From Matrix Vesicles to Miniature Rocks: Evolution of Calcium Deposits in Calf Costochondral Junctions

Abstract

Objective

Initial stages of cartilage matrix calcification depend on the activity of matrix vesicles. The purpose of the study was to describe how calcified matrix vesicles join into larger structures, to present their up-to-date undescribed 3-dimensional image, and to observe how calcified matrix relates to chondrocyte lacunae.

Design

Calcified cartilage was obtained from the zone of provisional calcification of calf costochondral junctions, then enzymatically isolated and studied by microtomography, scanning electron microscopy, atomic force microscopy and X-ray diffraction, and Fourier transform infrared spectroscopy.

Results

Hyaluronidase digestion released packets of granules surrounded by the cartilage matrix. Further digestion, with collagenase and trypsin, removed matrix and exposed granules with dimensions within 50 to 150 nm range, which we consider as equivalent of calcified matrix vesicles. Granules joined into larger groups with dimensions of 0.5 to 2 μm, which we call globular units. Certain matrix vesicles appeared well connected but contained globular units that had spaces filled with electron lucent material, presumably matrix or chondrocyte remnants. Globular units were organized into massive structures taking the shape of oval plates. Comparison of these plates with lacunae containing isogenous groups of chondrocytes from proliferative zone of costochondral junction suggests that the cells from a single lacuna were responsible for the formation of one plate. The plates were connected with each other and extended over provisional calcification zone.

Conclusions

The outcome showed how particular calcified matrix vesicles associate into globular units, which organize into massive structures assuming the shape of oval plates and eventually cover large areas of cartilage matrix.

Introduction

Epiphyseal cartilage growth plate serves as the structural template for bone formation and growth. It is divided into a reserve, proliferative, and hypertrophic zone. Chondrocytes in the proliferative and hypertrophic zones are arranged in rows parallel to the long axis of the cartilage and separated by longitudinal septa. The calcifying area of the growth plate is surrounded by bone mantel (bone collar), which arises by intramembranous ossification and protects the growth plate against mechanical damage.^1-3 Calcification of the cartilage matrix begins half-way into the middle of the zone of hypertrophic cells with the appearance of separate calcium deposits in or near calcifying globules of cellular origin, subsequently named matrix vesicles.^4-7 Matrix vesicles contain enzymes that promote calcification.^8-10 Calcium deposits accumulate within longitudinal septa and eventually fuse to large calcified areas in the bottom portion of the hypertrophic zone, a region frequently called the zone of provisional calcification. Then, in the bone metaphysis, calcium deposits serve as the substrate for early bone deposition.^1-3 In earlier works, calcium deposits associated with matrix vesicles were described as hydroxyapatite (HAp) crystals, eventually forming plate-like structures. ² Further studies demonstrated, however, that the calcium deposits represented poorly crystalline, carbonated apatite. ¹¹ Transmission electron microscope observations indicated that the calcified matrix vesicles formed aggregates of electron dense structures, often called calcification nodules. ¹² Calcium deposits observed by scanning electron microscopy (SEM) in cartilage from the rat rib costochondral growth plate (corresponding to epiphyseal plate of long bones) appeared, as structures with dimensions in the range of 0.4 to 1.5 µm, which we named globular units. In some areas, they fused with each other, but usually were separated from the neighbours by a slit measuring about 0.15 µm. ¹³ Thus, globular units seemed to be 3-dimensional structures representing an advanced stage of calcified matrix vesicle organization. Structures similar to globular units, named calcospherites, were previously described by Lester and Ash ¹⁴ in rat mandibular condyle.

Calcified deposits were, as far as we could establish, never isolated as the whole structure and studied by SEM and micro-computed tomography (µCT). Such visualization could show how particular calcified matrix vesicles associate into globular units and subsequently cover large areas of cartilage matrix in the zone of provisional calcification. Thus, in this work calcium deposits from costochondral junctions of calf cartilage were enzymatically isolated and observed by SEM, atomic force microscopy, and focused ion beam technology to observe calcified matrix vesicles’ aggregation into larger structures. The relationship between calcified matrix vesicle aggregates and cartilage matrix was also observed. Moreover, calcium deposits were analyzed by X-ray diffraction and Fourier transform infrared (FTIR) spectroscopy. Finally, calcium deposits were decalcified for histological observations on their organic components.

