Mechanical loading modulates glutamate receptor subunit expression in bone

Bone 37 (2005) 63 – 73 www.elsevier.com/locate/bone

Mechanical loading modulates glutamate receptor subunit expression in boneB Anna M. Szczesniakb, Robert W. Gilbertb, Maya Mukhidaa, Gail I. Andersona,b,c,* a Department of Surgery, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada B3H 3J5 Department of Pharmacology, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada B3H 3J5 c Department of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada B3H 3J5 b

Received 1 April 2003; revised 1 August 2003; accepted 1 October 2003 Available online 26 May 2005

Abstract The cellular mechanisms coupling mechanical loading with bone remodeling remain unclear. In the CNS, the excitatory amino acid glutamate (Glu) serves as a potent neurotransmitter exerting its effects via various membrane Glu receptors (GluR). Nerves containing Glu exist close to bone cells expressing functional GluRs. Demonstration of a mechanically sensitive glutamate/aspartate transporter protein and the ability of glutamate to stimulate bone resorption in vitro suggest a role for glutamate linking mechanical load and bone remodeling. We used immunohistochemical techniques to identify the expression of N-methyl-d-aspartate acid (NMDA) and non-NMDA (AMPA or kainate) ionotropic GluR subunits on bone cells in vivo. In bone sections from young adult rats, osteoclasts expressed numerous GluR subunits including AMPA (GluR2/3 and GluR4), kainic acid (GluR567) and NMDA (NMDAR2A, NMDAR2B and NMDAR2C) receptor subtypes. Bone lining cells demonstrated immunoexpression for NMDAR2A, NMDAR2B, NMDAR2C, GluR567, GluR23, GluR2 and GluR4 subunits. Immunoexpression was not evident on osteocytes, chondrocytes or vascular channels. To investigate the effects of mechanical loading on GluR expression, we used a Materials Testing System (MTS) to apply 10 N sinusoidal axial compressive loads percutaneously to the right limbs (radius/ulna, tibia/fibula) of rats. Each limb underwent 300-load cycles/day (cycle rate, 1 Hz) for 4 consecutive days. Contralateral, non-loaded limbs served as controls. Mechanically loaded limbs revealed a load-induced loss of immunoexpression for GluR2/3, GluR4, GluR567 and NMDAR2A on osteoclasts and NMDAR2A, NMDAR2B, GluR2/3 and GluR4 on bone lining cells. Both neonatal rabbit and rat osteoclasts were cultured on bone slices to investigate the effect of the NMDA receptor antagonist, MK801, and the AMPA/kainic acid receptor antagonist, NBQX, on osteoclast resorptive activity in vitro. The inhibition of resorptive function seen suggested that both NMDAR and kainic acid receptor function are required for normal osteoclast function. While the exact role of ionotropic GluRs in skeletal tissue remains unclear, the modulation of GluR subunit expression by mechanical loading lends further support for participation of Glu as a mechanical loading effector. These ionotropic receptors appear to be functionally relevant to normal osteoclast resorptive activity. D 2005 Elsevier Inc. All rights reserved. Keywords: Ionotropic glutamate receptors; Mechanical load; Osteoclast; Bone lining Cells; Immunohistochemistry; Rat

Introduction

B

The delay in the publication of this article was due to circumstances beyond the control of the authors. * Corresponding author. Department of Surgery, Faculty of Health Sciences, Flinders University, GPO Box 2100, Adelaide, 5001, South Australia. Fax: +1 618 8374 1998. E-mail address: [email protected] (G.I. Anderson). 8756-3282/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2003.10.016

Mechanical loading is a potent regulator of bone remodeling; however, characterization of the cellular and molecular events linking loading and bone matrix production is incomplete. Recent investigations have focused on the role of the excitatory neurotransmitter glutamate (Glu) as a potential cellular mediator of these events. Bone is innervated [6], and nerve fibers containing Glu [20] have

64

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

been identified in the vicinity of bone cells expressing receptors for this neurotransmitter [7,18]. As well, Bhangu et al. [3] have shown that osteoblasts express the molecules needed for glutamate exocytosis such that they may act in a similar fashion to presynaptic neurons by actively releasing glutamate [3]. Glutamate is the primary excitatory neurotransmitter of the central nervous system (CNS) and acts through two types of membrane receptors: the ionotropic receptors and metabotropic receptors. Ionotropic GluRs can be divided into two large families, the NMDA receptor family and the AMPA/kainate receptor family. Receptor cloning studies have shown a large number of receptor subtypes in both these families. GluRs may exist as pentamers of heterologous subunits with the precise function of a GluR being ultimately determined by its subunit composition [22]. A variety of ionotropic GluR subunits (NMDA [NMDAR1, NMDARs 2A – 2D, NMDAR3]; AMPA [GluRs 1– 4]; and kainate [GluRs 5 – 7 and KA1 and KA2]) have been described [8], and immunohistochemical investigations have identified a number of these subunits in bone, expressed in osteoblasts, preosteoblasts, osteoclasts and osteocytes [7,18,23]. Electrophysiological studies have confirmed the expression of functional NMDA receptors in human osteoblastic cell lines MG63 and SaOS-2 [14] and in primary mammalian osteoclasts [9]. Subsequently, a potential role for the NMDA receptor (NMDAR) has been demonstrated in osteoclasts whereby NMDAR channel blockers have been shown to inhibit in vitro bone resorption suggesting that the functioning NMDAR is required for osteoclast resorptive function [7,19]. The suggestion that Glu may serve as an intercellular communicator linking mechanical stimulation with bone remodeling stemmed from an investigation of the pattern of regulated gene expression in cortical bone in response to mechanical loading [15]. These studies identified a glutamate/aspartate transporter (GLAST) gene, the expression of which in osteocytes was down-regulated after mechanical loading. Since the GLAST protein serves as a regulator of glutamatergic transmission in the CNS [8], the authors proposed a paracrine mechanism for Glu in bone [10,15]. In this study, we investigated the influence of mechanical loading on ionotropic GluR subunit expression in the long bones of young adult rats and in acutely isolated bone cells. As well, we assessed the functional role of these ionotropic receptors using MK801 and NBQX in a resorption assay to inhibit NMDA and AMPA/kainic acid receptor subtype function.

