Osteocytes are derived from a select group of osteoblasts that have undergone a final differentiation. Due to their inaccessibility when embedded in the bone matrix, very little is known about the osteocyte cytoskeleton. This study provides an extensive analysis of the osteocyte cytoskeleton, based on the successful isolation of osteocytes from 16-day embryonic chick calvariae. We used OB7.3, a chicken osteocyte-specific monoclonal antibody, to confirm the osteocytic phenotype of the isolated cells and established culture conditions to promote growth of cells that most resemble osteocytes in vivo. Immunofluorescence staining with antitubulin, antivimentin, and antiactin showed the relative distribution of the microtubules, intermediate filaments, and actin filaments in both osteocyte cell body and processes. Field emission scanning electron microscopy revealed the three-dimensional relationships of the cytoskeletal elements and a unique organization of actin bundles that spanned the cell body and osteocyte processes. When combined with drug studies, these experiments demonstrate that actin filaments are crucial for the maintenance of osteocyte shape. Furthermore, we identified two actin-bundling proteins, alpha-actinin and fimbrin, in osteocyte processes. The prominence and unique distribution of fimbrin in osteocyte processes provides the possibility of its use as an intracellular marker to distinguish osteocytes from osteoblasts.
OSTEOCYTES ARE THE MOST numerous of bone cells, derived from a select group of osteoblasts that have undergone a final differentiation and are left behind, encased within the mineralized bone matrix.1,2 During this process, the prospective osteocytes undergo progressive changes in morphology such that the cuboidal shape of the osteoblast gives way to the stellate shape of the mature osteocyte, with its multiple, slender cytoplasmic processes.3–5
Recently, it has been shown that osteocytes are mechanosensory cells.6–12 With the discovery of gap junctions in the cytoplasmic processes of osteocytes,4,5,13 these cytoplasmic projections have now been recognized to comprise a complex intercellular communication network that links osteocytes to each other and to osteoblasts and lining cells at the bone surface.4,5,14 In addition, these connections are established early during osteoid (premineralization) matrix deposition and maintained throughout subsequent changes in cellular and matrix relationships.4,5 This persistence of intercellular connections has important implications for the architecture of bone. For reasons not known, the area around osteocytes is not mineralized; thus, maintenance of cellular connections allows for the formation of an open canalicular network among osteocytes through which bone tissue is perfused.1,4,5 Therefore, in addition to the osteocyte being the cytoplasmic conduit for the transmission of signals, osteocyte shape determines the physical outlines for patent, nonmineralized channels within the bone matrix.
Although the precise sequence of events by which the plump osteoblast transforms into the spider-like osteocyte is still unknown, it is predicted that changes in cytoskeletal organization must accompany this transition. Changes in cell shape are the result of regulated changes in the assembly and disassembly kinetics of the cytoskeleton, which is composed of three major classes of filament types: microfilaments, microtubules, and intermediate filaments.15–20 All three filament components form a highly integrated structural network which has been implicated directly or indirectly in different aspects of cell shape regulation in a variety of cell types.20–22 Many studies also suggest that the final form that a cell displays may well develop through successive stabilization of increasingly specialized cytoskeletal structures.23,24
The molecular mechanisms by which individual cells sense mechanical signals or how they transduce them into a chemical response are not well understood. However, it is generally agreed that the cytoskeleton is somehow involved.15–18,25,26 That the cytoskeleton is indeed sensitive to mechanical stimuli is demonstrated by the observed cytoskeletal rearrangements in cells subjected to mechanical strain or fluid shear stress.27–29
Therefore, understanding how the cytoskeleton is organized—identifying individual components responsible for the attainment and maintenance of osteocyte morphology and hence, cellular connections—will provide critical information as to how the cytoskeleton may play a role in osteocyte differentiation, the architecture of the osteocyte canaliculi network, and the mechanosensory capabilities of osteocytes. Very little, however, is known about the osteocyte cytoskeleton because morphological analysis of osteocytes has traditionally been difficult because of their embedment in the mineralized matrix. To date, all previous studies have been based on conventional transmission electron microscopy of demineralized bone tissue4,5,30–33 where information was confined to the thin sections (60–100 nm) utilized for this purpose. As a result, comprehensive, three-dimensional (3D) relationships between structures were either difficult to interpret or lost.
