Defective osteoblast function in ICAP-1-deficient mice

Cell anchorages to extracellular matrix (ECM) and surrounding cells control shape, migration, survival and proliferation. Integrins are a large family of adhesion receptors that mediate cell-matrix and cellcell interactions(Brakebusch et al., 2002; Hynes, 1992; Hynes, 2002). Integrins are bi-directional signaling molecules, which switch between a low (inactive) and a high affinity (active) state. The switch to the high affinity state is controlled by intracellular signals that act on the cytoplasmic domain of integrins and induce rapid and reversible changes in the conformation of the integrin extracellular domains (inside-out signal)(Calderwood, 2004). Following activation, integrins bind their ligands, merge into large clusters, recruit a multitude of proteins to form so called focal adhesions (FAs), and transmit signals to various subcellular compartments (outside-in signal).

ICAP-1 [integrin cytoplasmic domain associated protein-1; also known as integrin beta 1 binding protein 1 (ITGB1BP1)] is a ubiquitously expressed protein identified in a yeast two-hybrid screen as a β₁integrin cytoplasmic domain-interacting protein(Chang et al., 1997). Human cells express two ICAP-1 isoforms that are generated by alternative splicing. The large isoform associates with the cytoplasmic tail ofβ ₁ integrin. The small isoform lacks a C-terminally located integrin-binding site (Chang et al.,1997) and is therefore unable to interact with β₁integrin. Overexpression of ICAP-1 negatively regulates β₁integrin function by diminishing the strength of adhesion to, and enhancing cell migration on, fibronectin (FN)(Bouvard and Block, 1998; Bouvard et al., 2003; Zhang and Hemler, 1999). How ICAP-1 exerts its functional properties is still unclear. One study proposed direct competition with talin for binding to β₁ integrin(Bouvard et al., 2003). Talin is a large cytoplasmic protein that binds and activates several integrins, and links them to the actin cytoskeleton(Calderwood et al., 2002; Vinogradova et al., 2002). Recruitment of ICAP-1 on β₁ integrin would dislodge talin and thereby reduce the affinity state of β₁ integrins leading to FA disassembly (Bouvard et al.,2003). In line with this hypothesis is the finding that ICAP-1 is absent from FAs. A second study suggests that ICAP-1 might act as a guanine dissociation inhibitor (GDI) for the small GTPases RAC1 and CDC42(Degani et al., 2002). A reduced RAC1 and/or CDC42 activity could also explain the spreading defects of cells overexpressing ICAP-1 (Bouvard et al., 2003; Degani et al.,2002).

Finally, the identification of additional binding partners such as KRIT1 and the nucleotide diphosphate kinase NM23-H2 (also known as NME2)(Fournier et al., 2002; Zawistowski et al., 2002; Zhang et al., 2001) linked ICAP-1 to additional signaling pathways. Loss-of-function mutations in KRIT1 cause a human disease called cerebral cavernous malformation type I (Laberge-le Couteulx et al.,1999), which is characterized by abnormalities of the brain vasculature. KRIT1 has been shown to bind microtubules and the small GTPase RAP1A (Gunel et al., 2002; Serebriiskii et al., 1997). RAP1A can reverse the transformed phenotype of KRAS-overexpressing cells and modulate integrin-mediated cell adhesion on FN(Bos et al., 2001). NM23-H2 is a protein with nucleoside diphosphate kinase activity that has been linked to a variety of cellular activities including the suppression of metastasis and cell motility of tumour cells in vitro. NM23-H2 can bind to the promoter sequences of the PDGFA and c-Myc (MYC - Human Gene Nomenclature Database) genes, modulate the activity of small GTPases such as RAD (also known as RRAD) and RAC1, and localizes in cell ruffles upon integrin ligation (Fournier et al.,2002).

To directly test the function of ICAP-1 in vivo, we generated Icap-1-deficient mice. Most of the mutant mice are born and develop craniofacial dysmorphism and dwarfism caused by abnormal proliferation and differentiation of osteoblasts leading to a delayed closure of calvarial sutures. Furthermore, we show that ICAP-1 regulates β₁integrin activity and the condensation of preosteoblastic cells, an absolute requirement for proper bone development.

Five positive PAC clones were isolated from an RPCI21 library and used to generate the targeting construct (details available on request). Electroporation into passage-13 R1 ES cells was performed as previously described (Talts et al.,1999). Mice were genotyped by Southern blot or by PCR.

