Type XII collagen regulates osteoblast polarity and communication during bone formation

Osteoblast differentiation and maturation are crucial events in the formation of new bone and determination of bone quality (Nakashima et al., 2002; Yoshida et al., 2002; Komori, 2010). Bone formation begins with the differentiation of osteoblasts from pluripotent mesenchymal cells. The progenitors migrate to the sites of bone matrix deposition and differentiate into fully functional, bone matrix–producing osteoblasts (Imai et al., 1998). These events are regulated by the expression of runt-related gene 2 (Runx2, also known as Cbfa1) and sp7 transcription factor (Sp7, also known as osterix) and by canonical Wnt signaling followed by the secretion of specific bone matrix proteins such as type I collagen (Col1a1), secreted phosphoprotein1 (Spp1, also known as osteopontin), integrin-binding sialoprotein (Ibsp), and bone γ-carboxyglutamate protein (Bglap, also known as osteocalcin; Harada and Rodan, 2003; Koga et al., 2005). In addition to these changes in gene expression and matrix protein secretion, the mature osteoblasts become aligned along regions of bone deposition with an epithelioid arrangement. The mature, terminally differentiated osteoblasts are highly polarized relative to the site of bone matrix deposition (Ilvesaro et al., 1999; Nakashima et al., 2002; Stains et al., 2002). These features are common to periosteal and endosteal bone and both long and flat bones. Therefore, the alignment, polarization, and interaction of osteoblasts are important events necessary for production of a functional bone matrix.

Cell motility, communication, and shape are controlled via the local microenvironment where the extracellular matrix provides a substrate for cell anchorage, stores and presents growth factors or cytokines, and serves as a tissue scaffold (Rozario and DeSimone, 2010). The extracellular matrix interacts with the intracellular actin cytoskeleton via transmembrane proteins that affect cytoskeletal organization, cell motility and polarity, proliferation, and survival (Berthiaume et al., 1996; Geiger et al., 2001; Théry et al., 2005; Parsons et al., 2010; Kim et al., 2011). Thus, extracellular matrix–transmembrane protein–cytoskeleton interactions are important in the regulation of osteoblasts during bone formation. Integrins were shown to regulate osteoblast polarity and bone matrix deposition using mice expressing a dominant-negative truncated integrin β1 subunit (Zimmerman et al., 2000; Dobrowolski et al., 2008). In addition, extracellular matrix proteins, such as fibronectin (Globus et al., 1998) and osteopontin (Denhardt et al., 2001), are involved in osteoblast apoptosis and survival. Furthermore, the extracellular matrix is an important modulator of cell polarity and communication (Berthiaume et al., 1996; Geiger et al., 2001; Théry et al., 2005). However, the regulatory roles of the extracellular matrix in osteoblast organization, polarity, and communication during bone formation remain to be elucidated.

Type XII collagen is a member of the fibril-associated collagens with interrupted triple helices (FACITs) subfamily. FACITs are associated with striated fibrils assembled from fibrillar collagens such as type I collagen. In addition, they function at tissue interfaces, sites of cell-matrix communication (Wessel et al., 1997; Bader et al., 2009). Type XII collagen is a homotrimer of α1 chains and has three noncollagenous domains (NC1, NC2, and NC3) with two collagenous domains (COL1 and COL2). Two major splice variants with a molecular mass difference of 100 kD are generally coexpressed; however, the smaller form is predominant. The large splice variant contains sulfated glycosaminoglycans and is therefore also a proteoglycan. Although the functions of type XII collagen are not fully defined, evidence suggests that type XII collagen is a stress response molecule directly influenced by stretch and shear stress. It was reported that the Col12a1 gene could be directly stimulated by mechanical forces in fibroblasts (Nishiyama et al., 1994; Flück et al., 2003) and endothelial cells (Jin et al., 2003) as well as osteoblasts (Arai et al., 2008). The mechanical strain response element is conserved in the first intron of the Col12a1 gene in chicken, human (Chiquet et al., 1998), and mice (Arai et al., 2008).

Bone development and turnover/reorganization are influenced by mechanical stresses. Type XII collagen mRNA expression has been demonstrated in the periosteum, a site of active bone formation (Böhme et al., 1995). This suggests involvement of type XII collagen in the regulation of osteoblasts, including differentiation and maturation, alignment, polarization, and cell interactions. However, these potential regulatory roles of type XII collagen have not been defined. In this study, type XII collagen is demonstrated to be involved in the regulation of bone formation. Type XII collagen–null mice have fragile bones with a disorganized collagen fiber arrangement, decreased expression of bone matrix proteins, and decreased bone-forming activity associated with delayed terminal differentiation. These results are supported by morphological analysis demonstrating less organized and less polarized osteoblasts in vivo in the absence of type XII collagen. In addition, type XII collagen deficiency altered cell–cell interaction in osteoblasts and their derivatives and impaired gap junction communication through connexin43 (Cx43). Therefore, we hypothesize that type XII collagen regulates osteoblast organization and polarity as well as osteoblast–osteoblast and osteoblast–osteocyte communication required for normal bone deposition and the maintenance of bone quality and strength.