This study is based on the successful isolation and maintenance of osteocytes in culture which allowed us to conduct the first extensive analysis of the osteocyte cytoskeleton. Without the problematic autofluorescence intrinsic to the bone matrix, we used immunofluorescence microscopy to examine the relative distribution of individual cytoskeletal components in the osteocyte. This, combined with drug studies, allowed us to demonstrate the crucial role of actin filaments in osteocyte morphology. Complementation of our light microscopic analyses with high-resolution field emission scanning electron microscopy (FESEM) revealed the 3D organization of the osteocyte cytoskeleton and novel features of the actin filament distribution. Furthermore, we identified two actin-bundling proteins (alpha-actinin and fimbrin) in osteocyte processes. The prominence of the latter and its unique distribution in osteocytes provides the exciting possibility of its use as an intracellular marker to distinguish osteoblasts from osteocytes.
MATERIALS AND METHODS
Isolation of chick osteocytes:
Fertile White Leghorn eggs were obtained from Conway's Inc. (Indianapolis, IN, U.S.A.). Osteocytes were isolated from 16-day-old embryonic chicken calvariae, with modifications of the technique as described in Tanaka et al.34,35 Briefly, after stripping off the periosteum, calvariae were trimmed to remove the noncalcified and marrow-rich areas, as previously determined by histologic analysis.36 The remaining calvariae were dissected into small pieces and treated with 1 mg/ml collagenase type I (Sigma Chemical Co., St. Louis, MO, U.S.A.) in a bone cell isolation buffer (25 mM HEPES, pH 7.4, 10 mM NaHCO3, 70 mM NaCl, 60 mM sorbitol, 3 mM K2HPO4, 1 mM CaCl2, 30 mM KCl, 1 mg/ml bovine serum albumin [BSA], 5 mg/ml glucose, 300 nM μM α-tosyl-L-Lysyl chloromethane)37 for 30 minutes, followed by 5 mM EDTA in phosphate-buffered saline (PBS) containing 0.1% BSA (Sigma) for 30 minutes, and then again with collagenase for 40 minutes at 37°C. Cells released after the final collagenase digestion were passed through a 8.0 μm Nucleopore polycarbonate filter (Costar, Cambridge, MA, U.S.A.), collected, and preplated for 1 h in α-Minimum Essential Medium (α-MEM) supplemented with 10% fetal bovine serum (both from GIBCO BRL, Grand Island, NY, U.S.A.). During this period, most of the fibroblasts attached to the culture dish (Becton Dickinson and Company, Lincoln Park, NJ, U.S.A.), while the less adherent cells (osteocytes) were collected and seeded for the final culture. After 1 h, cells were washed with PBS and incubated 15–18 h in fresh medium with 1% fetal bovine serum.
Preparation of coated coverslips:
Rat tail type I collagen in 0.02 N acetic acid was used at 50 μg/ml. Both human fibronectin and poly-D-lysine were dissolved with double distilled water and then diluted with α-MEM to a final concentration of 20 μg/ml. All were purchased from Becton Dickinson Labware (Bedford, MA, U.S.A.). For single coatings, each substrate was applied to #1 glass coverslips (Corning Inc., Corning, NY, U.S.A.) and incubated for 1 h at 37°C, washed several times with double distilled water, and then air dried. For dual coating with poly-D-lysine and fibronectin, the same processes were repeated with several rinses in between.
The primary antibodies used included the monoclonal antibody (MAb) OB7.3, which specifically recognizes chicken osteocytes38 and was kindly provided by Dr. P.J. Nijweide, University of Leiden, The Netherlands. Pan-antifimbrin, a rabbit polyclonal antibody against chicken intestinal fimbrin (R163.3),39 was a gift from Dr. P. Matsudaira, Whitehead Institute, MIT, Cambridge, MA, U.S.A. Commercially purchased primary antibodies included anti-actin mouse immunoglobulin G (IgG) and anti-alpha-actinin mouse IgM (both from Sigma), anti-beta-tubulin mouse IgG (Amersham Corp., Arlington Heights, IL, U.S.A.), and antivimentin AMF-17b mouse IgG supernatant (Hybridoma Bank, Baltimore, MD, U.S.A.).
All secondary antibodies were commercially purchased and included anti-mouse IgG-fluorescein (FITC) conjugate and anti-rabbit IgG-FITC conjugate (both from Cappel Research Products, Durham, NC, U.S.A.), anti-mouse IgG-Texas-Red conjugate (Vector Laboratories, Burlingame, CA, U.S.A.), and anti-mouse IgM-FITC conjugate (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD, U.S.A.).