Six female 250 g Wistar rats were anesthetized using ketamine/xylazine given intraperitoneally (60 mg/kg of ketamine and 7.5 mg/kg of xylazine). The flexed right forelimbs (radius/ulna) of each animal were placed in brass loading grips (custom-made for rat forelimbs of this strain and body weight) at both the flexed elbow and carpal joints [24]. A Materials Testing System (MTS) (Materials System Testing Inc, 458.2 Micro Console) loading device was used to apply a static 3 N load prior to the application of cyclic compressive sinusoidal loads (Fig. 1). The carpal support cup was attached to the actuator of the MTS, and the elbow support cup was attached to the load cell of the MTS. The 3 N preload was applied with the MTS actuator in load control mode and permitted consistent grip of the limbs throughout the duration of the loading regime. Preload was applied such that the soft tissues of the limb were slightly compressed prior to the start of dynamic loading. Cyclic compressive sinusodial loads were then applied at a frequency of 1 Hz and a peak total force of 10 N for 5 min. In the absence of the 3 N preload, the 10 N load would decay over the 300 cycles due to creep of the soft tissues. This loading regimen was repeated daily for 4 days, the rats were allowed to rest for 4 days, then a second loading cycle was performed for a further 4 days. The rats were euthanized using barbiturate overdose intraperitoneally 4 days after their last load cycle. Although strains were not measured directly, extrapolating from the data published by Torrance et al. [24] for smaller animals, these four-times body weight loads applied sinusoidally would not have induced pathological strains. There were neither gross fractures nor any evidence of micro fractures seen on the histology sections examined after using this loading regimen [1]. Determination of the antebrachial stiffness in in vivo mechanically loaded rat forelimbs The effects of mechanical loading on global forearm stiffness was determined from MTS-derived load/displacement data. The stiffness data are reported as N/mm displacement. Load and displacement data for the right forelimbs were recorded at 4 equal time intervals during the

Materials and methods Characterization of the in vivo loading model and its anabolic response in bone An in vivo bone-loading model was evaluated for its ability to induce anabolic responses in the forelimbs of rats.

Fig. 1. Schematic diagram of the in vivo forelimb loading system. Here, the limb is held in the custom-made cups supporting the flexed carpal and elbow joints. Prior to hindlimb loading, however, the support cups were changed to those designed to accommodate the flexed stifle and hock joints to allow cyclic axial loading of the tibia.

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

300 load cycles. The means of the day 1 data, i.e. the first day of loading, were compared to the means of the final day of loading data to assess the effect of this loading regime on antebrachial stiffness [1]. The effects of in vivo mechanical load on cortical bone mineral apposition rate The six animals that underwent cyclic loading of their right forelimbs were also given tetracylines (25 mg/kg intraperitoneally) on day 1 of loading as well as on the last day of loading, 4 days prior to sacrifice. Left limbs were not mechanically loaded and served as controls. After euthanasia, both right and left forelimbs were processed for determination of mineral apposition rate. Briefly, the ulnae were removed, cleaned of soft tissues and the long bones placed in 70% ethanol and dehydrated via graded alcohol baths prior to being embedded in methylmethacrylate. Longitudinal tissue sections (8 Am) were taken from the right and left ulnae and examined along the caudal periosteal edge at two sites, one adjacent to the humeroulnar joint and the other at the midshaft of the ulna. Using fluorescent light microscopy and NIH Image software, the distances between the two parallel tetracycline labels were measured. The mean interlabel distance was calculated for each ulna, and paired t tests were used to compare the loaded vs. unloaded limbs. As the interlabel time intervals for all animals were the same, the interlabel data are reported in microns [1]. Investigation of the effects of in vivo mechanical load on glutamate receptor expression Four groups of three (n = 12) female Wistar Rats (250 g) were anesthetized with Somnotol (Sodium Pentobarbital, 26 mg/kg) injected intraperitoneally. The flexed right forelimbs (radius/ulna) of each animal were placed in brass loading grips (custom-made for rat forelimbs of this strain and body weight). The loading device was used to apply a static 3 N preload prior to the application of cyclic compressive sinusoidal loads via the grips at a frequency of 1 Hz and a peak total force of 10 N for 5 min. After forelimb loading, the MTS grips were changed and the ipsilateral hindlimb (tibia/fibula) was placed into custom grips for the hindlimbs and the same loading regimen repeated. The contralateral limbs were not loaded and served as controls. The 5-min loading regime was repeated as for the forelimbs on 4 consecutive days. Total displacement of the MTS actuator during load cycles did not exceed 1 mm during either forelimb or hindlimb loading. Two hours after the final loading session, the animals were euthanized. All experiments were carried out in accordance with protocols approved by the Animal Use Committee of the Faculty of Medicine, Dalhousie University in compliance with the guidelines of the Canadian Council for Animal Care.