Type XII collagen was localized in bone-forming regions of wild-type mice by immunofluorescence analysis. Immunoreactivity for type XII collagen was detected in cortical and trabecular bone from femur and calvaria (Fig. 1, A–F). Type XII collagen was localized to the periosteum of the femurs from postnatal day 30 (P30) mice. The reactivity for type XII collagen was stable during development and growth (P1–P30). These results indicate that type XII collagen is localized to the areas where osteoblasts actively secrete bone matrix, and its expression is stable compared with the other skeletal tissues such as tendon and muscle (unpublished data). In addition, quantitative real-time PCR and Western blotting analysis revealed that osteoblasts express type XII collagen (Fig. 1 H; Arai et al., 2008). These data support a regulatory role for type XII collagen in bone formation by osteoblasts.

To analyze the regulatory role of type XII collagen in bone formation, Col12a1-null mice (Col12a1^−/−) were generated by the targeted deletion of exons 2–5 (Fig. 1 G). Type XII collagen expression was completely absent in this mouse model (Fig. 1 H). The Col12a1^−/− mice demonstrated skeletal abnormalities (Fig. 1 I). They developed kyphosis with abnormally curved spines when compared with wild-type mice using x-ray analysis. An abnormality in vertebrae was present as double spinous processes. In addition, the Col12a1^−/− mice exhibited a smaller stature indicated by a significant reduction in body weight compared with wild-type mice at age P30 (Fig. 1 J). The femurs of type XII collagen–null versus wild-type control mice were analyzed using 3D micro-computed tomography (CT; Fig. 1, K and L). These data indicate significantly shorter femur lengths and a significant reduction in the percentage of total tissue area/bone length, i.e., more slender bones when normalized to body weight. Therefore, the presence of type XII collagen in bone-forming regions and abnormal skeletal development in the absence of type XII collagen support an essential role for type XII collagen in the regulation of bone formation.

The influence of type XII collagen on the structure of long bones and their mechanical properties was analyzed in Col12a1^−/− and wild-type mice (Fig. 2). The femurs from P30 Col12a1^−/− mice had smaller diameters and proportionally less cortical bone mass compared with wild-type mice when analyzed using 3D micro-CT. Statistical analysis of the micro-CT data normalized to body weight demonstrated an ∼10% reduction in cortical bone thickness compared with wild-type controls. In addition, the mechanical properties analyzed using a four-point bending procedure were significantly reduced in the type XII collagen–null femurs normalized for body weight. The maximum load to failure, bone robustness, and bone stiffness were all significantly decreased in Col12a1^−/− compared with wild-type mice. These data indicate that bones from Col12a1^−/− mice were not just smaller and thinner but also had decreased bone quality relative to body size compared with wild-type controls.

To analyze bone deposition, in vitro mineralized bone nodule formation analyses were conducted using bone marrow cells from P30 wild-type and Col12a1^−/− mice in osteogenic medium and analyzed after 14, 21, and 28 d in culture (Fig. 3 A). Mineralized nodules were present in wild-type osteoblasts after 14 d, whereas nodules were virtually absent from cultures of Col12a1^−/− osteoblasts under the same conditions. At day 21, nodules were formed in cultures of both genotypes, but the mineralized nodule area in the Col12a1^−/− cultures was significantly smaller than in wild-type cultures. At 28 d in culture, nodule formation was not as robust in Col12a1^−/− compared with wild-type osteoblasts. However, the difference between osteoblasts from the different genotypes was no longer significant, suggesting a delayed bone deposition and/or decreased bone-forming activity in the absence of type XII collagen.

The in vitro bone formation data were confirmed in vivo using calcein double labeling in P30 wild-type and Col12a1^−/− mice (Fig. 3, B–E). Nondecalcified cross sections demonstrated decreased calcein-labeling intervals in the periosteal region of Col12a1^−/− bone compared with wild-type controls. This interval in the endosteum is comparable in both genotypes (unpublished data). Based on this calcein double labeling, the mineral apposition rate, percentage of mineralized surface/total bone surface, and bone formation rate were analyzed. All indices were significantly decreased in Col12a1^−/− mice compared with wild-type controls. These data indicate that the absence of type XII collagen was associated with decreased bone formation in osteoblasts.

The potential involvement of bone resorption versus decreased bone formation in the absence of type XII collagen was investigated. The C-terminal telopeptide of collagen I (CTx), a biochemical marker of bone resorption, was analyzed, and there was no significant difference in serum levels between Col12a1^−/− and wild-type control mice at age P30 (Fig. S1 A). The expression of bone resorption markers, such as tartrate resistance acid phosphatase (Trap), nuclear factor of activated T cells cytoplasmic 1 (Nfatc1), receptor activator of NFκ-B (Rank), and rank ligand (Rankl), were analyzed from whole femurs by real-time PCR, and no significant differences between Col12a1^−/− and wild-type mice were observed (Fig. S1 B). Finally, osteoclast number per bone surface and osteoclast surface per bone surface (Fig. S1 C) were also comparable in Col12a1^−/− and wild-type controls as observed by Trap staining. These data indicate that the decreased bone mass in Col12a1^−/− mice was not a result of increased osteoclast resorption but resulted from abnormal osteoblast bone formation.