After 15–18 h in culture, isolated osteocytes were rinsed with PHEM (60 mM Piperazine-N, N′-bis[2-ethanesulfonic acid], 25 mM N-[2-Hydroxyethyl] piperazine-N′-[2-ethanesulfonic acid], 10 mM Ethylene glycol-bis [2-aminoethyl ether]-N,N,N′N′-tetraacetic acid, 2 mM magnesium chloride, pH 6.9) maintained at 37°C.40 Cells were then permeabilized with 0.15% Triton X-100 in PHEM for 90 s, followed by fixation for 10 minutes in 2% glutaraldehyde in PHEM at 37°C. To reduce free aldehyde groups and minimize autofluorescence, coverslips were incubated for 10 minutes with 1 mg/ml sodium borohydride (Sigma). To prevent nonspecific interactions, coverslips were immersed in a blocking solution of 1% BSA in PBS. For single immunolabeling experiments, cells were incubated with the primary antibody for 1 h at 37°C, followed by several rinses with PBS. After reaction with the appropriate second antibody for 30 minutes at 37°C, cells were again rinsed and then mounted with Aqua-poly mount (Polysciences, Inc., Warrington, PA, U.S.A.) containing 1 mg/ml p-phenylenediamine dihydrochloride (Sigma) to retard photobleaching.
Deviations from the above protocol are to be noted in three instances. The first example is immunolabeling with OB7.3 MAb. In contrast to other antibodies, OB7.3 is reacted with the sample prior to fixation. For example, isolated osteocytes on glass coverslips were first incubated for 30 minutes with OB7.3 MAb (diluted to 1:10 with α-MEM), rinsed with PBS, and fixed for 10 minutes in 3% paraformaldehyde in PHEM at 37°C. For in situ immunolabeling of osteocytes in calvariae fragments, the calvariae had to be stripped of their periosteum before incubating with primary antibody.
The second instance is double immunolabeling with anti-alpha-actinin and antifimbrin. Fixation was with 3% paraformaldehyde in PHEM for 10 minutes at 37°C. This protocol yielded the best results in retaining antigenicity of alpha-actinin for the antibody.
The third example is immunolabeling with antivimentin. Cell lysis with 0.15% Triton X-100 in PHEM was reduced to 30 s, and the best results for intermediate filament staining of osteocytes were obtained by fixing with Ethylene glycol bis-[succinic acid N-hydroxy succinimide ester] in dimethylsulfoxide (DMSO) for 15 minutes at room temperature.41
After rinsing the cells with PHEM, they were lysed in 0.5% Triton X-100 and then stained with a 1:40 dilution of FITC-Phalloidin (Sigma) in PHEM for 10 minutes (at room temperature). After rinsing with PBS, cells were viewed immediately.
Alkaline phosphatase staining:
After fluorescence micrographs of OB7.3-labeled cells were acquired, coverslips were rinsed and samples were incubated for 20 minutes with a mixture of 50 μg/ml Napthol ASMX (Sigma), 0.5% N,N-dimethylformamide, and 0.6 mg/ml of fast red violet LB salt in 0.1 M Tris-HCl, pH 8.5, at room temperature. Alkaline phosphatase (ALP)-positive cells were identified with a pink coloration throughout the cytoplasm and were easily distinguished from nonreactive cells. To obtain our data for Fig. 1, the same fields were analyzed for both OB7.3 and ALP staining.
Phase-contrast and fluorescence images:
Images of isolated osteocytes were obtained with an inverted Nikon Diaphot, using a ×100 (NA 1.4) objective (Nikon, Tokyo, Japan). Images were recorded with a 1024 × 1024 Photometrics CCD camera, KAF 1300 chip (Photometric Co., Tucson, AZ, U.S.A.) controlled by BDS 1.0 software (Biological Detection Systems, Inc., Pittsburgh, PA, U.S.A.).
Osteocytes on bone were visualized with a BioRad confocal scanning laser microscope (Bio-Rad Microscience, Hercules, CA, U.S.A.) controlled by Lasersharp software (Universal Imaging, West Chester, PA, U.S.A.). Images were then digitally processed using IP-Lab Spectrum 3.1 (Signal Analytics Corp., Vienna, VA, U.S.A.) and Adobe Photoshop 2.5.1 (Adobe Systems Inc., Mountain View, CA, U.S.A.). Negatives were made from the digital images using an Agfa film recorder controlled by Conductor software on Technical Pan Film (Kodak, Rochester, NY, U.S.A.).
