Osteopenia, decreased bone formation and impaired osteoblast development in Sox4 heterozygous mice

Bone mineral density (BMD) is considered an important determinant for bone strength and fracture risk in human bone disorders. Bones are renewed throughout life through bone remodeling, i.e. synthesis of bone matrix by osteoblasts (of mesenchymal origin) and bone resorption by osteoclasts (of hematopoietic origin) (Jilka, 2003). In osteoporosis, perturbations in local cytokines, growth factors, systemic hormones and transcription factors cause imbalance between bone formation and resorption, resulting in net bone loss (Karsenty, 1999). Peak bone mass, the result of net bone accrual in early life and with a maximum in the second and third decade of life, is greatly influenced by genetic factors (50-85%) (Giguere and Rousseau, 2000; Ralston, 2002; Nguyen et al., 2003; Ralston and de Crombrugghe, 2006). BMD depends on peak bone mass, but also on the rate of bone loss in later life. Although polymorphisms of several genes associated with reduced bone mass and increased fracture risk have been identified, e.g. bone morphogenetic protein-2 (BMP-2), LRP5 and Collagen I (Grant et al., 1996; Johnson et al., 2004; Mundy, 2006; Ralston, 2002), the mechanisms controlling the variation of bone loss rate are mostly unknown.

In skeletal tissue, mesenchymal bone cell development from a common progenitor cell (Pittenger et al., 1999) is under transcriptional regulation: Runx2 and osterix (Osx) (Ducy, 2000; Nakashima et al., 2002) are crucial for normal osteogenesis, whereas Sox5, Sox6 and Sox9 (Lefebvre et al., 1998) and PPARγ ligands (Lecka-Czernik et al., 2002) influence chondrocyte and adipocyte development, respectively. Several transcription factors in the Sox family [related to sex-determining region Y (SRY) proteins] are involved in skeletal development in addition to their roles in developmental processes in other tissues (Schilham et al., 1996; Hong and Saint-Jeannet, 2005). Sox4 contains the Sox family characteristic high mobility group (HMG) box, and is highly conserved in human, mouse, chicken (Maschhoff et al., 2003) and fish (Hett and Ludwig, 2005; Mavropoulos et al., 2005), and expressed in brain, gonads, lung, heart and thymus (Schilham et al., 1996). It is implicated in lymphocyte differentiation (Schilham et al., 1997; van de Wetering et al., 1993), cancer (Liu et al., 2006; Pramoonjago et al., 2006) and apoptosis (Ahn et al., 2002; Hur et al., 2004; Pramoonjago et al., 2006).

Our group first reported expression of Sox4 in skeletal tissue; Sox4 mRNA is highly expressed in normal and clonal osteoblasts of human and rat origin, where it is stimulated by parathyroid hormone (PTH), and the transcript is predominantly localized in hypertrophic chondrocytes in developing mouse hindlimbs (Reppe et al., 2000). We recently demonstrated elevated SOX4 mRNA levels in bone biopsies from patients with active primary hyperparathyroidism compared with levels after successful surgery and PTH normalization (Reppe et al., 2006). Sekiya et al. demonstrated that induction of chondrogenesis in human bone marrow stromal cells led to a transient, eightfold stimulation of Sox4, preceding the upregulation of Sox5, Sox6 and Sox9 mRNAs (Sekiya et al., 2002). Sox5, Sox6 and Sox9 regulate chondrocyte differentiation (Lefebvre et al., 1998), and Sox8 seems to regulate osteoblast differentiation through Runx2 inhibition (Schmidt et al., 2005).

In the present study we have examined the role of Sox4 in postnatal bone development in mice, with the hypothesis that the Sox4 gene affects normal bone formation. Preliminary investigations indicated that healthy Sox4^+/– mice developed osteopenia (Nissen-Meyer et al., 2004; Nissen-Meyer et al., 2005). Because ablation of both copies of the Sox4 gene in mice leads to circulatory failure in utero (Schilham et al., 1996), we designed a longitudinal study of Sox4^+/– mice and age- and gender-matched wild-type (WT) littermates, examining functional parameters of bone metabolism, histomorphometry, microcomputed tomography (μCT) and bone mineral densitometry related to age. To explore molecular and cellular mechanisms responsible for the observed osteopenia, we used primary cultures of differentiating Sox4^+/– and WT calvarial osteoblasts to assess proliferation, differentiation and mineralization properties. In addition, we studied whether silencing of Sox4 in WT osteoblasts using small interfering RNA (siRNA) treatment could mimic the haploinsufficient phenotype of Sox4^+/–-derived osteoblasts. Our data indicate an important role for Sox4 in the regulation of bone formation and homeostasis.

