Integrin receptor-mediated mobilisation of intranuclear calcium in rat osteoclasts

Integrins constitute a large family of transmembrane gly-coproteins. They consist of noncovalently linked 0’(3 het-erodimers and serve as the major cell surface receptors involved in cell-extracellular matrix (ECM) attachment, as well as cell-cell adhesion events (Hynes, 1992). Ligand specificity and integrin receptor recognition appears to be a function of the exact α- and β-chains involved, and their interaction with each other. Many integrins express a bind-ing site described by a tripeptide sequence Arg-Gly-Asp (RGD) that is present in a variety of adhesive proteins. There is growing evidence to suggest that, in addition to participating in adhesive interactions, integrin receptors can transmit signals into cells upon binding certain ligands. For example, in platelets, collagen binding to the α₂β₁ complex has been shown to activate several phospholipases, phos-phatidylinositol turnover, elevation of cytoplasmic pH and Ca²⁺, and activation of protein kinases (Shattil and Brugge, 1991); and adhesion of neutrophils via β₂ integrins has been shown to induce cell motility and Ca²⁺ transients (Jaconi et al., 1991; Ng-Sikorski et al., 1991).

The present study is aiming to investigate the functional role of integrin receptors in bone cells, specifically in osteo-clasts. Bone is a dynamic tissue, undergoing continual resorption and formation of the calcified matrix. One of the major functions of the bone resorbing cell, the osteoclast, is calcium homeostasis (Roodman, 1991). Osteoclasts are highly motile cells, which can mediate bone resorption by migrating to specific sites of attachment and subsequent cel-lular polarisation (Marks, 1984; Vaes, 1988). Adhesion molecules such as integrins are thought to be involved in processes of cellular migration and attachment (Hemler, 1990; Ruoslahti, 1991).

Studies of integrin expression in osteoclasts have revealed that the most highly expressed receptor is the vit-ronectin receptor, VNR, comprising 0’_v and β₃ chains (Davies et al., 1989; Horton and Davies, 1989). There is also evidence for the presence of α₂β₁, a laminin/collagen receptor, α2β1 and ICAM-1 (Horton and Helfrich, 1992; Nesbitt et al., 1993). VNR has the RGD binding motif in its structure, unlike ICAM-1 and α2β1, and has been shown to mediate osteoclast attachment to bone matrix (Helfrich et al., 1992a; Flores et al., 1992).

The study of osteoclasts in vitro is limited because of their relative inaccessibility, scarcity and fragility when iso-lated from bone. Freshly isolated osteoclasts from neonatal rat long bones comprise less than 1% of a heterogeneous cell population. Little, therefore, is known about signalling in osteoclasts. Studies have shown that osteoclast function may be mediated by changes in Ca²⁺ (Miyauchi et al., 1990; Zaidi et al., 1989). However, knowledge about signal trans-duction via integrins in mammalian osteoclasts is limited, and it was of interest to determine whether certain sig-nalling events mediate cytoskeletal organisation, and thus control osteoclast attachment and retraction. We have there-fore applied single cell dynamic ratiometric video imaging to study changes in intracellular calcium in freshly isolated rat osteoclasts, using simple linear peptides containing the RGD sequence as specific probes for VNR. These included BSP-IIA, an 18 amino acid RGD-containing synthetic frag-ment of bone sialoprotein, a candidate for serving as an endogenous VNR ligand. Additionally, we have used F11, a monoclonal antibody to the rat integrin (3₃ chain (Hel-frich et al., 1992b), and the disintegrin snake venom con-stituent echistatin (Musial et al., 1990; Savage et al., 1990). In this paper, we report the presence of a unique com-partmentalisation of intracellular calcium in rat osteoclasts, which can be mobilised by activation of the cell surface VNR. The calcium transient does not appear to be involved in regulation of cytoskeletal elements. While the presence of intranuclear calcium stores may have potentially impor-tant implications, the sequence of signalling events from the cell surface to the nucleus remains unclear and is under investigation.

