BMP2 and mechanical loading cooperatively regulate immediate early signalling events in the BMP pathway

Despite considerable advances in regenerative medicine and orthopaedic surgery, delayed fracture healing or non-unions of fractures still represent an important clinical concern [1]. Bone morphogenetic proteins (BMPs) are major and indispensable players during bone repair [2, 3]. After approval by the Food and Drug Administration in 2001 and 2002, recombinant human BMP2 and recombinant human BMP7 have been used clinically in different applications. However, roughly 1,000 times the normal physiological concentration has to be administered, and in many cases treatment is not superior to autologous bone grafting [4, 5]. If BMPs are to be widely used as powerful stimuli, a molecular understanding of their functionality under physiological and diseased conditions appears mandatory.

BMPs belong to the transforming growth factor-β (TGF-β) superfamily and were originally described as being able to induce bone formation. Today, it is known that there are about 25 different BMPs and they are capable of doing much more: they guide many other processes during organ development, tissue homeostasis and repair [6]. However, BMP2, -4, -6, -7 and -9 in particular play pivotal roles in bone morphogenesis [7].

BMP ligands signal by binding to heteromeric complexes of two types of Ser/Thr kinase receptors (BMP type I and type II receptors). Upon ligand binding, intracellular R-Smads (Smad1/5/8) become phosphorylated by BMPRI, followed by trimeric complex formation with Smad4 and subsequent nuclear translocation to regulate BMP target gene transcription. In addition, BMPs are known to activate several non-Smad pathways that involve signalling via mitogen-activated protein kinases (MAPK) (for example, p38, extracellular signal-regulated kinase (Erk) 1/2) or Akt/protein kinase B (PKB) [8]. The combination of both Smad and non-Smad signalling pathways and their respective intensities explain the pleiotropic and cell context specific effects of BMPs.

Each step of the BMP signalling cascade is tightly controlled by antagonists, co-receptors, intracellular associated proteins or by crosstalk to other growth factor pathways [6]. R-Smad molecules in particular constitute signalling platforms to other pathways by multiple post-translational modifications such as phosphorylation or ubiquitination in their linker region [9].

Similar to BMPs, mechanical boundary conditions are crucial for bone development, homeostasis and repair [10]. However, little is known about the impact of mechanical forces on the BMP signalling cascade. Such interaction might be highly relevant since in vivo administered recombinant BMPs (rBMPs) seem to be much less potent than in vitro. The link between macroscopic bone loading and cellular events is controlled by mechanotransduction pathways, which are still poorly understood. However, the impact of those mechanotransduction pathways on anabolic effects in bone tissue appears indispensable [11]. It is well-known that bone unloading leads to a loss in bone mass [12], or that the rigidity of fracture fixation critically influences the healing outcome [13, 14]. The process of mechanotransduction mainly involves three steps: mechanosensing, conversion of mechanical signals into biochemical ones and subsequent signal propagation [15]. Mechanosensing in osteoblasts likely includes multiple pathways involving signalling via integrins, G-protein coupled receptors or ion channels [16, 17]. In this context, mechanical signals that control cell fate decisions may comprise active forces, such as loading or shear forces, but may also be encoded by substrate characteristics like stiffness, geometry or ligand spacing [18, 19].

Therefore, the aim of this study was to provide further insights into the influence of mechanical forces on BMP signalling. We established a bioreactor system that allows cyclic compressive loading of osteoblast precursor cells in a three-dimensional environment with concurrent BMP2 stimulation. We could show that immediate early BMP signalling events are strongly potentiated by mechanical forces. We conclude that this effect is independent of autocrine BMP ligand secretion and thus gives striking evidence that mechanotransduction pathways directly target BMP signalling molecules without gene expression.

The application of recombinant BMPs to foster bone healing has turned out to be less potent than expected from in vitro studies. Effective delivery and high doses have been the most limiting factors for clinical treatments combined with the risk of side-effects [4]. There is great emphasis in the field to lower the concentrations of BMPs by approving delivery systems, increasing BMP's potency and, most of all, by understanding the molecular mechanism of supporting crosstalk pathways.

BMP signalling is a tightly controlled cascade that is regulated on different levels ranging from extracellular antagonists to receptor composition and intracellular interacting molecules [6]. On the tissue level, there exists strong evidence that BMP signalling and mechanical forces together regulate bone healing. However, little is known about the molecular mechanism of how mechanical boundary conditions might regulate BMP signalling. A better understanding of the crosstalk between both pathways seems essential to unravel their physiological interaction and to help to gain a better understanding towards an adequate use of both stimuli to improve patient treatment strategies.

