Professor Michael A. Horton, Bone and Mineral Centre, Department of Medicine, 5 University St., London WC1E 6JJ, UK. E-mail: email@example.com
Summary. Osteoclasts utilize αvβ3 integrin adhesion to bone matrix during bone resorption. We have generated osteoclasts from the peripheral blood of Iraqi-Jewish patients with Glanzmann thrombasthenia (GT) who are completely deficient in β3 integrin and exhibit a haemorrhagic diathesis resulting from the absence of platelet αIIbβ3. We show that, in contrast to osteoclasts generated from normal subjects or patients with αIIb integrin deficiency, GT osteoclasts lack αvβ3. These osteoclasts exhibited a two- to fourfold increase in α2 and β1 integrin expression, whereas other αv integrins, including αvβ5, were not significantly affected. An accompanying decrease in bone resorption was observed, with 44% and 59% declines in pit number and depth, respectively, and resorption lacunae showed abnormal morphology on scanning electron microscopy. However, osteoclasts from GT developed in similar numbers to controls and exhibited an otherwise ‘normal’ phenotype. We conclude that the observed rise in α2β1 expression compensates for the chronic genetic deficiency of αvβ3 in osteoclasts from patients with GT and is sufficient to enable bone resorption to proceed, albeit to a submaximal extent. This explains why Iraqi-Jewish patients with GT do not have osteopetrosis.
These features encouraged the pharmaceutical industry to develop non-peptidic RGD analogue drugs, currently in clinical trial (Hartman & Duggan, 2000; Miller et al, 2000; Horton, 2001), for use as inhibitors of osteoclast function in treating skeletal diseases such as osteoporosis or bone cancer. It has not been established whether chronic inhibition or lack of αvβ3 impacts upon osteoclast function or if there is a compensatory response in integrin expression by osteoclasts that may modify therapeutic efficacy.
One approach to examine the impact of chronic lack of an integrin is to study knockout mice that have a lifelong deficiency of the target gene. The phenotype of β3 integrin knockout mice has been reported recently (McHugh et al, 2000; Feng et al, 2001). The skeleton at birth was relatively unaffected by the lack of β3 integrin, but osteosclerosis developed in ageing mice. Limited information on the properties of mature β3 null osteoclasts has come from these murine studies – osteoclast adhesion to matrix and bone resorption are reduced (McHugh et al, 2000; Feng et al, 2001). Effects on other integrins, such as α2β1 expression, were not reported.
Glanzmann thrombasthenia (GT) is an autosomal recessive bleeding disorder defined by defective or quantitatively abnormal platelet αIIbβ3 receptors for fibrinogen. Diagnosis is based upon the clinical syndrome of mucocutaneous bleeding in association with platelet aggregation abnormalities in response to physiological stimuli. Earlier biochemical and molecular genetic characterization of the αIIb and β3 integrin subunits and their genes in patients with GT has shed considerable light on the aetiology of this genetic disorder. One particular form of GT is the extreme β3 null phenotype observed in the Iraqi-Jewish population in Israel (Coller et al, 1986; Newman et al, 1991; Rosenberg et al, 1998). Patients from this genetic isolate (Seligsohn & Rososhansky, 1984;Rosenberg et al, 1998) have very low levels of platelet αIIbβ3 on account of the complete absence of the integrin β3 protein. This is caused by an 11-bp deletion in exon 12 in the β3 (CD61/gpIIIa) gene, the mutation producing a frameshift leading to premature termination of translation at the trans-membrane domain of the β3 protein (Newman et al, 1991).
In this study, we generated osteoclasts from a series of GT patients and controls and analysed their integrin profiles and function. We show that lack of β3 protein results in a deficiency of αvβ3 in GT osteoclasts; the α2β1 integrin collagen receptor is upregulated, but other αv integrins are not, and bone resorption is reduced. We conclude that chronic (genotypic) deficiency of αvβ3 in human osteoclasts is accompanied by a compensatory rise in α2β1 expression. This is sufficient to enable bone resorption to proceed, but to submaximal extent, explaining why affected individuals do not present with osteopetrosis.
