and associated with] the high-molecular-mass fraction” [62]. Intriguingly, albumin and fetuin-A represent two of the most abundant non-collagenous proteins found in the bones. For instance, albumin is found at 0.7 mg/g of bone in bovines [78] whereas fetuin-A is usually found at 0.9 to 1 mg/g of bone in rats [79], bovines [78], and humans [80]. Both proteins are also found associated with protein-mineral complexes that form in the serum under conditions in which ectopic calcification is triggered. Such protein-mineral complexes have been shown for instance to form in the serum of animals treated with vitamin D, warfarin, etidronate, or adenine, all known to modulate or perturb calcium homeostasis and to induce calcification [69]–[72], [76]. Similarly, both albumin and fetuin-A are present in mineral complexes isolated from humans undergoing severe extra-skeletal calcification as seen in cases of calcifying peritonitis [74]. The concept that these proteins act as systemic calcification inhibitors is illustrated by the observation that fetuin-A-deficient mice spontaneously develop ectopic calcification when fed with a diet rich in mineral and vitamin D [67], [68]. On the other hand, evidence suggests that albumin may promote mineral formation in some experimental conditions, an observation that would confer to this protein a dual inhibition-nucleation role on nanoparticle formation. For instance, albumin at low concentrations (under 10 mg/ml) promote the deposition of calcium phosphate on collagen while higher concentrations (above 10 mg/ml) inhibit collagen mineralization [81], [82]. In addition, albumin has also been shown to promote the precipitation of calcium phosphate in metastable solutions when adsorbed onto HAP crystals [83], [84]. It seems important therefore to verify whether albumin and fetuin-A may lead directly to calcium nanoparticle formation under the culture conditions used to study NB and nanobacteria-like particles (NLP) in the past.
Through our earlier studies, we have shown that both bovine and human forms of albumin and fetuin-A are integral components of calcium nanoparticles formed from serum [1]–[3]. In fact, antibodies thought to be directed earlier against human NB [5], [8], [24], [26]–[29], [35], have been shown to cross-react across species against fetuin-A and albumin from both FBS and HS [1]–[3]. From our own antibody analysis, we have inferred earlier that both bovine and human antigens may be present in all the earlier NB studies done to establish the presence of nanobacterial antigens in human tissues; in those earlier studies, NB were supposedly “grown” in medium supplemented with untreated or gamma-irradiated FBS (described in more detail in refs. [2], [3], and the many references cited therein). Thus, for the purpose of demonstrating a role for proteins in the assembly of NB and NLP, we used both the bovine and human forms of albumin and fetuin-A in the current study. For brevity, however, only results pertaining to human serum albumin (HSA) and bovine serum fetuin-A (BSF) are shown in most experiments. These forms are not only more readily available in commercial quantities but they are used here to allow us to confirm that these proteins will bind strongly to nascent calcifying nanoparticles regardless of their species of origin. It should be noted that the use of the other two forms of these proteins (bovine serum albumin and human serum fetuin-A) have given comparable results.
Figure 4 (column marked by “None”) shows typical results obtained with inoculating either BSF, HSA, or both into DMEM and observing for turbidity changes over various periods of time (data shown up to 1 month). Surprisingly, no precipitation was seen with any of the three types of inoculation, with proteins added between 0.04 and 4 mg/ml and observed for extended periods of times of up to 2 months (data not shown). These results are representative of a total number of 15 separate experiments done over a course of two years by several independent researchers. Turbidity and precipitation that had been seen in some initial experiments were later traced to bacterial contamination associated with the protein solutions; after filtration through 0.2-µm membranes, proteins no longer produced precipitation in spite of prolonged incubation. To ensure that the results were not due to proteins being filtered out, the same solutions were inoculated without filtration but in the presence of 0.02–0.2% sodium azide added to the medium. No precipitation was seen under these conditions.
10.1371/journal.pone.0008058.g004 Figure 4  Seeding of NB-like particles by fetuin-A, albumin, and serum in metastable versus supersaturated medium.