Sox4^+/– offspring was morphologically indistinguishable from WT, with an overall normal life span. However, rectal prolapses were more frequently observed in the Sox4^+/– mice than in the WT; they also had somewhat reduced fertility. Internal organs and the skeleton were examined at time of sacrifice by an experienced pathologist (FPR) and apart from abdominal inclusion cysts in some females, no macroscopical differences were found between Sox4^+/– and WT controls.

Sox4^+/– mice were only slightly smaller than their littermates, as determined by total body mass (Table 1). Three-month-old Sox4^+/– mice had significantly shorter femurs than WT (–4.6% for males, P<0.001; similar for females), whereas the differences were non-significant for older mice (not shown).

We used repeated DXA (dual-energy X-ray absorptiometry) measurements to evaluate the age- and gender-related bone mineral content (BMC; Fig. 1A) and BMD (Table 1) in Sox4^+/– and WT mice. From 7 weeks to 12 months of age, male and female Sox4^+/– mice showed a significantly lower total bone area and whole body BMC and BMD (P<0.001 for all analyses of repeated measurements) compared with their WT littermates (Fig. 1A, Table 1). When specific regions of interest (ROI) were studied, Sox4^+/– mice had significantly lower BMD compared with WT in both the axial (spine) and appendicular (trochanter femoris and tibia) skeleton (Table 1). Overall, the impact of Sox4 haploinsufficiency on bone mass was manifest already by 10 weeks of age, and the differences remained throughout the observation period (corrected for bone area and body mass in both genders). There were no significant differences in body fat throughout the observation period (Table 1).

Three-dimensional μCT analyses of femoral bones showed alterations of cortical and trabecular bone microarchitecture in Sox4^+/– mice (Fig. 1B). By 3 months of age, cortical bone volume and cortical thickness at the femoral diaphysis were significantly lower in both male and female Sox4^+/– mice compared with WT (Fig. 1C and Table 2). Moreover, male Sox4^+/– mice showed significantly reduced total and medullary volume of cortical bone, indicating a decreased femoral bone size (diameter). These differences were maintained at 12 and 18 months of age (data not shown). At the distal femoral metaphysis, trabecular thickness was decreased and trabecular number slightly increased in 3-month-old Sox4^+/– mice compared with WT (Fig. 1C and Table 2). Trabecular bone volume fraction (BV/TV) was decreased in males (Fig. 1C), whereas no differences in connectivity were seen between WT and Sox4^+/– mice (Table 2). Altogether, the pattern of bone microarchitecture in heterozygous mice suggested a reduced bone formation during growth.

Data obtained by static histomorphometry of proximal tibiae from 3-month-old mice were similar to the μCT data from distal femur, showing lower values for trabecular BV/TV (%) and thickness in Sox4^+/– mice versus WT (data not shown). Osteoid volume/bone volume, osteoblast surface/bone surface (BS) and growth plate width were all significantly lower in tibiae from Sox4^+/– mice versus WT (Table 3). The zone of hypertrophic chondrocytes and the osteoclast surface:BS ratio were not significantly changed, but for both there was a trend towards reduction (Table 3). At 6 months of age, histomorphometric analyses of tibial bones from both genders corresponded to μCT data at this age, showing no significant differences between the WT and Sox4^+/– mice for the parameters mentioned (not shown).

Dynamic double fluorochrome labeling (using calcein and Alizarin Red) of bone accretion showed a marked (>50%) reduction of mineral apposition rate (MAR) in 3-month-old Sox4^+/– compared with WT mice (Fig. 1D, Table 3), consistent with the reductions in bone formation rate (BFR; Fig. 1E) and mineralized surface/BS (Table 3). Taken together, these results suggested that reduced bone mass in Sox4^+/– mice was because of a defect in bone formation rather than in bone resorption.

To evaluate the influence of Sox4 haploinsufficiency on cortical bone mechanical properties, the area moment of inertia (Ix) was derived from the geometrical measurements obtained by μCT, and the structural (ultimate force and stiffness) and material (Young's elastic modulus and ultimate stress) properties related to fracture resistance were directly evaluated by three-point bending of femurs. As shown in Table 4 and Fig. 1F, Ix, ultimate force and stiffness of femurs were significantly lower in male and female Sox4^+/– compared with WT mice. However, Young's elastic modulus and ultimate stress were similar in male and female Sox4^+/– and WT mice, indicating that the decreased bone strength in Sox4^+/– mice was due primarily to the smaller diameter of their diaphyseal bone.

The serum levels of the bone formation marker osteocalcin (OCN) were moderately reduced in 3-month-old female Sox4^+/– mice (165.3±24.0 ng/ml; mean ± s.d.) compared with WT (182.1±49.8 ng/ml; n=5 in both groups), but alkaline phosphatase (ALP) activity was not (132.9±9.6 mU/ml serum in Sox4^+/– versus 114.8±6.5 mU/ml serum in WT; P<0.01; n=5 and 7, respectively). Serum levels of PTH and total calcium measured at 6 and 12 months were similar in the WT and Sox4^+/– mice (data not shown).

