Inhibition of phosphodiesterase 5 reduces bone mass by suppression of canonical Wnt signaling

Inhibitors of phosphodiesterase 5 (PDE5) are widely used to treat erectile dysfunction and pulmonary hypertension in clinics. PDE5, cyclic guanosine monophosphate (cGMP), and protein kinase G (PKG) are important components of the non-canonical Wnt signaling. This study aimed to investigate the effect of PDE5 inhibition on canonical Wnt signaling and osteoblastogenesis, using both in vitro cell culture and in vivo animal models. In the in vitro experiments, PDE5 inhibition resulted in activation of cGMP-dependent protein kinase 2 and consequent inhibition of glycogen synthase kinase 3β phosphorylation, destabilization of cytosolic β-catenin and the ultimate suppression of canonical Wnt signaling and reduced osteoblastic differentiation in HEK293T and C3H10T1/2 cells. In animal experiments, systemic inhibition of PDE5 suppressed the activity of canonical Wnt signaling and osteoblastogenesis in bone marrow-derived stromal cells, resulting in the reduction of bone mass in wild-type adult C57B/6 mice, significantly attenuated secreted Frizzled-related protein-1 (SFRP1) deletion-induced activation of canonical Wnt signaling and excessive bone growth in adult SFRP1−/− mice. Together, these results uncover a hitherto uncharacterized role of PDE5/cGMP/PKG signaling in bone homeostasis and provide the evidence that long-term treatment with PDE5 inhibitors at a high dosage may potentially cause bone catabolism.

In the canonical Wnt (Wnt/β-catenin (β-cat)) signaling cascade, Wnt binds to Frizzled (Frz) receptors and the low-density lipoprotein receptor-related protein (LRP) 5 or 6, thereby activating dishevelled, suppressing the glycogen synthase kinase 3β (GSK3β) activity and inhibiting phosphorylation of β-cat at Thr41, Ser37, and Ser33 sites. The stabilized cytosolic β-cat enters the nucleus and consequently activates its downstream target genes via lymphoid enhancer-binding factor-1 (Lef-1) and T-cell factors. 1,2 This signaling is finetuned in part via a negative feedback mechanism involving secreted and transmembrane Wnt inhibitors and activators, secreted Frz-related proteins (SFRPs), and Dickkopf-1 (Dkk1). 3,4 Canonical Wnt signaling is critical not only to bone development in embryogenesis but also to the maintenance of bone mass during adult life. 5 The initial evidence came from the discoveries that in humans loss-or gain-of-function mutations in LRP5 were linked with the osteoporosispseudoglioma syndrome and a high-bone-density syndrome, respectively. [6][7][8] Subsequent studies in mice showed that Wnt signaling might promote ossification by inducing the differentiation of bone-forming osteoblasts, suppressing the development of bone-resorbing osteoclasts, and driving the differentiation of multi-potent stem cells toward an osteoblast cell fate. 9 Non-canonical Wnt signaling is β-cat independent and consists of two main pathways: the Rho small GTPasesmediated planar cell polarity pathway and the Wnt/Ca 2+ pathway, 10 involved in various aspects of cell fate differentiation and cell movement. Non-canonical Wnt signaling has profound effects on tissue morphogenesis in a variety of vertebrate species. 10 The potential role for non-canonical Wnt signaling in bone formation has been investigated recently in limited studies, which have shown that the non-canonical Wnt-Gαq/11-PKC pathway operates in mammalian osteoprogenitors to promote osteoblast development, and that Wnt16 exhibits a stimulatory effect on bone metabolism. [11][12][13] Nevertheless, the molecular events in the non-canonical Wnt signaling regulation of bone development and homeostasis have yet to be further elucidated.