Materials and Methods

Calcified Cartilage Preparation

Rib cages from calves ranging in age from 4 to 20 weeks were purchased at the local butchery. Cartilage was dissected and secured for further processing within 24 hours after the death of the animal. The costochondral junction (an equivalent of the cartilage growth plate of long bones) was cleared from the adhering tissues and hand broken at the level of metaphysis—the provisional zone of calcification. The whole calcified cartilage layer of the rib was either cut off and fixed in 70% ethanol or calcified cartilage was scraped from the exposed surface with a knife, collected in a dry dish, frozen at −80°C, lyophilized, and pulverized in liquid nitrogen. On average, 400 mg of dried cartilage powder was obtained from one animal.

Hyaluronidase Digestion

A calcified cartilage layer cut off from the whole rib preserved in ethanol and a 400 mg pellet of pulverized cartilage were digested in 10 mL of 0.1% hyaluronidase (hyaluronidase from bovine testes, Sigma-Aldrich) dissolved in phosphate-buffered saline (PBS). ¹⁵ The progress of digestion was controlled by staining with 0.1 aqueous solution of toluidine blue (Sigma-Aldrich). Digestion lasting 4 to 5 hours was sufficient to remove all traces of stainable material.

Collagenase, DNase, Trypsin Digestion

The organic material attached to calcified cartilage after hyaluronidase digestion was removed with 0.25% collagenase and 0.05% DNase solution in PBS (Sigma-Aldrich) at 37°C for 2 hours, followed by 0.25% trypsin in 0.1 mM Tris buffer, pH 8.2 (Sigma-Aldrich) at 37°C, for 1 hour and again by 2 hour exposition to collagenase and DNase. Calcified cartilage was rinsed twice with dH₂O for 15 minutes, and either left in water for µCT or spread on aluminum foil and lyophilized for SEM study. Digestion of the whole calcified cartilage layer of the rib yielded samples with dimension within micrometer to millimeter range, but they were mixed with fragments of bone collar and had to be hand-picked for further studies. Calcified cartilage fragments scrapped from the exposed rib surface were not contaminated by bone and after pulverization were considerably smaller than in the previous group. In some experiments, hyaluronidase digested material was left in a vertical position in 15-mL plastic tubes for 10 minutes to allow larger pieces of calcified cartilage to settle down and the supernatant was withdrawn leaving about 0.5 mL over the sediment. Then, the supernatant was spun down, the sediment was suspended in dH₂O, and spread on the aluminum foil for SEM observations. Samples from supernatant obtained after first collagenase digestion of cartilage that were already treated with hyaluronidase were prepared in the same manner.

Micro-Computed Tomography

Samples of calcified cartilage were scanned using a microfocused X-ray tomographic system (MICRO XCT-400, Xradia, Zeiss), at 40 kV and 200 μA. For each sample, 1500 projection images were recorded with an exposure time of 10 seconds and a magnification objective of 20×. The volume 0.4 mm³ was reconstructed with the instrument software and was then exported to Avizo Fire (FEI Visualization Sciences Group) for further 3-dimensional image analysis. Voxel size was 0.5 × 0.5 × 0.5 μm.

Scanning Electron Microscopy and Focused Ion Beam Cutting

Microstructure examinations of calcified cartilage samples were done using SEM cold emission Hitachi SU-8000 operating under 5 kV and with the working distance of between 3 and 15 mm. The resolution of this microscope at 5 kV was 1.2 nm. The images were taken in secondary electron (SE) mode using upper and lower detectors. A thin Ni-Cu layer with a thickness smaller than 5 nm was deposited before SEM studies, to dissipate charging artifacts and any resultant heat buildup, and minimize beam damage. Several cross-sections images of globular units were taken using a dual beam focused ion microscope (FIB) Hitachi NB-5000. The accelerating FIB beam voltage was 40 keV and a beam current was 0.01 nA. Before cutting, each particle was deposited by ion beam with a local protective tungsten layer.

