Bone is a highly dynamic, specialized connective tissue that is continuously renewing itself throughout life. In adults, old bone is replaced by new bone in an orderly fashion by interplay between osteoclasts and osteoblasts, a process known as bone remodeling. Osteoclasts resorb bone, whereas osteoblasts migrate to the resorbed area and form new bone matrix that subsequently mineralizes. Disturbance of this tight balance between bone resorption and bone formation in favor of resorption underlies the most prevalent bone disease, osteoporosis, which is characterized by reduced bone mass and deterioration of bone architecture resulting in increased bone fragility and fracture risk.1
Members of the transforming growth factor β (TGF-β) superfamily of cytokines were demonstrated to play decisive roles in the differentiation of mesenchymal stem cells (MSCs) into osteoblasts. TGF-β, the prototypical member of this superfamily, is one of the most abundant cytokines present in bone matrix and exerts a dual role on osteoblastic bone formation; TGF-β stimulates early osteoblast differentiation by promoting recruitment and proliferation of osteoblast precursors and the expression of matrix proteins; on the other hand, it inhibits late osteoblast differentiation and mineralization.2–6
Bone morphogenetic proteins (BMPs) comprise a subfamily within the TGF-β superfamily family and were identified originally by their ability to induce ectopic bone formation.7, 8 BMPs are potent regulators of osteoblast proliferation and differentiation, and several skeletal disorders such as fibrodysplasia ossificans progressiva, synostosis syndromes, and brachydactyly type B have been linked to genetic alterations in genes encoding components of the BMP signaling pathway.9–12 Several preclinical studies in primates and other mammals clearly have proven the effectiveness of BMPs in restoring large segmental bone defects. BMPs have demonstrated clinical utility for bone regeneration and are available commercially and used therapeutically in open fractures of long bones, nonunions, and spinal fusion.13 However, this process appears to be rather inefficient because large amounts of BMPs have to be applied to achieve beneficial effects. This might be due at least in part to the fact that BMPs induce expression of the BMP antagonist noggin,14, 15 providing a negative-feedback loop to hamper BMP activity and as a consequence clinical efficacy.
Members of the TGF-β superfamily exert their effects through distinct combinations of two different types of serine/threonine kinase receptors, that is, type I and type II receptors, that on ligand binding form a hetero-oligomeric complex.16, 17 The activated receptor complexes propagate the signal through C-terminal phosphorylation of receptor-regulated Smads (R-Smads) Smad2 and Smad3 in response to TGF-β, nodal, and activins and Smad1, Smad5, and Smad8 in the case of BMPs.18 The phosphorylated R-Smads associate with the Smad4 and translocate into the nucleus, where together with other transcription factors they bind promoters of target genes and control their expression.19–23 BMP-induced osteoblast differentiation requires Smad1 and Smad5 signaling, and overexpression of these Smad proteins in MSCs is sufficient to drive differentiation toward the osteoblast lineage.24 The inhibitory Smads Smad6 and Smad7 can inhibit BMP signaling, whereas Smad7 also can block TGF-β and activin signaling.25 In addition to Smad-dependent signaling, BMPs also activate Smad-independent pathways, such as phosphatidylinositol 3-kinase (PI3K) and p38 mitogen-activated protein (MAP) kinase, which also mediate BMP-induced osteoblast differentiation.26, 27
There exists a high degree of cross-talk between BMP signaling and various other signaling routes. The integrated signal induced by BMPs and other cytokines ultimately determines commitment of MSCs to differentiate toward the osteoblast lineage and the efficiency of bone formation. For example, TGF-β was found to inhibit BMP-induced osteoblast differentiation,6, 28, 29 and fibroblast growth factor (FGF) and Wnt signaling control the duration of BMP signaling by inducing Smad1 linker region phosphorylation by the extracellular signal-regulated kinase (ERK), p38 and c-Jun N-terminal kinase (JNK) MAP kinases, and glycogen synthase kinase 3 (GSK3), resulting in cytoplasmic retention and increased proteasomal degradation of Smad1.30–32 Targeting kinases that affect BMP signaling may lead to increased BMP-induced osteogenesis and promote BMP-mediated fracture healing. In order to identify novel kinases modulating BMP-induced osteoblast differentiation, we performed a kinase siRNA library screen. Unexpectedly, we discovered a novel mechanism as to how TGF-β signaling affects BMP-induced differentiation toward the osteoblast lineage.