Polyclonal anti-ICAP-1 antibodies were described previously(Bouvard and Block, 1998). Monoclonal antibodies against actin, vinculin (clone hVIN1) and talin (clone 8d4) were from Sigma-Aldrich (Germany). Polyclonal anti-β₁integrin serum was a gift from Dr Johansson (The Biomedical Center, Uppsala University, Sweden). Monoclonal β₁ antibodies 9EG7 and MB1.2 were from Pharmingen (France) and a gift from Dr Bosco (Robarts Research Institute, Ontario, Canada), respectively. Polyclonal anti-collagen I, III and osteonectin/BM-40 antibodies were from Dr R. Timpl (Max Planck Institute for Biochemistry, Martinsried, Germany). Polyclonal antibodies against cyclin D1,FGFR1 and FGFR3 were from SantaCruz (USA), against Ki67 from Novocastra (UK)and against 5-bromo-2′-deoxyuridine (BrdU) from Roche (Germany).

A primary mouse osteoblast-enriched cell population was isolated from newborn calvaria using a mixture of 0.3 mg/ml collagenase type I (Sigma) and 0.25% trypsin (Gibco BRL) as described previously(Bellows et al., 1986; Otto et al., 1996). Cells were grown in α-MEM medium containing 10% FCS.

In vitro differentiation of isolated osteoblasts was performed essentially as described (Globus et al.,1998). Briefly, 60,000 cells per well were plated in a 24-well tray. After 3 days of culture, when cells were confluent, the medium was switched to differentiation medium (α-MEM, 10% FBS, 50 μg/ml ascorbic acid, 10 mM β-glycerophosphate) and changed every second day. The differentiation process was visualized by alkaline phosphatase (AP) staining for osteoblast activity and by Alizarin Red S staining for calcium deposition.

For the adhesion assay, primary osteoblasts (passage 2) were seeded at 0.5×10⁵ cells in a 96-well tray coated with various concentrations of FN or COL1. The cells were incubated for 1 hour at 37°C and then washed three times with PBS before staining with a Crystal Violet solution (0.1% Crystal Violet, 20% methanol) for 1 hour at room temperature. After three washes in water, cells were lysed in 0.1% SDS for 1 hour. The absorbance was read at 550 nm with a Beckman Coulter AD 340 absorbance detector.

Cell proliferation was estimated by the BrdU assay as previously described(Fournier et al., 2005).

Primary osteoblasts (passage 2) were infected with a retrovirus expressing the large SV40 T antigen (Fässler et al., 1995), cloned and tested for their ability to induce AP upon differentiation (Mansukhani et al.,2000). Clone SV2.1 from an Icap-1-deficient mouse and clone SV6.5 from a wild-type animal were used. Rescue of ICAP-1 expression in SV2.1 cells was via retroviral infection using the pCLMFG-Icap-IRES-EGFP vector. A homogeneous cell population was sorted based on EGFP fluorescence with a MoFlo cell sorter (Dako Cytomation). ICAP-1 expression was checked by western blot, immunofluorescence and FACS using EGFP as a marker. This non-clonal cell population is referred to as SV2.1-Icap-1^resc hereafter.

Immortalized cells were harvested by trypsin digestion and washed twice in DMEM medium. Drops of 10 μl of DMEM-SVF medium containing 25,000 cells were spotted onto the cover lid of 10 cm Petri dishes, inverted and placed on a Petri dish containing 8 ml of PBS. Spheroid compaction was then followed over a 72-hour incubation period and images taken with a binocular microscope equipped with a digital camera.

Staining of whole-mount embryos with Alcian Blue/Alizarin Red(Aszodi et al., 1998) and X-Gal(Sakai et al., 2001) was carried out as described previously. X-ray images were obtained on a dual energy setup developed at CEA/LETI (Grenoble, France).

Whole-mount in situ hybridization was performed as described(Rice et al., 2000). Histochemistry and immunostaining on tissue sections were carried out as described (Aszodi et al.,1998). In vivo cell proliferation was analyzed using either the BrdU incorporation assay (Aszodi et al.,1998) or Ki67 immunohistochemistry. To detect cells with AP activity, calvarial cryosections were fixed for 10 minutes in 3%paraformaldehyde. After washing in PBS, the color reaction was developed in the BCIP/NBT substrate solution (Roche, France). Immunofluorescence staining and FACS analysis of primary osteoblasts were performed as previously described (Bouvard et al.,2003).