Materials and methods
GT patients and controls. Five unrelated patients (two males and three females) with GT, whose ages ranged between 10 and 67 years, had lifelong mucocutaneous bleeding manifestations, absent clot retraction and no platelet aggregation upon stimulation by ADP, collagen and epinephrine. The molecular basis of the disease in all five patients was homozygosity for an 11-bp deletion in exon 12 of the glycoprotein IIIa (β3) gene [online Mendelian inheritance in man (OMIM) 173470·0011] (Newman et al, 1991). This gene alteration is the predominant mutation causing GT in Iraqi Jews living in Israel (Rosenberg et al, 1998). The characteristic molecular phenotype in these patients is severe deficiency of αIIbβ3 and αvβ3 complexes in platelets (Coller et al, 1991) and absent αvβ3 complexes in Epstein–Barr virus (EBV)-transformed lymphocytes (Rosenberg et al, 1998). All patients from Israel consented to donate an extra 20 ml of blood collected under sterile conditions during routine diagnostic blood sampling (with ethical committee approval; Caim Sheba Medical Centre, Israel). Peripheral blood mononuclear cells (PBMNC) were separated from heparinized blood layered over Ficoll-Paque (Amersham Pharmacia Biotech, Little Chalfont, Bucks, UK) and centrifuged at 1400 g for 25 min at room temperature. After washing by dilution in PBS and centrifugation, PBMNC were rate frozen in 10% dimethyl sulphoxide, 90% fetal calf serum and sent on dry ice to the UK where they were stored in liquid nitrogen until used.
A similar procedure was followed for the collection and storage of control PBMNC from normal subjects (n = 4, average age 33·5 years) and patients with GT resulting from a defective αIIb component of αIIbβ3 (n = 2, aged 13 and 9 years) where platelet αvβ3 levels are normal. Here, blood samples were obtained with informed consent (St Mary's Hospital and Great Ormond St. Hospital ethics committee, London, UK).
Osteoclast generation from peripheral blood. Osteoclasts were generated as described previously (Lader et al, 2001) with modifications (Massey et al, 2001; Scopes et al, 2001). Briefly, frozen PBMNC were thawed and resuspended and washed once in minimum essential medium (MEM; Sigma-Aldrich, Gillingham, Dorset, UK). Approximately 1–2 × 105 cells in MEM supplemented with 15% heat-inactivated fetal bovine serum (Sera Laboratories International, Crawley, W. Sussex, UK), 2 mmol/l l-glutamine, 100 IU/ml benzylpenicillin and 100 µg/ml streptomycin (Gibco, Paisley, UK) were plated on devitalized bovine bone slices (d 1) in 96-well plates in a final volume of 200 µl of medium containing reagents. Cultures were performed under two different osteoclast-inducing conditions (Massey et al, 2001; Scopes et al, 2001) with similar results. Either interleukin 4 (IL-4; 1 ng/ml) or transforming growth factor β (TGFβ; 10 ng/ml) was added to the cultures on d 1 but at no other time points. On d 4 (the day of the first medium change), 180 µl of medium was removed from the wells, and RANKL (30–120 ng/ml; kindly provided by C. Dunstan, Amgen, CA, USA) was added for the first time. The cultures were then fed twice weekly by removal and replacement of half the medium and RANKL. All cultures were maintained in macrophage colony-stimulating factor (M-CSF, 25 ng/ml; kindly provided by Genetics Institute, Boston, MA, USA) throughout the experiments. The cultures were terminated between 10 and 21 d by removing bone slices from the wells and fixing them for analysis. Cultures were maintained at 37°C in 5% CO2/95% air.