Materials and Methods
Recombinant BMP-6 was kindly provided by Prof Dr S Vukicevic, recombinant human TGF-β3 by Dr K Iwata (OSI Pharmaceuticals, Melville, NY, USA) and recombinant noggin by Dr A Economides (Regeneron, Tarrytown, NY, USA). Other recombinant BMPs and the TGF-β blocking antibody were obtained from R&D Biosystems (Minneapolis, MN, USA). The Mouse ON-TARGET plus siRNA kinase library was purchased from Dharmacon (Lafayette, CO, USA). SB431542 was purchased from Tocris/Biotrend GmbH (Köln, Germany) and LY2157299 from Merck (Darmstadt, Germany).
Mouse pluripotent mesenchymal KS483 cells33 were cultured in α minimal essential medium (α-MEM; Gibco BRL, Carlsbad, CA, USA) and mouse premyoblast C2C12 cells in DMEM (Gibco) supplemented with penicillin/streptomycin (Invitrogen, Carlsbad, CA, USA) and 10% fetal bovine serum (FBS; Gibco). Human adult and fetal MSCs were cultured in α-MEM containing 10% FBS, 0.2 mM L-ascorbic acid-2-phosphate (Sigma-Aldrich, St Louis, MO, USA), and 1 ng/mL of basic fibroblast growth factor (bFGF; Peprotech, Hamburg, Germany).
Kinase siRNA screen
C2C12 cells were transfected with the Mouse ON-TARGET plus siRNA kinase library (Dharmacon) according to manufacturer's instructions. Briefly, C2C12 cells were plated in 96-well plates at 1750 cells/well. The next day, the cells were transfected with 100 nM siRNA using Dharmafect-3 and incubated at 37 °C for 2 days. Then the cells were stimulated with 125 ng/mL of BMP-6 for 3 days, and subsequently, alkaline phosphatase (ALP) activity was determined.
Transfections and luciferase reporter assays
C2C12 or KS483 cells were seeded in 12-well plates and transiently transfected with either the BRE-luciferase or the CAGA-luciferase reporter construct22, 23 using GeneJuice reagent (Merck) according to the manufacturer's protocol. Per well, 0.45 µg of firefly luciferase reporter construct together with 0.05 µg of a LacZ expression plasmid was transfected. After 2 days, the cells were stimulated for 16 hours with BMP-6 and/or TGF-β3 in serum-free medium. Lysis and determination of luciferase activity were carried out according to manufacturer's instructions (Luciferase Reporter Assay System; Promega, San Luis Obispo, CA, USA) using β-galactosidase activity as an internal control. The results are representative of at least three independent experiments, and reporter activity is shown as the mean ± SD of triplicates.
Alkaline phosphatase activity assay
Cells were seeded at a density of 15,000 cells/cm2 in 96-well plates, and agents were added when cells reached confluence. KS483 and C2C12 cells were incubated with the indicated agents in culture medium without FBS and the hMSCs in culture medium containing FBS with the addition of dexamethasone (10 nM). ALP activity was measured kinetically in the cell layer after another 2 to 4 days of culture for C2C12 and KS483 cells and after 7 days for hMSCs.
KS483 cells were seeded at a density of 15,000 cells/cm2 in 24-well plates, and when the cells reached confluence, stimuli were added for the first 4 days unless otherwise indicated. KS483 cells form mineralized bone nodules when cultured under osteogenic cultures conditions, that is, in α-MEM supplemented with 5% FBS, 0.2 mM of ascorbic acid (Sigma-Aldrich), and from day 7 onward 10 mM β-glycerolphosphate (Sigma-Aldrich, St Louis, MO, USA). Confluent fetal MSCs (fMSCs) were cultured in α-MEM supplemented with 10% FBS (Gibco), 1 ng/mL of bFGF, 0.2 mM of ascorbic acid, and 10 mM β-glycerolphosphate for the entire assay period; stimuli were added only during the first 4 days. The medium was renewed every 3 to 4 days. After 14 to 21 days, the KS483 cells or fMSCs, respectively, were washed with PBS, incubated briefly with a 2% alizarin red S solution, and washed with distilled water, and pictures were taken.