Phosphodiesterases (PDEs) are a large family of enzymes that cleave cyclic nucleotides. To date, 11 PDE subtypes have been identified, among which PDE5 has been most extensively studied. PDE5, cyclic guanosine monophosphate (cGMP), and cGMP-dependent protein kinase (PKG) are among the major components of the non-canonical Wnt signaling pathway and are involved in the regulation of intracellular Ca 2+ concentration. 14,15 It is now well established that PDE5 degrades 3'-5′-cGMP and its inhibition leads to an increase in intracellular cGMP levels and activation of protein kinase G (PKG), resulting in a decrease in Ca 2+ influx and consequent relaxation of smooth muscles, which produces the therapeutic effects in clinical erectile dysfunction (ED) and pulmonary hypertension (PH). 14 Currently, little is known regarding the involvement of PDE5 in Wnt signaling regulation of bone formation and homeostasis. The objective of this study was to determine the effect of PDE5 inhibition on canonical Wnt signaling and bone mass.

Results
Involvement of PDE5 and PKG in canonical Wnt signaling. To examine the possibility of PDE5 and PKG in canonical Wnt signaling, we determined the effects of PDE5 and PKG inhibitors, and PDE5-and PKG-specific siRNAs on Lef1 reporter activity in 293 T cells. The two PDE5 inhibitors (i.e., tadalafil and vardenafil) used inhibited the Lef1-luciferase activity in a dose-dependent manner; at 10 μM, tadalafil and vardenafil decreased the reporter activity by 58 and 30%, respectively, in the presence of control L cellconditioned medium (L), and by 42 and 60%, respectively, in the presence of Wnt3a-expressing cell-conditioned medium (Wnt3a) (Figure 1a and Supplementary Figure 1a). In contrast, KT5823, a potent inhibitor of PKG, dose dependently induced the Lef1-luciferase activity in 293 T cells; at 5 μM, it increased the reporter activity by 2.4-and 1.5-fold, respectively, in the presence of control L cell-conditioned medium and Wnt3a-expressing cell-conditioned medium  Figure 1b). PDE5a siRNA decreased PDE5a expression by~60% at the protein level and suppressed Lef1-luciferase activity by 15 and 45% in the presence of control L cell-conditioned medium and Wnt3aexpressing cell-conditioned medium, respectively (Figures 1c and e). PKG1 and PKG2 siRNAs, each lone, reduced PKG1 and PKG2 protein contents by 60 and 90%, respectively, whereas the combination of PKG1 and PKG2 siRNAs reduced PKG1 and PKG2 protein contents by 70 and 80%, respectively ( Figure 1f). Silencing of PKG2 but not PKG1 resulted in a decrease in Lef1-luciferase activities by 20 and 33% in the presence of control L cell-conditioned medium and Wnt3a-expressing cell-conditioned medium, respectively, and knockdown of both PKG1 and PKG2 decreased Lef1luciferase activities by~70 and 75% in the presence of the two conditioned media, respectively ( Figure 1d). Thus, inhibition of PDE5 negates canonical Wnt signaling, in line with this observation, and suppression of PKG, especially PKG2, induces canonical Wnt signaling in either the presence or absence of Wnt3a.
Effects of PDE5 inhibition on β-cat expression and stabilization. Given that β-cat is the central signal transducer of the Wnt signaling pathway, we evaluated the effect of PDE5 inhibition on β-cat messenger RNA (mRNA) abundance and protein content, and stabilization in 293 T cells, to further assess the possible involvement of PDE5 in Wnt signaling. PDE5 inhibitor tadalafil did not significantly change β-cat mRNA levels at either 5 or 10 μM but attenuated Wnt3a-induced increases in β-cat protein levels in whole-cell lysates, cytosolic, and nuclear fractions by 58, 60, and 65%, respectively, at 10 μM (Supplementary Figure 2a and Figure 2a). Immunostaining further confirmed the effect of tadalafil on β-cat protein content; at 10 μM, tadalafil decreased both cytosolic and nuclear β-cat protein levels in response to recombinant Wnt3a protein ( Figure 2b). Thus, inhibition of PDE5 negates canonical Wnt signaling through decreasing β-cat levels.
