Cyclophilin A (CypA) Plays Dual Roles in Regulation of Bone Anabolism and Resorption

CypA (Cyclophilin A) is a peptidyl-prolyl isomerase previously shown to be required for chondrogenic differentiation and endochondral ossification. However, the effects of CypA on osteoclast activity and bone maintenance are entirely unknown. Here, we show that Ppia−/− mice demonstrate low bone mineral density, reduced osteoblast numbers, and increased osteoclast numbers. When isolated from the calvaria, Ppia−/− osteoblasts demonstrate decreased osteogenic differentiation, whereas Ppia−/− osteoclasts derived from the long bones showed increased osteoclastic activity. Overexpression and gene silencing of CypA verified osteogenic and anti-osteoclastic effects. In osteoblasts, CypA is necessary for BMP-2 (Bone Morphogenetic Protein-2)-induced Smad phosphorylation. In osteoclasts, loss of CypA activates BtK (Bruton’s tyrosine kinase) and subsequently integrates with TRAF6 (TNF receptor-associated factor 6) and/or c-fos signaling to induce NFATc1 (nuclear factors of activated T cells, cytoplasmic 1). Collectively, CypA dually exerts pro-osteogenic and anti-osteoclastic effects. Thus, modulation of CypA may be useful in future efforts targeting osteoporosis.

accompanied by a reduction in the size of the skeleton, reduced osteoblast numbers, and increased osteoclast numbers. In vitro studies confirmed that CypA had dual pro-osteogenic, anti-osteoclastic effects. Moreover, CypA appeared to differentially regulate signaling pathways in osteoblasts and osteoclasts. In osteoblasts, CypA was required for BMP-2 induced Smad phosphorylation, while in osteoclasts loss of CypA resulted in BtK (Bruton's tyrosine kinase) activation and integrated with TRAF6 (TNF receptor-associated factor 6) and/or c-fos signaling to induce NFATc1 (nuclear factors of activated T cells, cytoplasmic 1).

Results
High-resolution microCT analysis of Ppia −/− mice. The skeletal phenotypes of wildtype (Ppia +/+ ) and knockout (Ppia −/− ) mice were analyzed by microCT at three and 13 weeks of age. The overall size of the Ppia −/− skeleton was reduced, including a 26% reduction in mean tibia length (Fig. 1a). Next, quantitative differences in trabecular and cortical Bone Mineral Density (BMD), and fractional Bone Volume (BV/TV) were assessed. At three weeks, marked reductions in trabecular bone indices (Fig. 1c) and cortical bone indices (Fig. 1d) were observed. Analysis of the Ppia −/− skeleton at 13 weeks of age showed similar findings (Fig. 1e-h).
Next, the lumbar spine of Ppia +/+ and Ppia −/− mice was examined ( Fig. 1i-m). The Ppia −/− vertebral body showed a 40% reduction in height at three weeks of age, along with a visibly apparent decrease in cortical thickness and mineralization (Fig. 1i). Trabecular bone indices showed a reduction in Ppia −/− mice (Fig. 1j). Similar observations were also found in the lumbar vertebrae at 13 weeks of age ( Fig. 1k-m), including a reduction in vertebral height (Fig. 1k), and reduction or non-significant trend toward reduction in all trabecular bone indices (Fig. 1l,m).
In summary, Ppia −/− mice demonstrate a reduction in measured parameters of trabecular and cortical bone. These findings were consistent across both the appendicular and axial skeleton. Differences were most notable during early skeletal development, but persisted into skeletal maturity.
Histologic, histomorphometric and immunohistochemical analysis of Ppia −/− mice. Histologic analysis of Ppia −/− mice was performed in comparison to wildtype littermates to confirm our radiographic findings. At three weeks of age, a reduction in trabecular bone was observed (Fig. 2a). Histomorphometric analysis confirmed a reduction in B.Ar (Bone Area), percentage B.Ar (% B.Ar), and trabecular B.Pm (bone perimeter) (Fig. 2a). Next, quantification of the numbers of osteoblasts and osteoclasts within the Ppia −/− skeleton was performed. Ocn (Osteocalcin) immunohistochemical staining permitted identification of bone-lining osteoblasts, and showed a reduction in osteoblasts per B.Pm (Ocn + cells/B.Pm), Fig. 2b. Next, Cathepsin K immunostaining revealed an increase in Cathepsin K + multinucleated osteoclasts among Ppia −/− mice (Fig. 2c). These findings were also confirmed in newborn mice, again showing a reduction in osteoblasts and increase in osteoclasts (Fig. 2d,e). Next, we extended these findings to the cranial skeleton. A reduction in the thickness of the calvarial bone was observed in Ppia −/− mice (Fig. 2f), accompanied by a reduction in immunoreactivity for Ocn among periosteal cells and osteocytes (Fig. 2g).