RNA isolation and quantitative real-time PCR
Total RNA was isolated using the NucleoSpin RNA II Kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's instructions. Reverse-transcriptase polymerase chain reaction (RT-PCR) was performed using the RevertAid H Minus First Strand cDNA Synthesis Kit (Fermentas, St. Leon-Rot, Germany) according to the manufacturer's instructions. Quantitative real-time PCR reactions were performed using the StepOne Plus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Gene transcription levels were determined with the comparative ΔΔCt method using GAPDH as reference, and the nonstimulated condition was set to 1.
Western blot analysis
Cells were seeded in 6-well plates and allowed to grow to confluence. Cells were washed with PBS and serum starved for 4 hours before stimulation for 60 minutes with 100 ng/mL of BMP-6 and/or 5 ng/mL of TGF-β3, washed with PBS, and lysed in SDS sample buffer. C-terminal Smad phosphorylation was detected using antibodies specifically recognizing phosphorylated Smad1 (pSmad1) or Smad2 (pSmad2). Western blotting was performed as described previously.12
The unpaired two-tailed Student's t test was used to determine the significance of differences between means. All relevant comparisons were significantly different (p < .05) unless otherwise indicated.
TGF-β signaling is required for optimal osteoblast differentiation
In order to identify novel mediators of BMP-induced osteoblast differentiation, we performed an siRNA screen to find kinases modulating BMP-6-induced ALP activity, which is an early marker for osteoblastic differentiation. For this purpose, C2C12 premyoblast cells were transfected with a Dharmacon Mouse ON-TARGET plus siRNA kinase library, stimulated with BMP-6, and after 3 days, ALP activity was determined. Transfection of Smad1-targeting siRNAs, which served as a control, resulted in reduced BMP-6-induced ALP activity. As expected, siRNAs directed against the BMP receptors ALK2, ALK3, and BMPRII inhibited ALP activity in response to BMP-6 stimulation (Fig. 1A). Knockdown of ALK6 appeared not to affect the BMP-6 response, consistent with the findings that BMP-6 does not bind to this BMP type I receptor on C2C12 cells.34 Transfection of siRNAs targeting the inhibitory Smad Smad7 resulted in increased ALP activity (data not shown). Surprisingly, although TGF-β generally is considered to antagonize BMP-induced osteoblast differentiation, upon transfection of siRNAs targeting the TGF-β type II receptor (TGFβRII) or the TGF-β type I receptor ALK5, BMP-6-induced ALP activity also was found to be reduced (Fig. 1B).
To confirm these findings, the effect of the selective ALK5 kinase inhibitors SB431542 and LY2157299 on BMP-6-induced ALP activity was investigated. Consistent with the results obtained from the siRNA screen, BMP-6-induced ALP activity in C2C12 cells was inhibited in presence of SB431542 or LY2157299 (Fig. 2A). This effect was not limited to BMP-6 or restricted to C2C12 cells because BMP-9-induced ALP activity in the pluripotent mesenchymal KS483 cells, which have the ability to differentiate toward the osteoblast, chondrocyte, and adipocyte lineages, also was inhibited by the ALK5 kinase inhibitors (Fig. 2B). To determine the effect of TGF-β signaling on the complete osteoblast differentiation process, mineralization assays were performed. Mineralization of KS483 cells in response to BMP-6 was reduced in the presence of SB431542 (Fig. 2C and Supplemental Fig. S1A). Inhibiting autocrine TGF-β signaling by means of TGF-β-blocking antibodies also resulted in reduced BMP-6-induced ALP activity and mineralization (Fig. 2D, E). Taken together, these results suggest that TGF-β signaling is required for optimal osteoblast differentiation.
