Inhibition of GDF8 (Myostatin) accelerates bone regeneration in diabetes mellitus type 2

Metabolic diseases like diabetes mellitus cause bone healing deficiencies. We found significant impairment of bone regeneration, osteogenic differentiation and proliferation in diabetic bone. Moreover recent studies suggest a highly underestimated importance of GDF8 (Myostatin) in bone metabolism. Our goal was to analyze the role of GDF8 as a regulator of osteogenic differentiation, proliferation and bone regeneration. We used a murine tibial defect model in diabetic (Leprdb−/−) mice. Myostatin-Inhibitor Follistatin was administered in tibial bony defects of diabetic mice. By means of histology, immunohistochemistry and QRT-PC osteogenesis, differentiation and proliferation were analyzed. Application of Myostatin-inhibitor showed a significant improvement in diabetic bone regeneration compared to the control group (6.5 fold, p < 0.001). Immunohistochemistry revealed a significantly higher proliferation (7.7 fold, p = 0.009), osteogenic differentiation (Runx-2: 3.7 fold, p = 0.011, ALP: 9.3 fold, p < 0.001) and calcification (4.9 fold, p = 0.024) in Follistatin treated diabetic animals. Therapeutical application of Follistatin, known for the importance in muscle diseases, plays an important role in bone metabolism. Diabetic bone revealed an overexpression of the catabolic protein Myostatin. Antagonization of Myostatin in diabetic animals leads to a restoration of the impaired bone regeneration and represents a promising therapeutic option.

factor 1 (IGF-1) increased 15 . In addition, inhibition of Myostatin led to an increase of osteogenesis and reduction of adipogenesis. Myo −/− mice showed less body fat, lowered blood glucose levels and a general increase in density, strength and bone mineralization 16,17 . Systemic effects of Myostatin knockout on parameters of the metabolic syndrome such as reduced blood glucose-, insulin-, and triglyceride values and a normalization of hyperphagia were confirmed in an additional study 11 .
Various models for the treatment of myositis and myopathy by application of Myostatin inhibition are currently undergoing intensive tests. Though Follistatin is not very selective and putatively inhibits GDF11 and Activin A, it is one of best available most cost-effective inhibitors of Myostatin [18][19][20][21][22] . It has been shown despite of a blockade of Myostatin target receptor ActRII (Activin Receptor 2) Follistatin also directly binds to Myostatin transforming it into an inactive form 23 . This has been shown in vivo 18 . For instance, the potent Myostatin inhibitor Follistatin affecting the Myostatin pathway has been shown to be clinically successful in treating muscular diseases. Additionally, transfection of Follistatin gene in human Phase 1/2a studies is a promising example of gene therapy for treating myopathies. However, these studies are exclusively limited to muscular diseases 20,[24][25][26] .
Thus, inhibition of Myostatin increases the osteogenic potential and bone mineralization, and reduces the impact of the metabolic syndrome. However, its utilization in increasing bone regeneration in general and in the diabetic condition in particular has not yet been investigated. The goal of this application is to use Myostatin inhibitors as a therapeutic approach to improve diabetic bone regeneration. Upon local application of Follistatin we aim to successfully block the Myostatin pathway and subsequently improve bone regeneration, osteogenic differentiation and proliferation in a murine diabetic model.