Confocal microscopy. Bone slices were fixed for 5 min in a 50:50 mixture of MEM with fixation buffer [3·5% paraformaldehyde and 2% sucrose in phosphate-buffered saline (PBS; Sigma-Aldrich)], washed in PBS, placed in ice-cold permeabilization buffer (20 mmol/l Hepes, 300 mmol/l sucrose, 50 mmol/l NaCl, 3 mmol/l MgCl2, 0·5% Triton X-100 and 0·5% sodium azide in PBS) for a further 5 min and finally washed in PBS, as described previously (Nesbitt & Horton, 1997). The bone slices were then incubated in phalloidin-tetramethyl rhodamine B isothiocyanate (TRITC) conjugate (red channel in the confocal microscope) in PBS (5 U/ml; Molecular Probes, Eugene, OR, USA) to enable resorbing osteoclasts to be identified by their characteristic F-actin ring structure (Fig 1). Integrin expression was analysed using monoclonal antibodies 23C6 for αvβ3 (Horton et al, 1985) 13C2, for all αv dimers (Horton et al, 1985) and PIF6 for the β5 chain of αvβ5. Monoclonal antibodies HAS6 and 4B7 that recognize α2 and β1, respectively, were kindly provided by Cancer Research UK, London, UK. Additional markers of osteoclast and monocyte differentiation included mouse monoclonal antibodies anti-human CD11b, CD14 and CD68 (Dako), anti-human CD13 (22A5; Horton et al, 1985) and cathepsin K (a kind gift from Dr M. Gowen, GSK, PA, USA). CD13 and CD68 markers are expressed highly on both osteoclasts and monocyte/macrophages, CD11b and CD14 by monocytes and osteoclast precursors but not mature osteoclasts (Massey & Flanagan, 1999; Lader et al, 2001), and cathepsin K is an enzyme found selectively in osteoclasts but not macrophages (Drake et al, 1996). Occasional reports of expression of α3 and α5 integrins by osteoclasts have appeared (see Helfrich & Horton, 1999; Horton et al, 2002); osteoclasts from two cases were additionally stained for α3 (clone PIB5) and α5 (clone PID6). The secondary antibody was fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (Dako) (green channel in the confocal microscope). Osteoclasts from normal subjects were defined as cells expressing the αvβ3, using monoclonal antibody 23C6, and/or F-actin rings and the presence of associated resorption lacunae (identified as ‘black’ regions in the ‘blue’ reflection image using the confocal microscope; seeFigs 1–3). A minimum of 30 osteoclasts from a minimum of three bone slices were examined in each experiment after staining with each antibody. Images were acquired for the three wavelengths sequentially (using Leica TCS NT confocal microscope software, Leica UK, Milton Keynes, UK) to avoid cross-talk of the red signal into the green channel. Images were collected as serial 1 µm xy slices throughout the cell and its underlying resorption lacuna; these are shown in the figures either as single xy slices or merged to produce the composite through-focus images.
Quantification of integrin expression. The FITC signal (green) for each integrin antibody (αvβ3, αv, β5, α2, β1) for single 1-µm sections was recorded at the level of the bone surface (identified in the reflection image, blue) from osteoclasts identified by their expression of their characteristic F-actin ring (red) (Nesbitt & Horton, 1997; Fig 1). Images for the three wavelengths were collected sequentially from a minimum of 30 randomly identified osteoclasts from at least three bone slices from each case and were analysed blind. Images from patient and relevant control cultures were all acquired using the same laser and photomultiplier tube settings to avoid erroneous variations in image brightness. Post hoc, the green (integrin-stained) images were downloaded, and the total pixel intensity (in arbitrary units) and pixel intensity per µm2 were ascertained for individual osteoclasts using Leica software.
Assessment of bone resorption and αvβ3 and TRAP expression. Devitalized cortical bovine bone slices (4 × 3 × 0·1 mm) were used as substrate for osteoclastic bone resorption, prepared as described previously (McSheehy & Chambers, 1986). At each time point, bone slices from both control and GT cultures were labelled with antibody 23C6, using conventional alkaline phosphatase immunohistochemistry, and tartrate-resistant acid phosphatase (TRAP), using previously described methodology (Sarma & Flanagan, 1996). The presence of qualitative bone resorption and TRAP- and αvβ3-positive cells was evaluated by reflected and transmitted light microscopy respectively (Sarma & Flanagan, 1996). The number and depth (in µm) of resorption pits were also ascertained by confocal reflection microscopy (Leica TCS NT, using × 63 water immersion objective). Pit depth was quantified after identifying the level of the bone surface and the deepest point of an identifiable bone resorption lacuna and reading their height difference with supplied software.
Scanning electron microscopy of bone surfaces. Scanning electron microscopy (SEM) was used to inspect the bone slices more closely for the qualitative nature of the resorption lacunae and for the presence of shallow areas of resorption (McSheehy & Chambers, 1986). The cells were removed from the bone surface by treatment with 0·25 mol/l ammonium hydroxide for 30 min. The bone slices were then washed in water and dehydrated in alcohol, critical point dried, mounted on stubs (Sigma-Aldrich) and sputter coated with gold. The entire surface of at least three bone slices per case was inspected for the presence of resorption events, and their morphology was evaluated by scanning electron microscopy (Cambridge S90, Cambridge Instruments, Cambridge, UK).