(A) Fetuin-A and albumin in both metastable and supersaturated media: bovine serum fetuin-A (BSF) and human serum albumin (HSA) were diluted into DMEM, separately or together, to the final concentrations indicated on the top. Where indicated, CaCl2 and NaH2PO4 were added to the final concentrations shown at the bottom (labeled as “Calcium+Phosphate Added”), followed by incubation in cell culture conditions for “1 Week” or “1 Month”, as indicated on the right of each panel. “None” refers to the absence of any exogenously added calcium and phosphate, that is, to metastable medium DMEM. Seeding of the proteins alone without additional ion input did not produce any turbidity change even following incubation for 1 month (“None” column). Low amounts of precipitation were noticed when precipitating ions were added (supersaturated solutions) that increased gradually over time through either bell-shaped (BSF, BSF and HSA) or straight, dose-dependent (HSA) relationships. (B) Serum in supersaturated medium: FBS (top panel) or HS (lower panel) were added to the final concentrations indicated on the top, followed by addition of precipitating reagents to the concentrations shown at the bottom. Addition of precipitating reagents at either 0.1, 0.3, or 0.5 mM into DMEM containing FBS or HS produced turbidities which increased steadily as seen after “1 Week” either through bell-shaped (0.1 mM calcium and phosphate) or straight, dose-dependent (0.3 and 0.5 mM calcium and phosphate) relationships. After further incubation, all treatments showed straight dose-dependent turbidities that continued to increase in a slow and gradual manner (“1 Month”).   In other experiments (Fig. 5), albumin was increased 10-fold up to 40 mg/ml so as to mirror its normal concentration in the serum, at 35–45 mg/ml in HS [92] and 23 mg/ml in FBS [93]. Likewise, BSF was tested up to 20 mg/ml in some experiments (not shown). Precipitation was not seen within this broad spectrum of protein concentrations. Precipitation was also not evident when proteins were tested against other cell culture media (Roswell Park Memorial Institute 1640 or RPMI-1640, F12 medium, medium 199, Glascow minimum essential medium, and Leibovitz L-15 medium, all obtained from Gibco; data not shown).
10.1371/journal.pone.0008058.g005 Figure 5  Effect of high concentrations of fetuin-A and albumin on the formation of NB-like particles in metastable versus supersaturated medium.
Stock solutions of BSF and/or HSA were diluted into DMEM to the concentrations indicated on the top, so as to mimick the concentration range of these proteins seen in HS. When indicated, the precipitating reagents CaCl2 and NaH2PO4 were added to the concentrations shown at the bottom (labeled as “Calcium+Phosphate Added”). A650 turbidity readings were performed following inoculation (“Day 1”) and after incubation in cell culture conditions for “1 Month”, as indicated on the right. Seeding of the proteins alone without additional precipitating reagents did not produce any turbidity increase even following incubation for 1 month (“None” column). Notice the bell-shaped turbidity increases seen with 0.3 mM and 0.5 mM calcium and phosphate columns, observed at “1 Month,” but not “Day 1.”   Since neither albumin nor fetuin-A could seed calcium nanoparticles in metastable solutions, inoculation was carried out in DMEM in the presence of small amounts of calcium and phosphate. Figure 4A shows results for three concentrations of calcium and phosphated added: 0.1 mM, 0.3 mM, and 0.5 mM. In the presence of 0.1 mM calcium and phosphate, a progressive, slow, dose-dependent turbidity increase was seen with subsequent addition of either BSF or HSA. In contrast, 0.1 mM calcium and phosphate alone, in the absence of proteins, did not produce any turbidity changes in the same period of observation, indicating that the turbidity seen with either HSA or BSF was not due to spontaneous or homogeneous mineral deposition associated with the presence of precipitating ions. The inoculation of both BSF and HSA in the same wells resulted in a bell-shaped turbidity increase, with peak turbidity induced by 1.2 mg/ml each of BSF and HSA added, and decreasing thereafter with larger doses of protein (4 mg/ml). It was obvious that the inoculation of both proteins together produced an additive inhibitory effect that reflected the summation of the individual effects associated with either BSF or HSA. In all three instances, however, the increase of turbidity was slow, taking at least two weeks to become noticeable.