TGF-β can stimulate BMP-induced osteoblast differentiation
We next chose to examine the effects of applying exogenous TGF-β on BMP-induced osteoblast differentiation. TGF-β enhanced BMP-6-induced ALP activity in KS483 cells in a dose-dependent manner (Fig. 3A). Since the induction of ALP activity during (BMP-stimulated) osteoblast differentiation is a transient process, we investigated the possibility that the stimulatory effect observed is due to a TGF-β-induced change in the kinetics of the response rather than enhanced differentiation. Therefore, ALP activity in KS483 cells on stimulation with BMP-6 with or without costimulation with TGF-β was measured. The stimulatory effect of TGF-β on BMP-induced ALP activity was detected at all time points tested (Fig. 3B). In addition to ALP activity, BMP-6-induced mineralization of KS483 cells also was enhanced upon costimulation with TGF-β (Fig. 3C and Supplemental Fig. S1B). In order to determine whether TGF-β also can positively affect osteoblast differentiation of primary cells, we studied the effect of TGF-β on the mineralization of primary fMSCs. Indeed, TGF-β also was found to increase both basal and BMP-6-induced mineralization of primary fMSCs (Fig. 3D and Supplemental Fig. S1C).
Most studies in which an inhibitory effect of TGF-β on BMP-induced osteoblast differentiation was found studied the effects on BMP-2-induced responses. Since BMP-2 uses ALK3 as a type I receptor and BMP-6 and BMP-9 mainly signal via ALK2,34, 35 we tested whether differentiation to the osteoblast lineage in response to BMP-2 is differentially affected by TGF-β costimulation compared with BMP-6. Similar to osteoblast differentiation in response to BMP-6, TGF-β also efficiently stimulated BMP-2–induced mineralization (Fig. 3C and Supplemental Fig. S1B).
When grown under osteogenic culture conditions in the presence of 10% FBS for 2.5 weeks, KS483 cells, similar to fMSCs (Fig. 3D and Supplemental Fig. S1C), display basal mineralization, which was shown to be induced by autocrine BMPs and BMPs present in serum.36, 37 This basal mineralization can be increased when the cells are stimulated by TGF-β (Fig. 3D, E and Supplemental Fig. S1C, D). To determine whether TGF-β can induce osteoblast differentiation on its own, independent of BMPs, mineralization of KS483 cells stimulated by TGF-β in the presence of the BMP antagonist noggin was examined. Noggin reduced both basal and TGF-β-stimulated mineralization (Fig. 3E and Supplemental Fig. S1D), suggesting that BMP signaling is required for TGF-β-stimulated osteoblast differentiation.
The biphasic effects of TGF-β depend on environmental conditions, timing, and duration
Although under our experimental conditions TGF-β seems to be required for efficient BMP-induced differentiation toward the osteoblast lineage, and addition of exogenous TGF-β is actually stimulating differentiation, several studies reported an inhibitory effect of TGF-β.6, 38 Notably, BMP-6 stimulation for only 1 day is already sufficient to promote mineralization in KS483 cell cultures, and stimulation with BMP-6 for 4 days results in mineralization to a similar extent as observed after 12 days of stimulation (Supplemental Fig. S2A). For this reason, in our mineralization assays, BMPs and TGF-β are present only during the first 4 days of the 16-days incubation period. When mineralization was assessed after culturing the cells in the continuous presence of both ligands, an inhibitory effect of TGF-β was observed (Fig. 4A and Supplemental Fig. S2B). This suggests that TGF-β stimulates early BMP-induced osteoblastic differentiation but inhibits it at later stages. To confirm this hypothesis, KS483 cells were stimulated with BMP-6 for 4 days, with TGF-β being present during the first 4 days, the last 12 days, or all 16 days before staining of the mineralized matrix. As before, stimulating the cells during the initial 4 days with TGF-β enhanced mineralization (Fig. 4B and Supplemental Fig. S2C). However, BMP-6-induced mineralization clearly was inhibited when TGF-β was continuously present during the assay period and when it was present only for the last 12 days. Together these data indicate that the stimulatory and inhibitory effects TGF-β exerts on the deposition of mineralized matrix by KS483 cells depend on the timing of TGF-β action.
Although in C2C12 cells TGF-β displayed mostly a stimulatory effect, sometimes TGF-β was found to act as an inhibitor of BMP-induced ALP activity (Supplemental Fig. S3), unlike KS483 cells, where TGF-β always stimulated BMP-induced ALP activity. We were unable to control this differential effect in C2C12 cells. In human MSCs (hMSCs), a frequently used model to study osteogenic differentiation, TGF-β inhibited BMP-6-induced osteoblastic differentiation, as determined by ALP activity, when stimulated in Lonza osteogenic medium (Fig. 4C, upper panel). However, when hMSCs were stimulated with both ligands in proliferation medium, TGF-β stimulated BMP-induced ALP activity (Fig. 4C, lower panel), indicating that TGF-β can under specific conditions also promote differentiation of these cells toward the osteoblast lineage.