In summary, Ppia −/− mice demonstrate a reduction in osteoblast numbers and a converse increase in osteoclast numbers. Moreover, this phenotype was observed across developmental timepoints and locations.
CypA silencing or forced overexpression was next performed in MC3T3-E1 osteoblastic cells. Two clones for CypA silencing are referred to as M1 and M3, while two clones for overexpression are designated P1 and P2. Western blot confirmed the efficacy of CypA silencing and overexpression (Fig. 3d). Next, examination of osteogenic genes was examined (Fig. 3e). Across all markers, CypA underexpression led to a reduction in osteogenic gene expression. Conversely, CypA overexpression led to an increase in osteogenic markers. ALP and Alizarin red staining confirmed our previous gene expression findings (Fig. 3f).
Next, CypA silencing or overexpression in RAW294.7 osteoclastic cells was performed. Two clones for CypA silencing are referred to as R1 and R2, while two clones for overexpression are designated P1 and P2. Western blot confirmed the efficacy of CypA dysregulation (Fig. 3j). Next, osteoclastic activity was assessed by TRAP staining (Fig. 3k). CypA silencing increased TRAP staining, while forced CypA overexpression reduced TRAP staining. Likewise, CypA silencing led to an increase in osteoclast markers (Fig. 3l). Conversely, CypA overexpression led to a reduction in osteoclast markers. Finally, CypA knockdown resulted in increased resorptive activity (Fig. 3m).
In summary, CypA had inverse effects on osteoblastic and osteoclastic activity. CypA enhanced osteoblastic differentiation while inhibiting osteoclastic activity. These findings were in agreement with the skeletal phenotype of the Ppia −/− mouse. or without CypA silencing) (Fig. 4a-c). CypA silencing resulted in a reduction in the degree and length of BMP-2 induced Smad1/5/8 phosphorylation (Fig. 4a). Similar findings were confirmed with Id1 (Inhibitor of differentiation)-luciferase activity (Fig. 4b). Next, co-immunoprecipitation assays were performed for Smad1 and Smad4 in the context of CypA silencing (Fig. 4c). Loss of CypA led to a reduction in Smad4/Smad1 binding (Fig. 4c). Thus, CypA expression is required for BMP-2 induced Smad1/5/8 phosphorylation, which subsequently recruits Smad4 into the complex to exert transcriptional activator to regulate Runx2 and results in Smad1/Smad4 binding (see Fig. 4d for a schematic representation).
Next, the downstream effects of CypA loss were examined in RAW297.4 osteoclastic cells ( Fig. 4e-h). First, phospho-BTK, c-fos and NFATc1 expression were examined across early osteoclastic induction (days 0-3) with CypA silencing (Fig. 4e). CypA knockdown resulted in an increase in BTK phosphorylation and NFATc1 expression (Fig. 4e). Next, co-immunoprecipitation assays were performed to evaluate CypA (myc-tagged CypA) binding to BTK (Fig. 4f). Results suggested that CypA interacts with BTK and hinders BTK phosphorylation. Finally, NFATc1 expression among the R1 clone was examined in the context of multiple signaling inhibitors, including the BTK inhibitor PCI-32765 (Fig. 4g). Results showed that PCI-32765 substantially reduced NFATc1 expression, while the NF-κ B inhibitor CAPE (Caffeic acid phenethyl ester) or JNK inhibitor partially reduced NFATc1 expression. Thus, loss of CypA resulted in BTK activation and integrated with TRAF6 and/or c-fos signaling to induce NFATc1 expression in osteoclastic cells (see Fig. 4h for a schematic representation). Our results are in agreement with the previously reported crucial roles of the tyrosine kinases Btk and Tec in RANKL-induced osteoclastogenesis, based on the genetic evidence obtained from Tec −/− , Btk −/− mice 15 .

Discussion
In summary, we have identified a novel dual function of CypA in the regulation of osteoblastic and osteoclastic activity. Although further study is warranted, CypA appears to differentially regulate signaling pathways in osteoblasts and osteoclasts. In osteoblasts, CypA is required for BMP-2 induced phosphorylation of Smad1/5/8.  In osteoclasts, loss of CypA activates BtK and induces NFATc1 expression via TRAF6 and/or c-fos signaling to induce NFATc1 expression.