65

Immunohistochemical investigation of GluR expression on bone tissue sections The bones were processed for immunohistochemical investigation of GluR expression. Briefly, the in vivo removal of blood and/or fixation of limbs were accomplished as follows. The thoracic cavity of each rat was exposed (3 cm transverse incision), and the left atrium was cannulated with a blunted 18 gauge needle. This procedure provided an infusion route for perfusion solutions. An efflux route for blood and perfusion solutions was established by making a 0.5 cm incision in the right ventricular wall. Perfusion of the animal with ice-cold phosphate-buffered saline (PBS, pH 7.4, 250 ml) facilitated the removal of blood from the tissues. Tissue sections were obtained from either decalcified bone or non-decalcified bone (described below) and assessed for suitability for immunohistochemical study. Preparation of decalcified bone tissue sections Tissue sections from decalcified bones (n = 36 long bones [bilateral radius/ulna and tibia from 6 animals]) were prepared in the following manner. Animals were initially perfused with 240 ml of PBS followed by perfusion with 240 ml ice-cold paraformaldehyde (PFA, 4%). Wellperfused limbs became stiff. Bilateral fore- and hindlimb long bones (radius/ulna, tibia) were then removed, dissected free of all soft tissues and decalcified with an EDTA – sucrose solution (240 g EDTA disodium salt dissolved in 1L of 5% phosphate-buffered sucrose, pH 7.3, 4-C) according to the technique of Bjurholm et al. [5] The mean decalcification time was assessed by radiography and determined to be ¨10 days. Following decalcification, bones were immersed in a 10% sucrose solution in 0.01 M phosphate buffer, (pH 7.3) for 2 days (for cryoprotection) then stored to 70-C until cryosectioning (described below). Preparation of non-decalcified bone tissue sections Non-decalcified bone tissue sections were prepared using a modification of the method of Hillam and Skerry [10]. Briefly, bones (n = 36 taken from 6 animals, as above) were perfused with ice-cold PBS, but not PFA, the limbs removed, dissected free of skin, but not soft tissue, and immediately cooled to 70-C prior to cryosectioning. Bone sectioning was performed on a Frigocut cryostat (model #2800). Bones were mounted onto cryostat chucks (either longitudinally or perpendicularly), embedded in OCT (Sakura) and cut into 7– 10 Am tissue sections. The cutting temperature of both the tissue and the knife at the time of sectioning was 30-C. Tissue sections were collected directly onto room temperature gelatin-coated (375 balloon, Sigma Chem. Co., St. Louis, MO) microscope slides (VWR Canlab, Dartmouth, NS, Canada).

66

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

Sections were stored at 20-C until immunohistochemical investigation. Non-decalcified bone sections were fixed immediately after sectioning (1 min in 4% PFA followed by PBS rinses). Immunohistochemical investigation of GluR subunit expression in bone tissue sections Tissue sections from both decalcified and non-decalcified bones were examined for the immunoexpression of ionotropic GluR subunits in the following manner. Immersing tissue sections in a 2% hydrogen peroxide solution (Sigma, St. Louis, MO) for 20 min depleted endogenous peroxide. Tissue sections were then incubated with 10% goat or horse serum (Vector Laboratories, Burlingame, CA) phosphate buffer (30 min) to block nonspecific antibody binding. Serial sections of tissue were incubated overnight (4-C) with primary antibodies directed against various GluR subunits. Antibodies directed against NMDA receptor 1 subunit (anti-NMDAR1), kainate receptor subunit (anti-GluR567) and AMPA receptor subunits (anti-GluR1, anti-GluR2, antiGluR2/3, anti-GluR4) were purchased from PharMingen (San Diego, CA), while antibodies directed against NMDAR2 receptor subunits (anti-NMDAR2A, antiNMDAR2B, anti-NMDAR2C) were obtained from Chemicon (Temecula, CA). Antibodies were diluted in PBS containing 0.002% Tween 20. Tissue sections were then incubated with biotinylated secondary antibodies (goat antirabbit IgG or horse anti-mouse IgG) (Vector Laboratories) for 1 h at room temperature. Antigen/antibody complexes were detected with a Vectastain Kit (Vector Laboratories) and developed with 3 – 3V diaminobenzidine-tetrahydrochloride (Sigma, St. Louis, MO). Rabbit or mouse immunoglobin (IgG) applied at the same concentrations (mg/ml) as the primary anti-GluR antibodies provided negative controls to the primary antiGluR antibodies (as recommended by PharMingen). Experiments performed in the absence of primary anti-GluR antibodies served as negative controls for all other steps in the immunohistochemical procedure. The identification of immunoexpression in rat brain sections to all of the antiGluR antibodies used in this study provided positive controls. Immunohistochemical investigation of GluR expression by acutely isolated osteoclasts All tissue culture reagents were obtained from Gibco (Gibco BRL, Mississauga, ON, Canada) unless otherwise stated. Osteoclasts from neonatal Long Evans rats (5 days old) were prepared according to the method of Arnett and Spowage [2]. Briefly, rat pups were euthanized with CO2, long bones (radius/ulna, tibia, femur and humerus) were removed, cleaned of adherent soft tissue then split and crushed (using a mortar and pestle) in a-MEM containing

10% heat inactivated fetal calf serum (FCS). The supernatant of cells was centrifuged at low speed (1200 rpm), resuspended in the above media and allowed to settle onto type 1 collagen coated glass coverslips (12 mm circle, Fisher Scientific, Mississauga, ON, Canada) at 37-C 5% CO2. After 1 h, nonadherent cells were removed by washing with PBS. A further 4– 8 h adhesion period followed, preparing the attached osteoclasts for the rigorous washing steps of the immunohistochemical procedure. Cells were then fixed (1 min, 3% PFA in phosphate buffer) in preparation for immunolabeling with anti-GluR subunit antibodies or for tartrate resistant acid phosphatase (TRAP) staining. Cells were always studied within 8 h of isolation. Prior to GluR immunolocalization, osteoclast preparations were preincubated for 15 min with 10% donkey serum (Pel-Freez Biologicals, Rogers, AK) in PBS to block nonspecific antibody binding. Osteoclasts were incubated for 1 h (37-C) with primary antibodies directed against GluR subunits (as above). Primary antibody incubation was followed by a series of PBS washes, and cells were then incubated with fluorescently conjugated secondary antibodies (CYTM3, CYTM2, Jackson Immuno Research Labs, Bar Harbour, ME) for 1 h at 37-C. A Nikon Eclipse E800 inverted phase-contrast microscope, equipped with epifluorescence, was used to visualize immunofluorescent expression of GluR subunits in these osteoclasts. Acutely isolated osteoclasts were also examined for the expression of TRAP (an indicator of osteoclastic phenotype) according to the method of Minkin (1982) [4,16]. Briefly, following fixation, osteoclasts were washed three times with PBS and then incubated with AS-BI phosphate (Sigma, St. Louis, MO) as a substrate in MichaelisVeronal Acetate Buffer at pH 5.0 in the presence of 20 mM l-tartaric acid (ICN, Montreal, QC, Canada) and with hexazonium pararosanilin as a coupling agent for a maximum of 30 min. TRAP positive cells stained ruby red. Immunohistochemical investigation of GluR expression by acutely isolated osteoblasts To assess the presence of GluRs in primary osteoblastlike cells, the bone chip explant method of Isheda et al. [11] was used to obtain osteoblastic cells from 125 g Wistar rats. The long bones were removed, cleaned of soft tissue and hand chopped into morsels of approximately 2– 3 mm diameter. These bone chips were then placed onto 35 mm dishes containing 10 Al droplets of a-MEM containing 20% FCS. After 14 days, the dishes were covered with cells originating from the bone chips. These dishes were treated with collagenase and trypsin to release the cells then with 30% serum to stop the enzyme activity. The cells were separated from the chips and plated onto new 35 mm dishes at 103 cells per dish. After 3 days in culture, the cells were