GRGDSP, GRGDS, GRGESP and GRADSP were from Peninsula Laboratories Europe, Ltd, (Merseyside, UK). BSP-IIA was either kindly provided by Dr P. Robey (NIDR, Bethesda, USA), or syn-thesised at the Peptide Synthesis Laboratory, Imperial Cancer Research Fund (London, UK), with the following sequence: Tyr-Glu-Ser-Glu-Asn-Gly-Glu-Pro-Arg-Gly-Asp-Asn-Tyr-Arg-Ala-Tyr-Glu-Asp. The anti-β₃ monoclonal antibody F11 was raised in our laboratory (Helfrich et al., 1992b). Echistatin was kindly pro-vided by Dr G. Rodan (Merck Sharp and Dohme Laboratories, USA). Salmon calcitonin and Fluo-3 AM were from Sigma Chem-ical Co. (UK). Thiazole orange was from Cambridge Biosciences (Cambridge, UK). Fura-2 AM was from Calbiochem Nova-biochem (Nottingham, UK). tBuBHQ was from Aldrich Chemi-cal Co. (UK).

Rat osteoclasts were obtained as previously described (Helfrich et al., 1992a). Briefly, limb bones, from 1-to 3-day neonatal rats, were dissected free of surrounding tissue. Osteoclasts were iso-lated by mechanical disaggregation of the bones. Cells were sus-pended in MEM containing 10% FCS and settled onto serum-coated glass coverslips (A.R. Horwell, 22 mm diameter, no.1.5) for 30 minutes at 37°C in 5% CO2. Cells were then loaded with 2 mM Fura-2 AM in MEM containing 10% FCS for 20 minutes at 37°C. All peptides tested were dissolved in buffer (see below) and were added in aliquots (10-20% chamber volume) when test-ing. Osteoclasts were identified on the basis of their size (>50 mm in diameter), and their multinuclearity (≥3 nuclei).

Video imaging was performed using the MagiCal system (Applied Imaging, UK), which utilises a software package TARDIS (Neylon et al., 1990; Mason, 1991) for the capture and analysis of digital image sequences. Osteoclasts were plated on thin glass coverslips (vide supra) and fitted into a thermostatically controlled perfusion chamber (37±0.5°C). The chamber was mounted on the stage of a Nikon Diaphot inverted microscope fitted with a ×40 fluor oil immersion objective and contained 0.5 ml buffer (com-position in mM: NaCl, 127; KCl, 5; MgCl₂, 2; NaH₂PO₄, 0.5; CaCl₂, 2; NaHCO₃, 5; glucose, 10; HEPES, 10; and 0.1% BSA). Epifluorescent images of cells were captured onto the face of an intensified CCD camera (Photonics Science, UK) operating at video frame rate. Cells were alternately excited at 340 nm and 380 nm wavelengths using a stepping filter wheel and emitted light was collected through a dichroic mirror and interference fil-ters at 510 nm. Background fluorescence at each excitation wave-length was obtained from a field of view on the coverslip with no cells and subtracted from subsequent images. The 340/380 ratio was obtained on a pixel-by-pixel basis and converted to calcium concentrations based on a calibration table found to be accurate for several cell lines including GH3 pituitary cells and Swiss 3T3 fibroblasts, using ionomycin (2 mM) in the presence of 10 mM Ca²⁺ (F_max) and 10 mM EGTA (F_min), and a K_d for Fura-2 bind-ing calcium of 225 nM.

Rat osteoclasts were loaded with 4 mM Fluo-3 AM for 30 min-utes. Cells were prepared and settled as described above. Images were captured by laser scanning confocal microscopy (Odyssey CLSM, Noran Instruments Inc.), from an optical slice of <1 mm thickness at the plane of the nuclei. Cells were excited at 488 nm and emitted light was captured at 510 nm. Images were captured approximately every 2 seconds.

