TGFβ is secreted by bone cells. Therefore, bone represents one of the biggest reservoirs for all three TGFβ isoforms (β1, β2 and β3) in the human body. In bone matrix, the TGFβ isoforms are present in their latent form, which become activated upon need (for example, during bone resorption by osteoclasts). The relevance of TGFβ for bone formation physiology is underlined by the finding that TGFβ1 knockout mice have a decreased tibia length of about 30% and a reduced bone mineral content . Furthermore, local injection of TGFβ1 under the periosteum stimulates the formation of cartilage and bone [19, 20] and systemic application of TGFβ2 leads to a general increase in osteoblast activity . In contrast, transgenic mice with continuous overexpression of TGFβ2 in osteoblasts show a dramatic, age-dependent loss of bone mass . Along the same lines, transgenic mice lacking functional TGFβ signaling in osteoblasts  or mice treated with the TGFβ type I receptor kinase inhibitor SD-208  have increased trabecular bone mass with tougher femurs and stiffer and stronger vertebral bodies. Contrary to earlier findings, these data suggest that continuous exposure to active TGFβ might harm bone physiology, as can be seen in patients suffering from chronic inflammation, whose active TGFβ1 serum levels are often constantly increased. In an earlier work, we showed that continuously elevated rhTGFβ1 levels inhibit osteoblast function, for example, alkaline phosphatase (AP) activity and formation of mineralized matrix. One possible mechanism by which TGFβ1 may exert its inhibitory effect on osteoblast differentiation is interfering with BMP signaling . Therefore, the aim of this study was to investigate possible regulatory mechanisms by which rhTGFβ1 inhibits rhBMP-2 and rhBMP-7 signaling in primary human osteoblasts. We demonstrated that rhBMP-2 and rhBMP-7 induce Smad1/5/8 signaling primary osteoblasts, isolated from femoral heads of patients undergoing total hip replacement. Upon a single stimulation with the cytokines, the signaling reached its peak after 72 h. Coincubation with only one-tenth of the amount of rhTGFβ1 completely abrogated the rhBMP-2-induced and rhBMP-7-induced Smad1/5/8 signaling in these cells. The opposite is seen in vivo in adult (rodent) kidney, where BMP-7 is expressed and can, when administered exogenously, reduce TGFβ-driven renal fibrogenesis during experimental chronic nephropathies . Expression analysis of the transcription factors (Smad1-5), receptors (Alk1-3, Alk5, Alk6, TGFβRII) and regulatory factors (Smad6, Smad7, Smurf1, Smurf2, SARA, BAMBI, Ski, SnoN, noggin, sclerostin) involved in BMP or TGFβ signaling, revealed that rhTGFβ1 downregulates the expression of Smad1, Alk1 and TGFβRII, both at mRNA and at protein level. This might explain the lack of Smad1/5/8 signaling observed in osteoblasts treated with rhBMP-2 and rhBMP-7 costimulated with rhTGFβ1.
Interestingly, expression of Smad6 was also downregulated by rhTGFβ1, which should enhance Smad1/5/8 signaling by reducing ubiquitination and degradation of Smad1/5/8 and the corresponding receptors by the E3 ubiquitin ligases Smurf1 and Smurf2 , where expression in osteoblasts was not affected in our setting. On the contrary, expression of the other inhibitors Smad6 and Smad7 was upregulated. As Smad7 binds to the activated receptors in competition with Smad2/3 , and thus serves as a negative feedback regulator for TGFβ1-dependent Smad2/3 signaling, its induction was not further investigated at this point. Expression levels of the other transcription factors and receptors were not significantly altered in our experimental setup. Similarly, expression levels of the majority of the investigated regulatory factors investigated were not significantly altered in the presence of rhBMP-2, rhBMP-7 or rhTGFβ1, the only exceptions being BAMBI and SnoN. The expression level of SnoN was strongly increased in the presence of rhTGFβ1. SnoN interferes with TGFβ signaling by interacting directly with Smad3 . Furthermore, SnoN is reported to antagonize TGFβ signaling on the transcriptional level via recruitment of HDACs . Consequently, we demonstrated increased general HDAC activity in rhTGFβ1-treated human osteoblasts, which might be responsible for the observed decreased expression levels of Smad1, Smad6, Alk1 and TGFβRII. HDAC activity was effectively blocked by the administration of two subtoxic doses (100 μM and 200 μM) of valproic acid. Blocking HDAC activity by valproic acid was able to abolish the rhTGFβ1-dependent inhibition of rhBMP-2-induced and rhBMP-7-induced Smad1/5/8 signaling in our setup. However, valproic acid does have severe side effects, thus detailed characterization of the specific HDACs regulated by TGFβ could identify a more specific HDAC inhibitor for use in patients with less side effects. Interestingly, BAMBI expression levels were slightly downregulated in the presence of rhTGFβ1 in our system. This should enhance rhBMP-2 and rhBMP-7 signaling as BAMBI, similar to noggin or sclerostin, has been reported to negatively affect bone formation in vivo by directly interfering with ligand-receptor binding, thus inhibiting both BMP and TGFβ receptor binding [27, 28]. In contrast to that, SnoN affects both TGFβ and BMP signaling via transcriptional regulation. This points towards a possible novel mechanism how rhBMP-2 and rhBMP-7 fracture therapies in patients could be optimized. Valproic acid is already in clinical use as one of the most common antiepileptic drugs with proposed off-label use as anticancer drug [29, 30]. However, it still lacks evaluation in vivo as valproic acid is reported to have severe side effects, for example, embryotoxicity. By identification of the specific HDACs regulated by TGFβ, an alternative HDAC inhibitor with fewer side effects could be chosen. Furthermore, as this study focused on primary human osteoblasts as the major target for BMP therapy, the effects of the chosen HDAC inhibitor on bone resorption by osteoclasts should be also investigated. The latter in particularly is limiting for the present study set, since little is known of how HDAC inhibitors may interact with osteoclasts or with a corresponding coculture system. Despite the overall positive results on the use of rhBMP-2 or rhBMP-7 on bone as an adjunct or as a replacement for autograft in compromised patients [2–7], several adverse events, for example, infections, hardware failure, pain, donor site morbidity, heterotopic bone formation and immunogenic reactions, have been reported nonetheless [8, 9]. In the present experiments addition of valproic acid not only abolished the inhibitory effect of rhTGFβ1 on rhBMP-2 and rhBMP-7 signaling, but even increased Smad1/5/8 signaling. This is supported by the findings of Schroeder and Westendorf that show that application of HDAC inhibitors, trichostatin A, sodium burtyrate, valproic acid and MS-275 favors osteoblasts maturation in MC3T3-E1 cells by upregulation of RUNX2 . Interestingly, patients with epilepsy show an increased fracture risk. Thus, it has be extensively studied how antiepileptic drugs affect bone turnover; however, no correlation between valproate medication and loss in bone mineral density was observed [32, 33]. This favors the use of this drug for improving BMP therapy. However, it should be further investigated in which way the drug should be administered for improving BMP therapy. An oral application may have advantages for the clinicians to treat patients that already show BMP therapy failure without an additional operation. However, oral administration of valproate may hold the risk for more side effects from the drug per se. This could be limited by a local application of the drug in combination with the BMP itself. Also, to further minimize possible adverse effects by the applied drug it should be further clarified which HDACs are involved in the observed gene regulation to possibly choose a more selective inhibitor with fewer side effects.