The effect of Sox4 on osteoblast development and function was evaluated in primary calvarial cell cultures derived from Sox4^+/– and WT mice. Real-time PCR analysis showed 42% reduction in levels of Sox4 mRNA in Sox4^+/– osteoblasts versus WT (Fig. 2A). By comparison, bone tissue from 3-month-old male Sox4^+/– mice also expressed lower levels of Sox4 mRNA (69%; P=0.056) relative to age- and gender-matched WT mice. Similarly, the mRNA levels for Osx, OCN, collagen type I A2 (Coll1A2) and ALP were significantly lower in the Sox4^+/– osteoblast cultures (P<0.05), whereas Runx2, BMP-2, osteopontin (OPN), PTH/PTHrP-receptor-1 (PTHR1), PTH-related peptide (PTHrP) and c-fos were unchanged (Fig. 2A). Osteoclast-regulating cytokines produced by primary osteoblast cultures were largely unaffected (Table 5), as was production of the adipocyte markers PPARγ and AP-2, and the estrogen receptor ERα (data not shown). A corresponding decline in Sox4 protein levels was observed in Sox4^+/– osteoblasts relative to WT, as assessed by western blot analysis (Fig. 2B), whereas ALP activity and the number of mineralized bone nodules (by von Kossa staining) were reduced by approximately 40 and 60%, respectively (Fig. 2C-F). The total cell number in Sox4^+/– calvarial cultures was markedly lower compared with WT (52±13 and 88±23, P<0.01; n=3). However, the relative fraction of ALP-positive cells was even further reduced (50±13% and 81±8% in Sox4^+/– and WT osteoblasts, respectively; P<0.001). Moreover, we assessed the proliferative capacity of Sox4^+/– osteoblasts by thymidine incorporation, and found that it was severely impaired, being only 27% relative to WT (Fig. 2G). Taken together, these data demonstrate Sox4^+/– osteoblast insufficiency in vitro.

No differences in typical osteoclastogenic properties were detected, as judged by the number of tartrate-resistant acid phosphatase (TRAcP)-positive multinucleated cells (96±33% in Sox4^+/– cells versus WT; n=4) and the ability to erode bone in vitro, shown as functional pit index assay (119±35.2% in Sox4^+/– cells versus WT; n=4). WT osteoclastic cells cultured in the presence of conditioned cell media from Sox4^+/– osteoblasts showed unchanged light microscopic morphology, phenotype and functional activity (not shown).

To study the effect of knockdown of Sox4, we treated primary calvarial osteoblasts from WT mice with siRNA against Sox4 or with scrambled siRNA as control. Real-time PCR demonstrated a decline in levels of Sox4 mRNA by 63% relative to control cells (Fig. 3A). Significant reductions in mRNA levels were found also for Osx, OCN, Coll1A2, PTHR1, PTHrP (29-52%) and ALP (75%), whereas Runx2 and OPN were not affected (Fig. 3A). Cell proliferation analysis replicated previous results in Sox4^+/– osteoblasts: Sox4 siRNA reduced [³H]thymidine incorporation by 54% compared with WT (Fig. 3B). Furthermore, the number of cells staining positive for ALP was reduced by 31% and ALP enzymatic activity by 33% (Fig. 3C-E). Similar to the Sox4^+/– osteoblast cultures described above, Sox4 siRNA treatment caused lower cell numbers (41±1.5 versus 61±6.3), and relative reductions in the number of ALP-positive cells (44±7% in Sox4 siRNA-treated cultures versus 64±14% in the control cultures, P<0.05). We did not find evidence that treatment of osteoblasts with Sox4 siRNA induced apoptosis (data not shown).

Our data showing reduced expression of two typical Runx2-dependent genes, Osx and OCN, both in Sox4^+/– and Sox4 siRNA-treated osteoblasts, might suggest that Sox4 and Runx2 are part of the same molecular pathway involved in regulating osteoblast development and/or function. Thus, gene silencing of Runx2 by siRNA technology was used to assess whether Sox4 is affected by decreased Runx2 levels in mouse primary calvarial osteoblasts. We found no effect of Runx2 knockdown on Sox4 mRNA expression, whereas Runx2 and OCN mRNA levels were downregulated by 65% and 62%, respectively (Fig. 3F). These results suggest that Sox4 action on osteoblastogenesis is independent of the Runx2 pathway (Fig. 4).

Sox4 belongs to a family of HMG box transcription factors regulating numerous developmental processes. In this study we have characterized the skeletal phenotype of Sox4^+/– mice and the functional role of Sox4 in osteoblasts. Our results establish that Sox4 is important for normal postnatal bone development, through effects on osteoblast development or function.