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

submitted to immunohistochemistry as for the osteoclasts and examined as above. Osteoclast resorption pit assays in presence of MK801 and NBQX Mature osteoclast isolation method Osteoclasts were harvested from the long bones of dayold New Zealand White rabbits according to the method of Arnett and Spowage [2]. Bones were chopped, and the endosteal surfaces vigorously flushed. The resulting cell suspension was washed, centrifuged at 1200 rpm, resuspended and the cells plated onto bovine cortical bone slices in a-MEM with l-glutamine, ribonucleosides and deoxyridonucleosides (Gibco BRL, Ontario, Canada) supplemented with 15% heat-inactivated fetal calf serum (FCS) (Gibco BRL), antibiotics (100 Ag/ml penicillin G (ICN Inc, OH), 50 Ag/ml gentamicin (Gibco BRL) and 0.3 Ag/ml fungizone (Gibco BRL) and 10 8 M 1, 25 vitamin D3 (1,25 Dihydroxycholecalciferol, Sigma). After 1 h at 37-C, in 5% CO2, nonadherent cells were removed, and the medium replaced every other day. Osteoclast resorption pit assays For assessment of osteoclast resorptive activity, 1.5 ml of medium was added on top of bovine bone slices placed into 24-well culture plates prior to the addition of 0.5 ml osteoclast cell suspension as above. Controls had no drug added, MK801 or NBQX were added at concentrations ranging from 0.01– 100 AM. The cultures were incubated for 4 days in 5% CO2 at 37-C prior to washing with PBS and gentle removal of the cells with a 0.25 N NH4OH solution-soaked cotton bud. After cell removal, the bone slices were subjected to immunohistochemical staining using mouse anti-bovine collagen type I antibody, secondary antibody and DAB (Sigma and Boehringer Mannheim, Quebec, Canada). The immunostained resorption pit areas per bone slice were then quantitated using a computer-assisted image analysis system (LECO Image Analysis System 2001, Montreal, QC, Canada) and an Olympus LHII microscope using 50  magnification. For each bone slice, 10 fields were assessed with four to six bone slices per treatment group evaluated for each experimental condition. The results of five repeat experiments are reported. To allow comparison of repeat experiments, the control values have been standardized to 100%, and the means and standard errors of the treatment groups expressed relative to their respective controls. Statistical analysis Statistical analysis was performed using Statview software (Abacus) on a Macintosh PC. For most assessments, ANOVA was performed for all groups, and if significant differences were found, individual comparisons were made using unpaired t tests.

67

Results Anabolic effect of in vivo mechanical loading on global forearm stiffness The effects of mechanical loading on global forearm stiffness were determined for the right forelimbs of 6 animals. The mean forelimb stiffness measured at day 1 was 10.86 T 0.95 N/mm, while the mean stiffness of the forelimbs on the final day of loading was 14.23 T 0.87 N/mm. These values were significantly greater than the day 1 values (P = 0.02) [1]. Effect of in vivo mechanical loading on bone mineral apposition rate The effects of mechanical loading on mineral apposition were determined using tetracycline labeling in the loaded vs. unloaded limbs. Loaded limbs had mean interlabel distances, significantly larger than those measured in the nonloaded (control) limbs. Mean interlabel distances measured in loaded limbs at the posterior periosteal edges adjacent to the humeroulnar joint were 42.6 T 4.1 Am in contrast to 31.6 T 2.55 Am in the control bones (P < 0.05). Mean interlabel distance measured at the periosteal surface of the midshaft of the ulna was 57 T 6.1 Am in loaded limbs and 36.2 T 1.9 Am in non-loaded limbs (P < 0.05). Expression of glutamate receptor subunits in tissue sections of adult rat long bones GluR subunit expression was investigated using antibodies directed against all commercially available NMDA, AMPA and kainate receptor subunits. Tissue sections obtained from the diaphysis, metaphysis and epiphysis of non-decalcified long bones were examined for the immunoexpression of anti-GluR antibodies. In these sections, osteoclasts, bone lining cells (osteoblasts or preosteoblasts), osteocytes, chondrocytes and vascular channels were identifiable. Osteoclasts were frequently identified in the cortical bone of the metaphyseal cortical bone region. Morphological and architectural characteristics including cell size (>50 Am) and their presence within cutting cones or Howship’s lacunae facilitated the identification of osteoclasts in these sections. Identifying osteoclasts on the basis of multinuclearity was not always feasible as a result of the thinness of tissue sections required for immunohistochemistry relative to osteoclast size. Cells of this morphology typically stained positive for TRAP, an indicator of osteoclastic phenotype [16]. Positive immunoexpression was recorded when the majority of cells of a particular type were positive for a given subunit in the region of interest. In this study, osteoclasts identified in the metaphyseal cortical bone demonstrated positive immunoexpression to the anti-NMDAR2A (Fig. 2), NMDAR2B, NMDAR2C, GluR2/3, GluR4 and GluR567 antibodies. The most intense immunolabeling was against the anti-

68

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

Fig. 2. Expression of GluR subunits on osteoclasts in rat bone sections. Section of cortical metaphyseal radial bone immunostained with the polyclonal NMDAR2A antibody. The osteoclast in its cutting cone exhibits staining which was noted to be localized to the cell membrane using confocal microscopy (not shown). All osteocytes seen were negative. Original magnification: 400; scale bar = 20 Am.