In the presence of both 0.3 mM and 0.5 mM calcium and phosphate, precipitation could be discerned earlier and, by one week, noticeable increases in turbidity were measured (Fig. 4, “0.3 mM” and “0.5 mM” columns). On the other hand, adding calcium and phosphate alone at these levels produced no increase in turbidity during the same periods of time (up to 1 month, wells 1 in each column, Fig. 4A). In the presence of added calcium and phosphate (0.3 mM and more noticeably with 0.5 mM), a bell-shaped dose-dependence was clearly seen with BSF, but not with HSA. HSA alone produced a simple, dose-dependent increase in turbidity without any noticeable inhibition seen with higher protein concentrations. In the presence of both BSF and HSA, however, the same bell-shaped dose-dependence could also be discerned, seemingly translating into an additive inhibitory effect produced by the individual BSF and HSA moieties. These results indicate that, in addition to both BSF and HSA displaying seeding capability in the presence of small amounts of calcium and phosphate added, only BSF showed clear inhibitory tendencies in the protein concentrations tested here.
To correlate these findings with those seen with unfractionated whole serum, we repeated the same experiments by inoculating FBS or HS into DMEM containing various amounts of calcium and phosphate. Figure 4B demonstrates again the slow, bell-shaped seeding that is seen with either FBS or HS inoculated into DMEM alone in the absence of any externally added calcium and phosphate (“None” column). In the presence of 0.1 mM calcium and phosphate, however, a much more prominent turbidity increase could be seen with either FBS or HS that was not apparent with calcium and phosphate alone. Again, turbidity appeared to peak at 1% FBS and 0.3% HS after 1 week, and at 10% FBS and 3% HS after 1 month (Fig. 4B, “0.1 mM” column). In the presence of higher amounts of calcium and phosphate (0.3 mM and 0.5 mM), the turbidity increase was enhanced, and by one week, precipitation had nearly reached maximal levels (Fig. 4B). It should be noted that, by one week, peak turbidity was seen at 3% FBS or HS, while 10% FBS or HS gave lower turbidities, indicative of a bell-shaped dose-dependence. However, with longer incubations, up to 1 month, some of this same inhibition seen at higher doses of serum appeared to have been subdued or overcome (Fig. 4B). Together, these results indicate a similar type of seeding-inhibition relationship seen with the whole serum that can be mimicked at least in part by the two purified serum proteins albumin and fetuin-A.
That albumin and fetuin-A should differ in their effects on mineral precipitation (Fig. 4A) may be explained by their differential affinities for calcium and apatite. In fact, compared to albumin, fetuin-A is known to bind much more strongly to calcium. For instance, fetuin-A was described to have 6 calcium-binding sites, with one of them being associated with a binding constant of 0.95×10−4 M in the case of BSF and 1.42×10−4 M for HSF [65]. On the other hand, bovine serum albumin (BSA) also has multiple binding sites, but they are associated with a higher calcium-binding constant of 7×10−3 M [65]. Fetuin-A is also known to have unique apatite-binding sites ([94]; reviewed also by Jahnen-Dechent et al. in ref. [75]). In earlier studies on nanoparticle formation, it had also been shown that, compared to albumin, fetuin-A is a much more potent inhibitor of calcification [2]. This inhibition, however, appears to be transient, being overcome gradually through incubation, a phenomenon that seems to correlate with what has been seen with whole serum, as shown here and elsewhere [2].
In addition to the inherent differences in calcium and apatite-binding affinities between fetuin-A and albumin, it is also known that, compared to fetuin-A, albumin is found in much higher amounts in HS. At 35–45 mg/ml in HS [92], albumin is 50–56x more abundant than fetuin-A (0.7–0.8 mg/ml [95]; see also ref. [2]). In FBS, on the other hand and as noted earlier, albumin is reportedly found at 23 mg/ml [93] while fetuin-A has been measured at 10–21 mg/ml [96]. The differences in the concentrations of these two proteins in HS versus FBS as well as the differences in their respective affinities for calcium and apatite indicate that fetuin-A and albumin may exert markedly different seeding or inhibitory influences insofar as nanoparticle formation is concerned.