Even though Sox4^+/– mice thrived and exhibited normal gross anatomy from birth, bone mass indices were significantly reduced (compared with WT littermates) already by 7-10 weeks of age. Histomorphological analyses showed reduced thickness of cortical and trabecular bone, in addition to markedly lower (>50%) BFR and MAR, in 3-month-old Sox4^+/– mice versus WT. Importantly, biomechanical bone strength (fracture resistance) in femur was also affected, most likely because of a smaller diaphyseal diameter. In light of the striking reductions in MAR and other histomorphometric parameters (Table 3), we cannot exclude the possibility that the relatively modest quantitative effects of Sox4 haploinsufficiency on bone mass parameters in vivo may involve compensatory mechanisms. However, dynamic parameters for bone formation velocity at a given time (MAR, BFR) are not directly comparable to BMC, which integrates net mineralized bone tissue formed over the lifetime of the animal. Nevertheless, our results clearly demonstrate that Sox4^+/– mice exhibit an osteopenic phenotype, including impaired bone structure and biomechanical strength, mainly caused by suppressed bone formation and/or mineralization in early life.

The differences between Sox4^+/– and WT mice appear most pronounced at 3 months of age; in addition to bone mass, the differences in femoral length and bone structure (by μCT and static histomorphometry) peaked at 3 months and lost significance in older age. Moreover, our histomorphometry data showing reduced osteoid volume in bones from Sox4^+/– mice do not support a mineralization defect (e.g. osteomalacia) as a major responsible factor for the observed low bone mass. These findings strongly indicate that Sox4 is a limiting factor for normal bone development during periods of active and intense growth. Following peak bone mass and skeletal maturation, the osteopenia did not worsen but was maintained in both genders during adult life, suggesting that in later stages of adulthood, reduced levels of Sox4 are sufficient to sustain normal bone homeostasis. Importantly, our results also indicate that the bone phenotype in Sox4^+/– mice cannot be fully rescued by other members of the Sox family.

The suboptimal skeletal development and osteopenic phenotype in Sox4^+/– mice are interesting in relation to genetic factors regulating early stages of bone formation. We have previously shown that Sox4 mRNA is highly expressed in hypertrophic chondrocytes of the mouse embryonic growth plate during early phases (ED15.5) of endochondral bone development, and also by human and rodent osteoblasts in vitro (Reppe et al., 2000). Whereas cRNA in situ hybridization was not sufficiently sensitive to detect Sox4 mRNA in osteoblasts in vivo, real-time PCR enabled us to verify reduced Sox4 expression in bones from Sox4^+/– mice relative to WT (in vivo), as well as in the respective calvarial osteoblast cultures derived from these animals (in vitro).

Another group has reported that highest levels of SOX4 mRNA expression were seen in the proliferative phase during the course of human osteoblast development (Billiard et al., 2003). In the present study we found severely impaired proliferation of Sox4^+/– osteoblasts, as assessed by a decreased thymidine incorporation rate of 73%, and reduced numbers of cells. In support of these observations, preliminary cell-cycle analyses indicated delayed S-phase progression and slower passage into the G2 phase in Sox4^+/– osteoblasts (data not shown), suggesting that it takes mutant cells longer to traverse the cell cycle. Delayed osteoblast maturation was also demonstrated in Sox4^+/– calvarial cultures, with Osx, OCN, Coll1A2 and ALP mRNA levels selectively reduced (25-70%), whereas the other mRNAs investigated were not changed. These changes in the osteoblast gene expression pattern in Sox4^+/– cells were almost entirely mimicked by siRNA treatment of WT osteoblasts (with the exception of PTHR1 and PTHrP, see below). By contrast, no significant effect of the Sox4 mutation was observed on osteoclast development and function in vitro. Hence, our data support a predominant effect of Sox4 deficiency on bone formation, and implicate this transcription factor as an important regulator of osteoblast proliferation. Whether the delay in osteoblast maturation, at least partly, is secondary to the defect in proliferation remains to be established.

The localization of SOX4 to chromosome 6p22 (in humans) and 13 (in mice) (Critcher et al., 1998) coincides with a mapped chromosomal region recently predicted to comprise a gene with pleiotropic effects on osteoblast activity, number or recruitment in baboons (Havill et al., 2006) and to affect BMD in mice (Shimizu et al., 2002). In light of our data in the present study, it will be important (in future research) to clarify whether this gene is identical to Sox4. The upregulation of human SOX4 mRNA in patients with primary hyperparathyroidism (Reppe et al., 2006), together with our results that SOX4 is downregulated in postmenopausal primary osteoporosis (K.M.G., S.R., V.T.G., R.J., F.P.R., L.S.H.N.-M. and O. K. Olstad, unpublished), strengthen the concept that SOX4 has a functional role in human bone metabolism.