GluR2/3, and anti-NMDAR2B antibodies. No bone cells were immunolabeled when the control immunoglobulins (IgG mouse or IgG rabbit) were used as primary antibodies or in the absence of primary anti-GluR antibodies. Immunohistochemical investigation of osteoclasts in decalcified bone tissue sections demonstrated immunoexpression for anti-NMDAR2A and 2B, GluR2/3, GluR4 and GluR567, but unlike the non-decalcified tissue, failed to express labeling with the anti-NMDAR2C antibody. Immunolabeling to anti-NMDAR1, GluR2 and GluR1 antibodies was either extremely faint or absent regardless of the bone preparation technique employed. Since osteoclasts were rarely identified in the diaphyseal bone sections the immunoexpression of GluR subunits in diaphyseal bone will not be described due to the low numbers of cells available for examination. We also characterized the expression of GluR subunits on bone lining cells (osteoblasts and preosteoblasts). Tissue sections obtained from decalcified bone exhibited irregularly preserved trabecular bone architecture. Therefore, we restricted our investigations to sections obtained from nondecalcified tissue. Bone lining cells were identified on the basis of their location on bone surfaces and their flattened appearance. Positive immunoexpression was recorded when the majority of cells of a particular type were positive for a given subunit in the region of interest. Bone lining cells of the metaphyseal/epiphyseal region expressed immunolabeling to anti-NMDAR2A anti-NMDAR2B, anti-NMDAR2C, anti-GluR2/3, anti-GluR4 (Fig. 3A) and anti-GluR567 submit antibodies. The most intense labeling was observed to NMDAR2B and GluR2/3. Immunolabeling to antiNMDAR1, GluR2 and GluR1 antibodies was either extremely faint or not observed. The diaphyseal periosteum exhibited rare or very faint immunolabeling when compared to that observed in metaphyseal/epiphyseal regions. This may reflect an artefactual loss of cells from these bone surfaces during the removal of soft tissues. We observed no

Fig. 3. Expression of GluR subunits on bone lining cells in rat bone sections. (A) Section of radial metaphyseal bone immunostained with the polyclonal GluR4 antibody. Note that immunostaining is localized to bone lining cells. (B) IgG (rabbit) labeled control in the same region of radial metaphyseal bone. Osteocytes always showed negative immunolabeling for the GluR4 antibody. Original magnification: A and B, 200; scale bar = 20 Am.

difference in the expression of GluR subunits between limbs (radius/ulna versus tibia). Table 1 summarizes the immunoexpression of the various GluR subunit antibodies on both osteoclasts and bone lining cells. These results pertain

Table 1 In vivo expression of GluR subunits in metaphyseal regions of nondecalcified bone sections GluR subunit type

Cell type examined Osteoclast

NMDAR1 NMDAR2A NMDAR2B NMDAR2C GluR1 GluR2 GluR2 & 3 GluR4 GluR567

Osteocytes

Bone lining cells

+ + +

+ + +

+ + +

+ + + +

+ denotes a positive immunoexpression for GluR subunit antibody. denotes no immunoexpression for GluR subunit antibody.

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

69

bone resorption. Acutely isolated osteoclasts were examined for their expression of GluR subunits. In this study, all identified osteoclasts demonstrated positive immunoexpression to anti-GluR subunit antibodies. Immunoexpression to numerous GluR subunit antibodies was observed and included NMDAR2A and 2B, GluR1 (Fig. 4A), GluR2, GluR2/3 (Fig. 4B) and GluR567. The most intense staining was observed for GluR2/3, while expression of NMDAR1, NMDAR2C and GluR4 was extremely faint or absent from these rat osteoclasts. Osteoclasts immunolabeled with the control antibodies did not demonstrate fluorescence. While the immunoreactivity for NMDAR1 was not observed in either acutely isolated rat osteoclasts or in sections of rat bone, we have identified expression of this antibody in acutely isolated neonatal rabbit osteoclasts and in sections of rat brain using the same antibody (data not shown). Expression of glutamate receptors on acutely isolated rat osteoblasts None of the cells examined was positively labeled for any of the anti-GluR subunits when osteoblasts were grown using the bone chip derived outgrowth method.

Fig. 4. Immunolabeling of selected ionotropic GluR subunits on acutely isolated neonatal rat osteoclasts. (Left panels) Osteoclasts under phase contrast light microscopy. Large multinucleated cells were the dominant cell type in these preparations, although there were significant numbers of mononuclear TRAP+ cells. Original magnification: 400; Scale bar: 20 Am. (Right panels) Corresponding fluorescent photographs demonstrating the immunoexpression of GluR1 (A) and GluR2/3 (B) subunits respectively. Immunofluorescent labeling utilized a CY3-conjugated secondary antibody. Negative controls (lacking the primary antibody) were performed but were uniformly nonfluorescent and as such are not illustrated. Original magnification: 400.

to cells of the metaphyseal/epiphyseal regions of nondecalcified tissue. No bone cells were immunolabeled when the control immunoglobulins (IgG mouse or IgG rabbit, [see Fig. 3B]) were used as primary antibodies or in the absence of primary anti-GluR antibodies. A number of other bone cell types were also examined for expression of GluR subunits. Osteocytes were not stained in any of the limbs examined. None of the cells of the vascular channels, growth plate or articular cartilage chondrocytes expressed immunoreactivity to any of the antiGluR antibodies utilized in these studies (data not shown). Expression of glutamate receptors on acutely isolated neonatal rat osteoclasts Acutely isolated neonatal rat osteoclasts had typical multinucleated morphology and stained positive for the lysosomal enzyme, TRAP. When cultured for 3 –4 days, these cells develop into large multinucleated cells capable of

Fig. 5. Immunolabeling of longitudinal section of rat metaphyseal cortical bone with NMDAR2A antibody. (A) An osteoclast within a cutting cone showing immunoreactivity for NMDAR2A subunit. (B) Loss of NMDAR2A expression subsequent to mechanical loading. Original magnification: 1000; scale bar = 20 Am.