Phosphorylation of β-cat at Ser45 by casein kinase-1 (CK1) and at Thr41, Ser37, and Ser33 by GSK3β regulates its stabilization in the cytosol, and the phosphorylated β-cat is recognized by E3 ubiquitin ligase and undergoes proteolytic degradation. 16 To determine the role of PDE5 in the stabilization of β-cat, we assessed changes in β-cat phosphorylation in 293T cells treated with tadalafil at 10 μM, in either the presence or absence of Wnt3a. Wnt3a slightly decreased the phosphorylation of β-cat at Ser45, whereas tadalafil showed no significant effect (Supplementary Figure 2b). Phosphorylation of β-cat at Ser33, Ser37, and Thr41 was reduced by Wnt3a (80%) but induced by tadalafil (1.4-fold), and tadalafil was capable of reversing the negative effect of Wnt3a on β-cat phosphorylation (3-fold) (Figure 3a). In contrast, Wnt3a increased (2.5-fold) but tadalafil decreased (40%) GSK3β phosphorylation at Ser9, and tadalafil was effective to attenuate the Wnt3a-induced GSK3β phosphorylation (60%) (Figure 3a).
Finally, we performed reporter assays and co-immunoprecipitation to confirm GSK3β signaling as a downstream event of PDE5/cGMP/PKG2 signals in the regulation of β-cat stabilization. GSK3β inhibitor, SB216763, increased Lef1-luciferase activity but showed no significant effect in the presence of tadalafil in 293T cells (Figure 3d). However, the presence of tadalafil had no effect on Wnt3a induction of Lef1luciferase activity in 293T cells expressing GSK3β siRNA (GSK3βsi) where GSK3β protein content was decreased by 60~80% (Figures 3e and f). In cells treated with Wnt3a alone, PDE5 and bone homeostasis Y Gong et al protein complex precipitated with a GSK3β antibody contained a considerable amount of PKG2; however, in cells treated with other test agents, protein complex contained a small amount of PKG2 in addition to GSK3β, as expected ( Figure 3g). Taken together, these observations suggest that activation of PKG2 by inhibition of PDE5 may suppress canonical Wnt signaling through inducing GSK3β-mediated phosphorylation of β-cat at Thr41, Ser37, and Ser33 sites. In contrast, tadalafil reduced not only Wnt3a-induced increases in mRNA levels of Lef1 (72%) and DKK1 (75%) but also of osteoblastic markers including AP (74%), OSX (86%), and Runx2 (67%), and AP activities (33%) and formation of mineralized nodules (30%) (Figures 4b-i). Thus, in C3H10T1/2 cells, inhibition of PDE5 suppresses not only the canonical Wnt signaling but also the osteoblastogenesis in response to Wnt3a.
Decreases in bone mass after systemic inhibition of PDE5 in vivo. To assess the potential role of inhibition of PDE5 in bone mass in vivo, we first evaluated the effect of tadalafil on osteoblastogenesis and Wnt/β-cat signaling in wild-type adult (2 months) C57BL/6 mice. Orally administered for 2 months at 45 or75 mg/kg daily that was 20-and 32-fold higher than the clinical dosage for ED (10 mg daily) and We next assessed the effect of tadalafil on osteoblastogenesis and Wnt signaling in 2-month-old SFRP1 knockout (SFRP1 − / − ) mice. At the baseline level, SFRP1 − / − mice had a significantly higher mass of cancellous bone but not cortical bone as compared with SFRP1 +/ − mice, as revealed by histological analysis of the longitudinal sections of the distal femur. At an oral dose of 75 mg/kg daily for 2 months, tadalafil robustly decreased the mass of cancellous bone but not in cortical bone in SFRP1 − / − mice (Figure 6a). Threedimensional reconstruction of the distal femur using μCT further confirmed that there were significant increases in BMD (2.1-fold), BMTV (2.3-fold), and TbN (1.5-fold) but not in TbTh and TbSp in SFRP1 − / − mice as compared with SFRP1 +/ − mice, and that tadalafil treatment resulted in significant decreases in BMD (38%), BMTV (28%), and TbN (47%), but not in TbTh and TbSp as compared with the vehicle treatment (Figures 6a-c, and Supplementary Figure 4c and d). Likewise, BMSCs from SFRP1 − / − mice formed much more mineralized nodules than those from SFRP1 +/ − mice (1.4-fold), and tadalafil attenuated SFRP1 knockout-associated excessive formation of mineralized nodules by~25% over vehicle (Figures 6d and e). Moreover, BMSCs from SFRP1 − / − mice exhibited higher transcription activities of Lef1 and Dkk1 (13-and 14-fold, respectively) than those from SFRP1 +/ − β-catenin Tada.