Although the Ppia −/− mouse shows a low BMD skeletal phenotype throughout its life, the manifestations are most severe during early development. Unfortunately, a skeletal analysis at later ages is not possible due to postnatal lethality. One potential explanation for these age-dependent effects may be attributed to the activity of other isomerases which could compensate for the loss of function of CypA in older age. This result is similar to the previous report that Phd3 (prolyl hydroxylase 3) compensates for the loss of Phd2 (prolyl hydroxylase 2) in the vertebrae 16 . Further studies are needed to investigate if a similar compensation is observed in the Ppia −/− mouse skeleton.
One of the surprising findings of the current study was the anti-osteoclastic effect of CypA. Our in vitro findings suggest that CypA has direct effects on osteoclasts, although secondary paracrine changes produced by CypA deficiency within the bone marrow milieu cannot be excluded. The manifestations of CypA deficiency are distinct from the results of Pin1 (peptidyl-prolyl cis-trans isomerase NIMA-interacting 1) deficiency 17 . Pin1, another peptidyl-prolyl isomerase, was recently reported to promote osteoblastic activity 17 . For example, Pin1 deficient mice exhibit a reduced BMD skeletal phenotype accompanied by decreased osteoblastic activity and osteoblast-associated BMP signaling. Moreover, Pin1 appears to regulate BMP signaling, potentially via interaction with Smad5 17 . Despite these similarities between CypA and Pin1, their effects on osteoclasts are distinct. While CypA negatively regulates both the number and activity of osteoclasts, Pin1 deficiency seems to have no effect 17 . These similarities and differences suggest that CypA and Pin1 exert distinct molecular mechanisms to ultimately affect bone anabolism and bone resorption.
Although CypA clearly exerts important skeletal effects, the precise cell type most affected by CypA deficiency is unknown. We previously reported on the effects of CypA on cartilage, showing that CypA acts on committed chondrocytes primarily via regulation of NF-kB and the transcription factor Sox9 6 . In the present study it is clear that CypA acts on committed osteoblasts via regulation of Smad1/5/8 phosphorylation, subsequent Smad4 recruitment, and regulation of Runx2. The potent effects of CypA on cells of the osteochondral lineage suggest the possibility that CypA exerts its effects on mesenchymal stem cells and their progeny. Overall, the relative importance of CypA on differentiated versus undifferentiated osteochondral cells is an interesting topic of future investigation.
The clinical relevance of peptidyl-prolyl isomerase regulation largely comes from studies examining the effects of CsA (cyclosporin A) on bone. CsA has been recognized to have inhibitory effects on PPIase activity 18,19 . While not entirely agreed upon, most investigators have reported that CsA has detrimental effects on bone metabolism [20][21][22] . For example, high dose CsA treatment has catabolic effects resulting in reduced BMD in mice in vivo 20 . These studies were further confirmed by Chen et al., who reported that CsA treatment of rats resulted in reduced BMD, reduced biomechanical strength, and increased rate of bone fracture 21 . However, investigators have found that effects of CsA may be bimodal. For example, Yeo et al. found that while high dose CsA has catabolic effects, low dose CsA appears to have anabolic effects in mice 20 . Similarly, CsA was reported to have anabolic effects in rat alveolar bone formation 23 . Of note, however, upregulation of CypA has been associated with alveolar bone destruction in an experimental model of periodontitis 24 . Thus, it is important to consider the role of PPIase activity in not only the bone, but also the surrounding soft tissues. In summary, these findings support the concept that inhibition of PPIase activity is detrimental to bone health, whereas promotion of PPIase activity may have anabolic effects.
Of course, CypA is not alone in the dual regulation of osteoblastic and osteoclastic activity. Denosumab, a human monoclonal antibody to RANK ligand, is a newer anabolic agent that is administered by subcutaneous injections at 6-month intervals 25 . Denosumab has been shown to inhibit osteoclast activity, decrease bone resorption, and increase bone density. An ongoing Phase II clinical trial (NCT02049866) is examining the use of Denosumab for the prevention of post-teriparatide bone loss in premenopausal women 26 . Anti-Sclerostin antibody (also known as Romosozumab) is another dual anabolic, anti-osteoclastic treatment. In Phase I and II trials, anti-Sclerostin antibody increased bone density of both healthy men and postmenopausal women. Phase III trials are expected to end by late 2015 27,28 . Thus, modulation of CypA activity may represent a novel option to regulate both osteoblast and osteoclast activity in order to combat osteoporotic bone loss. Three-dimensional Micro-computed Tomography Evaluation. Long bones and vertebrae were analyzed from mice at three and 13 weeks of age by microCT (micro-computed tomography), as previously described 6 . Six WT and five KO mice in three weeks of age, and six WT and four KO mice in thirteen weeks of age were processed for microCT analysis. WT and KO mice were processed for microCT analysis. Briefly, whole bodies were imaged using high-resolution microCT (SkyScan 1172, Bruker MicroCT N.V., Kontich, Belgium) at an image resolution of 27.4 μm (39 kV and 246 μA radiation source, using a 0.5 mm aluminum filter). The quantitative and structural parameters follow the nomenclatures described by the American Society for Bone and Mineral Research Nomenclature Committee 29 . Common sites of interest were examined, including the tibia, and individual lumbar vertebral bodies (the three terminal vertebral levels).