70

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

Effects of in vivo mechanical loading on GluR expression in bone tissue sections The effects of mechanical loading on the expression of GluR subunits was assessed on osteoclasts and bone lining cells of non-decalcified rat long bone sections. In the control, non-loaded bones, cells expressed immunolabeling towards a variety of GluR subunits (see Table 1). In the contralateral loaded limbs of the same animal, immunohistochemical expression for some, but not all GluR subunit antibodies, was lost. Those GluR subunits lost included GluR2/3, GluR4, GluR567 and NMDAR2A (Figs. 5A and B) in osteoclasts, while the expression of GluR2, GluR2/3, GluR4, NMDAR2A and NMDAR2B was lost in mechanically loaded bone lining cells. Immunoexpression for NMDAR2B and NMDAR2C was not affected in osteoclasts following mechanical loading, while NMDAR2C expression was unaffected in bone lining cells. Expression of no new GluR subunits was observed after mechanical loading. Table 2 summarizes the effects of mechanical loading on GluR subunit expression in the bone cells in non-decalcified metaphyseal/epiphyseal tissue sections. Effects of adding either MK801 or NBQX on osteoclast resorptive function in vitro We assessed the effect of adding the NMDA receptor antagonist, MK801, or the AMPA/kainic acid receptor antagonist on osteoclast resorptive function in vitro using rabbit and rat osteoclasts on bovine bone slices. The controls had a mean area of osteoclast resorption of 5.58 T 0.52%. MK801 decreased osteoclast pit area resorbed when added at concentrations of 10 AM or more (P < 0.05) compared to drug-free controls cultured in this system (Fig. 6). At a concentration of 100 AM, MK801 decreased pit area by 85.3% (P < 0.001).

Fig. 6. The effects of the NMDA receptor channel blocker, MK801, on rabbit osteoclast resorptive function in vitro. MK801 added at concentrations of greater than 10 AM significantly decreased osteoclast resorptive activity compared to controls. Data represent the means + standard error of the means of 10 fields from five bone slices per concentration.

The addition of NBQX at concentrations above 10 AM also decreased the pit area of bone resorbed (P < 0.05) (Fig. 7). At concentrations of 100 AM NBQX, the osteoclast pit area was reduced to 48.3% of control (P < 0.05). In this concentration range, the NBQX is acting to block both AMPA and kainic acid receptor functions, whereas at concentrations below 1 AM, NBQX acts as a selective AMPA receptor blocker. Interestingly, at 1 AM NBQX, there was a significant increase in osteoclast pit area relative to controls. No effect was seen with NBQX concentrations less than or equal to 0.1 AM.

Discussion A growing body of evidence lends support for the potential role of Glu as an effector in the pathway linking mechanical loading to the process of bone remodeling [15,21]. Our results provide clear evidence for the expres-

Table 2 GluR subunit expression in osteoclasts and bone lining cells in vivo: effect of mechanical loading on receptor subunit expression GluR subunit type

NMDAR1 NMDAR2A NMDAR2B NMDAR2C GluR1 GluR2 GluR23 GluR4 GluR567

Osteoclasts

Bone lining cells

Non-loaded bone

Loaded bone

Non-loaded bone

Loaded bone

+ + +

+ +

+ + +

+

+ + + +

N/E

+ + +

denotes no immunoexpression to GluR subunit antibody. + denotes a positive immunoexpression for the GluR subunit antibody. N/E denotes that effect of mechanical loading on this subunit was not investigated.

Fig. 7. The effects of the AMPA and kainate receptor antagonist, NBQX, on rabbit osteoclast resorptive function in vitro. Data are expressed as the means + standard error of the means of 10 fields from five bone slices per concentration.

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

sion of a variety of ionotropic GluR subunits (both NMDA and non-NMDA types) in rat bone cells and for the selected modulation of GluR subunit expression by mechanical loading. With respect to the expression of GluR subunits in bone cells, our immunohistochemical studies show that osteoclasts and bone lining cells (preosteoblasts and osteoblasts) express GluR subunits and that each cell may express several different types of GluR subunits. We found clear expression of selected NMDAR subtype subunits. NMDARs consist of two distinct types of subunits, including the main subunit, NMDAR1, and modulatory subunits, NMDAR2A – 2D. While the NMDAR1 subunit is both necessary and sufficient to form functional NMDAR channels, the NMDAR2 subunits alone do not form functional channels but act to modulate properties of heteromeric channels composed of both NMDAR1 and NMDAR2 subunits [8]. While our studies did not provide conclusive evidence for the immunohistochemical expression of NMDAR1 subunits, they clearly demonstrate the expression of NMDAR2A – 2C subunits in osteoclasts and bone lining cells. This work provides the first description of NMDAR2 subunit expression in rat osteoclasts. The lack of identifiable NMDAR1 subunits in these rat studies does not preclude their existence, however, as electrophysiological and pharmacological investigations have provided evidence of functional NMDARs in rabbit osteoclasts [12] and human osteoblastic cell lines [14]. Other investigators have identified NMDAR1 subunit expression using immunohistochemical techniques [7,18]. One explanation for this discrepancy may lie in the use of control antibodies. In our studies, we did observe bone cells expressing immunoreactivity to the primary anti-NMDAR1 antibodies used, however, when the pattern of antiNMDAR1 antibody staining was compared with the appropriate IgG control (matched for both concentration and species), an unequivocal positive expression could not be ascribed. The integrity of the GluR subunit antibodies used in this study was confirmed through immunohistochemical investigation of rat brain slices. We also describe, for the first time, the immunoexpression of selected non-NMDA receptor subunits. We provide evidence for the expression of anti-GluR567 subunits on both osteoclasts and bone lining cells. The anti-GluR567 antibody binds to three subunits of the kainate acid GluR subtype. While this antibody does not identify the specific subunit(s) expressed by these cells, our demonstration of immunoreactivity indicates that at least one of these subunits is expressed. Our results also demonstrate that bone cells possess selected subunits of AMPA GluRs. We demonstrated immunoreactivity towards anti-GluR2/3 and anti-GluR4 antibodies in osteoclasts and towards anti-GluR2, 2/3 and 4 in bone lining cells. The anti-GluR2/3 antibody used in these experiments binds to two of the AMPA GluR subunits, GluR2 and GluR3. In osteoclasts, the observation of immunoreactivity to anti-GluR2/3 but not to anti-GluR2