The delayed/decreased mineralized nodule formation in osteoblast cultures from type XII collagen–null mice suggests altered or delayed osteoblast differentiation (Fig. 4). To investigate the involvement of the type XII collagen in osteoblast differentiation, the expression of osteoblast differentiation markers was used. Runx2 was used as an early stage marker, type I collagen (Col1a1) was used as a middle-stage marker, and osteocalcin (Ocn) and osteopontin (Opn) were used as late-stage markers. Primary osteoblasts from neonatal calvaria of both genotypes were cultured in osteogenic medium and analyzed after 0, 7, 14, and 21 d using quantitative real-time PCR. Comparable expression patterns were detected for Runx2 and type I collagen in wild-type and Col12a1^−/− osteoblasts (Fig. 4, B and C). In contrast, expression of osteocalcin and osteopontin, markers for late-stage osteoblast development, was significantly decreased in Col12a1^−/− osteoblasts compared with wild-type controls (Fig. 4, D and E). The mRNA expression data were validated using immunoblot analyses of type I collagen, osteocalcin, and osteopontin from P30 femurs (Fig. 4 F). There was no difference in type I collagen in the two genotypes; however, both osteopontin and osteocalcin were decreased in the type XII collagen–null femurs. This was analyzed further using short hairpin RNA (shRNA) for a functional Col12a1 gene knockdown in murine preosteoblastic cells (MC3T3-E1). Lentivirus transduction with the specific shRNA demonstrated an ∼40% reduction of Col12a1 mRNA expression (Fig. S2, A–D). In this system, the Col12a1 knockdown had no effect on Runx2 or type I collagen mRNA expression. However, a reduction in Ocn mRNA expression was observed (Fig. S2, B–D). Protein levels for type I collagen were unaffected with both osteopontin and osteocalcin showing reduced protein expression in the type XII collagen knockdown cultures (Fig. S2 E). These data are comparable with that obtained from the type XII collagen–null mice. The comparable expression patterns observed in both systems fully support a role for type XII collagen in the regulation of osteoblast terminal differentiation.

The in vivo phenotype of femurs from Col12a1^−/− mice was analyzed for bone matrix organization as well as osteoblast/osteocyte organization. Osteoblasts in wild-type bone were localized on the surface of cortical bone, where they were aligned and well organized in an epithelioid manner when analyzed in hematoxylin and eosin (H&E)–stained sections (Fig. 5 A). In contrast, Col12a1^−/− osteoblasts were poorly aligned, with a disorganized arrangement relative to the surface of cortical bone (Fig. 5 A). Collagen fiber organization was analyzed in cross sections of the diaphysis of femurs from P14 wild-type and Col12a1^−/− mice using Sirius red staining with polarized light microscopy (Fig. 5 B). Large birefringent collagen fibers were arranged in parallel and were well organized in the wild-type cortical bone matrix. In contrast, birefringence was sparse and disorganized in Col12a1^−/− bone matrix, indicating that collagen fibers were smaller and/or less organized in Col12a1^−/− cortical bone compared with controls. These results indicate that type XII collagen is involved in determining osteoblast organization and normal bone matrix quality.

To further analyze osteoblast organization and polarity along the bone deposition interface, osteoblasts in the periosteum were analyzed using confocal microscopy with nuclear, actin cytoskeletal, and Golgi markers (Fig. 6). Osteoblasts in wild-type bone were well organized in an epithelioid manner along the bone interface. These osteoblasts exhibited a well-organized circumferential actin cytoskeleton with a peripheral localization. In addition, the actin fibers were localized in osteoblast–osteocyte processes interacting in the bone matrix to create an extensive osteoblast–osteocyte network. Overall, the nuclear and Golgi markers were localized in a polarized manner, in which Golgi usually locate toward the cortical bone opposite the nuclei. In contrast, Col12a1^−/− osteoblasts were poorly organized along the bone interface. The actin cytoskeleton was also poorly organized, composed of clumps of actin fibers showing random distribution around the cells. In addition, fewer osteoblast/osteocyte processes were observed in the bone matrix, indicating a less extensive osteoblast–osteocyte network than present in the wild-type bone. The nuclear and Golgi markers were not as polarized as in the control mice. These data indicate a role for type XII collagen in the organization of osteoblasts into well-organized and polarized layers necessary for regulated bone matrix deposition.

Because Col12a1^−/− osteoblasts showed a disorganized arrangement at the bone surface and a less extensive osteoblast–osteocyte network, a defect in osteoblast to osteocyte transition was hypothesized. To address this hypothesis, osteocyte density was analyzed as osteocyte number/100 µm² in H&E sections. A significant increase (∼60%) in the number of osteocytes in Col12a1^−/− compared with wild-type cortical bones was observed (Fig. S3). These data indicate that type XII collagen deficiency results in abnormal arrangement and polarization of osteoblast linage cells in the periosteum and suggests a dysfunctional, premature osteoblast to osteocyte transition.

To further analyze the morphological phenotype of osteoblasts and osteocytes of Col12a1^−/− and wild-type mice, transmission EM was conducted during an active bone-forming period (P14; Fig. 7, A–C). The wild-type osteoblasts were attached to the surface of cortical bone and were aligned along the region of bone formation. In addition, the osteoblasts were well polarized relative to the bone matrix. In contrast, the Col12a1^−/− osteoblasts were less organized and poorly polarized (Fig. 7 A). This was consistent with the data observed by confocal immunofluorescence microscopy. The analysis of adjacent osteoblasts in the periosteum revealed that wild-type osteoblasts had numerous processes that connected osteoblast–osteoblast and osteoblast–osteocyte. In contrast, Col12a1^−/− osteoblasts had large intracellular spaces with fewer processes (Fig. 7, B and C). This was consistent with the disorganized actin cytoskeleton observed by confocal microscopy. This indicates altered cell–cell interaction and/or communication in the absence of type XII collagen in bone. Similar to the osteoblast phenotype, osteocytes had processes that interact with other osteocytes and/or toward the cortical bone surface connecting to other osteocytes or osteoblasts in wild-type bone. However, Col12a1^−/− osteocytes had fewer and less developed processes (Fig. 7 D). This is consistent with the altered osteocyte number/distribution observed (Fig. 6, A and B; and Fig. S3). Thus, type XII collagen–null bones are associated with alterations in cell–cell interactions that may lead to altered cell–cell communication.