To address these inherent differences between fetuin-A and albumin, we repeated the same seeding experiments shown in Figure 4A, however this time using higher concentrations of albumin (up to 40 mg/ml). We reasoned that increasing the amounts of albumin used would mirror more closely the normal concentrations of albumin found in both HS and FBS. Figure 5 shows one such series of experiments done in the presence of 0.3 mM and 0.5 mM calcium and phosphate. Either fetuin-A or albumin at the concentrations tested (BSF up to 1 mg/ml; HSA up to 40 mg/ml) produced similar bell-shaped, dose-dependent seedings indicative of seeding-inhibition seen with the higher doses of albumin used. In this particular experiment, BSF and HSA concentrations above 0.1 mg/ml and 5 mg/ml, respectively, gave progressively lower turbidities. When added together, fetuin-A and albumin produced additive inhibitory effects (Fig. 5). Again, in the absence of externally added calcium and phosphate, neither fetuin-A (BSF, 0.1–1 mg/ml) nor albumin (HSA, 5–40 mg/ml) produced any noticeable precipitation after 1 month. Adding the two proteins together (BSF up to 1 mg/ml, HSA up to 40 mg/ml), without any externally added ions, also failed to produce any precipitation (Fig. 5).
The seeding-inhibitory relationships displayed by both fetuin-A and albumin varied with the protein lots as well as the concentrations tested, which gave a significant margin of error. Nonetheless, the basic seeding and inhibitory influences delineated by these proteins remained unchanged. For instance, Figure 6 illustrates another experiment in which BSF was adjusted between 0.7–70 µg/ml and HSA between 0.04–4 mg/ml. In the presence of 0.3 mM, 0.5 mM, and 0.7 mM calcium and phosphate, these protein ranges gave prominent bell-shaped, dose-dependent increases in turbidity. Turbidity increases could be seen within 4 days of incubation and increased only slightly up to 1 month (Fig. 6). With BSF, at 4 days, turbidity increased with BSF concentrations up to 2.1 µg/ml, decreasing thereafter. Likewise, for HSA, turbidity peaked between 120 to 400 µg/ml of protein concentration and similarly decreased with additional protein input. The protein concentrations seen to produce peak turbidities in this experiment are surely different from those seen elsewhere using different lots of proteins (compare Fig. 6 with 4A). Remarkably, in the experiments depicted in Figure 6, as little as 0.7 µg/ml of BSF was sufficient to induce increased mineral precipitation in the presence of submillimolar amounts of calcium and phosphate (0.3 mM, 0.5 mM, and 0.7 mM; see wells 2 in all 3 columns of Fig. 6).
10.1371/journal.pone.0008058.g006 Figure 6  Seeding of NB-like particles by fetuin-A and albumin at low concentrations in supersaturated ionic solutions.