-   Figure 3a and 3b). Together, these results further confirm that systemic inhibition of PDE5 specifically reduces the excessive bone growth derived from forced activation of canonical Wnt signaling in adult SFRP1 − / − mice.

Discussion
By using wild-type C57BL/6 and SFRP1 knockout mice, we have uncovered that the PDE/cGMP/PKG2 signaling operates in conjunction with GSK3β-mediated β-cat stabilization to regulate canonical Wnt signaling in maintenance of bone mass in adult mice in vivo. Using 293T and C3H10T1/2 cells, we have shown that inhibition of PDE5 induces cGMPdependent PKG2, which activates GSK3β and thereby destabilizes β-cat in the cytosol, resulting in inhibition of canonical Wnt signaling and consequent reduction of osteoblastic differentiation in vitro (Figure 7). In one of our previous studies, we have demonstrated that Wnt3a through LRP5/6, Dvl, and most likely Frz receptors, activates a signaling module composed of Gαq/11βγ-PI3K-Rac1-JNK2, resulting in stabilization of β-cat through phosphorylation at Ser191 and Ser605, and consequent localization to the nucleus. 16 Here we show that PDE5 inhibition leads to activation of cGMP/PKG2, which destabilizes β-cat and consequently suppresses canonical Wnt signaling, supporting a role for the non-canonical pathway in the regulation of canonical Wnt signaling. 19 The mechanisms underlying the activation of canonical versus non-canonical pathways by Wnt ligands are not fully understood. 11 It has been suggested that the Frz2/Gαi/PDE5/cGMP/PKG signaling module is an important component of non-canonical Wnt signaling in the activation of intracellular Ca 2+ transient. 15 Here we report that PDE5/cGMP/PKG signals are also involved in canonical Wnt signaling. It seems that non-canonical Wnt signaling and canonical Wnt signaling operate interactively and the binary distinctions of these two Wnt pathways have come under scrutiny. As overexpression of Frz4 in HEK293 cells or Frz5 in Xenopus embryos is sufficient to transduce canonical signaling by Wnt5a, 20,21 the specificity of the underlying signaling complexes may be dictated by the Frz receptor(s).
A previous study has observed that activation of Frz2 induces the cGMP-dependent phosdiesterase and decreases the intracellular cGMP levels, 15 suggesting that Wnt3a fails to activate Frz2, and that PDE5/cGMP/PKG signals regulate canonical Wnt signaling independently of Wnt3a. The current study has shown that PDE5/cGMP/PKG2 signals regulate the stabilization of β-cat at the protein but not the mRNA level. Inconsistent to this observation, Li et al. 22 and Tinsley et al. 23 have shown that both non-steroidal anti-inflammatory drug, sulindac sulfide (SS), and PDE5 inhibitor, tadalafil, inhibit PDE5 activity and increase cGMP levels, resulting in PKG activation, decreased proliferation, and apoptosis through transcriptional suppression of β-cat, but not its proteolytic degradation in colon cancer cells, and more recently, Tinsley et al. 24 have demonstrated that SS may trigger β-cat protein degradation and suppress β-cat gene transcription through increased phosphorylation in human breast cancer cells. The inconsistency is possibly due to the use of different cell types in the previous studies (i.e., cancer cells) and this study (i.e., normal cells).