Based on the morphological findings that indicate altered cell–cell interaction, the expression of gap junction proteins and cell adhesion molecules was investigated to determine the relationship between type XII collagen and cell–cell communication. Gap junction proteins and cadherins are required for coordination of mechanical and chemical signal transduction in bone development and formation (Civitelli, 2008; Matsuo and Irie, 2008). A quantitative analysis of Cx43 and cadherin 2 and 11 (Cad2 and Cad11) was performed using primary osteoblasts after 0, 7, 14, and 21 d in culture (Fig. 8, A–E; and Fig. S4 A). Cx43 mRNA expression was significantly decreased at each time point in Col12a1^−/− osteoblast cultures versus wild-type control cultures. As expected, type XII collagen mRNA expression was stable (Fig. 4 A). Interestingly, the difference between genotypes was detected only in early stage osteoblasts (0 and 7 d in culture) in Cad11 expression but not in Cad2 and Cad11 expression in late-stage osteoblasts in Col12a1^−/− cultures versus the wild-type controls. The mRNA expression was consistant with immunoblot analysis in primary osteoblasts (Fig. 8, D–F; and Fig. S4). The expression of Cx43 protein was decreased in Col12a1^−/− primary osteoblasts as well as in short hairpin Col12a1 (shCol12a1)–treated MC3T3-E1 cells with differentiation and P30 whole femurs, consistent with the mRNA expression results. Therefore, the absence of type XII collagen leads to decreased Cx43 expression at both mRNA and protein levels. Type XII collagen protein increased with osteoblast development. However, a relationship between the increase of type XII collagen and Cx43 expression was not detected in normal osteoblast development because Cx43 expression was stable in both mRNA and protein in cultured wild-type osteoblasts in osteogenic medium (Fig. 8 D). To investigate the possible interaction of type XII collagen and Cx43, coimmunoprecipitation assays were performed in wild-type primary osteoblasts and MC3T3-E1 cells cultured in osteogenic medium. Immunoprecipitation with anti–type XII collagen antibody did not pull down Cx43 and vice versa (unpublished data). These results suggest that the type XII collagen may indirectly regulate Cx43 expression during bone formation, perhaps by altering cell behavior during transition to the organized and polarized osteoblastic layer.

Based on the decreased expression of Cx43 in type XII collagen–null osteoblasts, a dye coupling assay was used to clarify whether type XII collagen can regulate gap junction–mediated intercellular communication (Fig. 8, G and H; and Videos 1 and 2). Osteoblasts were labeled with calcein acetoxymethyl ester (AM) dye and seeded onto unlabeled confluent osteoblasts. Dye transfer was examined and quantitated using confocal microscopy. The wild-type dye-carrying osteoblasts continuously transferred dye to unlabeled osteoblasts (host cells) after the dye-carrying cells attached to the host. In contrast, in the majority of Col12a1^−/− osteoblasts, the dye was not transferred and remained within the cells after attachment to Col12a1^−/− host cells. The percentage of dye-transferred cells was significantly decreased when either dye-carrying or host cells are Col12a1^−/−. A reduction of ∼40% was observed compared with dye-carrying wild-type cells on wild-type host cells. A similar reduction in dye transfer was observed using Col12a1 knockdown MC3T3-E1 cells (Fig. S5). These results indicate that type XII collagen deficiency impaired gap junction communication possibly through altered Cx43 expression.

These results indicate a critical role for type XII collagen in bone formation. The absence of type XII collagen resulted in decreased bone quality associated with altered osteoblast organization, decreased polarity, and altered bone matrix deposition. In addition, type XII collagen–null osteoblasts had altered cell–cell interactions that impaired gap junction communication and Cx43 expression.

Cell shape and polarity are essential for cell proliferation, development, and function. This is regulated by extracellular matrix–transmembrane protein–cytoskeleton interactions (Geiger et al., 2001; Burgstaller et al., 2010; Kim et al., 2011). In bone formation, osteoblasts become highly polarized, align along the bone matrix, and develop intercellular interactions during maturation. However, the regulation of osteoblast shape and cell–cell interaction during maturation and bone formation remains unclear. Here, we demonstrate that type XII collagen regulates osteoblast differentiation and maturation, alignment, interaction, and polarization. The genetic deletion of type XII collagen results in abnormal osteoblast differentiation, decreased bone matrix deposition, and decreased bone quality. This is related to impaired terminal differentiation of osteoblasts.