TGF-β inhibits BMP-induced noggin expression and prolongs BMP activity
BMP-induced osteoblast differentiation requires Smad signaling (Fig. 1A). Interestingly, BMP-Smad-mediated transcription, determined by means of BRE-luciferase reporter activity,23 was inhibited by TGF-β in C2C12 cells and KS483 cells (Fig. 5A, B), suggesting that the positive effect of TGF-β on BMP-induced osteoblast differentiation seems not to be due to increased Smad1, -5, and -8 signaling. Consistent with these findings, BMP-induced BRE-reporter activity was found to be increased on addition of SB431542 and LY2157299 (data not shown). Notably, although BMP-induced reporter activity always was reduced significantly by TGF-β costimulation, it was never blocked completely, and residual transcriptional activity was observed.
It has been demonstrated that BMPs induce expression of their antagonist noggin, thereby providing a negative-feedback loop. For this reason, we investigated the effect of TGF-β on BMP-6-induced noggin transcription. Consistent with its effect on BMP-induced transcriptional activity, TGF-β inhibited the BMP-6-induced noggin expression in KS483 cells and primary fMSCs (Fig. 5C). We hypothesized that reduced noggin expression on stimulation with both BMP-6 and TGF-β compared with BMP-6 alone might result in prolonged BMP signaling. Therefore, KS483 cells were stimulated with BMP-6 and/or TGF-β, and the levels of phosphorylated Smad1 were monitored (Fig. 5D). BMP-6 stimulation resulted in increased active Smad1 levels, which were reduced 2 days after stimulation. However, when the cells were stimulated with BMP-6 together with TGF-β, the pSmad1 levels remained high after 2 days, suggesting that TGF-β indeed prolonged BMP signaling. In agreement with the idea that this prolonged BMP activity contributes to osteoblastic differentiation, expression of osteoblast differentiation markers osteocalcin and bone sialoprotein in KS483 cells was strongly enhanced on costimulation with BMP-6 and TGF-β (Fig. 5E).
TGF-β is abundantly present in the bone matrix in complex with the latency-associated protein, rendering it inactive.2 On bone resorption, osteoclasts generate an acidic microenvironment and release proteases causing activation of TGF-β.39 Subsequently, the activated TGF-β induces migration of bone MSCs to bone-resorptive surfaces, where local factors, such as BMPs, induce their differentiation toward the osteoblast lineage.3 Thus, during this initial differentiation stage, the recruited MSCs receive both BMP and TGF-β signals, which, based on our results, potentiate the differentiation process. When differentiation is proceeding, the concentrations of cytokines in the bone microenvironment and the responsiveness and signaling outcome of the differentiating cells change. At this stage, TGF-β is likely to inflict inhibitory effects on BMP-induced signal transduction and continuing osteoblast differentiation.
It is evident that TGF-β can exert opposite effects on bone formation. Osteoblast-specific overexpression of TGF-β2 leads to an osteoporotic phenotype, whereas blocking TGF-β signaling yields increased bone formation.38, 40, 41 On the other hand, mice deficient for Smad3 are osteopenic, and in primates implanted TGF-β2 induces heterotopic endochondral bone formation.42, 43 Our studies indicate that TGF-β also can have, besides the well-acknowledged inhibitory effects,6, 28, 29 a stimulatory effect on BMP-induced osteoblast differentiation. Consistent with our findings, Ripamonti and colleagues44 showed that TGF-β can enhance BMP-induced in vivo bone formation in primates. Interestingly, we found that short-term TGF-β stimulation enhanced basal mineralization and that this could be blocked by noggin, indicating that TGF-β actually stimulates osteoblast differentiation induced by BMPs produced by the cells themselves and/or present in serum.36, 37 Moreover, we showed that blocking TGF-β signaling by means of siRNA-mediated knockdown of its receptors, pharmacologic inhibition of ALK5 kinase activity, or applying TGF-β neutralizing antibodies impedes BMP-induced osteoblast differentiation. These results suggest that TGF-β signal transduction is required for optimal osteoblast differentiation. In agreement with these findings, mice in which TGF-β signaling was inhibited in osteoblasts via osteocalcin promoter–driven expression of a truncated TGF-β type II receptor displayed decreased osteocyte density,41 indicative of reduced osteoblast differentiation. It has to be noted that SB431542 and LY2157299, besides blocking the kinase activity of ALK5, also inhibit activity of ALK4 and ALK7. Thus it is possible that myostatin or activins, which signal via these receptors, also can positively regulate BMP-induced osteoblast differentiation. Similar to TGF-β, these factors have been described to inhibit osteoblast differentiation.45, 46 It would be of interest to see whether these cytokines under certain conditions also have the ability to stimulate BMP-induced osteogenesis.