Activation index (AI) of β₁ integrin was estimated essentially as previously described(Calderwood et al., 2004). Briefly, primary osteoblasts were isolated and passage-2 cells were aliquoted into two pools containing either Tyrode's buffer alone or Tyrode's buffer supplemented with 5 mM EDTA. After a 15-minute incubation at 4°C, cells were incubated with or without the FITC labeled Fn7-10 fragment for 45 minutes at 4°C in the presence or absence of 5 mM EDTA, washed in ice-cold Tyrode's and analyzed on a FACScan (Becton Dickinson) flow cytometer. The collected data were analyzed using CellQuest software (Becton Dickinson). In parallel, cells were analyzed for β₁ expression using the MB1.2 monoclonal antibody to detect the level of β₁ integrin on the cell surface. The AI was calculated as follows: each specific mean intensity fluorescence (MFI) was calculated by subtracting the background obtained with Fn7-10 fragment incubation in the presence of EDTA or without the primary antibody in the case of the MB1.2 labeling. AI=[(MFI Fn7-10)-(MFI Fn7-10+EDTA)]/(MFI MB1.2)-(MFI MB1.2 control).

Total RNA was isolated from adult kidney using TRIzol reagent (Gibco BRL)according to the manufacturer's recommendations. For northern analysis, 10μg of total RNA was separated on a 1.2% agarose-2.2 M formaldehyde gel,transferred to Hybond+ membrane (Amersham) and probed with a ³²P-labeled Icap-1 cDNA.

For biochemistry, brains of adult mice were homogenized in RIPA buffer (10%w/v) and used for western blotting as described(Bouvard et al., 1998).

Three PAC clones containing the mouse Icap-1 gene (also known as Itgb1bp1 and Bodenin) were isolated and characterized. The Icap-1 gene spans over 20 kb and comprises 7 exons(Fig. 1A). The transcription initiation and stop codon are located in exons 2 and 7, respectively. In human, two ICAP-1 isoforms have been identified: ICAP-1α corresponding to the full-length protein (200 amino acids) and ICAP-1β representing a shorter, 150-amino acid protein. The short isoform results from alternative splicing of exon 6, which contains the β₁ integrin-binding site (Chang et al., 1997). In mice, we were unable to detect the short isoform using RT-PCR amplification of RNA isolated form different adult tissues (data not shown). Furthermore,searches of EST and UCSC genome annotated databases also did not provide evidence of a mouse Icap-1β. By computer BLAST search, the gene encoding ICAP-1 was localized on mouse chromosome 12 and on human chromosome 2.

To study the in vivo function of ICAP-1, we generated an Icap-1-null allele by homologous recombination. The targeting strategy made use of a lacZ gene inserted in frame with the endogenous ATG and deleted exons 2 and 3 preventing the expression of a functional ICAP-1 protein (Fig. 1). Three correctly targeted embryonic stem (ES) cell clones were used to generate germline chimeric males. The null mutation was confirmed by Southern, northern and western blot analyses(Fig. 1B-D). Neither the Icap-1 mRNA nor the ICAP-1 protein were detected in tissues derived from homozygous mutant (Icap-1^-/-) mice.

To determine the expression pattern of ICAP-1, heterozygous animals(Icap-1^+/-) were collected at various embryonic and adult stages and subjected to β-galactosidase (β-gal) histochemistry(Fig. 1E,F). At embryonic day 8.5 (E8.5), whole-mount staining demonstrated a faint β-gal activity in the developing heart and facial mesenchyme(Fig. 1E). At later stages,β-gal activity became gradually visible all over the embryo with the exception of the liver (Fig. 1F). On tissue sections, only a moderate β-gal activity was observed in liver, spleen, thymus and intestinal epithelial cells, whereas other tissues expressed high β-gal levels (data not shown). These results are in agreement with previously published expression data(Faisst and Gruss, 1998).

Icap-1^+/- mice appeared normal. Southern blot genotyping of newborn mice from heterozygous intercrosses resulted in a normal number of homozygous mutants, suggesting that ICAP-1 has no rate-limiting function until birth. However, when the mendelian ratio of 4-week old litters from Icap-1^+/- × Icap-1^+/- and Icap-1^+/- × Icap-1^-/-intercrosses was evaluated, 20% of Icap-1^-/- mice were missing (Table 1). The reason for this perinatal lethality is unknown.