Inhibition of PDE5 and activation of cGMP/PKG2 induce the phosphorylation of β-cat at Thr41, Ser37, and Ser33 but not at Ser45, suggesting that GSK3β but not CK1 may be downstream of the PDE5/cGMP/PKG2 signaling module. This notion is further supported by the findings that knockdown of PKG2 is sufficient to activate the phosphorylation of GSK3β and the stabilization of β-cat. Finally, PKG2 and GSK3β coexisting within one immunoprecipitated complex prompts us to speculate that PKG2 directly phosphorylates GSK3β in the complex. In line with this speculation, several recent reports have shown physical interactions between PKG2 and GSK3β. In ATDC5 cells, PKG2 directly phosphorylates GSK3β to promote the hypertrophic differentiation. 25 In UMR106 cells, PKG2 also phosphorylates the GSK3β at Ser9, and the  26 Thus, the present study has identified a mechanism in which activation of cGMP/PKG2 by inhibition of PDE5 results in GSK3βmediated destabilization of β-cat and consequent suppression of canonical Wnt signaling.
The role for canonical Wnt signaling in the embryonic development of bone is well established, but the function in maintaining the bone homeostasis in adulthood is less clear. 27,28 With regard to non-canonical Wnt signaling, little is known regarding its role in bone development and homeostasis in both embryogenesis and adulthood. 11 Consistent with the role of PDE5/cGMP/PKG2 signaling in Wnt3ainduced osteoblastogenesis in vitro, systemic administration of a PDE5 inhibitor, tadalafil, reduces bone mass not only in normal but also in SFRP1 − / − knockout adult mice in this study. The severe phenotype after tadalafil treatment reflects the importance of PDE5 activities in the maintenance of bone homeostasis. Endogenous Wnt signaling plays an important role in bone formation through stimulating the osteoblastogenesis and suppressing the adipogenesis and osteoclastogenesis. 29 Our in vivo analyses show that: (1) inhibition of PDE5 reduces the trabecular bone but not the cortical bone, which is identical to the phenotypes of SFRP1 or LRP5 gain-and loss-of-function; [30][31][32] (2) inhibition of PDE5 leads to the robust inhibition of Lef1 and Dkk1 transcription in BMSCs; (3) inhibition of PDE5 almost abolishes the SFRP1 knockout-producing overgrowth of bone; and (4) neither inhibition of PDE5 nor SFRP1 gain-of-function affects the osteoclastogenic activities. Together, these in vivo observations imply that inhibition of PDE5 attenuates not only the osteoblastogenic differentiation in response to Wnt3a but also the activities of canonical Wnt signaling, suggesting that inhibition of PDE5 reduces osteoblastogenesis possibly through suppressing canonical Wnt signaling.
Nitric oxide synthase (NOS) is the enzyme responsible for producing NO and NO activates soluble guanylyl cyclases to increase the intracellular cGMP levels and activate PKG. 14 Consistent with our findings, previous studies have demonstrated that mice lacking endothelial NOS exhibit profound abnormalities in bone formation and a significant delay in osteoblastic differentiation, 33,34 and that mice lacking either PKG2 or PKG1 may develop severe phenotypes; however, analyses of the skeletons of PKG2 but not PKG1 mutants have revealed the obvious defects in chondrogenesis but not in osteoblastogenesis. 25,[35][36][37][38][39] It is likely that PKG2 and PKG1 play a functional redundancy in osteoblastogenesis.
In summary, the present study has demonstrated that PDE5 inhibition may cause bone mass loss involving non-canonical Wnt pathway, which operates interactively with canonical Wnt signaling in a way to destabilize β-cat through a cGMP/PKG2/ GSK3β signaling-dependent mechanism. Given that PDE5 inhibitors are commonly used in patients with ED and PH, our findings may have significant clinical implications in alerting physicians of putative adverse effect of PDE5 inhibitors.