Acute Sox4 knockdown (>50% by siRNA) in normal calvarial osteoblasts in vitro reduced the mRNA levels for PTHR1 and PTHrP, which is interesting also in light of our previous observation that Sox4 is a PTH-responsive gene (Reppe et al., 2000). Thus, Sox4 is not only a target for PTHR1 signaling, but may also be part of a regulatory loop that modulates the PTH/PTHrP-receptor system (Kronenberg, 2006). This effect of Sox4 gene silencing on PTHR1 and PTHrP mRNAs was clearly distinct from the pattern found in Sox4^+/– osteoblasts, and the reason for the apparent discrepancy between long-term versus acute effects of Sox4 deficiency remains to be clarified.

The transcription factors Runx2 and Osx play crucial roles in osteoblastogenesis, as indicated by severe defects in bone development following knockout of these genes (Komori et al., 1997; Nakashima et al., 2002). Runx2, an important regulator of the osteoblast-related protein OCN (Ducy, 2000; Gaur et al., 2005), also modulates ALP and Osx expression, whereas Osx is not required for expression of Runx2 (Nakashima et al., 2002). Thus, Osx acts downstream of Runx2 to regulate osteoblast differentiation (Nakashima et al., 2002), but also mediates other signaling pathways including BMP-2 and IGF-1 (Celil et al., 2005), independent of Runx2. Our results from Sox4 siRNA experiments are in agreement with a Runx2-independent mechanism for Sox4 action in osteoblast development. Whereas mRNA for Osx was markedly decreased in Sox4 siRNA-treated WT osteoblasts, as well as in Sox4^+/– osteoblast cultures, Runx2 was not affected. Furthermore, treatment of WT osteoblasts with Runx2 siRNA did not reduce Sox4 expression. It is possible that the effect of Sox4 on osteoblast development, as assessed by mRNAs for ALP and OCN, is mediated by Osx, but independent of Runx2 (Fig. 4). Analysis of the promoter sequences of OCN (data not shown) and Osx (Lu et al., 2006) revealed several putative binding sites for Sox-like transcription factors in these genes; therefore, direct transcriptional regulation through Sox4 cannot be excluded.

Growth plate width was significantly reduced in limbs from Sox4^+/– mice, suggesting that chondrocytes are also affected by Sox4 deficiency (Table 3). Because there was no evidence for alterations in the number or activity of osteoclasts/chondroclasts in bones from Sox4^+/– mice, the primary responsible mechanism for reduced growth plate width is most likely related to reduced ability of columnar chondrocytes to proliferate and subsequently undergo hypertrophy (Vanky et al., 1998). The zone of hypertrophic chondrocytes was slightly diminished, although not significantly, in the same limbs (Table 3). It is also noteworthy that (activated) PTH/PTHrP receptors delay chondrocyte hypertrophy through both Runx2-dependent and -independent pathways (Guo et al., 2006). Whether reduction in growth plate width in Sox4^+/– mice relative to WT is associated with disturbances in PTHR1-mediated signaling in chondrocytes, and the potential involvement of Runx2 (through dependent and independent pathways) in this process, are interesting topics that need to be addressed in further studies.

In conclusion, we have characterized the skeletal phenotype of Sox4 haploinsufficient mice and investigated the functional role of Sox4 deficiency in osteoblastogenesis in vitro. Our results demonstrate that Sox4 influences bone formation both in vivo and in vitro, by modulating osteoblast maturation and function. These results implicate Sox4 as an important regulator of osteoblast proliferation and differentiation, and suggest that Sox4 action is mediated at least partly by Osx, but is independent of Runx2. Finally, we find that Sox4 deficiency results in reduced growth plate width, with the capacity to modulate skeletal growth.

Sox4^+/– heterozygous mice from a C57Bl/6 background were a gift from Hans Clevers (Hubrecht Laboratory, Netherlands Institute for Developmental Biology, Utrecht, The Netherlands) (Schilham et al., 1996), and all mice used in the present study have been backcrossed in the same strain for >10 consecutive generations. The animals were kept and bred in local animal facilities in Oslo, Norway and L'Aquila, Italy, following FELASA guidelines, and all animal procedures were approved by the committees for animal research in Norway and Italy. Pups were separated and earmarked at 3 weeks of age, and the resulting earpieces were used for genotyping using a DNeasy tissue purification kit (Qiagen) and PCR with two sets of specific primers: NEO749 (forward) 5′-ACAAGATGGATTGCACGCAGG-3′, NEO1119 (reverse) 5′-GAATGGGCAGGTAGCCGGATC-3′; Neo-f1 (forward) 5′-AGGATCTCCTGTCATCTCACCTTGCTCCTG-3′ and Neo-rev1 (reverse) 5′-AAGAACTCGTCAAGAAGGCGATAGAAGGCG-3′ (Invitrogen). Before preparation of osteoblast and osteoclast cultures (postnatal day 8-10), genotyping was performed on earpieces obtained at day 6-8.