Osteoblasts and osteocytes express several different integrins that connect the cytoskeleton with the extracellular matrix. The functional roles of integrins in osteoblast development and bone formation have been examined. Interestingly, less polarization of mature osteoblasts and impaired process formation in osteocytes were demonstrated using transgenic mice with a dominant-negative β1 integrin driven by the osteocalcin promoter, which is expressed only in mature osteoblasts and osteocytes (Zimmerman et al., 2000). These data are comparable with our current results, suggesting a potential type XII collagen–integrin regulatory interaction. Type XII collagen has an RGD (Arg-Gly-Asp) peptide sequence (Oh et al., 1992), which is a cell adhesion motif suggesting that type XII collagen may regulate osteoblast polarization via an RGD site interacting with a β1 integrin during maturation. However, other indirect interactions mediated by intermediary matrix molecules should also be considered. Type XII collagen also has a mechanical strain response element in the Col12a1 gene (Arai et al., 2008), and evidence indicates an involvement of type XII collagen in the mechanical response. Cell shape, alignment, and organization are also regulated by mechanotransduction. Osteoblasts respond to cyclic mechanical bending, and osteoblast/osteocyte communication via gap junctions transmits mechanical signals (Duncan and Turner, 1995; Jiang et al., 2007; Shapiro, 2008) that result in osteoblast differentiation, growth, and bone matrix deposition. Therefore, type XII collagen may be expressed in response to mechanical forces and expressed at the osteoblast surface where it transduces the signal to the cytoskeleton through integrins, thereby stimulating osteoblast terminal differentiation in bone-forming regions.

Communication in mature osteoblasts and osteoblasts–osteocytes is controlled via gap junctions. Cx43 is the most abundant gap junction component in osteoblast cell coupling and the osteocyte network within the bone matrix. The functional disruption of Cx43 causes delayed osteoblast differentiation with decreased expression of osteocalcin and delayed ossification in vivo (Lecanda et al., 2000; Civitelli, 2008; Dobrowolski et al., 2008). These data support our hypothesis that type XII collagen regulates cell–cell communication and mediates gap junction function. The function of type XII collagen in mediating cell shape changes and cell organization may regulate Cx43 expression and gap junction function. This is consistent with data indicating that the extracellular matrix, especially laminin, regulates connexin expression and functions through gap junctions in neural progenitor cells (Imbeault et al., 2009). In addition to a fibril-associated localization, type XII collagen is localized to basement membrane zones in many tissues (Wessel et al., 1997; Bader et al., 2009). Our dye coupling experiment revealed that gap junction communication was impaired in the absence of type XII collagen even when osteoblasts were in direct contact with each other. Moreover, Cx43 is involved in transmitting signals that induce bone formation in response to in vivo mechanical loading (Grimston et al., 2008). Our coimmunoprecipitation assays revealed no direct interaction in type XII collagen and Cx43. Thus, type XII collagen may regulate gap junction communication and Cx43 expression in osteoblast linage cells, but this may be an indirect regulation.

Alteration of cell shape and mobility also influences cell matrix production. Maintenance of bone quality is important in osteoporosis, bone fracture healing, and bone implantation. Bone matrix mainly consists of type I collagen. Collagen fibrils regulate mineral deposition and growth (Murshed et al., 2005). FACIT collagens, including type XII collagen, play a role in collagen fibril formation by providing specific molecular bridges between fibrils and other matrix components (Gordon and Olsen, 1990; van der Rest and Garrone, 1991; Gordon and Hahn, 2010; Birk and Bruckner, 2011). In this study, although the expression of type I collagen did not change, the absence of type XII collagen decreased bone matrix proteins such as osteocalcin and osteopontin. A recent study demonstrated that osteopontin regulates bone toughness by mediating matrix heterogeneity in bone matrix (Thurner et al., 2010). This supports a role for fibril-associated type XII collagen in modification of collagen fibril properties, resulting in the regulation of bone matrix formation and mineralization. Type XII collagen may provide a target to develop treatment modalities for bone fracture healing and transplantation.

In conclusion, we propose a model in which type XII collagen is involved in the regulation of bone formation (Fig. 9). Type XII collagen surrounding immature osteoblasts, either alone or as a complex, responds to mechanical stimuli that initiate osteoblast maturation. This is associated with increased type XII collagen synthesis and secretion as well as localized deposition of type XII collagen along the maturing osteoblast boundary. This results in well-organized polarized osteoblasts driving osteoblast organization, cell–cell interaction, and maturation. Osteoblast organization, cell–cell interaction, and polarization along bone deposition interfaces drive Cx43 expression, thereby establishing gap junction–mediated cell communication resulting in terminal differentiation, increased bone matrix protein production, and organized deposition of a functional bone matrix. In contrast, when type XII collagen is absent, there is a dysfunctional response to the mechanical stimuli and an alteration in the osteoblast maturation sequence. This defect delays/alters osteoblast organization, polarity, and cell–cell contacts, and, therefore, the downstream pathways are compromised, including up-regulation of Cx43 and bone matrix proteins. In addition, altered cell interactions result in an abnormal regulation of the osteoblast to osteocyte transition. Together, this results in decreased bone formation and altered quality. This is a novel finding indicating that type XII collagen in the extracellular matrix plays important roles in osteoblast development. Furthermore, determination of osteoblast shape, organization, and communication are crucial events in osteoblast terminal differentiation and the high quality production of bone matrix. Because type XII collagen is distributed in all the bone-forming areas, this regulation is crucial for general bone formation.

Col12a1 knockout mice were generated on a mixed background with C57BL/6 and 129/SvJ. Homologous recombination was performed using a construct including neomycin-polyadenylation signal (poly A) under the control of phosphoglycerolkinase, which was inserted into exons 2–5 of the Col12a1 gene and was not removed from the line studied. Mice were backcrossed into the C57BL/6 background and demonstrated increased perinatal lethality. To circumvent this, the line was bred with 129/SvJ mice and the offspring used for this work. Both mutant and control mice were in mixed C57BL/6-129/SvJ backgrounds. Mutant mice in both backgrounds had comparable phenotypes. Parameters defining the phenotype for the organism (i.e., body weight), bone function (i.e., biomechanical parameters including the maximum load to failure), and bone structure (i.e., fiber organization and amount) were examined in wild-type, heterozygous, and homozygous mice. There were no significant differences observed in these parameters between wild-type and heterozygous mice. All animal studies were performed in compliance with the Institutional Animal Care and Use Committee–approved protocols.