Homozygous mutant mice were slightly smaller at birth compared with wild-type littermates (Fig. 2A), and this disparity in size increased progressively postnatally. Fourteen days after birth (P14), Icap-1^-/-mice were 5-10% shorter and weighed 20-50% less than control mice. These differences were maintained throughout life(Fig. 2B and data not shown). At around 3 weeks of age, Icap-1^-/- mice developed obvious craniofacial abnormalities characterized by a domed skull, a shortened and broadened snout and bulged eyes (Fig. 2C). X-ray analysis of 5.5-month-old mutant mice revealed several additional abnormalities (Fig. 2D). The vault of the skull was shortened and rounded and the processus spinosus of the vertebrae was poorly ossified(Fig. 2D and inset). In the appendicular skeleton, the long bones had apparently normal bone density but were about 12-23% shorter than in control mice. These morphological defects were observed in outbred (mixed C57Bl6/Sv129J) as well as inbred (CD1 and C57Bl6) lines indicating that they occurred independently of the genetic background.

To analyze cartilaginous and bony tissues, the skeletons of homozygous mutant and wild-type animals were stained at various developmental stages with Alcian Blue (which stains cartilage) and Alizarin Red (which stains calcified tissue). The Alcian Blue staining was indistinguishable between mutant and control embryos throughout development, suggesting that cartilage formation is not grossly affected in Icap-1-deficient mice(Fig. 2E,F). The Alizarin Red staining, however, was reduced in the skeleton of Icap-1^-/- mice as early as E14.5, indicating a defect in ossification. The reduced Alizarin Red staining was most pronounced in the parietal and frontal bones of the calvaria and in the maxillary and mandibular components of the facial skeleton (not shown). At E16.5, the reduction in Alizarin Red staining became even more prominent in the skull and in the bony collar surrounding the long bones of the appendicular skeleton(Fig. 2E).

At the newborn stage, Alizarin Red staining of long bones was similar between control and Icap-1^-/- animals(Fig. 2F), suggesting that the ossification of the collar started late in Icap-1^-/- mice,but was catching up at later stages. In the skull region, the postnatal development of the chondrocranium (the majority of the base of the skull) was normal in Icap-1^-/- mice. This was shown by the normal ossification of the exoccipital, basioccipital, basisphenoid and presphenoid bones (Fig. 3A) and the normal formation and development of the synchondrosis of the skull base in newborn,P15, P21 and P60 control Icap-1^-/- animals(Fig. 3A and data not shown). The ossification defect of calvarial bones, however, was obvious in newborn Icap-1^-/- mice (Fig. 3B). The frontal, parietal and interparietal (supraoccipital)bones were reduced in size, giving rise to enlarged anterior and posterior fontanelles and to widened sagittal and metopic (interfrontal) sutures(Fig. 3B). At the age of 2 months, control mice had completed the ossification of the metopic sutures but still had patent lambdoid, sagittal and coronal sutures(Fig. 3C,F). In mutant mice of the same age, the posterior part of the metopic suture was still unossified(open) (Fig. 3D,E,G,H). In some Icap-1^-/- mice, non-ossified, Alcian Blue-positive areas were observed extending from the posterior metopic suture to the frontal bones(Fig. 3E,H). Furthermore, the parietal bones were hypoplastic in mutant calvarias leading to shortened sagittal and V-shaped coronal sutures. The bone defect was also evident in other parts of the skeleton, such as in the vertebrae, where 15-day-old mutant mice showed non-fused vertebral arches(Fig. 3I) and in the pelvic bone, where mutant mice showed delayed fusion of the pubis and the ischial bone (Fig. 3J).

Since the skull vault is severely affected in Icap-1^-/-mice, we decided to focus our analysis on the development of the calvaria. To test where ICAP-1 is expressed in the developing skull bones, we performedβ-gal staining (Fig. 4A). Whole-mount staining of mutant calvariae revealed strong lacZexpression at the leading edges of the calvarial bones and in the suture regions (Fig. 4A). Frontal sections through the parietal region of the skull showed that the paired parietal bones from normal mice were separated at the midline by a narrow sagittal suture (Fig. 4B). However, in mutant mice the distance between the two parietal bones was wider without the typical suture organization(Fig. 4C). The sagittal suture is a fiberrich stripe of mesenchyme flanked by condensed bulges of tissue called osteogenic fronts that contain the osteoprogenitors that differentiate into osteoblasts and lay down bone (Fig. 4D). In mutant mice, the osteogenic fronts were thin and contained very few cells resulting in impaired growth of the parietal bones and widening of the intervening mesenchyme (Fig. 4C,E). β-gal staining revealed strong Icap-1promoter activity in mature osteoblasts of the parietal bone, in osteoprogenitors of the osteogenic fronts, and in mesenchymal cells between the two fronts (Fig. 4G). As expected, comparable tissue sections from wild-type mice showed no lacZ expression (Fig. 4F).