Atomic Force Microscopy (AFM)

Topography visualization of calcium deposits in calf costochondral junctions was performed using MFP 3D BIO microscope (Asylum Research/Oxford Instruments). Stable and controllable experiment conditions were ensured by closing AFM in a custom-made chamber that precludes temperature fluctuations. The microscope was operating in semi-contact mode. To register high-resolution topographical maps, a silicon AC160TS-R3 (Olympus) scanning probe was used. According to the producer, the probe is characterized with spring constant approximately 26 N/m. The radius of the tip (~9 nm) used for the imaging was estimated with calibration grating provided by the microscope’s producer. Samples were imaged over an area of 5 × 5 μm at 0.3 to 0.4 Hz. IgorPro (version 6.17), a professional, dedicated software provided by the microscope’s producer, was used for data analysis.

Fourier Transform Infrared Analysis

Calcified cartilage samples for infrared (IR) analysis were deposited on the aluminum foil. FTIR spectra were recorded on Nicolet 8700 spectrometer (Thermo Scientific) in attenuated total reflection–FTIR mode (ATR-FTIR) with diamond crystal. For each sample, 256 scans in the range of 400 to 4000 cm⁻¹ and with resolution of 4 cm⁻¹ were collected. For more detailed analysis, the peak resolve, preceded by the Fourier self-deconvolution (FSD), for ν₄PO₄ + HPO₄ band was performed with the OMNIC software. The FSD procedure was done with a bandwidth of 22 cm⁻¹ and sensitivity coefficient (enhancement) of 2. Spectral decomposition was done after linear base line subtraction and a Lorentzian shape of the bands was used.

X-Ray Diffraction

The crystalline phases of calcified cartilage samples were studied by X-ray diffractometry (XRD) using a Rigaku Miniflex II diffractometer and a Cu Kα radiation source. The diffractograms were recorded from 2θ = 3° to 55° with a step size of 0.02° and a scan time of 10 seconds per step. Peak identification for crystalline phases was achieved using the International Centre for Diffraction Data (ICDD).

Histological Staining

Fragments of the whole rib cartilage were fixed in 70% ethanol, dehydrated, embedded in paraffin, sectioned at 6 µm, and stained with hematoxylin-eosin. Fragments of the enzymatically treated calcified cartilage were rinsed in calf serum to inactivate trypsin and decalcified in 10% ethylenediaminetetraacetic acid in 4% formaldehyde to expose cartilage matrix masked by calcium deposits. After rinsing in dH₂O, decalcified fragments of matrix were used for preparation of smears on Silane-Prep slides (Sigma-Aldrich). Smears were dried at 37°C and used for hematoxylin-eosin, glycosaminoglycans, and collagen type II staining. Glycosaminoglycans were stained with 1% Alcian blue 8GX (Sigma-Aldrich) solution in acetic acid, pH 2.5 for 60 minutes. Collagen type II was detected in smears and in sections of formalin-fixed and paraffin-embedded rib cartilage, used as a positive control. Mouse monoclonal anti-collagen II antibody (in a 1:100 dilution) (ABCAM) served as the primary antibody and biotinylated F(ab′)2 fragments of rabbit anti-mouse immunoglobulins (Dako A/S) as the secondary antibody. ExtrAvidin-peroxidase (Sigma-Aldrich) was used for the detection of biotinylated antibodies and the peroxidase reaction was developed using 3,3′-diaminobenzidine (DAB) (Sigma-Aldrich). Slides left unincubated in the primary antibody served as the negative control. Some smears and rib sections were digested with hyaluronidase and prepared as in hyaluronidase digestion for 30 minutes. Demonstration of alkaline phosphatase was undertaken with Vector Blue alkaline phosphatase substrate kit (Vector Laboratories, Inc) according to the manufacturer’s protocol.

Results

Morphological Evaluation

Hyaluronidase digestion released several groups of granules embedded in matrix proteins ( Fig. 1a ). They were collected after larger deposits settled at the bottom of the tube. After removal of the matrix by collagenase, the granules became clearly visible ( Fig. 1b ). The granules had diameters in the range of 50 to 150 nm and usually appeared in small groups. We assume that the single granules correspond in size with calcified matrix vesicles observed by Bonucci ¹² in transmission electron microscopy studies.

Figure 1.

Scanning electron micrographs of calcium deposits released from matrix by enzymatic digestions. (a) Granules released from pulverized cartilage by hyaluronidase. They form small groups embedded in matrix proteins. (b) Matrix removed by collagenase. Granules, presumably equivalent to mineralized matrix vesicles, described in transmission electron microscopy studies, ¹² occurred in small groups.