Our findings show that the effect of TGF-β on BMP-induced osteoblast differentiation depends on timing and context. This might explain why thus far in particular the inhibitory effects of TGF-β have been reported. Osteogenic medium for hMSCs is optimized for driving efficient osteoblast differentiation. However, under these conditions, TGF-β acts as an inhibitor. When hMSCs were allowed to differentiate in proliferation medium, TGF-β enhanced BMP-induced osteoblast differentiation, yielding higher ALP levels than was observed when cells were stimulated with BMPs in osteogenic medium. Thus it is important to realize that besides TGF-β, possibly also other cytokines will exert diverse effects on hMSC differentiation toward the osteoblast lineage depending on the specific culture conditions.
The positive effects of TGF-β on BMP-induced osteoblastic differentiation are reminiscent to what is found in cartilage, where BMPs and TGF-β cooperate in chondrocyte differentiation. TGF-β-activated kinase 1 (TAK1) controls chondrocyte differentiation by mediating C-terminal phosphorylation of Smad1, -5, and -8 in response to BMPs.47 Interestingly, in the siRNA screen performed, knockdown of TAK1 resulted in reduced BMP6-induced ALP activity in C2C12 cells (data not shown). Although TAK1 also can be activated by TGF-β, it is unlikely that TGF-β-activated TAK1 stimulated BMP-induced osteoblast differentiation because TGF-β-induced TAK1 activation is independent of ALK5 kinase activity and consequently is insensitive to SB431542 and LY2157299.48 Moreover, TGF-β was shown to reduce the amplitude of BMP-induced Smad-mediated transcriptional activity. Noggin behaves as a direct BMP target gene, transcription of noggin is regulated by BMPs, and the noggin promoter contains several putative Smad-binding elements and BMP-responsive elements (Olexandr Korchynskyi, unpublished observations). Consequently, TGF-β inhibits BMP-induced noggin transcription. We found that consistent with the reduced noggin expression, when cells were costimulated with BMP-6 and TGF-β, phosphorylated Smad1 levels were maintained for a longer period of time. Thus, although TGF-β prevents maximal BMP-induced transcription, it enables prolonged BMP signaling, which ultimately might contribute to increased osteoblast differentiation.
Based on the studies reporting solely the inhibitory effects of TGF-β on (BMP-induced) osteoblast differentiation, it was suggested that to increase fracture healing, TGF-β signaling should be inhibited. Our findings imply that in the case of BMP-based therapies, this might not necessarily be beneficial. In addition, Zimmermann and colleagues49 showed that TGF-β serum levels in patients were increased after fracture, at the time of soft callus formation and when osteoblasts, chondroblasts, and their progenitors are recruited to the fracture area. Interestingly, they found that the TGF-β serum levels declined earlier in patients suffering from delayed fracture healing compared with patients with normal fracture healing, substantiating the fact that TGF-β might fulfill an important role in fracture healing. In support of this, Gazit and colleagues50 demonstrated that systemic application of TGF-β increases trabecular bone volume, bone formation, and the mineral apposition rate and enhances fracture healing in osteopenic mice. Combining these findings with our results, we hypothesize that it might be beneficial to initially combine BMP and TGF-β treatment followed by application of TGF-β antagonists to enhance the bone fracture-healing process.
All the authors state that they have no conflicts of interest.
We would like to thank Dr E Pardali for critically reading the manuscript and Dr RL van Bezooijen and Prof Dr CWGM Löwik for valuable discussions. This work was supported by the Dutch Organization for Scientific Research (NWO 918.66.606) and the Centre for Biomedical Genetics.