The morphogenesis of the calvaria depends on the number of osteoprogenitors and their differentiation into osteoblasts at the margins of the suture. To test whether the reduced number of osteoprogenitors in mutant osteogenic fronts was caused by aberrant cell survival and/or proliferation we performed apoptosis assays, BrdU incorporation assays and Ki67 immunostaining in newborn mice. The TUNEL assay and staining of activated caspase 3 showed very few apoptotic cells and similar in number for normal and mutant osteogenic fronts(data not shown). The labeling indexes for Ki67 and BrdU, however, were reduced in mutants by 40% and 64%, respectively, indicating a pronounced proliferation defect in the Icap-1-deficient osteoprogenitor cell population (Fig. 5A,B). Proliferation of osteoblast progenitor cells is tightly regulated by ECM interactions and growth factors. In order to determine whether the diminished proliferation at the osteogenic fronts are cell autonomous defects or caused by impaired secretion of growth factors and/or ECM components by the surrounding tissue, we isolated primary osteoblasts from the calvaria of control and mutant mice, immortalized them by retroviral transduction of the SV40 large T antigen, and tested their proliferation rates. Similar to the in vivo observations, Icap-1^-/- cells(SV2.1-Icap-1^-/-) displayed a lower BrdU incorporation rate compared with control cells (SV6.5-Icap-1^+/+) in vitro (Fig. 5C). Retroviral infection of the full-length Icap-1 cDNA into SV2.1-Icap-1^-/- osteoblasts(SV2.1-Icap-1^resc) restored normal proliferation(Fig. 5C), corroborating the link between ICAP-1 loss and the proliferation defect. To further characterize the proliferation defect at the molecular level, we performed western blot analysis of cyclin D1 expression. Adhesion of wild-type or rescued osteoblasts on FN led to increased cyclin D1 expression 5 hours after seeding. Conversely, Icap-1^-/- preosteoblasts cultured under identical experimental conditions showed a significant reduction in cyclin D1 expression(Fig. 5D). These results indicate that the decreased (pre)osteoblast proliferation observed in Icap-1^-/- mice is cell autonomous.

The differentiation of osteoprogenitors into mature osteoblasts is characterized by the deposition of `bone-specific' ECM molecules, as well as the spatially and temporally coordinated expression of growth and transcription factors. To study the cranial skeletogenic differentiation in Icap-1^-/- mice, we first evaluated the expression of osteogenic markers such as alkaline phosphatase (AP), osteonectin (SPARC -Mouse Genome Informatics) and collagen type I (COL1) on frontal sections of newborn calvariae (Fig. 6). In the wild type, AP activity was confined to the osteoid region of the developing parietal bones and to preosteoblasts at the osteogenic front of the sagittal suture (Fig. 6A). Osteonectin expression overlapped with sites of AP activity(Fig. 6B,C) and was strong along the bone surface and weak in cells at the osteogenic front. In mutant tissue, the cell number with AP activity was reduced at the bone-suture margins, defining a smaller osteogenic front area committed to osteoblastic differentiation (Fig. 6A). Osteonectin deposition was also reduced at the osteoid surface and was almost absent at the osteogenic front (Fig. 6B,C). Similarly, COL1 immunostaining was markedly reduced on the bone surface and at the suture margin in the mutant compared with the wild type (Fig. 6D,E).

Since the expression patterns of FGF-receptors correlate with the osteogenic differentiation process (Iseki et al., 1999; Rice et al.,2000), we compared the expression of FGFR1 and FGFR3 in wild-type and mutant newborn calvariae. In control tissue, both proteins were detected in the osteoblasts of the calvarial bones and in cells of the osteogenic front and weakly in the sutural mesenchyme (Fig. 6F,G). In Icap-1^-/- calvariae, FGFR1 and FGFR3 immunolabeling was fainter than in wild type, and this difference was particularly pronounced at the osteogenic fronts(Fig. 6F,G). In addition,whole-mount in situ hybridization on E17.5 heads revealed that the expression of both the early bone marker Runx2 (Cbfa1) and the later marker bone sialic protein (Bsp; Ibsp - Mouse Genome Informatics) were reduced in mutant mice(Fig. 6H,I). Again, a weaker signal was observed at the edge of the bony region, reflecting the marked reduction of cells committed into the osteoblastic lineage within the osteogenic front. Altogether, these results indicate that the osteogenic front is not normally formed in mutant animals.