For analysis of osteocyte morphology under different substrate conditions, cultures were first reacted with OB7.3 MAb, and fluorescence images were taken of OB7.3-positive cells. In addition to ensuring that we included only osteocytes in our analysis, the fluorescence images provided a more accurate analysis of osteocyte morphology because the osteocyte processes are very thin and often not easily discerned by phase microscopy. For each experiment, more than 30 cells in each substrate were analyzed for the number of cell processes per cell body, the number of branches per process, and the length of the longest process. A total of four experiments were performed. Values are expressed as means (double measurements in three or more individual experiments) ±SD. p values were calculated by t-tests, and differences were considered significant when p < 0.05.
Treatment of osteocytes with nocodazole, latrunculin B, and cytochalasin D
Nocodazole (used at 5 μg/ml), latrunculin B (0.2 μg/ml), and cytochalasin D (2 μg/ml) were purchased from LC Laboratories (Woburn, MA, U.S.A.). Isolated osteocytes were seeded on glass coverslips photoetched with a locator grid pattern (Bellco Glass, Inc. Vineland, NJ, U.S.A.). The location of cells was recorded using BDS software and a computer-controlled stage. Images before and after drug treatment were taken every 20 minutes up to 1.5 h. After the experiment, cells were lysed, fixed, and processed for immunocytochemistry as described above.
Whole cell preparation:
Isolated osteocytes were plated on #2 glass coverslips (Fisher Scientific Co., Pittsburgh, PA, U.S.A.) that had been coated with poly-D-lysine and fibronectin. After 15 h in culture, cells were treated with latrunculin B (dissolved in 0.01% DMSO) or just DMSO. After 40 minutes, cells were fixed with 2% glutaraldehyde in 0.1 M HEPES buffer (pH 7.4) for 10 minutes at room temperature. To preserve the membrane surface, they were postfixed in McDonald's osmium-potassium ferricyanide mixture: 0.5% OsO4 plus 0.8% K3Fe (CN) for 15 minutes, washed in buffer, and immersed in 0.15% tannic acid for 1–2 minutes.42 After rinsing in water, the preparations were dehydrated in ethanol and critical point dried.43 They were then Argon ion sputter coated with a thin layer of platinum and imaged with the high-resolution in-lens FESEM (S-900; Hitachi, Tokyo, Japan) at the Madison Integrated Microscopy Resource at 1.5 kV accelerating voltage.
Osteocyte cytoskeletal preparations:
To visualize the underlying cytoskeleton of osteocyte before and after the drug treatment, cells plated on coverslips as described above were treated with 0.15% Triton X-100 in PHEM for 90 s, to remove cell membranes, and fixed for 10 minutes with 2% glutaraldehyde in PHEM at 37°C. After washing in buffer for 30 minutes, they were postfixed with 0.1% OsO4 for 10 minutes and stained with 1% uranyl acetate. The cells were then dehydrated with ethanol and critical point dried.43 After argon ion sputter coating with a thin layer of platinum, the cells were imaged with the Hitachi S-900 FESEM.
An important advance in the development of methods for identification and isolation of osteocytes was generation of osteocyte-specific antibodies. One such antibody, the OB7.3 generated in Dr. P. Nijweide's laboratory, specifically recognizes antigenic sites of yet unidentified molecules associated with the extracellular side of osteocyte cell membranes. The specificity of this antibody for chicken osteocytes was demonstrated in vivo and in vitro.38,44 OB7.3-positive reactivity is observed in cells with the stellate shape characteristic of osteocytes, with little or no ALP activity, a marker often used for the osteoblast phenotype.44,45
For this study, we were fortunate to have a generous gift of the OB7.3 antibody from Dr. Nijweide for our use. In contrast to the isolation protocol described by Van Der Plas and Nijweide,44 where OB7.3-coated magnetic beads were used in a “panning method” to isolate osteocytes, we omitted this step because the binding of the antibody-coated beads to the cells can be difficult to remove, rendering the population less amenable to morphological analyses.
An important aspect of our modified protocol for isolation of osteocytes is the selection of calvarial areas that are primarily calcified and relatively free of marrow development.34–36 By day 16, embryonic chicken calvariae were sufficiently calcified, and by dissecting away noncalcified regions along the suture side, as well as marrow-rich paraorbital areas of the parietal bone, we are left with a select area that is osteocyte rich. Subsequent collagenase and EDTA digestions (see Materials and Methods) yielded a population of cells that contain about 90% osteocytes, as determined by subsequent immunocytochemistry with OB7.3.