For x-ray analysis, 5-mo-old Col12a^+/+ and Col12a1^−/− mice were anesthetized, and images were taken with an x-ray generator (Polymat 70; Siemens) using 44 kV and 6.4 mA. After x-ray analysis, whole mount mouse skeletal staining was performed as previously described (McLeod, 1980). In brief, the animals were sacrificed, skinned, eviscerated, and fixed in 96% ethanol for 48 h. The cartilage was stained with alcian blue (0.015% in 80% ethanol and 20% acetic acid) for 48 h followed by incubation in 96% ethanol for 3 d. Samples were cleared by immersion in 2% KOH overnight. Bones were counterstained with alizarin red for 5 h (50 mg/liter alizarin red in 1% KOH). After transfer of the stained skeletons to 2% KOH for 5 d to dissolve the soft tissue, the samples were preserved in 1% KOH and glycerol (20:80).

Diaphyseal cross-sectional morphology and tissue mineral density (TMDn) of the femur were measured using an eXplore Locus SP Pre-Clinical Specimen micro-CT system (GE Healthcare) according to previously published methods (Jepsen et al., 2007). 3D images of the mid-diaphysis, which was located immediately distal to the third trochanter, were obtained at an 8.7-µm voxel size. The scan protocol consisted of 3,600 image acquisitions over a 5-h scan (acquisition parameters: 80 kVp, 80 µA, 3-s exposure time [∼69 kJ], and 0.010 in. of aluminum beam filter). Femurs were individually thresholded (Otsu, 1979) to segment bone and nonbone voxels. Morphological traits describing the amount of tissue (cortical area, marrow area, total area, and cortical thickness) and the spatial distribution of tissue (moment of inertia, J) were quantified. Total bone area was defined as the sum of the cortical and marrow areas. The relative cortical area (cortical area/total area) provided a measure of the proportion of the total area that was occupied by bone. These traits were quantified for each cross section, and the values were averaged over the analysis region. TMDn, which is the mean mineral value of the bone voxels only expressed in hydroxyapatite (HA) density equivalents, was calculated by converting the grayscale output of bone voxels in Hounsfield units (HUs) to mineral values (milligrams/cubic centimeters of HA) through the use of a calibration phantom containing air, water, and HA (SB3; Gammex RMI). TMDn was defined as the mean bone voxel HU value divided by the mean of the HU value of the HA phantom multiplied by 1,130 mg/cm³ (HA physical density). The same calibration phantom was included in all scans to adjust mineral density measurements for the variability in x-ray attenuation inherent to independent scan sessions (Jepsen et al., 2007).

After micro-CT analysis, femurs were loaded to failure in four-point bending at 0.05 mm/s to assess whole bone mechanical properties (Jepsen et al., 2003). Load deflection curves were analyzed for stiffness (the slope of the initial portion of the curve), maximum load, postyield deflection, and work to failure. Femurs were tested at RT and kept moist with PBS during all tests. Because body weight differed for Col12a^+/+ and Col12a1^−/− mice, differences in mechanical properties were not only determined by comparing group means but also by testing whether the femurs were less stiff and less strong relative to body size (Jepsen et al., 2003). For the latter analysis, whole bone stiffness and maximum load were plotted against body weight, and the slopes and intercepts of linear regressions were compared by analysis of covariance.

To analyze in vivo bone formation, calcein labeling was conducted to determine the levels of newly formed bone within a unit time period according to the methods previously described (Izu et al., 2009). In brief, calcein (10 mg/kg of body weight) was injected intraperitoneally 2 and 7 d before sacrifice. Femurs were fixed in 4% paraformaldehyde and rinsed in 0.1 M phosphate buffer. Femurs were embedded undecalcified in polymethylmethacrylate, and 40-µm-thick cross sections were prepared from the mid-diaphysis. The analysis site was standardized by examining the cross section located immediately distal to the third trochanter. The histomorphometric indices of osteoblastic activity included percentage of labeled perimeter (L.Pm/B.Pm,%), bone formation rate, and mineral apposition rate. The histomorphometric index of osteoclastic activity included percentage of eroded perimeter (Er.Pm/B.Pm,%). To analyze bone turnover, osteoclast number per bone surface (N.S/BS, N/mm) and osteoclast surface per bone surface (Oc.S/BS,%) were analyzed by TRAP staining of 5-µm cross sections of femurs as previously described (Izu et al., 2009).

Mineralized nodule assays were performed as previously described (Izu et al., 2009). Bone marrow cells were obtained from P30 Col12a1^+/+ (n = 5) and Col12a1^−/− (n = 6) mice and seeded into 12-well plates (2.0 cm²/well) at a density of 2 × 10⁶ cells/well. The bone marrow cells were cultured in a standard growth medium containing 50 µg/ml ascorbic acid and 10 mM β-glycerophosphate. The medium was changed every 2 d. The cultures were harvested on days 14, 21, and 28 and stained with alizarin red solution. At each time point, cells from each individual mouse were analyzed in triplicate. The area of mineralized nodules/total dish was measured with the ImageJ Analysis Program (National Institutes of Health).