In recent years, multiple studies have shown the importance of mechanical forces for cellular differentiation [27, 28]. But many in vitro studies focusing on osteogenic differentiation were performed in two-dimensional culture systems and few of them on a molecular basis. To better mimic the in vivo cellular environment, a three-dimensional culture system is indispensable. Therefore, in this study we investigated early events during osteoblastic differentiation induced by BMP2 under mechanical loading in a three-dimensional environment. hFOBs were seeded on open porous collagen scaffolds (average bulk stiffness of 8.5 ± 0.9 kPa) and mechanically loaded with up to 10% straining. hFOBs properly adhered to collagen fibres, and collagen scaffolds exhibited a suitable and physiological stiffness range for initial osteoblastic differentiation [27]. These cells further showed similar signalling dynamics in three-dimensional when compared to two-dimensional monolayer cultures (Figure 1). The bioreactor setup is tuned to mimic the early phase of bone healing events during tissue formation, keeping culture conditions, oxygen supply and mechanical loading parameters constant [29].

To unravel the molecular mechanism comprising this crosstalk, we analysed BMP-induced signalling at different time points. We investigated early phosphorylation events directly downstream of the activated BMP receptors as well as transcriptional responses at different time points (early and late).

We found that BMP2 stimulation and mechanical load synergistically regulate immediate early phosphorylation events in the BMP pathway (Figure 3). BMP2 stimulation with concurrent mechanical loading resulted in the strongly enhanced C-terminal phosphorylation of Smad1/5/8 followed by an increased nuclear translocation when compared to cells stimulated with BMP2 only. This effect was observed as early as 15 minutes after stimulation and was maintained up to several hours (Figure 3 and Additional file 1).

Based on these findings, we postulate that mechanical signals directly influence immediate early BMP signalling events without the involvement of autocrine ligand secretion. The fact, that loading alone did not show significant differences in Smad1/5/8 phosphorylation or Id1 expression further proves this hypothesis. This is in contrast to previous studies where mechanical load was reported to activate the BMP pathway [30, 31]. This may be explained by different types of mechanical forces and study design that included pre-cultivation on scaffold matrices prior to loading for up to 7 days. In this case, BMP pathway activation by mechanical loading might be due to autocrine ligand secretion during culture. In fact, Wang et al. demonstrated that Noggin addition during mechanical stimulation abolished BMP pathway activation induced by mechanical loading [32].

The first step during mechanotransduction comprises the sensing of extracellular mechanical signals by a mechanoreceptor, such as integrins or ion channels [16]. Especially integrins crosstalk to TGFβ and BMP signalling pathways [33]. Similarly there exists increasing evidence that integrin expression and signalling is also important for BMP-induced signalling during osteogenic differentiation [34–36]. It was demonstrated that both BMP type I and type II receptors co-localize with αvβ integrins [34]. Furthermore, many proteins associated with integrin signalling complexes, such as c-Src or Rack1, are also interacting with the cytoplasmatic tail domain of the BMP type II receptor [37, 38]. We hypothesise that integrin activation under loading conditions might lead to altered conformational changes of BMP receptors, which modulate their interactome and alter their signalling properties. Recently, it has been shown that endocytosis of integrin receptors depends on extracellular matrix stiffness and that this altered endocytosis also affects BMP receptor endocytosis and signalling [39]. The route of BMP receptor endocytosis itself critically determines the signalling outcome [40]. We have previously shown that blocking endocytosis inhibits BMP-induced Id1 expression while having no effect on Id2 [41]. Similarly, mechanical load enhanced BMP-induced expression of Id1 but not Id2 (Figure 5). Since receptor endocytosis is strongly related to the membrane lipid composition, it is likely that membrane raft microdomains may play an important role as mechanosensing platforms.

Chang et al. proposed that integrins might mediate Smad activation under shear stress conditions [42]. In our system, ligand independent Smad1/5/8 activation (that is, C-terminal phosphorylation) was not observed as indicated by load-only treatment (Figure 3). However, ligand independent integrin mediated signalling might be involved in the activation of non-Smad pathways and their target genes.

After 15 minutes of stimulation, mechanical loading led to the strong induction of p38, Akt and Erk1/2 phosphorylation (Figure 4). Erk1/2 and p38 have, in particular, been described as important players during mechanotransduction in mesenchymal precursor cells [23, 43, 44]. Furthermore, signalling pathways via MAPK might be involved in regulating Smad signal intensity and duration. The Smad1 linker region comprises several sequential phosphorylation sites for cyclin dependent kinases (CDKs), MAPK and glycogen synthase kinase three beta (GSK3β) that regulate their transcriptional capacity and prime Smad molecules for degradation via the ubiquitin proteasome pathway [9, 45]. In contrast to the Smad pathway, we did not observe synergistic effects of mechanical load and BMP2 on the non-Smad target proteins. But gene regulation under loading conditions of osteogenic marker genes likely involves the interplay of Smad and non-Smad pathways.