Methods
The optimal threshold for each developing skeletal site was determined based on the normal bone in WT mice. For both three weeks and 13 weeks old animals, threshold ranges of 60-255 and 80-255 were used for trabecular analysis of tibias, and lumbar vertebrae, respectively, and 110-255 was used for cortical analysis of femurs and tibias. Such threshold ranges were selected to optimally represent the normal structures in WT bones, while including slightly less mineralized woven bone structures in the trabecular threshold, and excluding trabecular structures in the cortical threshold.
Briefly, 800 μg of cell lysate was incubated with anti-myc or BTK antibodies at 4 °C overnight in a clinical rotor. Protein A/G PLUS-agarose beads (Santa Cruz) were then added to the mixture and rotated for 3 hours at 4 °C. The tubes were briefly centrifuged at 1000 g at 4 °C for 30 seconds to precipitate the beads bound to the Ag-Ab complex, which was subjected to Western blot analysis probing with anti-myc or anti-BTK antibody.
Knockdown and overexpression of CypA in MC3T3 and RAW264.7 cells. Generation of stable knockdown cell lines was performed as previously described 6,14,32 . To knockdown CypA in MC3T3 and RAW264.1 cells, the cells were first transfected with the pSilencer-CypARNAi vector 12 (synthesized by Invitrogen, Carlsbad, CA, USA) using lipofectamine 2000 (Invitrogen) for 48 hours, followed by screening in 500 mg/ml of G418 (Sigma-Aldrich) for 48 hours. Twelve single clones of each cell line were selected and maintained in DMEM (Dulbecco's Modified Eagle's Medium) with 10% FBS and 200 mg/ml of G418 for two weeks. For CypA overexpression, cells were transfected with CypA-pcDNA3.1 plasmid 12 using lipofectamine 2000 for 48 hours. Western blot was performed to determine the efficiency of knockdown or overexpression. Based on their efficiency, M1 and M3 clones (MC3T3-E1 cells), and R1 and R2 single clones (RAW264.7) were selected and compared to SC (scrambled clones) as well as wild-type parental cells.
Mineralization assay was performed as previously described 6 . Briefly, primary mouse osteoblasts or MC3T3-E1 cells were cultured in 24-well plates and osteogenic differentiation was performed for three weeks. Cells were stained with Alizarin Red S solution (PH 4.2, Sigma-Aldrich) for 10 min at room temperature and washed three times with water. TRAP (Tartrate-resistant acid phosphatase) staining. Primary mouse osteoclasts and RAW264.7 cells were seeded in 24-well plates and differentiated for seven days. After induction of differentiation, cells were washed with phosphate buffered saline (PBS) and fixed in paraformaldehyde 4% in PBS for 10 min, following by incubating for 20 min at 37 °C in the TRAP staining solution (Sigma-Aldrich) according to the manufacturer's instructions.
Resorption assay and assessments. Bone resorption assay was performed as previously described 33 .
Briefly, RAW 264.7 cells (5 × 10 3 cells per well) were seeded in Corning Osteo-Assay Surface 96-well plates (Corning Life Sciences) in 500 μl of complete α -MEM containing 10% FBS, 10 −7 M PGE2, 10 ng/ml M-CSF and 10 ng/ml recombinant RANKL. Cells were incubated at 37 °C in a humidified 5% CO 2 incubator and medium was changed every day. The cells were incubated for seven days prior to processing for Von Kossa staining.
To perform Von Kossa staining, plates were stripped with 10% bleach for 5 min to remove cells, rinsed with distilled water, and air-dried until further use. Plates were treated with 500 μl/well 1% (w/v) silver nitrate solution and placed in front of a 75 watt light bulb for 1 hour. Wells were then washed for 5 min in distilled water. Un-reacted silver was removed with 5% (w/v) sodium thiosulfate (500 μl/well, 5 min). The plates were then washed and air dried prior to imaging. All microplate wells were imaged with six fields/well (10× ). Percentage of the resorbed surface was carried out with NIH ImageJ software.

Statistical analysis.
All values are expressed as mean ± SEM (standard error of the mean), unless otherwise stated. One way ANOVA analysis of variance and Student's t-test was analyzed using SPSS 20.0 for windows. P values of less than 0.05 were considered to be significant. All experiments were performed in triplicate and repeated at least twice.