Isolated osteocytes were distinguished by their stellate shape. Each osteocyte has a small cell body, occupied primarily by the nucleus, and multiple slender processes (Fig. 1A). Variations in cell body size do occur, ranging roughly from 5–20 μm in diameter, which could be a consequence of osteocytes caught at different stages of maturation. Using OB7.3 MAb, up to 90% of the cells in this culture reacted positively, while larger cells with morphologies similar to fibroblasts did not cross-react with this antibody (arrowheads in Fig. 1B).
The OB7.3-positive osteocytes were negative to ALP staining. However, OB7.3 negative cells in the population reacted positively for ALP activity (arrowhead in Fig. 1C).
Plated osteocytes show morphology that closely resembles osteocytes in vivo
Figure 2 shows stereopair fluorescence images taken of calvarial fragments treated with OB7.3. These images of osteocytes in vivo show that the osteocyte cell body is small, about 5–20 μm in diameter. Numerous slender processes leave the cell body, with lengths that approximate one and a half to two times that of the cell body diameter. Often these processes are branched, with connections to neighboring cells, within the same plane as well as in adjacent planes.
To find culture conditions that would facilitate growth that best mimics osteocyte morphology in vivo, we prepared coverslips coated either with nothing, type I collagen, fibronectin, poly-D-lysine, or poly-D-lysine and fibronectin. Two hours after seeding, differences in cell spreading could be observed with the various substrates. Most cells remained rounded on coverslips with nothing or type I collagen, whereas many were fully spread on coverslips with fibronectin, poly-D-lysine, and poly-D-lysine with fibronectin (data not shown). After 15 h, those nonspreading cells in nothing and type I collagen eventually died, suggesting that cell attachment to the underlying substrate is crucial for osteocyte viability.
To analyze the influence of different substrates on osteocyte morphology, cultures were reacted with OB7.3 MAb to first identify the osteocytes. The OB7.3-positive cells were then analyzed for the number of cell processes per cell body, the number of branches per process, and the length of the longest process. Figure 3A shows the representative morphology of cells on each substrate. Although the number of processes per cell did not vary significantly between culture conditions (Fig. 3B), osteocyte processes in cultures of poly-D-lysine and poly-D-lysine with fibronectin showed more branching (Fig. 3C). In addition, when the longest process of each cell was measured, cells plated on poly-D-lysine with fibronectin generally show processes about twice as long as the osteocyte cell body (Fig. 3D). Overall, osteocyte process growth was promoted by a combination of poly-D-lysine and fibronectin and, we concluded that cells of this culture seemed the closest in morphology to osteocytes in vivo.
Actin, microtubules, and intermediate filaments
Anti-actin immunofluorescence showed actin to be a prominent component both in the cell body and cell processes of osteocyte (Fig. 4A). A strong perinuclear concentration of actin filaments was often evident, with some filamentous bundles continuing into cell processes. Comparison with corresponding phase images (data not shown) showed these actin bundles to be present throughout the entire lengths of osteocyte processes. When stained with FITC-phalloidin, which binds specifically to actin filaments, a more intense pattern of actin distribution was often observed (Fig. 4B). In cell processes, strong actin signals were observed throughout their entire lengths as well as in thin branches. In the cell body, small bundles of actin filaments can be observed to crisscross the cytoplasm, often resulting in star-shaped patterns (arrowheads).
To localize microtubules, cells were analyzed with anti-beta-tubulin immunofluorescence. Microtubules were observed to radiate from a perinuclear focus to fill the entire cell body (Fig. 4D). However, microtubules extended only into the proximal one third or one half of cell processes, as can be seen by comparing fluorescence and corresponding phase images (Figs. 4C and 4D). Anti-vimentin staining showed the presence of intermediate filaments in the cell body, extending partially into the proximal region of primary osteocyte processes but not the branches (compare Figs. 4E and 4F).
Actin bundling proteins, alpha-actinin, and fimbrin are abundant in osteocyte processes
To characterize the prominent actin bundles in the osteocyte processes, we examined the presence of two actin-bundling proteins, alpha-actinin and fimbrin (Figs. 5A and 5D are phase images). Double-label immunofluorescence using an anti-alpha-actinin (Fig. 5B) and anti-pan-fimbrin (Fig. 5C) showed that although both proteins were abundant in the osteocyte processes, alpha-actinin appeared in short, linear fragments and dots, whereas fimbrin was localized throughout the processes. Comparison of the distribution of these two proteins in osteocyte and stromal cells revealed some interesting differences (Figs. 5D, 5E, 5F). In an adjacent stromal cell, anti-alpha-actinin showed a linear and periodic distribution, consistent with their location in stress fibers (Fig. 5E). In contrast, fimbrin was mostly diffusely distributed in the cytoplasm (Fig. 5F).