To test whether committed cells differentiate normally, primary osteoblasts were isolated from newborn wild-type and Icap-1^-/-calvariae and incubated in medium supplemented with ascorbic acid andβ-glycerophosphate to induce osteoblast differentiation and the formation of mineralized bone nodules (Fig. 7A). After 2 weeks in the differentiation medium, AP activity was evident in almost all cells derived from control calvariae indicating their commitment to the osteoblast lineage. Icap-1^-/- calvarial cells also started to express AP, albeit at lower levels (data not shown). After 4 weeks of culture, differentiating osteoblasts from Icap-1^-/- calvariae contained fewer and smaller mineralized nodules as visualized by Alizarin Red(Fig. 7A) and von Kossa staining (data not shown). Similarly, immortalized Icap-1^-/- cells (SV2.1-Icap-1^-/-)showed markedly reduced AP staining and mineralized nodule formation relative to wild-type (SV6.5-Icap-1^+/+) and rescued cells(SV2.1-Icap-1^resc)(Fig. 7B and data not shown). These data show that ICAP-1 loss impairs osteogenesis and identifies a role of ICAP-1 as a cell-autonomous factor in osteoblast differentiation.

Since we routinely induced differentiation after cells derived from calvariae or immortalized osteoprogenitors had reached confluence in vitro,the numbers of neither the mutant (SV2.1-Icap-1^-/-) nor the wild-type (SV6.5-Icap-1^+/+) or rescued cells(SV2.1-Icap-1^resc) significantly increased during the differentiation period (data not shown). Altogether, these findings suggest that the differentiation block occurs in addition to the cell proliferation defect.

β₁ integrins have been proposed to play a crucial role during osteoblast proliferation and differentiation(Moursi et al., 1997; Zimmerman et al., 2000). Since ICAP-1 interacts with the cytoplasmic tail of the β₁ integrin chain and modulates integrin function in vitro(Bouvard et al., 2003), we analyzed β₁ integrin expression in wild-type and mutant calvariae in vivo. At newborn stage, frontal sections through the sagittal suture and the parietal bones were immunostained with an anti-β₁ polyclonal antibody and with the monoclonal antibody 9EG7, which recognizes the ligand-bound form of β₁ integrins(Fig. 8A and data not shown). In sections from control mice, both antibodies strongly labeled the osteogenic fronts and the bone surfaces, whereas the intervening mesenchyme was faintly labeled. Conversely, sections from Icap-1-deficient mice showed clearβ ₁ integrin staining of the bone surface but very faint staining of cells at the osteogenic front. This apparently diminishedβ ₁ staining was mainly due to the reduced osteogenic front population, rather than reduced expression on the cell surface of individual cells, because FACS analysis showed only a slight reduction ofβ ₁ integrin expression on Icap-1^-/-osteoblasts (Fig. 8B, see below).

To further analyze the consequence of ICAP-1 expression loss for the adhesion of osteoblasts, assays were carried out using primary osteoblasts isolated from calvarial tissues. Despite a small decrease inβ ₁ expression (Fig. 8B), adhesion of Icap-1^-/- osteoblasts to FN or COL1 was moderately but significantly increased compared with wild-type osteoblasts (Fig. 8C). This suggests that increased adhesion to ECM substrates resulted from the activation, rather than an increased cell surface expression, ofβ ₁ integrins on Icap-1^-/- osteoblasts.

To investigate whether the loss of ICAP-1 expression interferes with integrin activation, we estimated the ligand-binding affinity of the FN receptor α₅β₁ integrin both in wild-type and Icap-1^-/- primary osteoblasts. The cell-binding domain of FN, corresponding to the type III repeats 7-10 (Fn7-10), was expressed,purified and FITC-labeled. The capability of both mutant and wild-type primary osteoblast cells to interact with Fn7-10 at a non-saturating concentration was analyzed by FACS. As shown in Fig. 8D, we consistently observed an increase in Fn7-10 binding to Icap-1^-/- osteoblasts compared with wild-type cells. The activation index, which normalizes the specific binding of FITC-Fn7-10 to the total β₁ surface expression level, was elevated by approximately three-fold in Icap-1^-/- as compared with wild-type cells (Fig. 8E).