Material and Methods
Animal surgery. All animal experiments were approved by the IACUC LANUV NRW (The Ministry for Environment, Agriculture, Conservation and Consumer Protection of the State of North Rhine-Westphalia; 84-02.04.2016.A045). All methods were performed in accordance with the relevant guidelines and regulation by the IACUC LANUV NRW. Heterozygous Lepr db (db + /db − ) mice were obtained from Jackson Laboratory (#000697) and kept with unlimited access to water and standard laboratory chow. Heterozygous db + /db − mice on a C57BL/6 J background were mated to obtain WT, db + /d−and db − /db − mice. Genotyping for breeding was performed on genomic DNA by restriction enzyme digest after PCR (forward primer: ATGACCACTACAGATGAACCCAGTCTAC; reverse primer: CATTCAAACCATAGTTTAGGTTTGTC) according to Horvat and Bügner 27 . Female littermates at the age of 16 to 20 weeks were used for all experiments. All surgical procedures were performed under inhalation anesthesia with isoflurane and buprenorphine (Abbott GmbH). An established murine tibial defect model was performed as previously described 9,10 . Briefly, after shaving and disinfecting the left leg, an incision was made on the proximal anterior skin surface over the tibia. After splitting the anterior tibial muscle, the tibia was properly exposed. A 1 mm unicortical defect was created on the anterior tibial surface (approx. 3 mm under the knee joint) and filled with a 1 mm circular collagen sponge coated with a reagent according to the animal group. The 3 animal groups included (1) WT control mice with PBS, (2) diabetic control animals with PBS, (3) diabetic animals treated with 2 µg Follistatin. Wound closure was performed with 6-0 Prolene interrupted sutures. Each group consisted of at least 7 animals. Euthanasia was performed according to national and international laws and guidelines. Briefly, cervical dislocation was performed after thorough anesthesia to harvest tissue.
Isolation of mouse adipose-derived stem cells (mASCs). mASCs were harvested from inguinal fat pads of Lepr db−/− Black 6 mice and Black 6 WT mice and labeled similarly as previously described 28 . Inguinal fat pads were excised and washed in betadine/PBS. Thereafter, fat pads were minced under sterile conditions and digested with 0.1% Collagenase A (Roche Diagnostics, Mannheim) for 30 min at 37 °C in a shaking water bath, centrifuged down, washed and plated. Culture conditions were performed with DMEM GlutaMAX, 10% FBS and 100 IU/ml penicillin/ 100 IU/ml streptomycin for in vitro experiments. At passage 1 and 80% confluence, osteogenic differentiation was initiated. Osteogenic differentiation medium (additional β-Glycerolphosphate 1%, ascorbic acid 0.1% modificated from Neuhuber et al. 29 ) was changed every third day with glucose concentration accordingly. Cells were incubated at 37 °C with a carbon dioxide level at 5%.
Neutralizing (ND 50 ) antibodies for Activin (7.5 µg/ml -Goat, polyclonal, R&D Systems, AF338) and Myostatin (3 µg/ml -Goat, polyclonal, R&D Systems, AF788) were used as described before 30 . Quantification of osteoid formation. Tibiae were harvested at a given time and fixed in 4% paraformaldehyde (Sigma Aldrich) overnight at 4 °C and decalcified in 19% EDTA (Applichem) for five days with daily changes of the solution. Samples were then dehydrated and embedded in paraffin and cut into serial sagittal sections (thickness 6 µm).
Every 6 th section was used to characterize bone formation with aniline blue (Roth) staining as previously described 31 . Images were taken with a bright field microscope (Zeiss Axiovert 100, Zeiss) and following settings with Axiovision 4.8: objective 2.5x, exposure time 614 ms, dimensions 3900 × 3090 Px, scanned color. Histomorphometric measurements of aniline blue stained sections were performed in Adobe Photoshop with modifications 9 . Additionally, Goldner stainings were performed according the standard protocol 32 . Briefly, a 2000 × 2000 Px dimensioned selection box was placed to cover the entire defect area. Utilizing the Adobe Magic Wand Tool (settings: tolerance 60%; no-contiguous) new osteoid formation was selected semi-automatically. Tonal separation in two steps and deselecting existing cortical bone resulted in highlighted pixels reliably corresponding to bone formation area.
Trabecular Area (Tb.Area) and Osteoblast Surface (Ob.S) were calculated following the 2012 published recommendations of the American Society for Bone and Mineral Research 33 . Osteoid Area per Total Area (OA/TA) was measured as described before 34 .
Analogous to quantification of bone formation, a region of interest was selected (2,000 × 2,000 Px). Immunohistochemically positive stained pixels were automatically selected by using the Adobe Magic Wand Tool (settings: tolerance 60%; noncontiguous).
Osteocalcin was immunodetected with Osteocalcin antibody (rabbit, polycloncal, SantaCruz Biotechnology, sc-30045, 1:100, Osteocalcin, AB_653627). Similar to immunofluorescence sections were incubated with primary and secondary antibody. After washing with PBS, a rabbit biotinylated secondary antibody followed by the AB reagent and NovaRed (Vector Laboratories) was used for detection. Images were taken with an AxioImager M2 Imaging System (Zeiss). For histormorphometry of Osteocalcin Osteocalcin positive cells were counted in randomly detected boxes (100 µm × 100 µm) within the defect area.
RNA preparation and cDNA synthesis. The homogenization of harvested bone was achieved with an in 1 ml Trizol reagent (Life Technologies, Darmstadt, Germany) on ice. Tissue war processed as described before 36 . Subsequently, RNA purification was done with RNeasy Mini Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. Synthesis of cDNA was performed by means of the High Capacity cDNA Reverse Transcription Kit with RNase inhibitor (Life Technologies) using 200 ng total RNA per reaction.