RatLaps ELISA (Immunodiagnostic Systems) was used for bone resorption assay in vivo by quantitatively determining the C-terminal telopeptide α1 chain of type I collagen in mouse serum. Serum was obtained from wild-type mice (n = 4) and Col12a1^−/− (n = 5) mice at P30 and analyzed in duplicate.

Femurs were dissected from wild-type and Col12a1^−/− mice at P14 and P30 and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, at 4°C overnight. The tissues were decalcified in 15% EDTA in 0.01 M phosphate buffer, pH 7.4, at 4°C for 1 wk and then embedded in an optimal cutting temperature compound, frozen by liquid nitrogen, and stored at −80°C. Cross sections were stained with H&E to observe bone histological phenotypes, with Sirius red to observe collagen arrangement under polarized microscopy, and with immunostaining to detect cell polarization under confocal microscopy. Osteocyte density was measured in H&E cross sections of P14 wild-type (n = 3) and type XII collagen–null (n = 3) mice. Five different areas per individual mice were randomly selected, and the numbers of osteocytes in cortical bone areas were calculated.

Immunostaining was performed as previously described (Izu et al., 2009). In brief, 5-µm cross and longitudinal sections of the femur diaphysis from P30 wild-type and Col12a1^−/− mice were prepared. The cryosections were blocked by 5% goat serum and then were incubated with rabbit anti–type XII collagen antibodies (Veit et al., 2006) using a 1:1,000 dilution at 4°C overnight. The secondary antibody was Alexa Fluor 488 goat anti–rabbit IgG (Invitrogen) used at 1:500. Vectashield mounting solution with DAPI (Vector Laboratories) was used as a nuclear marker. Images were captured using a fluorescence microscope (DM5500; Leica). Identical conditions and set integration times were used to facilitate comparisons between samples. To investigate cell polarity, 10-µm cross sections of P30 Col12a1^+/+ and Col12a1^−/− femur diaphysis were incubated with rabbit monoclonal anti-GM130 antibody (EP892Y; Abcam) using a 1:50 dilution. Alexa Fluor 555 goat anti–rabbit IgG (Invitrogen) and Alexa Fluor 488–phalloidin (Invitrogen) were used at 1:200. Images were captured using a confocal laser-scanning microscope (FV1000 MPE; Olympus) with a 60× 1.42 NA oil immersion lens. To avoid bleedthrough between fluorescence emissions, samples were scanned sequentially with 405-, 488-, and 543-nm lasers, and emissions were collected with appropriate spectral slit settings.

Femur samples from Col12a^+/+ and Col12a1^−/− mice at P14 were analyzed by transmission EM (Birk and Trelstad, 1984). In brief, femurs were dissected and fixed in 4% paraformaldehyde, 2.5% glutaraldehyde, and 0.1 M sodium cacodylate, pH 7.4, with 8.0 mM CaCl₂ at 4°C overnight and then decalcified in 15% EDTA in 0.01 M phosphate buffer, pH 7.4, at 4°C for 1 wk. The femurs were postfixed with 1% osmium tetroxide and dehydrated in an ethanol series followed by propylene oxide. The tissue samples were infiltrated and embedded in a mixture of EMbed 812, nadic methyl anhydride, dodecenylsuccinic anhydride, and DMP-30 (Electron Microscopy Sciences). 90-nm thin sections were cut using an ultramicrotome (UCT; Reichert) and poststained with 2% aqueous uranyl acetate and 1% phosphotungstic acid, pH 3.2. Cross sections from the midshafts of the femurs were examined at 80 kV using a transmission electron microscope (1400; JEOL) equipped with a digital camera (Orius; Gatan, Inc.).

Primary mouse osteoblastic cells were obtained by sequential enzyme digestion of excised calvarial bones from neonatal mice using 2% dispase and 1% collagenase in PBS. The cells were resuspended in α-MEM (Invitrogen) supplemented with 10% FBS (Invitrogen) and 1% antibiotics (Invitrogen). Osteoblasts were initially seeded at a density of 2 × 10⁵ cells/6-cm dish. At confluency, the culture medium was changed to the growth medium containing 50 µg/ml ascorbic acid and 10 mM β-glycerophosphate. The medium was changed every 2 d. The cells were collected on day 0, 7, 14, and 21 after changing to growth medium. The collected cells were used for the analysis of real-time PCR, Western blotting, and dye coupling assays.

Lentiviral constructs carrying shRNA-targeting mouse Col12a1 were purchased from Thermo Fisher Scientific. Five mouse Col12a1 shRNA clones (V2LMM_15532, V2LMM_17805, V3LMM_459147, V3LMM_459151, and V3LMM_459152) were screened. The nonsilencing GFP lentivirus was used as a control. Plasmid DNA was amplified and purified using a HiSpeed Plasmid Midi kit (QIAGEN) and then transfected along with packaging plasmid into HEK293T cells to generate lentiviruses by using the Trans-Lentiviral GIPZ packaging system (Thermo Fisher Scientific). The cells were refed with DME medium supplemented with 10% FBS and 1× antibiotic–antimycotic (Invitrogen) 24 h after infection. The culture supernatant was harvested 72 h after transfection. The supernatant-containing lentivirus was centrifuged or filtered and used to infect the mouse preosteoblast cell line MC3T3-E1 purchased from American Type Culture Collection. The MC3T3-E1 cells were cultured with α-MEM supplemented with 10% FBS and 1× antibiotic–antimycotic (Invitrogen) as well as 2 µg/ml puromycin to select target cells. After selection for 7 d, cells were analyzed for in vitro differentiation and dye coupling assay.