During in vitro differentiation of both primary and immortalized osteoblasts, we consistently observed that the bone nodules formed by Icap-1^-/- cells were fewer, smaller and less compact than those formed by wild-type cells (Fig. 9A). As differentiation of bone cells requires an initial step of cell condensation (Globus et al.,1998; Ornitz and Marie,2002), a defect in this step might lead to altered or delayed differentiation. To address this question, we cultured immortalized osteoprogenitors in suspension using the hanging drop technique. Under these conditions, wild-type or rescued cells aggregated and formed compact spheroids within 48 hours, whereas spheroids formed by Icap-1^-/-cells were less compacted (Fig. 9B and data not shown).

Since ICAP-1 loss increases integrin affinity(Fig. 8), we investigated whether blocking of β₁ integrins in their activated state would mimic ICAP-1 deficiency. To this end, we complemented the culture medium with the integrin-activating monoclonal antibody 9EG7 and then formed spheroids (Fig. 9B). Whereas spheroidtriggered compaction of Icap-1^-/- cells was not affected by treatment with the 9EG7 antibody, compaction of control or rescued cells was consistently delayed. These findings indicate that modulation of integrin affinity by ICAP-1 is required for proper compaction of osteoblastic cells.

We report the molecular analysis of mice carrying a disrupted Icap-1 gene. Icap-1-deficient mice suffer from defective osteogenesis characterized by mild growth retardation and severe craniofacial dysmorphism. We focused our analyses on the calvarial abnormalities to obtain a clear view of osteoblast differentiation.

Bone formation (osteogenesis) involves the conversion of mesenchymal tissue into bone, either directly (intramembranous ossification) or via cartilaginous intermediates (endochondral ossification)(Zelzer and Olsen, 2003). The first mechanism is responsible for the generation of the flat bones of the skull vault as well as for the formation of the bony collar around the diaphysis of the long bones. In Icap-1^-/- mice,intramembranous ossification is affected, as both periosteal and calvarial osteogenesis are delayed. Alcian Blue staining revealed that the formation of cartilaginous templates of endochondral bones is grossly undisturbed in mutant mice. Since ICAP-1 is also expressed in chondrocytes (data not shown), we cannot fully exclude mild differentiation and/or proliferation defects of mutant chondrocytes, which might also contribute to the small stature of Icap-1^-/- mice. We focused our investigations on intramembranous ossification and show for the first time the important role of ICAP-1 during osteoblast differentiation. The ossification of calvarial bones(frontal, parietal and interfrontal) was dramatically delayed during embryonic development, leading to open fontanels and wide sutures at birth. These symptoms mirror the calvarial abnormalities of mice carrying mutations for the transcription factors Runx2 (Otto et al., 1997) and Atf4(Yang et al., 2004), which have been identified as key regulators of osteoblast differentiation. The remarkable phenotypic similarities between the Icap-1^-/-mice and these mouse models suggest that ICAP-1 is a newly recognized and important regulator of osteoblast development. Cranial sutures regulate bone expansion during postnatal life and remain non-ossified as long as the brain is growing (Opperman, 2000). Osteogenesis at the suture involves the differentiation of mesenchymal cells into preosteoblasts and into osteoblasts at the sutural margins and the subsequent deposition of a collagenous matrix along the bony plates. Increased growth of the calvarial bones leads to premature closure of the sutures(craniosynostoses), whereas delayed bone growth results in suture latency. ICAP-1 is expressed at the osteogenic fronts and its deficiency leads to a dramatic reduction in the osteoblast population at the osteogenic fronts,which in turn prevents correct temporal fusion of cranial sutures. The reduced osteoblast number at the osteogenic fronts is due to reduced proliferation and impaired osteoblast differentiation.

We have previously reported that ICAP-1 has a dual localization and is found in the cytoplasm/membrane and in the nucleus(Fournier et al., 2005). Loss of ICAP-1 expression in immortalized osteoblasts considerably reduced cell proliferation and cyclin D1 expression. Therefore, the reduced proliferation in the osteogenic front is likely to result from a lack of ICAP-1 in the nucleus, which would lead to reduced cyclin D1 expression. This might explain the severe decrease in the cell population committed to the osteoblast lineage in the sutural region that we observed in the Icap-1-deficient animals.