Statistical analysis. The results were analysed using paired t-tests, where significance was accepted as P < 0·05. Results are displayed as mean ± standard error of the mean (SEM).
Control and β3 null GT osteoclast cultures
PBMNC from normal subjects, as reported by us previously (Massey & Flanagan, 1999; Lader et al, 2001; Massey et al, 2001; Scopes et al, 2001), and GT caused by αIIb deficiency cultured in the presence of RANKL, M-CSF and TGFβ or IL-4 developed into osteoclasts. Large multinucleated TRAP-positive cells (not shown) formed numerous typical resorption lacunae (SEM features illustrated in Fig 4A and B) as seen with osteoclasts isolated directly from bone; these were associated with dense F-actin rings (Figs 1A2 and 2A2) and areas of resorption identified by confocal reflection microscopy (Figs 1A3 and 2A3). The area of bone surface resorbed by control osteoclasts fell within that observed for historical controls, where this ranged from 10% to 80%. Confocal microscopy showed that control osteoclasts expressed αvβ3 and α2β1 but not αvβ5 integrins (see Figs 1 and 2 and quantified in Table I). The cells were CD13 and CD68 (Fig 3A and E) positive and did not express CD11b (Fig 3C) or CD14 (not shown). The two latter molecules were detected in the mononuclear cells (monocytes) present in the osteoclast cultures. Typically for osteoclasts, they had abundant intracellular cathepsin K (Fig 3G), polarized to the ruffled border adjacent to the bone surface and MMP9 (not shown) enzymes.
Table I. Expression of αvβ3 and α2β1 integrins in β3 null Glanzmann thrombasthenia and normal controls, assessed by confocal microscopy and image analysis.
Cultured β3 null GT osteoclasts shared many of the features of osteoclasts generated from normal healthy individuals. Aside from their altered integrin expression (below; see Figs 1 and 2, Table I), they differentiated into mature, polarized, bone-resorbing cells (although suboptimally vide infraFig 4) (Figs 1B and 2B) with a normal immunophenotype (Fig 3B, D, F and H).
Integrin expression by β3 null GT culture-derived osteoclasts
Normal osteoclasts expressed high levels of αvβ3 and moderate amounts of α2β1 (Figs 1 and 2, quantified in Table I); similar levels to control osteoclasts were seen in those derived from patients with the αIIb deficiency form of GT, in which platelet αvβ3 expression is normal.
In contrast, osteoclasts from β3 null GT lacked αvβ3, detected with the complex specific antibody 23C6 (> 96% reduction; see images in Fig 1B1 versus 1A1 and quantified in Table I). Other αv integrin dimers were not upregulated to compensate for the β3 deficiency – there was no increase in αv chain expression (using αv chain selective antibody, 13C2, that recognizes all αv integrin dimers; Table I), nor was β5 expression (i.e. the level of αvβ5) increased from its normally, relatively low level found in mature osteoclasts (Table I) (Nesbitt et al, 1993). However, a significant,2·4- to 4·3-fold increase in expression of the α2 and β1 integrin chain expression, respectively, was observed (Fig 2B1 versus 2A1, Table I).
We also examined two of the β3 null GT cases for α3 and α5 integrin expression, two integrin subunits that have previously, but not consistently, been reported to be expressed in osteoclasts (see Helfrich & Horton, 1999; Horton et al, 2002). Neither integrin was found in control, culture-derived osteoclasts, nor were they upregulated in GT osteoclasts.
Altered bone resorption in β3 null GT
Bone slices were examined by confocal reflection microscopy (example shown for control culture in Figs 1A3 and 2A3 and β3 null GT in Figs 1B3 and 2B3) and by SEM (Fig 4); these showed that, although resorption was evident in cultures from all cases of β3 null GT, this was abnormal in amount and character. Specifically, a 43·6% decrease in the number of resorption pits was seen in β3 null GT when compared with controls, and these were significantly shallower (59·0%) (quantified in Table II). The confocal data were confirmed by SEM (Fig 4), which showed that resorption lacunae were shallower and less well defined (compare Fig 4D with B); additionally, some areas of minimal resorption (surface erosion) were seen (arrows in Fig 4C). Despite this, there appeared to be no obvious imbalance in resorption of collagenous material versus the mineral component of bone. Normal bone resorption was observed in αIIb deficiency GT (Table II).