Calcium deposits released from matrix by enzymatic digestions displayed a variety of forms, from single mineralized matrix vesicles to large conglomerates, globular units ( Fig. 2a-c ) or still larger massive structures ( Fig. 2d ). The fully formed globular unit is composed of numerous matrix vesicles and had a diameter in the 0.5 to 2 µm range. It seems to be formed by association of mineralized matrix vesicles into pairs, triplets, larger groups of particles and finally, into globular units. Electron lucent spaces remained inside globular units ( Fig. 3 ). Atomic force microscopy visualization of globular units ( Fig. 2e and f ) also supported the conclusion that they arose by fusion of smaller elements corresponding in size to matrix vesicles.

Figure 2.

Scanning electron micrographs showing examples of formation and arrangement of globular units: (a) General appearance of pulverized, enzymatically dispersed calcified cartilage from calf rib costochondral junctions. Calcified particles were suspended in dH₂O, dropped on the aluminum foil, frozen, and lyophilized. Numerous calcified particles of varied shape were present. (b) Matrix vesicles occurring singly, in pairs, triplets, or larger groups. (c) Globular unit with attached matrix vesicles. (d) Globular units organized into massive structure assuming the shape of a plate. Numerous units had distinct indentation, usually in the central part. (e, f) Globular units visualized with atomic force microscope (AFM). Picture suggests that globular units arose by fusion of numerous small elements with dimensions less than 0.1 µm, presumably representing mineralized matrix vesicles.

Figure 3.

Demonstration of electron lucent spaces within globular units. (a) Intact. (b) Part of the unit was cut off by focused ion beam (FIB). The whole structure looks as the miniature rock with several caves. (c) Conglomerate of globular units—broken line indicates the place of FIB sectioning. (d) Large cave in globular unit (scanning electron micrograph).

Globular units, associated into massive agglomerates ( Fig. 4 ), were sometimes connected by bands of fibrous calcified material, which we named “suspension bridges.” They can be over 1 µm long, had diameter about 100 nm, smooth outlines, and homogenous appearance. Such fibers could be entrapped within globular units as suggested also by Figure 2c .

Figure 4.

Scanning electron micrographs of architecture of clusters formed by globular units. (a) Groups of clusters of globular units. (b-d) Surface architecture of clusters. Mountain landscapes with suspension bridges over valleys separating individual globular units. Suspension bridges were over 1 µm long, had diameter about 100 nm, smooth outlines, and homogenous appearance. Their ends were connected to globular units, but matrix vesicles were not attached to their middle parts.

Massive structures formed by globular units had an impressive appearance, but their real position was visible only in preparations made from nonpulverized cartilage. As shown in Figure 5 , globular units were associated into plate-like structures ( Fig. 5a ) with long extensions. The shape of a single plate corresponded to that of a single lacuna housing isogenous chondrocytes ( Fig. 5b ). The position of plates shown in Figure 5a presumably corresponds to the zone of provisional calcification with calcified longitudinal septa marked in Figure 5c in a fragment of the whole rib. In some places, 2 or 3 septa were connected by calcium deposits overshadowing chondrocyte lacunae located in between.

Figure 5.

Localization of massive structures formed by globular units. (a) Enzymatically isolated fragment of intact (nonpulverized) calcified cartilage presumably corresponded to the area marked in (c). Calcium deposits assumed form of oval plates with long extensions. Their surface was covered by regularly distributed globular units. (b) Large chondrocyte lacunae from the proliferative zone with isogenous group of chondrocytes. Chondrocytes are contracted due to fixation in 70% ethanol. Lacunae, arranged in rows, are separated by longitudinal septa, which calcified in the zone of provisional calcification. Borderlines of 2 lacunae (defined as the territory occupied by isogenous group of chondrocytes) are marked by dashed lines. The shape of lacunae corresponds to that of the oval plates in (a). (c) Three-dimensional visualization of the central part of calcification zone from the calf rib costochondral junction. Visible single calcified septa and calcified areas extending over two or three lacunae. (a, c) Micro-computed tomography. (b) Hematoxylin-eosin.