Cell culture and conditioned medium preparation. C3H10T1/2 cells, Wnt3a-expressing and control L cells, and HEK293T cells were all obtained from ATCC (Manassas, VA, USA). For maintenance, C3H10T1/2 cells and HEK293T cells were cultured, respectively, in Basal Medium Eagle (Life Technologies) and Dulbecco's modified Eagle's medium (Life Technologies) supplemented with 10% fetal calf serum (FCS, Life Technologies). Primary bone marrow stromal cells (BMSCs) were isolated from the femur and tibia, and cultured in alpha Minimum Essential Media (Life Technologies) with 15% FCS as described previously. 11 Conditional media were prepared from Wnt3a-expressing and control L cells, respectively, as described previously. 16 Transient transfection and dual-luciferase assay. SV40 large T-antigen-expressing HEK cells (293T cells) were plated onto 24-well plates. Next day, they were transfected with 2 μg Lef1 reporter construct using Lipofectamine 2000 reagent and 0.02 μg Renilla luciferase construct (Promega, Madison, WI, USA) for 6 h in the absence of serum. 40 In some cases, the cells were cotransfected with siRNA oligonucleotides targeting the test genes PDE5a, PKG1, PKG2, and GSK3β (see Supplementary Table 1 for the specific sequence information), Lef1 reporter construct and Renilla construct. Transfectants were then cultured in a 1 : 2 diluted conditioned medium prepared from control L cells or Wnt3a-expressing cells in the presence of tadalafil and/or vardenafil for 48 h. At the end of the designated culture, cell lysates were prepared and dual-luciferase assay was performed according to the manufacturer's instructions (Promega). The firefly luciferase activity was normalized to the Renilla luciferase activity.
Western blotting, immunoprecipitation and immunocytochemistry. 293T and C3H10T1/2 cells were plated 1 × 10 4 and 1.5 × 10 4 cells/cm 2 overnight, after a variety of treatments. Cytosolic and nuclear fractions of cells were prepared by using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific, Waltham, MA, USA) as per the manufacturer's instructions. β-Actin and Lamin-B were used as the internal standards for the cytosolic and nuclear fractions, respectively. Western blot analysis and immunoprecipitation were performed using standard protocols and the immunoreactive signals for the proteins of interest were quantitated using the LI-COR Odyssey Infrared Imaging System (LI-COR Bioscience) and ImageJ (http://rsb.info.nih.gov/ij/). 293T cells seeded on chamber slides (Nalge Nunc International, Rochester, NY, USA) at 0.75 × 10 4 /cm 2 and cultured in regular medium overnight and in serum-free medium for 12 h. After pretreatment with tadalafil for 2 h, serum-starved cells were further stimulated with recombinant murine Wnt3a protein at 100 ng/ml in fresh serum-free medium for 30 min. Cells were then immunostained with primary antibodies against β-catenin and subsequently counterstained with 4',6-diamidino-2-phenylindole. Immunoreactive protein signals were examined by confocal microscopy.
Osteoblast differentiation assay and quantitative RT-PCR. C3H10T1/2 cells were seeded onto six-well plates. At confluence, cells were stimulated with conditioned medium prepared from control L or Wnt3aexpressing cells for 48 h in the presence of vehicle or tadalafil. AP activity was determined and expressed as nanomoles of p-nitrophenol formed per minute per milligram of protein as previously described. 41 . For mineralization assays, confluent cells were incubated in the presence of 50 mg/ml ascorbic acid and 50 mM β-glycerophosphate for 21 days. Total RNA was extracted using TRIzol reagent (Takara Biotechnology (Dalian) Co., Ltd., Dalian, China). mRNA levels of osteoblast markers and target genes of canonical Wnt signaling including Dkk1, Lef1, AP, Runx2, and Osx were determined by real-time PCR as previously described. 42,43 GAPDH was included as an internal control and the relative levels of mRNA species of interest were calculated by the 2 − ΔΔCt method. 44 The primers used in our PCR analysis are presented in the Supplementary Information (Supplementary Table 2).