Quantitative real-time PCR. Quantitative determination of relative gene expression was performed on
Applied Biosystems 7900HT Fast Real-Time PCR System (96 well plates) using TaqMan ® gene expression assays (MSTN, Hs00976237_m1; ACVR1C, Hs00899854_m1) and TaqMan ® universal master mix (Applied Biosystems, Darmstadt, Germany). For each reaction, 2 ng cDNA were used. Data were analyzed according to the manufacturer's ΔΔCt method (Applied Biosystems). 18 S was used as a reference gene.

Radiography.
Standardized digital radiographs (exposure parameters: 40 kV, 0.5 mAs) of the operated tibiae were completed 28 days after surgery to detect mineralization of the defects. The methods was described before. To further determine the ossification additional luminance assays of the region of interest (ROI: 40 × 20 Px) were performed 37, 38 . Statistical analysis. Results of the study are presented as mean ± standard error of the mean (SEM) of at least three independent experiments. P-values were calculated by student's t-test comparing two groups and ANOVA if comparing more than two groups. Statistical significances were set at a p-value < 0.05.

Results
Myostatin expression is significantly increased in diabetic bone tissue. Unicortical defects were performed on diabetic and wildtype animals. First we analyzed Myostatin and Activin A expression in uninjured (Sham) and 3 days after setting a unicortical tibial defect (Fig. 1). A group with Follistatin treatment (2 µg Follistatin) was included to observe an inhibition of Myostatin and Activin A gene expression by Follistatin. A 9.4 fold increase of Myostatin expression was observed in uninjured diabetic bone (p < 0.0001), while a 2.1 fold increase of Activin A (p = 0.03) was detected. While injured bone has a 2 fold increase of Myostatin expression in wildtype animal (p = 0.018) there is a decline in diabetic injured bone by 34%. Placing a tibial defect lead to non-significant increase of Activin A both in wildtype and diabetic animals. Uninjured diabetic bone showed elevated Activin A levels compared to their wildtype correlate (p = 0.024).