Physicochemical Evaluation

Spectra obtained from FTIR analysis were typical for crystalline forms of calcium phosphate ceramics ( Fig. 6a ). Sharp peaks with divided maxima, which do not occur in the case of amorphous calcium phosphates, were present. ¹⁶ The main bands characteristic for ν₃PO₄ and ν₄PO₄ + HPO₄ were in the range of 900 to 1200 cm⁻¹ and 500 to 630 cm⁻¹, respectively. In the ν₄PO₄ + HPO₄ domain were the 2 maxima placed at 558 and 599 cm⁻¹. The existence of these 2 maxima/subpeaks prove the crystalline form of analyzed calcium phosphates and can be used for determination of their percentage of crystallinity. ¹⁷ The results of the FSD procedure revealed 3 subpeaks located at 558, 575, and 599 cm⁻¹, which suggested the presence of apatitic forms of Ca-PO₄ compounds.^16,18 Deeper analysis of the ν4PO₄ + HPO₄ domain ( Table 1 ) have shown that the area of apatitic bands constituted about 80% of whole domain. In the spectrum, the small peak at 3570 cm⁻¹ was detected, which is characteristic for –OH stretching in HAp.

Figure 6.

Crystal and chemical structure of calcified deposits from costochondral junction cleared of adhering organic matter by enzymatic digestion. (a) Fourier transform infrared (FTIR) spectra, with indicated characteristic bands OH, PO₄, and HPO₄. (b) X-ray diffractogram: broad peaks appeared only within the hydroxyapatite region.

Table 1.

Results of Peak Separation for ν4PO₄ + HPO₄ Band ^a .

	Wave number (cm−1)	Area (%)
PO band	513	6.0
n.apat. HPO42−	528	12.8
apat. HPO42−	542	18.9
apat. PO43−	558	37.7
apat. PO43−	575	7.8
apat. PO43−	598	15.0
n.apat. PO43−	609	1.7

apat, = apatite; n.apat. = non-apatite.

a

Deeper analysis of the ν4PO ₄ + HPO ₄ domain indicates that the area of apatitic bands constituted about 80% of the whole domain.

The XRD patterns obtained for calcified deposits are shown in Figure 6b . The peaks were indexed based on a hexagonal HAp crystal of space group P63/m (a = b = 9.418 Å; c = 6.884 Å) with reference to the standard Powder Diffraction File (card no. 09-0432). The broadening of the XRD pattern was typical for the HAp nanocrystalline structure (due to the very small crystal size). No secondary phases were detected besides hydroxyapatite.

Histological Evaluation

Material left after decalcification of the pulverized, enzymatically purified calcium deposits was acellular and contained glycosaminoglycans and collagen type II. Clear demonstration of collagen was possible only after the removal of glycosaminoglycans by hyaluronidase ( Fig. 7a-c ). The activity of alkaline phosphatase was not detected.

Figure 7.

Smears of organic material left after ethylenediaminetetraacetic acid (EDTA) decalcification of calcium deposits. (a) Hematoxylin-eosin, (b) Alcian blue, (c) immunochemical demonstration of collagen type II. Black dots represent calcium salts left after EDTA treatment.

Discussion

Calcium deposits could be easily released from the adhering matrix by enzymatic treatment. This new approach enabled SEM demonstration of calcium accumulations occurring as single granules, doublets, triplets, and larger groups of granules, which eventually coalesced into globular units (Figs. 1-4). Globular units, in turn, joined into massive structures extending over the zone of provisional calcification. Comparison of our SEM observations with those reported in TEM studies^5,6,8,19,20 indicated that the dimensions of granules and matrix vesicles were within the same range, thus their identity appears unquestionable.

The comparison of oval plates ( Fig. 5 ) from SEM visualizations with light microscopy appearance of cartilage lacunae suggests that the plate was produced by an isogenous group of chondrocytes from one lacuna. The whole packet of calcified granules (mineralized matrix vesicles) released by hyaluronidase was probably produced by one chondrocyte. Taking into consideration that a single lacuna houses no more than 20 chondrocytes and an oval plate is composed of numerous globular units, it seems reasonable to assume that one chondrocyte is responsible for the formation of numerous packets. Moreover, packets of matrix vesicles cannot be firmly attached to the matrix collagen fibrils since they are easily released by hyaluronidase. Chemical analysis of the packet capsule formed by matrix components could probably elucidate this point. Formation of packets must proceed rapidly, since, as estimated in the 4-week-old Yucatan swine, the growth plate chondrocytes spend an average only 4 days in the proliferative zone and 48 hours in the hypertrophic zone. ²¹ Biochemical analysis of hyaluronidase released packets could be obtained in a sufficient amount for analysis

Schenk et al. ²² observed that about two-thirds of the longitudinal septa in the tibial epiphyseal plate were partially or completely mineralized, while the remainder were essentially nonmineralized. Similarly, noncalcified matrix in close apposition to globular units was observed within costochondral junction of rat ribs. ¹³ Thus, it appears that the nonmineralized matrix remained in regions too remote from chondrocytes to deposit packets of matrix vesicles.