Actin filaments are important for integrity of osteocyte processes
To determine the relative importance of these cytoskeletal elements for osteocyte morphology, cells were treated with 5 μg/ml nocodazole to depolymerize microtubules and 0.2 μg/ml latrunculin B or 2 μg/ml cytochalasin D to depolymerize actin filaments. Comparison of cell morphology before (Fig. 6A) and after (Fig. 6B) treatment with nocodazole showed no obvious changes. Confirmation that microtubules were indeed depolymerized in these cells was obtained by anti-beta-tubulin immunofluorescence, which showed only diffuse staining (Fig. 6C).
Treatment with actin depolymerizing drugs showed significant changes. Within 20 minutes of treatment with latrunculin B, most of the cell processes developed varicosities or were retracted, and branches of the processes often disappeared. The cell bodies appeared to decrease in size, collapsing to approximately the size of the nuclear region (Fig. 6E). Phalloidin staining of these drug-treated cells showed remnants of filamentous actin in small, dispersed areas (Fig. 6F). Cells treated with cytochalasin D showed similar responses after depolymerization of actin filaments (Figs. 6G, 6H, 6I). These results indicate that actin filaments, not microtubules, are important for the integrity of osteocyte cell processes.
A unique organization of actin filaments in the osteocyte
Low voltage (1.5 kV) FESEM, with a resolution of about 2–3 nm, is a powerful tool for the study of the cytoskeleton. The images are easy to interpret because they resemble the everyday experience of objects illuminated by diffuse light from above.
FESEM images show smooth membrane surfaces at the osteocyte cell body and processes (Figs. 7A and 7B). In contrast, the retracted processes of latrunculin B-treated cells show longitudinal wrinkles due to actin depolymerization (Fig. 7D).
To visualize the underlying cytoskeleton by FESEM, the cell membrane of osteocytes was removed by Triton X-100 lysis. Straight filamentous bundles were predominant along the lengths of processes (Figs. 8A and 8B), and, consistent with anti-actin immunofluorescence, the cell body contained filamentous bundles that often converged toward focal regions (Figs. 8C and 8D). Some of these may be located within a thin layer of cytoplasm above the nucleus (Figs. 8E and 8F). At the perinuclear site, a dense meshwork of filaments surrounded the nucleus (Fig. 8G). The diameter of these filaments suggested that they were primarily actin and intermediate filaments.
The contributions of actin filaments to the cytoskeletal organization of osteocyte became more apparent by FESEM analysis of cells treated with latrunculin B. In these drug-treated cells, the straight filamentous bundles characteristic of osteocyte processes were absent (Figs. 9A and 9B). Instead, only a loose meshwork of primarily 10 nm and 25 nm filaments was left (Fig. 9B). In contrast to the cell bodies of nontreated cells, which contain the unique aster-like convergences of straight filamentous bundles (Fig. 9C), the cell bodies of drug-treated cells were devoid of these bundles, and only a loose meshwork of 10 nm and 25 nm filaments remained (Fig. 9D).
Isolated osteocytes and process outgrowth
In vivo observations indicate that the spatial distribution of osteocytes and their morphology is highly responsive to their external environment. The cell bodies of osteocytes are oriented with their long axis parallel to the surrounding collagen fibers,46 and process formation, while asymmetrical, is not random since processes radiating toward the vascular side of bone are longer than those radiating toward the mineralizing surface.3–5
In vitro, isolated osteocytes respond with differential adhesivity to a variety of bone proteins, including collagen I and fibronectin.21 We extend these observations by analyzing osteocyte process outgrowth in response to these and other substrates (Fig. 3). Our results show that a combination of fibronectin and poly-D-lysine promotes a pattern of process formation and overall osteocyte morphology that best mimics calvarial osteocytes in vivo. Comparative analyses of these cells with cells plated on other substrates indicate that the differences in morphology are not due to significant variation in the number of processes per cell, but rather to differences in the lengths and branching complexity of the processes. These observations suggest that while the intrinsic factors for process initiation are present in isolated osteocytes, substrate factors play a role in influencing the lengths and pattern of formation of osteocyte processes. In addition, it should be noted that osteocytes in vitro are grown in an environment that facilitates growth of processes in only two dimensions, i.e., parallel to the plane of support. It is therefore possible that the cytoskeletal organization of osteocytes in vivo may be even more complex, because processes in this 3D matrix sprout uniformly from the entire cell body (Fig. 2). In contrast, osteocytes in vitro do not show process outgrowth from the “dome” of the cell body. (Fig. 7A).