Sox4^+/– mice and WT littermates were at regular time intervals anaesthetized with subcutaneous (s.c.) injections of Hypnorm (Cilag) and Dormicum (Roche; 0.05-0.075 ml/kg, working solution: 1.25 mg/ml midazolam, 2.5 mg/kg fluanisone and 0.079 mg/kg fentanyl citrate), and subjected to BMD/BMC measurements by DXA in a PIXImus densitometer (GE Medical systems/Lunar Corp.). Calibrations were performed with a phantom mouse with a defined value, and quality assessments were performed before each use. The coefficient of variation for total BMD is 0.59%.

ROIs were analyzed using the PIXImus software (v1.46) and include total body (the calvarium, mandible and teeth were excluded), three lumbar vertebrae (ca 17×55-59 pixels), trochanter femoris, femoral midshaft and proximal tibia of right hindlimb (all 17×11 pixels). Age- and sex-matched groups of mice were measured repeatedly at ∼6-week intervals for 12 months, starting at 7-10 weeks. Measurements were performed twice with animal repositioning between scans, and mean values were analyzed using Microsoft Excel. Repeated measurements (ex vivo) of identical bone areas with repositioning gave a coefficient of variation of ±1.4%.

10-week-old Sox4^+/– and WT mice were injected subcutaneously with Alizarin Complexone (Sigma; 20 mg/kg body weight) and calcein (Fluka; 30 mg/kg body weight) 10 and 3 days before sacrifice, respectively (Marzia et al., 2000). Both tibiae were dissected and fixed in 4% formaldehyde followed by embedding in a methyl methacrylate resin (Technovit 9100 New; Heraeus Kulzer GmbH, Wehrheim, Germany) for dynamic histomorphometry. The following parameters were calculated: MAR (μm/day), BFR/BS (μm³/μm²/year) and mineralizing surface (MS/BS, %).

Histomorphometric measurements were performed as previously described (De Benedetti et al., 2006; Marzia et al., 2000) and with the suggested nomenclature (Parfitt et al., 1987). Briefly, one tibia from each animal was sectioned longitudinally through the frontal plane. Undecalcified sections (∼2 μm-thick) were stained with Methylene blue/azure II for quantitative analysis of structural variables of trabecular metaphyseal bone, and osteoblasts. Osteoclasts were evaluated in adjacent sections treated for the cytochemical demonstration of TRAcP. The following variables were measured in the proximal tibia: (1) trabecular bone volume/tissue volume (BV/TV, %); (2) trabecular thickness (μm), trabecular number (no./mm) and trabecular separation (μm), derived according to the parallel plate model (Parfitt et al., 1983) and measured in the same zone as BV/TV; (3) growth plate width (μm) and size of the hypertrophic chondrocyte zone (μm); (4) osteoid volume (OV/BV, %) and osteoblast surface (%); (5) osteoclast surface (%). Osteoid volume, osteoblasts and osteoclasts were measured in a metaphyseal region extending at least 100 μm away from the distal end of the growth plate and excluding the endocortical surfaces.

We assessed trabecular and cortical bone architecture using μCT (μCT40; Scanco Medical AG, Basserdorf, Switzerland), employing a 12-μm isotropic voxel size. Specifically, trabecular bone architecture was evaluated at the distal femoral metaphysis, whereas cortical bone morphology was evaluated at the femoral midshaft, as previously described (Bouxsein et al., 2005; Ferrari et al., 2005). Bones from mice aged 3, 6, 12 and 18 months were examined.

For all μCT evaluations, we used a nominal isotropic voxel size of 12 μm. Morphometric parameters, including BV/TV (%), trabecular number (mm^–1), trabecular thickness (μm), trabecular separation (μm), structure model index (SMI) and connectivity density (mm^–3), were computed without assumptions regarding the underlying bone architecture (Hildebrand and Ruegsegger, 1997). At the femoral midshaft, 50 transverse CT slices were obtained and used to compute the total volume within the periosteal envelope (mm³), cortical bone volume (BV, mm³), medullary volume (mm³) and cortical thickness (μm). We also used the CT images to measure the bone inner and outer diameter (r_i and r_o, mm). The femurs were approximated as perfect tubes and the area moments of inertia (I_x, mm⁴) at the midshaft approximated by I_x=π(r_o⁴–r_i⁴)/4.

Bones from 3-month-old mice were rehydrated at room temperature in phosphate-buffered saline (PBS) and femoral biomechanical properties were assessed by three-point bending (Brodt et al., 1999; Jepsen et al., 2003), using the Instron Microtester 5848 (Instron, Norwood, MA) equipped with a 100-N gauge and custom bone supports with a 7-mm distance. Load was applied at a constant rate (0.02 mm/sec) until failure and the force-displacement data sampling was set to 100 Hz. We measured the ultimate force (N) and bending stiffness (N/mm) from the load-displacement curve and computed the Young's elastic modulus, E (GPa), and ultimate stress (MPa) using the relevant mid-femoral cross-sectional geometry measured from μCT and following the method described by Schriefer et al. (Schriefer et al., 2005).