Following the BMP pathway further downstream, we analysed the transcriptional regulation of several BMP target genes. Earlier studies tried to elucidate gene expression profiles in osteoblast precursor cells under mechanical stress [46, 47]. It was postulated that mechanical load induces osteogenic differentiation [48] and that mechanical forces exert synergistic effects on osteogenic differentiation together with BMP2 [49]. However, these studies are hardly comparable due to different cellular systems, including osteogenic and non-osteogenic cell types, and mechanical stimulation devices in two dimensions and three dimensions. In addition, most studies focused on long-term differentiation events that potentially include feedback signalling loops.

We showed that the transcriptional network mediating early osteogenic differentiation events includes genes regulated by mechanical forces or the BMP ligand only, as well as genes that are synergistically affected by both triggers. This reflects multiple levels of potential crosstalk between the BMP and mechanotransduction pathway. BMP2 stimulation with concurrent mechanical loading led to synergistic regulation of the early BMP target gene Id1, a key regulator in BMP-induced osteoblastic differentiation (Figures 3 and 5). We also confirmed this in primary human mesenchymal stem cells (Additional file 3). This is of particular interest, because Id1 transcription is not only under the control of Smads but also of early growth response protein one (Egr-1), a transcription factor rapidly induced by mechanical stress [50]. c-fos, known to be a major target of mechanotransduction [22], was strongly induced by mechanical loading, while BMP treatment had negligible effects (Figures 3 and 5). However, Smad4 was shown to interact with c-fos, which modulates activating protein one (AP-1) activity [51]. Whether different strain amplitudes trigger different responses or whether there exists a certain strain threshold remains to be elucidated.

Autoregulation of BMP ligand or antagonist expression is one possibility to modulate the signalling pathway endogenously. It has been shown that mechanical loading of osteoblasts leads to a transcriptional up-regulation of several BMP ligands, such as BMP2, -4, -6, and -7 [32, 52–54]. We instead found that different BMP ligand subtypes are differentially affected by loading. While BMP4 and BMP7 tend to be down-regulated under loading conditions, BMP6 expression was positively affected by mechanical loading, even more so when BMP2 was present (Figure 5). These findings are in line with in vivo data obtained during fracture healing and distraction osteogenesis [55, 56]. Different BMP ligands not only exhibit a distinct spatiotemporal expression pattern but also respond differently to mechanical forces. Interestingly, BMP4 and -6 also differ in their susceptibility to the BMP inhibitor Noggin, with BMP6 being not inhibited by this antagonist [25]. Expression analysis revealed that Noggin mRNA was significantly up-regulated by BMP2 and this up-regulation was further enhanced by mechanical loading. Thus Noggin regulation is a crucial event during osteogenic differentiation to balance signalling intensity and is also sensitive to mechanical stimulation. Also other TGFβ -superfamily antagonists, such as sclerostin, gremlin and follistatin, are regulated by mechanical forces [53, 57, 58]. The BMP antagonists may represent an important target to improve bone healing when inter-related to adequate mechanical boundary conditions. Furthermore, other growth factor pathways, such as Wnt or insulin-like growth factor (IGF) signalling, are influenced by mechanical loading. They share many downstream partners and target genes with the BMP pathway and might be also involved in BMP pathway regulation [16].

This study highlights the complex interaction of mechanical forces with the BMP signalling cascade. We demonstrated that BMP signalling is directly regulated by mechanotransduction pathways, without the involvement of autocrine ligand secretion. We also gave evidence that crosstalk of both pathways over longer time periods might occur on several signalling levels. A hypothetical model on the interplay between both pathways has been proposed (Figure 6). Direct crosstalk is possible as early as at the receptor level at the plasma membrane, in the cytosol or in the nucleus by altering transcription factor properties. Finally, mechanosensing by inner nuclear membrane proteins, which have been shown to also anchor Smad proteins, may participate in this relationship [59–61].

Although the fine-tuned course of osteogenic differentiation during bone healing still remains unclear, the present work is the first to illustrate the tight interaction of BMP- and mechanical stimuli-associated signalling cascades. These cascades are spatiotemporally highly balanced and are fine-tuned processes that need further analyses for a deeper understanding of their interplay. The general principles, however, have been illustrated and are discussed in the present work. To transfer that knowledge into improvements in bone healing, such as the required stiffness of bone fixations in BMP-treated cases, requires further in vivo analyses and in vitro characterization. Such knowledge will ultimately help to improve treatments in the majority of clinical cases and, thus, avoid failures of BMP-initiated stimulation of healing.