Actin filaments form the axial core of the osteocyte shape
At the superficial level of cell shape, the osteocyte shares similarities with neurons; it has a round or spheroid cell body from which cytoplasmic processes emanate. However, while microtubules form the axial core of neurons and microtubule depolymerization results in retraction of neuroneal processes, it is the actin filaments that play such a role in osteocytes.
Our light microscopic staining of actin filaments (Figs. 4A and 4B) provides the first evidence of the integrative nature and continuity of the actin filaments between the cell body and the osteocyte processes. Although bundles of microfilaments were previously observed in osteocytes, this organization of actin filaments was not fully appreciated from analysis of conventional transmission electron micrographs.30–33,47 In this study, our FESEM images portray the 3D complexity of the cytoskeletal organization (Fig. 8). We observe unique focal convergences of actin bundles within the osteocyte cell body, some of which overlie the nucleus, possibly tethering it down, while other bundles continue into the osteocyte processes (Fig. 8C). Although similar focal convergences of actin filaments were reported in some fibroblasts,21,23 they were usually observed only at cell peripheries and do not occur in the uniform abundance as observed in osteocytes.
Comparison of actin filament distribution with that of other cytoskeletal components confirms the predominant distribution of actin filaments in osteocytes. While microtubules and intermediate filaments are also present in osteocyte cell bodies and processes (Figs. 4D and 4F), their distribution in the processes extends only to the proximal regions. In contrast, actin filaments extend to the ends of the cytoplasmic processes (Figs. 4A and 4B). In light of the mechanosensory function of osteocytes and the possible role of the cytoskeleton in mechanotransduction, this places actin filaments in a strategic position to mediate physical deformations of the plasma membrane, from the reaches of the distal portions of the cytoplasmic processes to the cell center.
Vesicles within osteocyte processes48,49 suggest transport activity. Microtubules are known to serve as tracks for vesicle transport20,50–53; however, the general paucity of microtubules and their absence at distal portions of osteocyte processes point to the actin cytoskeleton as the most likely candidate for organelle transport in osteocytes. In neurons, actin filaments have been demonstrated to perform this function at the very tips of growth cones, a region that is often microtubule free.54,55
Actin filaments and osteocyte shape
The importance of actin filaments as major support elements in osteocytes is demonstrated by depolymerization of actin filaments using latrunculin B and cytochalasin D. Significant changes in osteocyte morphology are observed: retraction of processes, decrease in overall size of the cell body to a cytoplasmic rim around the nucleus, and loss of membrane tension (Figs. 6, 7, and 9). Given that intercellular communication via gap junctions implicitly requires maintenance of cell-cell connections, these experiments show that the actin cytoskeleton is thus a necessary component in the maintenance of this circuit by providing a stable framework for the osteocyte processes.
Phase and electron micrographs indicate that nuclei in cells after removal of actin filaments tend to “bulge” (Figs. 6E and 6H), presumably due to the loss of the restraining bundles of actin filaments that overlay the nucleus (Figs. 8C and 8F). This direct influence of actin filaments on nuclear shape could be an important aspect of the osteocyte's mechanosensory capability, since it allows for the direct transmission of physical stimuli from the periphery to the nuclear region, possibly influencing activities within the nucleus. During the past several years, cytoskeletal control of gene expression has become increasingly evident, even though the signal transduction pathways responsible for these interactions are still unclear.15–17,56,57 Disruption of both actin and microtubule cytoskeletons by a variety of agents, such as 12–0-tetradecanolyphorbol-13 acetate, cytochalasin B, nocodazole, and colchicine, have resulted in changes of cell shape and phenotypic gene expressions in many cell types.19,58–63
Furthermore, we also provide evidence that actin filaments keep the osteocyte membrane in a state of tension, and depolymerization of actin filaments disrupts axial tensions, leaving the plasma membrane in folds and wrinkles (Fig. 7D). This may be functionally significant to the mechanosensory capability of osteocytes; recent evidence that osteocytes are stimulated by relatively small fluid shear stresses10–12 would predict that in order for the osteocytes to sense fluid flow over the cell surface, the membranes must be kept in a state of tension. Moreover, stretch-activated channels are proposed to be a component sensitive to mechanical stimulation,64,65 and failure to maintain tension along the membrane may inhibit ion channel activity.66,67
Osteocyte processes are a novel class of actin-rich projections
To date, well characterized actin-rich projections in animal cells include the dendritic spines of neurons, the microvillus border of columnar epithelia, the stereocilia of cochlea hair cells, and filopodia and lamellopodia of migrating cells in culture.20 In this study, we add our characterization of osteocyte processes to this list.