Serum was collected from WT and Sox4^+/– mice sacrificed at the same time of the day to avoid diurnal fluctuations of serum markers. Measurements of total serum calcium and PTH were performed using the QuantiChrom Calcium Assay Kit (DICA-500; BioAssay Systems, Hayward, CA) and the Mouse Intact PTH ELISA Kit (Immunotopics, San Clemente, CA), following the respective protocols. Serum OCN was measured using the Mouse Bone Panel 1B Lincoplex kit (Cat# MBN1B-41K) following the manufacturer's protocol (LINCO Research, St Charles, MO), and analyzed with a BioPlex luminometer (Bio-Rad Laboratories). ALP activity was measured as described (Dimai et al., 1998).

Osteoblast cultures were derived from 8-day-to 10-day-old Sox4^+/– and WT mice of both genders as described (Marzia et al., 2000). Initial experiments using separate osteoblasts from male and female mice showed similar results for all functional parameters tested. We therefore used gender-mixed osteoblast cultures for most experiments. Briefly, dissected calvariae were sequentially digested with 1 mg/ml Clostridium histiolyticum type IV collagenase (Sigma) and 0.025% trypsin (Becton Dickinson) in Hank's buffered solution. Cells from second and third digestions were grown in Dulbecco's modified Eagle's medium (DMEM) with antibiotics and 10% fetal bovine serum (FBS). At confluence, cells were trypsinized, counted and plated in appropriate vessels for further experiments. All experiments were performed on 7-day cultures or as indicated. Cell culture media and supplements were purchased from Invitrogen (Carlsbad, CA).

Cells were lysed in RIPA buffer (50 mM Tris HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitors. 100 μg proteins were resolved under reducing conditions by 12% SDS-PAGE and transferred to nitrocellulose membranes. Following blocking of the blot with 5% non-fat milk in TBS-T buffer (20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.2% Tween 20), the anti-Sox4 primary antibody (Chemicon International, cat. #AB5803) was diluted 1:200 (in 1% non-fat milk in TBS-T) and incubated with the blot for 1 hour at room temperature. Next, the filter was washed 3×10 minutes in TBS-T and incubated with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibody for 1 hour at room temperature. Protein bands were revealed by ECL detection. The filter was then stripped and reprobed with anti-actin antibody (Santa Cruz Biotechnology, Heidelberg, Germany, cat. #SC-1616) for normalization.

The procedure was modified from Marzia et al. (Marzia et al., 2000). Briefly, osteoblasts were plated in 24-well multiplates (5000 cells per well) and grown to ∼70% confluence. Following incubation for 24 hours in DMEM with antibiotics and 0.2% bovine serum albumin (BSA), the cells were incubated in 1 μCi/ml [³H]thymidine (specific activity 5.0 Ci/mmol; Amersham) overnight. Cells were then washed twice and solubilized in 0.1% SDS. 10 μl 10 mg/ml BSA was added to each sample as a carrier protein, and precipitated by adding 100 μl 100% trichloroacetic acid (TCA) and incubating for 30 minutes on ice. Following centrifugation, the precipitate was resuspended in 0.5 ml 0.1% SDS. 5 ml Insta-Gel II scintillation fluid (Packard Instrument Company, Groningen, The Netherlands) was added and the samples were counted in a β-scintillation counter (Packard Tri-Carb 1900TR).

Differentiation was evaluated by histochemical and biochemical analyses of ALP activity using reagents and protocols from Sigma kit 104-LS. Total numbers and numbers of ALP-positive cells were mean counts of three microscopic fields from three cultures of WT and Sox4^+/– mice (magnification 20×). For siRNA experiments (see below), cells were counted from two cultures (two experiments) of WT osteoblasts treated with Sox4 and control siRNA.

Osteoblasts were plated in 6- or 24-well multiplates and grown to 90% confluence (∼7 days following plating). Media were then replaced with mineralizing media (DMEM supplemented with 10% FBS, 50 μg/ml ascorbic acid and 10 mM β-glycerophosphate) and cultured for 3 weeks with medium change every 3 days, as described (Marzia et al., 2000). Mineralization was evaluated by von Kossa staining, counting and quantification of positive nodules in 10 representative microscopic fields.

Total RNA was isolated from mouse bones and cell cultures using Trizol® (Life Technologies/Invitrogen, MD) and further purified by RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA quality was checked using a Matrix 2100 Bioanalyzer (Matrix AS, Oslo, Norway) and results analyzed by Agilent 2100 expert software (Agilent Technologies, Blacksburg, VA).