While in the earlier work, the character of calcium deposits in matrix vesicles was a matter of controversy,^18,23 more recent studies shown that matrix vesicles isolated from chondrocyte formed hydroxyapatite-like crystals, demonstrated by XRD ²⁰ and by FTIR with peaks at 960, 1025, 1102, 1065, and 1157 cm^−1,24 reviewed in Kirsch. ²⁵ Our material consisted of whole calcium deposits rather than isolated matrix vesicles and we observed similar patterns in FTIR and XRD spectra.

According to Kim et al., ¹¹ apatite crystals isolated from epiphyseal plates assumed the form of wide, very thin rectangular plates gathered in small clusters. They had an average length of 103 ± 26 and average width of 68 ± 16 nm. Electron microscope observations of Bonucci and Gomez ²⁶ indicated presence of crystals of a needle- or filament-like shape with the length of 40 to 160 nm and thickness from 2 to 5 nm. Therefore, each mature matrix vesicle contains considerable a number of crystals kept together by some components of cartilage matrix. Thus, matrix vesicles may be considered organic-inorganic hybrids, each consisting of an organic filament with attached calcium and phosphate ions.^12,26 Recently, the specific role in matrix vesicles’ mineralization was ascribed to annexins, proteins present in the nucleational core of native matrix vesicles.^9,25,27

The question arises how singular matrix vesicles are joined into globular units and how the latter form massive structures. All these structures resisted mechanical treatment during cartilage milling and enzymatic digestion removing the noncalcified matrix, thus connection between the particular calcified elements must be strong. Conceivably, their aggregation resulted from the activity of the tissue-nonspecific isozyme of alkaline phosphatase (TNAP). The role of this enzyme in cartilage and bone calcification is well established.^9,28-30 Phosphatases catalyze the hydrolysis of phosphomonoesters with the release of inorganic phosphate. ³¹ In mineralizing tissues, TNAP may hydrolyze inorganic pyrophosphate or pyridoxal phosphate to generate inorganic phosphate promoting mineralization.^30,32 TNAP is linked to the outside of the plasma membrane via glycosylphosphatidylinositol ³² and has the crown domain with the collagen binding loop. ³³ Thus, TNAP still bound to cell membrane of chondrocytes or released from the membrane and attached to collagen could provide phosphate groups needed to bridge the particular matrix vesicles into globular units and larger structures present in the zone of provisional calcification. We have not been able to detect TNAP by histochemical staining, but, as quoted above, its activity in mineralizing tissues is well established and, in our material, could be lost or inactivated during preparatory procedures.

Suspension bridges, shown in Figure 4 , do not appear to be formed by fusion of matrix vesicles since it is difficult to imagine the string of several matrix vesicles arranged in line and without visible contours, prominent in the neighboring globular units. Thus, they must arise by deposition of calcium salts within collagen fibers. Collagen type II dominates in cartilage matrix and seems unsuitable as the core of suspension bridges, since it forms fibril networks and not particular fibers. ³⁴ Poole et al ³⁵ found, however, that at the surface of isolated chondrons, some collagen fibers with suitable diameter and arrangement serve as the suspension bridge core. Comparing figure 3B in their article with Figure 4c and d in this work makes such a suggestion plausible. On the other hand, atomic force visualization of globular units seems to exclude collagen fibers as the essential element for their formation.

To summarize, while the beginning of cartilage calcification due to the activity of matrix vesicles is well documented, the further steps leading to the formation of massive plate-like structures in the zone of provisional calcification is less known. Enzymatic digestion of calcified cartilage may begin to help to clarify the unknowns. This method releases single calcified granules, allows the observation of their associations into larger structures, and may form the basis for studies on the molecular mechanism responsible for their formation.