Table II. Quantification of bone resorption in β3 null Glanzmann thrombasthenia and normal controls.
Percentage decrease (P-value) (β3 null GT vs. normal control)
43·6 (P < 0·001)
59·0 (P < 0·001)
We have evaluated the impact of β3 integrin deficiency in Iraqi-Jewish patients with GT on osteoclasts. Deficiency of β3, but not of αIIb, leads to absence of the osteoclast integrin, αvβ3, which shares its β-chain with αIIbβ3 of platelets. Other αv integrins, for example αvβ5, which is normally expressed at low levels in mature osteoclasts (Nesbitt et al, 1993), were not upregulated to compensate for lack of αvβ3. In contrast, the integrin α2β1 collagen receptor, normally expressed at an intermediate level in osteoclasts, was upregulated two- to fourfold. Human PBMNC–derived osteoclasts from β3 null patients formed in equivalent numbers to controls – not in increased numbers as seen in vivo in the β3 knockout mouse (McHugh et al, 2000; Feng et al, 2001). Otherwise, they were ‘phenotypically normal’, as shown in Fig 3 (Massey & Flanagan, 1999; Lader et al, 2001; Massey et al, 2001; Scopes et al, 2001). Likewise, β3 null osteoclasts polarized on bone as in controls, judged by the formation of F-actin rings of apparently normal morphology by confocal microscopy. There was some evidence, however, for abnormalities in β3 null osteoclast function. Osteoclastic resorption of bone proceeded at ≈ 50% of control levels, assessed by pit number and depth, and this was corroborated by SEM analysis of resorption lacunae.
Our data on integrin expression are supported, to some extent, by our previous studies on platelets and EBV-transformed lymphocytes in the face of the profound β3 deficiency in GT (Coller et al, 1991; Rosenberg et al, 1998). In both cell types, αvβ3 deficiency accompanied the lack of platelet αIIbβ3. These cells exhibited a decrease in adhesion to vitronectin, suggesting that no compensatory increase in αvβ5 occurred; this contrasted with findings in αIIb deficiency GT where αv expression was increased (Coller et al, 1991; Rosenberg et al, 1998). Normal adhesion to collagen was observed, which is compatible with the retention or increase in α2β1 levels, although this was not tested directly.
The phenotype of two mouse knockout models are pertinent to our understanding of skeletal function in GT. The skeleton develops normally in mice with deletion of αv that survive to full term (Bader et al, 1998), and this model only points to a redundant role for αv in skeletal modelling during development. The β3 knockout skeleton is of ‘normal’ structure at birth, but bone resorption is compromised, and osteosclerosis develops with increasing age (McHugh et al, 2000; Feng et al, 2001). Control levels of β5 integrin were seen on Northern analysis of bone marrow macrophages suggesting, as in our studies, that αvβ5 is not upregulated to compensate for the absence of β3. Other integrins have not been analysed in the mouse model, however, and in contrast to GT osteoclasts, a more marked effect upon cell adhesion and spreading was observed.
What are the implications of our studies, and those in mice, for skeletal function and calcium metabolism in β3 null GT patients? It may be predicted that GT patients would develop an age-dependent osteosclerosis, as in mice lacking β3 integrin. However, a preliminary study of bone mineral density in five Iraqi-Jewish women with GT showed that values for spine and femur were within the normal range (Coller et al, 1994). Thus, these data support the concept that a lifelong, genetic lack of αvβ3 integrin may not be especially deleterious to the skeleton. Despite this, it would be of considerable interest to investigate bone and calcium metabolism in this group of patients in more detail, as it could provide significant information upon the effects of chronic treatment with αvβ3 integrin antagonists in humans.
Does the information provided by our studies inform the clinical development of integrin antagonist drugs? If the findings on integrin expression and osteoclast function in mouse and human β3 gene deletion are also replicated upon chronic exposure to αvβ3 antagonist drugs, then the compensatory response could curtail clinical efficacy. However, such a controlled inhibitory effect in the skeleton might actually be advantageous as complete inhibition of bone turnover is deleterious. Early work with chronic dosing in ovariectomized rodent models (Hartman & Duggan, 2000; Miller et al, 2000) predicted that these issues may not be a major consideration in the clinical development of this new class of bone-active drugs.