71

suggests the presence of GluR3 subunits in these cells. A similar conclusion could not be drawn in bone lining cells where GluR2 subunits were expressed. Selective antibodies towards the GluR3 subunit are not yet commercially available. The demonstration of immunoexpression for the GluR2/3 subunit in these cells is in contrast to results of others. Recently, Chenu et al. [7] concluded GluR2/3 subunit expression to be undetectable in rat bone cells. The lack of GluR2/3 subunit expression in their model may reflect (1) a loss of GluR2/3 antigenicity subsequent to the decalcification and embedding methodologies used in their tissue section preparation or (2) a difference in the age of rats (125 g rats used in Chenu study vs. 250 g used in these studies) [17]. As such, one additional aim of our study was to elaborate a method for the preparation of bone that would permit immunohistochemical staining of GluR subunits in bone cells. Previous studies have employed prolonged fixation, demineralization and embedding steps when processing bone for tissue section immunohistochemistry [5,7,18]. To address the influence of these processes on GluR subunit antigenicity, we examined two bone preparation procedures: (1) preparation of tissue sections from demineralized bone and (2) preparation of tissue sections from non-demineralized bone. In our study, tissue sections prepared from non-demineralized bone proved to be more appropriate with respect to preservation of bone cytoarchitecture and GluR subunit antigenicity. We report a loss of antigenicity to the NMDAR2C subunit in osteoclasts within demineralized sections. These findings lend support to the suggestion that the lack of immunoreactivity to the anti-GluR2/3 antibody reported by Chenu et al. [7] may reflect a loss of antigenicity to that subunit antibody during the tissue processing procedure [17]. In our experiments, identification of GluR subunits on osteocytes was inconclusive. While intense anti-GluR subunit staining of osteocytes within their lacunae was occasionally observed, these findings were deemed non-specific as similar intensity of labeling could be observed in the nonprimary antibody-containing controls probably reflecting lacunar antibody trapping rather than specific labeling. Furthermore, the majority of osteocytes examined (>95%) did not exhibit detectable immunolabeling for any of the antiGluR antibodies employed. This finding contradicts the results of others that have identified selective expression of NMDA and AMPA receptor subunits in rat osteocytes [7,18]. We investigated the expression of GluR subunits on acutely isolated neonatal rat osteoclasts. Acutely isolated cells were identified as osteoclasts on the basis of their cell morphology (large multinucleated cells which stain positive for TRAP) [2,16]. Osteoclasts were always examined for GluR expression within 8 h of isolation (i.e. giving the cells long enough to attached sufficiently to withstand the numerous washing steps but not so long as to undergo phenotypic drift). However, these same cells if left in culture for longer periods went on to become much larger, increasingly nucleated and capable of resorbing bone. Our early

72

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73

examination reduced the likelihood of phenotypic change, a potential consequence of extended culture conditions. Acutely isolated neonatal osteoclasts expressed a different subset of GluR subunits from those observed in the adult tissue sections. In comparison with adult osteoclasts, the expression in neonatal osteoclasts to NMDAR2A and B, GluR2/3 and GluR567 remained the same, while expression to NMDAR2C and GluR4 were absent, and GluR1 and GluR2 were gained. The differences observed may reflect the different age of animals from which bone cells were obtained (adult versus neonate). This interpretation is consistent with studies in rat brain which demonstrated changes in GluR subunit composition during development [25]. These experiments failed to identify GluR subunit expression on osteoblasts in vitro. The lack of GluR expression on isolated osteoblasts may result from the fact that the in vitro growth from bone chips takes a number of weeks and that during this culture period the phenotype of these cells may have changed. Furthermore, the osteoblastic cells investigated in our studies were derived from cells of bone chips, which are predominantly populated by cells of osteocytic phenotype. To proliferate in culture, these cells de-differentiate into osteoblast-like populations, proliferate and then develop bone nodules. It is possible that the osteoblastic cells obtained by this method did not develop the characteristics of the in vivo osteoblast lining cells but retained characteristics closer to those of osteocytes. Our in vivo results demonstrated that the majority of osteocytes did not express immunoreactivity to any of the anti-GluR subunit antibodies examined. Our MK801 experiments in rabbit osteoclasts confirm those of Chenu et al. [7] (also in rabbit), and our rat osteoclast observations suggest that these receptors are functional in more than a single species. Itzstein et al. [12] suggest that, by antagonizing NMDA channel function, osteoclast resorptive activity may be inhibited through disruption of actin ring formation. While NMDA receptors have received some attention, our investigations of the functional role of non-NMDA receptors are novel. The inhibition of osteoclast resorptive activity with concentrations of NBQX higher than 10 AM suggests that it is the kainic acid receptor function that is being blocked. The increase in osteoclast resorptive activity with the 1 AM dose may imply a role for AMPA receptors in inhibiting basal osteoclast function. This study has demonstrated our in vivo mechanical loading model to be capable of inducing an anabolic response in rat long bones. The significant change in both bone stiffness and bone mineral apposition rates measured at metaphyseal and diaphyseal periosteal sites confirms the anabolic effect of this loading regimen. Our demonstration that mechanical loading leads to selective changes in GluR subunit expression suggests a novel mechanism for regulating bone cell activity. In the CNS, the conductance properties of ionotropic GluRs are ultimately determined by their subunit composition [8,11,13]. We propose that

mechanical loading induced changes in GluR subunit expression that ultimately translates into changes in GluR subunit composition, potentially leading to altered GluR channel function. Since we and others, using pharmacological studies, have demonstrated a role for GluR receptormediated ionic conductance in the control of osteoclast resorptive activity [7,12,18], it is interesting to speculate that mechanical loading may induce changes in GluR subunit composition (and therefore conductance) and that these changes may serve in the physiological regulation of osteoclast activity. As for the role of GluR in bone lining cells, the juxtaposition of glutamatergic nerves and GluR possessing bone cells of the trabeculae [7,20] suggests a potential synapse-like glutamatergic regulation of the bone lining cell function. The functional relationship between Glu from these nerve endings and mechanical modulation of GluR expression on juxtaposed lining cells remains unclear. Further studies are necessary to determine the precise role of selected GluR subunits in bone cell activity. Not only are the NMDA receptors functionally relevant to osteoclast function but, in addition, kainic acid receptors, and possibly AMPA receptors, also appear to modulate osteoclast resorptive activity in vitro. Nevertheless, the presence of GluR subunits in bone cells, as well as the demonstration that mechanical loading can alter their protein expression, may suggest a novel paradigm for the modulation of bone remodeling.