Stock solutions of BSF and/or HSA were diluted into DMEM to the relatively low concentrations shown on the top. The precipitating reagents CaCl2 and NaH2PO4 were then added at either 0.3, 0.5, or 0.7 mM, as indicated at the bottom (labeled as “Calcium+Phosphate Added”). Turbidity was monitored by A650 turbidity reading following inoculation (“Day 1”) and after incubation in cell culture conditions for the time indicated on the right side of each panel. Bell-shaped curves of precipitation were noticed at “Day 4” for the three different combinations of proteins as well as the three different concentrations of added precipitating reagents. Precipitation in these conditions increased in a time-dependent manner with further incubation. When 0.3 mM of precipitating reagents was used, bell-shaped increases in turbidity were seen for the three protein combinations after 1 month of incubation. With calcium and phosphate added to 0.5 and 0.7 mM, the initial bell-shaped increase in turbidity observed at “Day 4” was seen to shift to the right at “1 Month” reading.   The lack of any noticeable inhibitory effect on mineral precipitation seen with low doses of fetuin-A (0.7–2.1 µg/ml of BSF) further confirms the notion proposed here that, at low concentrations of fetuin-A, this same calcification inhibitor binds avidly to nucleating apatite complexes, but it is eventually overwhelmed by excess calcium and phosphate, at which point it becomes a “nidus” for further calcification. In fact, as seen in Figure 6, the peak turbidity can be seen to shift progressively toward the right following an increase in the calcium and phosphate concentrations added (compare the various columns). Thus, in the presence of 0.3 mM calcium and phosphate, as little as 2.1 µg/ml of BSF produced peak turbidity, whereas with 0.7 mM calcium and phosphate, the peak turbidity was seen with BSF at 21 µg/ml, a result seemingly confirming the workings of an optimal stoichiometric balance between fetuin-A and its mineral-seeding capability as a function of the total amount of precipitating ions present in any given medium. Beyond these peak concentration levels of turbidity, inhibitory influences could be clearly discerned in both these experiments (see, for instance, wells 4–6 in the 0.3 mM calcium and phosphate column and well 6 in the 0.7 mM calcium and phosphate column).
Assuming further a molecular weight of 48 kDa for BSF, it can be calculated that at 0.7 mM of calcium and phosphate ions, the optimal turbidity produced by 21 µg/ml, or 0.44 µM, corresponds to the binding of 58 apatite units by each BSF molecule. The number of apatite crystals bound to each protein molecule was estimated by dividing the number of phosphate ions present in the well producing maximal precipitation by the number of protein molecules present in the same well as described in the  Materials and Methods . Phosphate ions were considered to be the limiting factor in the formation of apatite crystals since they were present in lower amounts than calcium ions. We assumed further that every phosphate ion would be part of a crystal unit of apatite under these conditions and that each crystal unit would contain an average of 63 phosphate ions based on earlier estimations [94]. On the other hand, by using an average number of 105 calcium ions for each apatite unit [94] as the basis of our calculation, an estimate of 54 apatite units bound to each BSF molecule was obtained. Either way, the calculations given here correspond at best to only rough approximations since we did not attempt to define more precisely the dose-dependence relationship for BSF insofar as mineral precipitation is concerned. In general, however, our calculated number supports the earlier estimates made by Heiss et al. [94], who established that each molecule of BSF could bind to 10 apatite units for early soluble calciprotein particles (CPPs) or to around 22 apatite units for their so-called “precipitating late” CPPs.
Similar stoichiometric relationships could be defined for albumin, with peak turbidities seen around 0.4 mg/ml (5.88 µM assuming a molecular weight of 68 kDa for albumin) in the presence of 0.7 mM of calcium and phosphate, which would correspond to a stoichiometric binding ratio of 4 apatite units per molecule of albumin. Compared to fetuin-A then, the binding affinity of albumin for apatite is some 14-fold lower. Compared however to the protein concentrations that produce maximal turbidity/precipitation here in the presence of 0.7 mM calcium and phosphate, both proteins are known to be present in the serum at levels that are at least 100-fold higher. Given their predominance in the serum, both proteins are expected to exert a strong inhibitory influence on mineral growth and precipitation even under conditions when the serum is loaded with excess calcium and phosphate. In fact, our own earlier study [3] had already revealed the need for an unusually high amount of calcium or phosphate added to the serum in order for mineral precipitation to ensue.
The turbidity profiles seen in Figure 6 could be sustained largely unchanged for several weeks of incubation except for small shifts in dose-dependence seen with longer incubation times (Fig. 6, see “1 Month” reading). It should also be noted that that these same experiments showed a stronger inhibitory effect associated with albumin as compared with earlier results (Fig. 4A). Nonetheless, both seeding and inhibitory influences are readily apparent for both proteins from the turbidity data shown here. As before, the presence of both fetuin-A and albumin in the same culture wells appeared to produce additive inhibitory effects.