The common feature shared by osteocyte processes and other actin-rich projections is the presence of prominent actin bundles that form the core of these projections. By immunocytochemistry, we provide the first identification of two actin-bundling proteins, alpha-actinin and fimbrin, in osteocyte processes. Of the two, fimbrin appears to be the predominant component in osteocyte processes; it is localized throughout the extent of most processes, while alpha-actinin distribution is more limited (Figs. 5B and 5C). At present, we do not know the significance or relative contributions of these two proteins to the formation of the osteocyte processes. However, it is known that fimbrin bundles actin filaments that have parallel polarity while alpha-actinin bundles antiparallel filaments.68 The polarity of actin filaments in osteocyte processes is not known; the micrographs of one study31 using heavy meromyosin binding to identify actin filaments were not of sufficient magnification to discern the direction of arrowhead structures that define actin filament polarity.
Osteocyte processes possess features that are unique. They are longer than most of the other projections, and unlike any of the other cytoplasmic protrusions, osteocyte processes have the ability to branch. This suggests that in addition to lateral associations of actin filaments required for bundling, some mechanism must exist in these processes that splits actin bundles into separate portions in response to some yet unknown signal.
Fimbrin is an intracellular marker of osteocytes
Even though FESEM images show that actin bundles from the cell body continue into the cell processes (Fig. 8C), the prominent distribution of fimbrin appears to be associated only with actin bundles in cell processes. Anti-fimbrin did not label the actin bundles in the cell body, showing instead a diffuse, cytoplasmic staining (Figs. 5C and 5F).
Anti-fimbrin staining of cells resembling osteoblasts in our culture show only diffuse, cytoplasmic immunoreactivity, even though slender stress-fiber-like bundles are present, as evidenced by their characteristic periodic staining when labeled with alpha-actinin antibody (Fig. 5E). The prominence of fimbrin in actin bundles of osteocyte processes and its absence in stress fibers affirm the unique organization of actin filaments in the osteocyte processes and distinguish them from actin filaments in stress fibers. This specificity of fimbrin to osteocyte processes may be potentially useful for its use as an intracellular marker for distinguishing osteocytes from osteoblasts.
Although this study provides the first report of fimbrin in osteocyte processes, this observation is consistent with observations in other cell types, where fimbrin is present in surface structures such as microvilli, microspikes, membrane ruffles, and stereo cilia of auditory hair cells.39 In general, actin-bundling proteins such as fimbrin and alpha-actinin stabilize actin filaments through increased lateral interactions. In osteocytes, this would serve to enhance the stability of osteocyte processes that is necessary for the maintenance of a communication network within bone tissue. Moreover, by the time the osteocyte is fully embedded in the calcified matrix, the need for motility and elongation of osteocyte processes would diminish, because stable contacts are established and the canaliculi themselves permit little flexibility.
In summary, the successful isolation and maintenance of osteocytes in culture has enabled us to conduct an extensive analysis of cytoskeletal organization in these cells. We determined that actin filaments are crucial for the maintenance of osteocyte shape and identified two actin bundling proteins, fimbrin and alpha-actinin, in the unique, actin-rich processes. The specificity of fimbrin to osteocyte processes may be useful for its use as an intracellular marker to distinguish osteocyte from its precursor cell, the osteoblast.
S.S.L. thanks the Trustees of Meikai University for providing a Visiting Professorship to the laboratory of Dr. M. Kumegawa. During this visit, S.S.L. was introduced to osteocyte cell culture and thanks Dr. Kumegawa for his hospitality and generosity in sharing information. The authors are grateful to Dr. Simon Atkinson for critical reading of the manuscript and Michael Bowling and Michael Babb for technical assistance in preparing the figures. The contributions by Dr. Hans Ris were supported by grant DRR-570 from the Biomedical Research Technology Program, NCRR-NIH, to the Madison Integrated Microscopy Resource.