Inhibition of Myostatin by Follistatin enhances osteogenic differentiation and calcification in vitro.
mASC of Lepr db−/− Black 6 mice were differentiated in osteogenic medium as described before for 7 days to evaluate the pro-osteogenic effect of Follistatin application 35 . Differentiation medium combined with Follistatin or Myostatin neutralizing antibody (3 µg/ml) showed significantly enhanced osteogenic differentiation, while Activin neutralizing antibody (7.5 µg/ml) showed a less osteogenic differentiative effect. Myostatin was most significantly inhibited by Follistatin (23 fold decrease) and Myostatin neutralizing antibody (244 fold decrease) administration. Activin A levels were significantly reduced in Follistatin (3 fold decrease), Myostatin (11 fold decrease) and Activin A (147 fold decrease) neutralizing antibody treated groups. SMAD 2/3 level was significantly reduced by Follistatin (130 fold decrease) and Myostatin neutralizing antibody treatment (126 fold decrease) (Fig. 2).
SCientifiC RepoRts | 7: 9878 | DOI:10.1038/s41598-017-10404-z Osteogenic differentiation of mASC of WT and diabetic origin for 28 days showed a decline of calcification in mASC of diabetic origin compared to WT mASC in the Alizarin red staining. This negative effect was eliminated upon Follistatin application (9 µg/ml).
Unbounded Myostatin levels and Myostatin signaling are significantly decreased by local application of Follistatin. Next, we sought to investigate the effects of Myostatin inhibition on bone regeneration.
In order to verify sufficient inhibition of Myostatin by local application of Follistatin into the defect, immunohistochemical stainings for Myostatin and subsequent downstream targets p38 and SMAD 2/3 were performed on day 3 postoperatively. Accordingly, the selectivity of Follistatin application was determined by immunofluorescence for Activin A, SMAD 1 (downstream target of BMP 7) and GDF 11. Immunohistochemistry revealed significantly increased levels of unbounded Myostatin (5 fold), p38 (4 fold) and SMAD 2/3 (2.5 fold) in diabetic tibial defects as compared to wildtype, while locally applied Follistatin in diabetic animals showed a significant decrease of all three proteins (unbounded Myostatin: 31 fold, p38: 17 fold, SMAD 2/3: 10.6 fold) compared to diabetic control animals. Activin A positive signal was decreased upon Follistatin treatment in diabetic animals in compared to WT but not compared to the diabetic control group. GDF 11 was not significantly increased in diabetic bone defects, however Follistatin application lead to a further decrease of GDF 11 in diabetic bone compared to diabetic defects (2.8 fold, p = 0.008). SMAD 1, the downstream target of BMP 7 was decreased in diabetic defects (2.3 fold, p = 0.04) but not significantly increased by Follistatin application (Fig. 3).
Follistatin enhances osteoid regeneration. To analyze osteoid formation after local application of Follistatin in vivo, aniline blue and Goldner-staining were utilized to detect osteoid formation 7 days postoperatively. Histomorphometry of the stainings revealed a significant decrease of osteoid regeneration in diabetic animals compared to WT animals ( Fig. 6A-D, H/I). Aniline Blue staining of diabetic defects demonstrated 8.2 fold decrease in osteoid area (p < 0.001) compared to WT animals, while local application of Follistatin (2 µg) rescued osteoid formation (6.5 fold, p < 0.001). Follistatin treatment of wildtype animals did show a slight, but not significant improvement in osteoid formation compared to wildtype control animals (Fig. 6A-D). To further validate bone regeneration quality a Golder staining was performed. Again, diabetic defects showed a 9.5 fold decrease in (B) qPCR showed increased level of Activin A in diabetic bone compared to WT bone. Follistatin treatment did not alter Activin A expression compared to control animals (diabetic bone with PBS). Results are shown as means ± SEM. P-value: * < 0.05; ** < 0.01; ** < 0.001 (two sample t-test).
osteoid area compared to WT (p < 0.001). The administration of Follistatin locally in the defect enhanced diabetic osteoid formation by 11 fold (p < 0.001) compared to control diabetic animals (Fig. 6D/E).
Calculation of osteoblast surface per trabecular osteoid area showed decreased osteoblast count in diabetic animals (3 fold, p < 0.001), while Follistatin treated defect showed an increase in osteoblast count (1.7 fold, p = 0,008) compared to diabetic bone without treatment (Fig. 6H). Evaluation of osteoid area per tissue area confirmed these findings. The OA/TA (osteoid area per tissue area) in the diabetic defect area was reduced 8.8 fold (p = 0.005) compared to WT bone. Follistatin treated defects showed a 6.8 fold increase of BA/TA compared to diabetic controls (Fig. 6I).
With conventional digital radiography of the tibiae 28 days post-operatively bone regeneration was further described. With measuring the luminance of the region of interest (40 × 20 Px). We detected decreased luminance, corresponding with mineral density in diabetic operated tibiae compared to WT operated tibiae (−21%, p = 0.01). Upon single time administration of Follistatin into the tibial defect we demonstrated a restoration of luminance to 105% of the WT level and + 33% compared to the diabetic PBS control defects (p = 0.004) (Fig. 7).