1 μg of RNA was subjected to a 20 μl reverse transcriptase (RT) reaction by M-MLV Reverse transcriptase (Promega, Madison, WI) according to the manufacturer's procedure, and diluted 5× before samples (in triplicates) were subjected to real-time PCR analysis in a LightCycler (Roche Diagnostics, Penzberg, Germany). LightCycler™ Fast Start Master SYBR Green (Roche Diagnostics) or Brilliant® SYBR Green QPCR master mix (Stratagene) kits were used. PCR conditions and primer pairs used are listed in Table S1. To distinguish cDNA and genomic DNA, primers were placed on corresponding exons at the junction of an intron. In addition to this, we also used primers for ALP, Coll1A2 and PTHR1 as published (Huang et al., 2004), OCN (zur Nieden et al., 2003), c-fos (Tanaka et al., 2004) and AP2 (Jiang et al., 2004).

Cycle threshold (Ct) values were obtained graphically (Roche Diagnostics, software version 3.5). Gene expression was normalized to β-actin or GAPDH and ΔCt values calculated. Comparison of gene expression between two samples (WT and Sox4^+/– bones or osteoblasts) was obtained by subtraction of ΔCt values between the two samples to give a ΔΔCt value, and relative gene expression calculated as 2^–ΔΔCt normalized to WT.

Differentiated primary osteoclasts were obtained from the bone marrow of 5-7-day-old Sox4^+/– and WT mice by a modification of the method described by David et al. (David et al., 1998; Teti et al., 1999). Pups were sacrificed by cervical dislocation, and long bones were dissected free from soft tissues and cut into small fragments. Bone marrow cells were released by gently pipetting the fragments in DMEM supplemented with 100 IU/ml penicillin, 100 μg/ml streptomycin, 2 mM L-glutamine and 10% FBS. Cells were plated in culture dishes and allowed to attach for 24 hours before non-adherent cells were removed by aspiration and extensive washing. The total adherent cell fraction was cultured for up to 7 days in the presence of 10^–8 M 1,25(OH)₂vitamin D₃.

Cultures were fixed in 3% paraformaldehyde in 0.1 M cacodylate buffer, and positivity for the osteoclast marker enzyme TRAcP was detected histochemically using the Sigma-Aldrich kit No. 85 (Sigma), according to the manufacturer's instructions.

Osteoclasts were also grown on bone slices, differentiated as described above, and fixed in 3% paraformaldehyde in 0.1 M cacodylate buffer. Cells were then removed by ultrasonication in 1% sodium hypochlorite, and slices were stained with 0.1% toluidine blue. Pits were counted and the pit index computed according to Caselli et al. (Caselli et al., 1997).

Four siGENOME™ SMART pool® siRNA duplexes specific for mouse Sox4 were designed by and purchased from Dharmacon (Lafayette, CO). Osteoblasts from calvariae of 7-day-old WT mice were trypsinized and plated in 24-well plates or in 3.5 cm culture dishes. At approximately 50% confluence, cells were transfected with the annealed siRNA-Sox4 (siRNA final concentration 100 nM) using oligofectamine (Invitrogen, Carlsbad, CA) in Opti-MEM (Invitrogen). Cells were treated with siRNA-Sox4 for 48 hours, then the RNA was extracted, reverse transcribed and subjected to amplification for the Sox4 gene in order to evaluate its downregulation. The same procedure was followed for RNAi knockdown of Runx2.

Results are generally expressed as means ± s.d. and analyzed by SPSS v12.0.1. Analyses of bone density time courses were performed using SPSS Mixed Models, taking into account the repeated measurements of each mouse parameter and the dependency of DXA parameters within each individual. Time, genotype and gender were analyzed statistically as fixed effects, with bone area, per cent fat and body mass as covariates for analyses with total BMC and BMD as dependent variables, respectively. Variables obtained from the same bone (DXA-/μCT-results, histomorphometry, biomechanics) were considered as dependent data sets and analyzed by multivariate analyses (MANOVA). Independent data sets were analyzed with unpaired Student's t-test, using Welch correction when variances were unequal. For all statistical tests, P<0.05 was considered significant.

We are grateful to Hans Clevers for providing the Sox4^+/– mice for these studies. Thanks to Ole Petter Fraas Claussen and Paula deAngelis for help with flow cytometry, to Åse-Karine Fjeldheim and Ole Kristoffer Olstad for technical advice, and to Joseph Sexton for statistical consulting. D. Pioletti is gratefully thanked for providing resources and guidance for bone biomechanical testing, and for advice regarding evaluation of the results obtained. We are indebted to Fanny Cavat for technical assistance with μCT measurements and Stefania De Grossi and Aileen Murdoch Larsen for technical assistance in histological preparations. The study is part of the EU collaboration Life Sciences, Genomics and Biotechnology for Health OSTEOGENE, project no. 502941. This research has also been supported by a Marie Curie Fellowship of the European Community programme Quality of Life and Management of Living Resources under contract number QLK3-CT-2001-60040 (to R.P. and D.F.), by Helse Øst (project no. 29750104), and by the Norwegian Research Council where S.R. is a research fellow.