GRGDSP (IC₅₀ for osteoclast retraction, 210 ± 14.4 mM; Horton et al., 1991) produced an elevation of [Ca²⁺]_i within 10 seconds in osteoclasts (Fig. 1A). Low resting levels of 20-80 nM rose to peak levels of 500 nM in the nuclear regions. [Ca²⁺]_i levels returned to pretreatment levels within 90 seconds. The calcium signal was observed in 63% of the cells tested (n=66). A similar response was seen with BSP-IIA (Fig. 1B), 1 mg/ml (400 μM) (n=53) (IC₅₀ for osteoclast retraction, approximately 400 mM; M. A. Horton, unpublished observations). The cells retained their ability to respond to a second addition of peptide, with (Fig. 1B) or without (Fig. 1A) a wash with buffer in between the two additions. Since osteoclasts can range from actively resorb-ing, motile cells to non-resorbing immotile cells in vivo, freshly isolated osteoclasts would not be expected to reflect a 100% responsiveness in vitro. Similar responses were seen with the pentapeptide GRGDS (800 mM) (data not shown).

The RGD analogues GRGESP and GRADSP were tested in order to establish the specificity of the RGD sequence in evoking a calcium signal. GRGESP caused a less than twofold increase in [Ca²⁺]_i in one out of five cells at 400 mM, while GRADSP caused a similar rise in [Ca²⁺]_i in 15% of the cells tested (n=20, 400 μM) (data not shown). These peptides have been shown to have no effect on osteoclast retraction at these concentrations (Horton et al., 1991; M. A. Horton, unpublished observations). Analysis of the cal-cium signal indicated that the spatial distribution of the cal-cium transient observed with the RGD-containing peptides was heterogeneous, localised to regions within the cell that corresponded to the nuclei as observed under bright field (Fig. 1C, images 2–4) and to their thin cytoplasmic skirts. It was also observed that the calcium signal preceded retrac-tion, which was not evident for at least two minutes fol-lowing addition of the peptides. Fig. 1C, image 6, shows a retracted osteoclast at the end of the experiment at approx-imately 7 minutes.

To confirm further the role of the VNR 0’_vβ₃ dimer in the GRGDSP-induced increase in [Ca²⁺]_i, we tested the anti-β₃ monoclonal antibody F11 (Helfrich et al., 1992b). This antibody caused no increase in [Ca²⁺]_i but blocked the GRGDSP response (Fig. 2A). Similarly, echistatin, a venom from the saw-scaled viper Echis carinatus, which may serve as a VNR-specific ligand as suggested by the colocalisa-tion with 0’_v in osteoclasts (Sato et al., 1990), inhibited the GRGDSP response while having no effect on [Ca²⁺]_i by itself (Fig. 2B). At similar concentrations, both echistatin and F11 also inhibited the calcium signal induced by 1 mg/ml BSP-IIA (data not shown).

In order to establish the source of calcium causing the increase in [Ca²⁺]_i, we tested the effect of GRGDSP in nom-inally calcium-free buffer containing 1 mM EGTA. As shown in Fig. 3, following a wash with EGTA-containing buffer, GRGDSP again produced a calcium signal, demon-strating that the response is due to mobilisation of intra-cellular stores. The response to BSP-IIA also remained unaffected in the absence of extracellular calcium (data not shown).

Examination of the video images during the calcium response to RGD-containing peptides showed a distinct compartmentalisation of the calcium transient with the max-imum response appearing to be localised in or around the nuclei of the multinucleated osteoclasts (see, for example, images 2-4, Fig. 1C). The following experiments were designed to define the exact localisation of the calcium tran-sient. Osteoclasts that displayed an increase in nuclear free calcium ion concentration in response to RGD-related pep-tides were subsequently stained with the nuclear dye thia-zole orange (Makler et al., 1987), and the cellular localisation of the stain was compared with that of the changes in [Ca²⁺]_i. The results from a representative experiment are shown in Fig. 4A, where a clear colocalisation of the cal-cium signal in response to BSP-IIA and the nuclear stain can be seen.