Discussion
The aim of the study was to investigate the pro-osteogenic effect of Myostatin inhibition by Follistatin in diabetic bone. We could demonstrate that Myostatin is increased in uninjured Lepr db−/− as compared to WT bone in mice. We were able to show no change of Myostatin mRNA expression upon Follistatin treatment. Targeting overexpression of Myostatin in diabetic bone, local application of Follistatin -a potent Myostatin inhibitor -showed a sufficient inhibition of Myostatin and the major downstream targets SMAD 2/3 and p38 in diabetic bone. Concordantly, osteogenic differentiation (RUNX-2), proliferation (PCNA) and osteoid formation were significantly enhanced in diabetic bone upon local application of Follistatin. Moreover, an elevation of PECAM was detected in Follistatin treated bone though not significant.
Although the impact of Myostatin is mostly attributed to muscle tissue, we showed Myostatin activation and subsequently its inhibition upon administration of Follistatin in bone tissue, however there might be an additional impact of Myostatin inhibition through a hypertrophic effect on surrounding muscle. It has been shown that access of bone to muscle is supportive for osseous healing not only by contributing blood supply but also by muscle cells serving as osteoprogenitor cells 39,40 . Furthermore mechanical forces of muscles may facilitate bone healing 41,42 . It has been shown that bone mass and strength highly correlates with body weight, whereas lean body mass is particularly associated with bone strength and mineralisation due to mechanical stimuli 43 . Furthermore a novel role of muscle as an endocrine organ is postulated. The secreted cytokines and growth factors from muscle are referred as myokines. From a clinical perspective, fractures covered with a muscle flap heal more quickly 44 . Studies have shown that placing a porous barrier between the muscle coverage and the bone enhances the healing capacity when the pores are large enough for muscle derived factors to diffuse 45 . Myokines such as Interleukin 6 and leukemia inhibitory factor have been shown to stimulate bone formation, while Follistatin has not been determined as a potential pro-osteogenic myokine yet. Hence, there seem to be multiple interaction between muscle and bone upon regeneration, and we just started to understand this reciprocal relationship However Myostatin might be a key factor linking muscle atrophy with bone loss and therefore acts as anti-osteogenic myokine. We were able to show the pro-osteogenic capacity of Follistatin through Myostatin blockade in vivo and in vitro 46 .
Mice expressing a dominant-negative Myostatin receptor (activin receptor type IIB) in muscle showed reduced blood glucose, serum insulin, triglyceride levels, and the rate of triglyceride synthesis, and improved insulin sensitivity, while elevated insulin sensitivity was not due to reduced fat mass 11 . Myostatin inhibition re-evaluates the role of muscle metabolism in diabetic disease. The underrated potential of bone and muscle as organs in treating diabetes systemically must be re-evaluated; several glucose transporter types are described to be expressed in different bone cells implicating an important role in glucose-metabolism 47,48 . Our work emphasizes the underestimated role of Myostatin in bone metabolism and therefore the potential implication in treating systemic diabetic disease.
The role of Follistatin may be central in bone remodelling, as Follistatin-like 1 was demonstrated to be upregulated and secreted during osteoclast differentiation and positively regulates osteoclast formation induced by RANKL 49 . Follistatin is also considered as a partial antagonist of BMPs -however we have demonstrated a similar effect of Follistatin compared to BMPs on bone formation. A common downstream target of BMPs and Follistatin is SMAD 1 20,22 . BMP-7 for example promotes the expression of RUNX-2 and Osterix directing them in osteoblast differentiation. Accordingly, in our experiments Follistatin significantly enhanced RUNX-2 in diabetic bony defect. Activin A was shown to be involved in positively regulating bone formation, however we were able to show that osteogenic differentiation and SMAD 2/3 decrease were mediated primarily through inhibition of Myostatin by Follistatin. On the other hand it has been shown that Follistatin may interfere with Activin A, consequently inhibiting osteogenic precursor cells to differentiate but enhancing proliferation of osteoblasts 50 . However another study demonstrated an inhibitory effect of Activin A on osteoblast differentiation 16 . Thus, the exact role of Activin A remains elusive and has to be further investigated. GDF 11 as agonist at the Activin receptor 2 was inhibited by Follistatin. Still, no elevation in diabetic bone compared to WT bone was found in our in vivo experiments. Interestingly, investigations by Gregorevic and colleagues suggest Follistatin mediated Myostatin independent muscle hypertrophy by SMAD 1/3 and mTOR suggesting a strong involvement of Follistatin on BMP7 activation 22,26 . However we detected no significant elevation of BMP 7 downstream target SMAD 1 upon Follistatin treatment in vivo. Though we found decreased levels of SMAD 1 in diabetic bone compared to WT. Inhibins including Follistatin are important regulators of bone turnover as well and have a bimodal effect on bone metabolism 51 . Inhibin A for example was also demonstrated to stimulate bone formation during distraction osteogenesis 52 . Further experiments are highly demanded to differentiate the strong bimodal role of Follistatin in bone metabolism as partially antagonizing of BMP-2/4 and Activin A. In our study, local application of Follistatin onto injured diabetic bone demonstrated a significant pro-osteogenic and proliferative component of Follistatin.
The interaction of Follistatin, Myostatin and recently discovered GDF 11 and BMP 7 is puzzling due to their individual role in bone remodelling and requires separate experimental setups to further characterize their role.

Conclusion
Our study demonstrated that local inhibition of Myostatin in diabetic bony defects can lead to a significant enhancement of osteogenesis, osteogenic differentiation and proliferation. Our study suggests that potential applications of Follistatin may be expanded and further investigation of Follistatin as local and potential systemic treatment of diabetes and diabetes associated bony defects are warranted.