Isolated Icap-1-null primary osteoblasts displayed an increased cell adhesion to FN and COL1. We have previously shown that overexpression of ICAP-1 in HeLa cells disrupts focal adhesion, probably by inhibiting the association of the cytoplasmic tail of β₁ integrin with talin(Bouvard et al., 2003). It has been shown that talin binding to the β integrin domain is a key step in the regulation of integrin activation(Calderwood et al., 2004; Calderwood et al., 2002). In good agreement with this, we found an increase in Fn7-10 binding to integrin in mutant cells that was not accompanied by any upregulation ofβ ₁ integrin expression. This confirms that ICAP-1 regulatesβ ₁ integrin affinity.

Osteoblast differentiation is a multistep process that first required an initial condensation of the mesenchymal cells to form the osteogenic front(Fig. 10)(Hall and Miyake, 2000). The cell population at the osteogenic front is visible during calvaria bone development at the edge of the expanding bone within the intervening mesenchyme. This condensed cell population further differentiates and expresses different osteoblastic markers including RUNX2, AP and BSP. Our knowledge of how integrins or cell adhesion is implicated into this process is poorly documented, but recent reports suggested an unexpected role for matrix stiffness in controlling early osteoblast differentiation(Engler et al., 2006; McBeath et al., 2004).

It has also been reported that β₁ integrins play an important role in osteoblast differentiation in vitro. It was shown that COL1 interaction with α₂β₁ integrin regulates osteoblast-specific gene expression and osteoblast differentiation(Takeuchi et al., 1996; Xiao et al., 1998). Similarly,α ₅β₁ integrin interaction with FN promotes differentiation (Moursi et al.,1997) and survival (Globus et al., 1998) of cultured calvarial osteoblasts. Finally, transgenic overexpression of a dominant-negative form of β₁ integrin in osteocytes blocks β₁ integrin function, reduces adhesion to collagen and FN, and diminishes bone mass in cortical and flat bones(Zimmerman et al., 2000). Although these studies identify an important role for β₁integrins for osteogenesis, they do not address the role of integrin affinity modulation in bone development. Our data demonstrate for the first time that not only β₁ integrin affinity increases, but also affinity decreases, are of paramount importance for proper bone development both in vivo and in vitro.

We consistently observed a marked reduction of osteoblast cells in Icap-1-deficient animals. Even the expression of the very early marker RUNX2 at the osteogenic front region was significantly reduced. Furthermore, the in vitro differentiation assays revealed that the bone nodules formed by Icap-1-null immortalized osteoprogenitors were smaller and less compacted as compared with nodules formed from wild-type or rescued cells. This reduced compaction of Icap-1-deficient cells was further confirmed by the hanging drop technique, suggesting that ICAP-1 is required for osteoblast condensation, which is a crucial and early step during osteoblast differentiation.

Our data suggest that the proliferation and the differentiation defects occur independently and that they both contribute to the abnormal osteogenesis. In support for this notion, we observed that limited ICAP-1 expression in Icap-1-deficient preosteoblast cell lines fully restored their proliferation rate, but only partially their potential to differentiate and form nodules in vitro. Furthermore, in our differentiation assay, we routinely use confluent cells to rule out the possibility that the proliferation defect is influencing the formation of mineralized nodules. Indeed, we observed only a slight increase in cells during the differentiation period and this increase was similar in control and Icap-1-deficient cultures. Our data indicate that the condensation defect of the Icap-1-deficient preosteoblasts further limits the number of progenitors that will finally differentiate into mature osteoblasts(Fig. 9). The functional importance of the condensation of progenitors for osteoblast development has also been observed in connexin 43 (GJA1)-deficient mice(Lecanda et al., 2000).

Previous work implicated β₁ integrins in cell compaction(Robinson et al., 2004; Robinson et al., 2003). Our data suggest that ICAP-1 might control β₁ integrin function during this process by regulating activation/deactivation cycles. To our knowledge this is the first direct evidence reporting that integrin affinity is important for cell cohesion and differentiation.

We thank Drs R. Timpl, S. Johannsonn, B. Nieswandt, C. Bosco and G. Karsenty for generously providing antibodies and in situ probes; C. Robert-Coutant and J. M. Dinten for X-ray imaging of mice; V. Collin for cell sorting and P. Marie, M. Pfaff and D. Pearton for valuable discussions and critical reading of the manuscript. D.B. was supported by an E.C. Marie Curie longterm fellowship (QLGA-CT-2000-52076), the Association pour la Recherche contre le Cancer (ARC), the CNRS and the Max Planck Society. R.F. is supported by the DFG, BMBF and the Max Planck Society. This work is dedicated to the memory of Rupert Timpl, Günter Kostka, Martin Pfaff and Christine Robert-Coutant.