To establish whether ligands known to increase cytoso-lic calcium via calcium entry would induce a signal arte-factually localised to nuclei in osteoclasts, an increase in [Ca²⁺]_i was induced via treatment with the calciotropic hor-mone calcitonin (Malgaroli et al., 1989). Salmon calcitonin (sCT) (Fig. 4B, image 3) either induced a wave of increased [Ca²⁺]_i propagated through the cytosol or caused an evenly distributed increase in [Ca²⁺]_i throughout the cytosol, but did not induce a response localised to the nuclei, even in a cell that showed a response clearly localised to its nuclei after exposure to the VNR agonist BSP-IIA (Fig. 4B, image 2). 2,5-Di(tert-butyl)-1,4-benzohydroquinone (tBuBHQ), which has been shown to specifically inhibit ATP-dependent sequestration of calcium into the endoplasmic reticu-lum and release calcium from this pool without affecting nuclear calcium uptake (Kass et al., 1989; Nicotera et al., 1990), caused an increase in [Ca²⁺]_i in an area of the cell distinct from that caused by BSP-IIA (Fig. 4B, image 4). The calcium ionophore ionomycin, like calcitonin, caused an even increase in [Ca²⁺]_i throughout the cell (data not shown). These findings clearly demonstrate a unique com-partmentalisation of calcium stores in osteoclasts, with dif-ferent stores being sensitive to different ligands.

Using laser scanning confocal microscopy, rat osteoclasts loaded with Fluo-3 AM and stimulated with BSP-IIA con-firmed the nucleoplasmic localisation of the induced increase in [Ca²⁺]_i (Fig. 4C). Confocal images in a sectional plane through the centre of the nuclei of rat osteoclasts loaded with Fluo-3 AM showed an increase in nuclear Fluo-3 fluorescence following addition of BSP-IIA (Fig. 4C,_*). This was followed by an increase in the cytosolic region (Fig. 4C,_**). The increase in calcium in the region of the cytoplasmic skirt was, as expected, not apparent, due to the absence of information from this region in the particular optical slice under examination.

In osteoclasts, calcium has been shown to control cellular function, such that the high calcium levels in resorption lacunae mediate an inhibitory feed-back mechanism. Thus, increases in extracellular calcium have been shown to lead to rises in intracellular calcium and interfere with bone resorption by disassembling the osteoclast cytoskeleton (Miyauchi et al., 1990; Zaidi et al., 1989). Our data show that [Ca²⁺]_i is also increased when certain ligands bind cell surface receptors. Peptides bearing the RGD sequence, such as GRGDSP and BSP-IIA, recognise integrin receptors that have the RGD binding motif in their structure. In osteo-clasts, this receptor is presumably VNR, since the calcium signal in response to these peptides is blocked by F11, a monoclonal antibody to the rat β₃ integrin chain. The dis-integrin, echistatin, also inhibited the RGD-stimulated cal-cium signal. Echistatin was first described as a potent inhibitor of fibrinogen-dependent platelet aggregation (Musial et al., 1990; Savage et al., 1990), but in bone it has been shown to be a potent inhibitor of bone resorption and to colocalise with the 0’_v integrin chain in osteoclasts (Sato et al., 1990). The ability of GRGESP and GRADSP also to increase [Ca²⁺]_i in some osteoclasts, although lacking the RGD sequence, indicates that these ligands may bind VNR, but may not induce retraction at the concentrations that evoke calcium signals. While the ability of GRADSP to bind VNR has not been investigated, GRGESP, at high con-centrations, can induce osteoclast retraction in a manner similar to GRGDSP (Horton et al., 1991; Horton et al., 1993).

There are several lines of evidence from the present study to suggest that the calcium signal may not directly mediate cellular retraction in osteoclasts: first, the ability of GRADSP to mimic GRGDSP in eliciting a calcium signal in some osteoclasts while having no effect on osteoclast retraction; second, the ability of echistatin and F11 to induce retraction in the absence of an effect on [Ca²⁺]_i; and third, the observation that there is no correlation between the potency of the peptides to induce osteoclast retraction and to evoke a calcium transient. In fact, the most potent induc-ers of osteoclast retraction (F11 and echistatin) were inef-fective in causing a calcium signal, and even inhibited the calcium signal produced by BSP-IIA or GRGDSP. These data suggest that cellular retraction is not necessarily depen-dent on intracellular calcium. Alternatively, VNR occupancy by ligands such as F11 and echistatin might activate a cal-cium-independent signalling pathway for promoting osteo-clast retraction, while ligands such as GRGDSP and BSP-IIA might activate a distinct calcium-dependent pathway.