Acknowledgments This work was supported by grants from NSERC (Canada) as well as The Arthritis Society of Canada. The authors thank Dr. Ying Fang Chen for her technical assistance. The authors also wish to thank Kay Ferguson from the Department of Pharmacology for her extensive assistance in the trouble shooting the cryosectioning methods.

References [1] Anderson GI, Fleming M, Orlando K, Chaudhary D, Waldman S, Lee JM. Effect of indomethacin, verapamil, estrogen and alendronate on bone’s response to controlled loading. Transactions 44th Annual Meeting of the Orthopaedic Res Society, New Orleans, LA, USA, March, vol. 2; 1998. p. 949. [2] Arnett TR, Spowage M. Modulation of the resorptive activity of rat osteoclasts by small changes in extracellular pH near the physiological range. Bone 1996;18:277 – 9. [3] Bhangu PS, Genever PG, Spencer GJ, Grewal TS, Skerry TM. Evidence for targeted vesicular glutamate exocytosis in osteoblasts. Bone 2001;29:16 – 23. [4] Biskobing D, Fan X, Rubin J. Characterization of MCSF induced proliferation and subsequent osteoclast formation in murine marrow culture. J Bone Miner Res 1995;17:1025 – 32. [5] Bjurholm A, Kreicbergs A, Schultzberg M. Fixation and demineralization of bone tissue for immunohistochemical staining of neuropeptides. Calcif Tissue Int 1989;45:227 – 31.

A.M. Szczesniak et al. / Bone 37 (2005) 63 – 73 [6] Calvo W, Forteza-Vila J. On the development of bone marrow innervation in new-born rats as studied with silver impregnation and electron microscopy. Am J Anat 1961;126:355 – 71. [7] Chenu C, Serre CM, Raynal C, Burt-Pichat B, Delmas PD. Glutamate receptors are expressed by bone cells and are involved in bone resorption. Bone 1998;22:295 – 9. [8] Dingledine R, Borges K, Bowie D, Traynelis F. The glutamate receptor ion channels. Pharmacol Rev 1999;51:7 – 61. [9] Espinosa L, Itzstein C, Cheynel H, Delmas PD, Chenu C. Active NMDA glutamate receptors are expressed by mammalian osteoclasts. J Physiol 1999;518:47 – 53. [10] Hillam RA, Skerry TM. Inhibition of bone resorption and stimulation of formation by mechanical loading of the rat ulna in vivo. J Bone Miner Res 1995;10:683 – 9. [11] Isheda Y, Tertinegg I, Heershe JN. Progesterone and dexamethasone stimulate proliferation and differentiation of osteoprogenitors for adipocyte and macrophages in cell populations derived from rat vertebrae. J Bone Miner Res 1995;11:921 – 30. [12] Iszstein C, Cheynel H, Burt-Pichat B, Espinosa L, Delmas PDP, Chenu C. Specific antagonists of NMDA receptors prevent osteoclast sealing zone formation required for bone resorption. Biochem Biophys Res Commun 2000;268:201 – 9. [13] Jonas P, Burnashev N. Molecular mechanisms controlling calcium entry through AMPA-type glutamate receptor channels. Neuron 1995;15:987 – 90. [14] Laletic-Ljubojevic I, Suva LJ, Mathuis FJM, Sanders D, Skerry TM. Functional characterization of N-methyl-d-aspartic acid-gated channels in bone cells. Bone 1999;25:631 – 7. [15] Mason DJ, Suva LJ, Genever PG, Patton AJ, Steuckle S, Hillam RA. Mechanically regulated expression of a neural glutamate transporter in bone: a role for excitatory amino acids as osteotropic agents? Bone 1997;20:199 – 205. [16] Minkin C. Bone acid phosphatase: tartrate resistant acid phosphatase as a marker of osteoclast function. Calcif Tissue Int 1982;34:285.

73

[17] McKee MD, Nanci A, Landis WJ, Gotoh Y, Gerstenfeld LC, Glimcher MJ. Effects of fixation and demineralization on the retention of bone phosphoproteins and other matrix components as evaluated by biochemical analyses and quantitative immunocytochemistry. J Bone Miner Res 1991;6:937 – 45. [18] Patton AJ, Genever PG, Birch MA, Suva LJ, Skerry TM. Expression of an N-methyl-d-aspartate-type receptor by human and rat osteoblasts and osteoclasts suggest a novel glutamate signaling pathway in bone. Bone 1998;22:645 – 9. [19] Peet NM, Grabowski PS, Laketic-Ljubojevic L, Skerry TM. The glutamate receptor antagonist MK801 modulates bone resorption in vitro by a mechanism predominantly involving osteoclast differentiation. FASEB 1999;13:2179 – 85. [20] Serre CM, Farlay D, Delmas PD, Chenu C. Evidence for a dense and intimate innervation of the bone tissue, including glutamate-containing fibers. Bone 1999;25:623 – 31. [21] Skerry TM. Identification of novel signaling pathways during functional adaptation of the skeleton to mechanical loading: the role of glutamate as a paracrine signaling agent in the skeleton. J Bone Miner Metab 1999;17:66 – 70. [22] Slotboom DJ, Konings WN, Lolkema JS. Structural features of the glutamate transporter family. Microbiol Mol Biol Rev 1999;63: 293 – 307. [23] Szczesniak AM, Gilbert RW, Anderson GI. Expression of glutamate receptor subunits in skeletal tissue: a novel pathway in bone remodeling? X-Symposium of the Polish Osteoarthrology Society, Krakow, Poland; 1999. [24] Torrance AG, Mosley JR, Suswillo RF, Lanyon LE. Non-invasive loading of the rat ulna in vivo induces a strain-related modeling response uncomplicated by trauma or periosteal pressure. Calcif Tissue Int 1994;54:241 – 7. [25] Watanabe M, Jnove Y, Satimura A, Mishina M. Developmental changes in distribution of NMDA receptors channel subunit mRNA. NeuroReport 1992;3:1138 – 40.