Total RNA was isolated from primary osteoblasts obtained from neonatal mouse calvaria, lentivirus-treated MC3T3-E1 cells, and P30 mouse femurs by using an RNeasy Lipid Tissue Mini kit (QIAGEN) according to the manufacturer’s protocol. Reverse transcription was performed using 1 µg of total RNA with random primers (High-Capacity cDNA Reverse Transcription kit; Applied Biosystems). Quantitative real-time PCR analysis was performed using StepOnePlus (Applied Biosystems). The reaction was performed in a 25-µl reaction mixture containing 200 nM cDNA samples, 50 nM of sense and antisense primers, and 12.5 µl SYBR green PCR Master Mix (Applied Biosystems). The primer sequences are shown in Table I. The PCR conditions were initially 95°C for 20 s followed by 40 cycles at 95°C for 3 s and 60°C for 30 s. The mRNA expression levels were normalized by calculating the ratios against Gapdh expression levels.

Flexor digitorum longus, tibia, muscle, skin, and cornea were removed from P30 Col12a1^+/+ and Col12a1^−/− mice, and primary osteoblastic cells obtained from neonatal calvaria described in the previous section were used for Western blotting analysis. The whole tissue and cell lysate preparation and immunoblot were performed as previously described (Izu et al., 2009) using rabbit anti–type XII collagen (1:1,000 dilution; KR33), rabbit anti-Cx43 (1:10,000 dilution; Sigma-Aldrich), rabbit anti–mouse collagen I (1:1,000 dilution; Millipore), rabbit anti-Cad11 (1:250 dilution; Invitrogen), rabbit anti–N-cadherin (1:1,000 dilution; Cell Signaling Technology), 0.1 µg/ml of goat anti–mouse osteopontin (R&D Systems), goat anti-osteocalcin (1:100 dilution; Santa Cruz Biotechnology, Inc.), mouse anti–β-actin (1:5,000 dilution; Millipore), and anti–mouse or anti–rabbit HRP-conjugated secondary antibodies (GE Healthcare).

Dye coupling analysis was used as previously described (Lecanda et al., 2000) to assess cell to cell diffusion of negatively charged dyes. Primary osteoblasts and lentivirus-induced knockdown MC3T3-E1 cells were used. In brief, primary osteoblasts obtained from each genotype group were preloaded with calcein AM (BD) and Cell Tracker (Invitrogen) for 15 min and mobilized by trypsin digestion and then were added on top of a monolayer of each cell type. Cell trace calcein red-orange AM (Invitrogen) was used for lentivirus-treated cells because cells were transduced shGFP. The “parachuted” cells were allowed to settle onto the unlabeled monolayer and returned to the incubator for 1 h. Because calcein dye is able to transfer through gap junctions but Cell Tracker remains in the plasma membrane, the transfer of calcein dye from donor to acceptor cells (host cells) was detected using a confocal laser-scanning microscope (SP2 DM IRE2; Leica). The images were taken with a 40× 1.25 NA oil immersion lens. Calcein AM and calcein red-orange AM signals from cells were collected with 488- and 561-nm lasers and appropriate spectral slit settings. Six different experimental groups were established, donor and host cells were either wild-type or Col12a1^−/− cells, and donor cells of each genotype were seeded onto plates with no host cell as a control. The number of dye-transferred cells per total parachuted cells was determined in 10 random images acquired from each well (24-well plate), and 6 wells were used per experimental group. Images were captured using a spinning disk confocal system (Ultraview Vox; PerkinElmer) attached to an inverted microscope (Ti; Nikon). The images were taken with a 40× 1.3 NA oil immersion lens and an electron-multiplying charge-coupled device camera (C9100-13; Hamamatsu Photonics) using Velocity version 5.3.2 software. Calcein AM and Cell Tracker were excited with 488- and 561-nm laser lines, and emissions were collected with 527/55- and 587/125-nm band pass filters, respectively. Cells were maintained in α-MEM supplemented with 10% FBS and 1× antibiotic–antimycotic at 37°C during image capture.

Statistics were conducted as Student’s t tests unless noted otherwise. Asterisks indicate P < 0.05, and double asterisks indicate P < 0.01. All of the numeral data in the results were presented as means ± SD.

Fig. S1 shows that altered bone matrix deposition observed in type XII collagen–null mice is not the result of altered bone resorption. Fig. S2 shows delayed terminal differentiation in type XII collagen knockdown MC3T3-E1 cells comparable with that seen in the type XII collagen–null mouse model. Fig. S3 shows increased osteocyte density in type XII collagen–null cortical bone, indicating dysfunctional regulation of osteoblast to osteocyte transition. Fig. S4 shows no alteration in Cad2 and Cad11 expression in the absence of type XII collagen in osteoblasts. Fig. S5 shows impaired gap junction function in Col12a1 gene knockdown MC3T3-E1 cells comparable with that seen in the type XII collagen–null mice. Video 1 shows normal gap junction communication in Col12a1^+/+ osteoblasts, and Video 2 shows impaired gap junction communication in Col12a1^−/− osteoblasts.

We thank Sheila M. Adams and Amanda Garces for technical assistance with the EM and Chinda Hemmavanh and Semra Oezcelik for their valuable technical assistance in the laboratory. We thank Dr. Marian F. Young for many helpful discussions.

This work was supported by grants from the National Institutes of Health (DEB AR44745, EY05129, KJJ AR44927, and AR056639).