The ability to evoke a calcium signal in the absence of extracellular calcium clearly implicates mobilisation of intracellular sources of calcium. The potentiation of the GRGDSP response in the presence of EGTA (Fig. 3), a consistent observation in osteoclasts, can be explained on the basis of data from another integrin, LFA-1, binding to its ligand ICAM-1. It was shown that calcium, when bound to the divalent cation binding site, served to keep the recep-tor in an inactive state (Dransfield et al., 1992). Removal of calcium, as with EGTA, activated the receptor. It is likely, based on this and other findings (Conforti et al., 1990; Elices et al., 1990; Kirchhofer et al., 1990), that such a functional regulation by divalent cations is characteristic of other integrins as well, including VNR.

Our data are in contrast to those of Miyauchi et al. (1991), which show that RGD-containing peptides produce a decrease in [Ca²⁺]_i. Their studies differed in two key respects. First, they utilised polykaryons derived by culture of avian bone marrow mononuclear cells, whereas we have used freshly isolated osteoclasts, which are more likely to reflect the in vivo situation. Second, they studied avian osteoclasts, which are known to differ in several funda-mental ways from mammalian osteoclasts; for example, by lacking a response to and receptors for the calcium-regulating hormone calcitonin (Nicholson et al., 1987).

Perhaps most interesting is the spatial localisation of the calcium signal observed in our studies of rat osteoclasts. The signal is localised to two areas: the largest response in the nuclear area, and a smaller increase in the region of the cytoplasmic skirt of the osteoclast. While both these spatially distinct signals may have important, and potentially different, functional consequences, the ‘edge’ effect needs to be interpreted with some caution. This is because the edge of the cytoplasmic skirt is so thin that artificially high 340/380 ratios could result from low levels of dye that are beyond the range of the camera’s sensitivity, resulting in aberrant high calcium values in random pixels, a technical limitation that cannot be ignored.

Although dye loading conditions (see Materials and Methods) resulted in an otherwise even dye distribution, it may be argued that the nuclear localisation of the calcium signal is due to accumulation of Fura-2 in the perinuclear endoplasmic reticulum. Several observations demonstrate the contrary, including the colocalisation of the nuclear stain thiazole orange with the Fura-2 signal, and the abil-ity to elicit spatially distinct signals in the same osteoclast following exposure to different ligands. Conclusive evi-dence for the nuclear localisation of the calcium signal was obtained using the technique of laser scanning confocal microscopy. This approach allows the visualisation of events occurring in a thin (0.7-1 mm) ‘optical slice’ of the osteoclast taken through the plane of the nuclei, excluding potentially artefactual information from above and below the plane of focus. Under these conditions, it was clear that an immediate calcium transient in the nucleus was followed by a slower and smaller increase in the cytosolic regions, either due to calcium efflux from the nucleus to the cytosol or due to mobilisation of other intracellular stores in the cytosol.

It has been established that free calcium ion concentration gradients can be maintained between nuclear and cytosolic cell compartments (Williams et al., 1985; Way-bill et al., 1991). The demonstration of nuclear phos-phatidylinositol bisphosphate (PIP₂) in Friend ery-throleukaemia cells (Cocco et al., 1987) suggests that nuclear PIP₂ hydrolysis could lead to the formation of, for example, the calcium-mobilising metabolite inositol 1,4,5-trisphosphate, and the protein kinase C activator diacyl-glycerol (Berridge and Irvine, 1989; Irvine, 1992). Further, there is evidence to suggest that nuclear metabolism of PIP₂ may actually be regulated by cell surface receptors (Divecha et al., 1991), and that the calcium binding protein calmodulin and the calmodulin binding protein are present in the nucleus (Bachs and Carafoli, 1987). The present find-ings provide strong evidence for the presence of nuclear calcium stores activated by a cell surface receptor, the VNR. Our data thus propose a possible route whereby inter-actions between the cell membrane and adhesion molecules of the extracellular matrix may influence the activity of cal-cium-dependent processes within the nucleus and hence cel-lular function. Studies are underway to address the poten-tial implications of changes in [Ca²⁺]_i localised to the nucleus, including characterisation of second messenger pathways and regulation of immediate early gene expression. We are also examining whether a nuclear cal-cium transient in response to integrin-ligand interaction is characteristic of a subset of specialised, non-proliferative cells, or whether it is a more widespread phenomenon, seen in other cell types.