Introduction
Metformin is widely prescribed as a first-line therapy for patients with type 2 diabetes mellitus (T2DM) as an anti-hyperglycaemic agent which acts primarily by suppressing glucose production by the liver [
1]. In contrast to thiazolidinediones (TZD), another T2DM therapy which has adverse effects on the skeleton [
2,
3], several studies have documented that metformin is osteogenic in vitro. It was reported that metformin can induce MC3T3-E1 osteoblastic cells differentiation and bone matrix synthesis via adenosine 5′-monophosphate-activated protein kinase (AMPK) activation and subsequent induction of endothelial nitric oxide synthase (eNOS) and bone morphogenetic protein-2 (BMP-2) expression [
4,
5]. Metformin was also found to regulate Small Heterodimer Partner (SHP) in MC3T3-E1 cells, an orphan nuclear receptor which stimulates osteoblastic bone formation by interacting with the transcription factor Runx2 [
6]. Similarly, metformin increased osteoblast proliferation, alkaline phosphatase activity and the number of mineralised nodules formed in rat primary osteoblasts, possibly via stimulation of Runx2 and IGF-1 production [
7,
8]. The action of metformin on bone marrow mesenchymal cell progenitors (BMPCs) has also been investigated and metformin caused an osteogenic effect, suggesting a possible action of metformin in promoting a shift of BMPCs towards osteoblastic differentiation [
9]. In contrast, two in vitro studies have shown no effect of metformin on the osteogenic differentiation of bone marrow-derived mesenchymal stem cells (MSCs) [
10] and matrix mineralisation of both MC3T3-E1 cells and primary osteoblasts [
11]. A high concentration of metformin (2 mM) even clearly inhibited osteoblast differentiation [
11].
Less work has investigated the effect of metformin on bone in vivo, and the data are more supportive also of an osteogenic effect of metformin. It was reported that 2 months of treatment with metformin prevents the bone loss induced by ovariectomy in rats [
12,
13], suggesting protective effects of metformin against bone loss. In agreement with these studies, a 2-week treatment with metformin in rats was shown to increase trabecular volume, osteocyte density and osteoblast number in femoral metaphysis [
14]. Furthermore, when administered together with the TZD rosiglitazone, metformin prevented the anti-osteogenic effects of rosiglitazone on bone [
14]. A very recent study performed in insulin-resistant mice also showed that metformin given for 6 weeks protects femoral bone architecture compared to rosiglitazone, although metformin had no effect on lumbar spine [
15]. However, few clinical studies have shown beneficial effects of metformin on bone health. Metformin was shown to reduce the association between diabetes and fractures in human patients [
16]. More studies have confirmed that rosiglitazone therapy alone or combined rosiglitazone and metformin therapies were associated with a higher risk of fractures compared to metformin as a monotherapy [
17‐
20]. Interestingly, markers of bone formation were decreased in the metformin group compared to the rosiglitazone one in T2DM patients from the ADOPT study [
21].
The aim of our study was to confirm the osteogenic effect of metformin in vivo on bone architecture in basal conditions (control rats) and in osteopenic bone, using a model of bone loss induced by ovariectomy (ovariectomised mice) to mimic the case of post-menopausal women. For each model, we used different modes of metformin administration that have both been utilised in previous rodent studies; while ovariectomised mice had metformin administered orally by gavage, control rats received metformin in the drinking water. We also wanted to explore the hypothesis that metformin promotes fracture healing in a rat model of mid-diaphyseal, transverse osteotomy in the femur, stabilised via a precision miniature external fixator. Surprisingly, we show in this study that 1- to 2-month treatment with metformin, regardless of the administration route, has no significant effect on rodent bone architecture and fracture healing in vivo, and that metformin significantly reduces bone formation rate in osteopenic trabecular bone.
Materials and methods
Animals and experimental procedures
Experimental procedures used 3-month-old female Wistar rats (Charles River Laboratories, Inc., Margate, UK) and 3-month-old female mice that were in a mixed C57BL/6-129Sv genetic background. These mice were bred in our animal facilities and housed in groups of five in polypropylene cages. Wistar rats were allowed to acclimatise for 1 week after transport before the start of experiments and were housed individually. Both rats and mice were subjected to a 12 h light/dark cycle with room temperature maintained at 21 °C. For mice, metformin (Sigma-Aldrich Company Ltd, Dorset, UK) was given by gavage 100 mg/kg/daily. For rats, metformin was given in the drinking water at a concentration of 2 mg/ml for 8 weeks. On average, water consumption in rats is 10–12 ml per 100 g body weight daily and metformin did not affect the drinking volume. These metformin doses were previously shown to give similar plasma concentrations in rodents than those found therapeutically in humans. The drinking water, along with food, was available ad libitum. The water bottles were replenished twice a week. All animal experimentation procedures were in compliance with Home Office approval and were performed under the threshold of the UK Animals (Scientific Procedures) Act 1986.
The first experiment was designed to investigate whether metformin could protect against the bone loss induced by ovariectomy. Eighteen female C57BL/6-129Sv mice aged 3 months were all ovariectomised, as previously performed by us [
22,
23]. Four weeks after ovariectomy, mice were divided randomly into two groups, one (
n = 9) receiving saline while the other one (
n = 9) receiving metformin (100 mg/kg) daily by gavage for 4 weeks. At days 6 and 3 prior to euthanasia, mice were intraperitoneally injected with calcein (Sigma-Aldrich) and alizarin red complexone (Sigma-Aldrich), respectively, to label bone-forming surfaces in trabecular bone. At the end of the experiment, mice were sacrificed, the serum collected for measurement of metformin concentration, the tibia dissected for micro-CT analysis of cortical and trabecular bone parameters and bone histomorphometry while the femora were used for protein isolation and RT–PCR analysis. Since we did not have a SHAM group, the success of ovariectomy was evaluated by uterine atrophy observations during dissection.
The second experiment was designed to investigate the effect of metformin on basal bone mass. For this study, we used the right contra-lateral tibia of non-ovariectomised female rats which underwent a fracture in the left femur. Twenty female Wistar rats of approximately 200 g were divided randomly into two groups, one (
n = 10) having access to drinking water alone while the other (
n = 10) receiving metformin in the drinking water (2 mg/ml) for 8 weeks. Four weeks after the beginning of treatment, all the rats (
n = 20) underwent a mid-diaphyseal transverse osteotomy in the left femur as described previously [
24]. Surgery was performed under general anaesthesia (ketamine 75 mg/kg and xylazine 10 mg/kg) and appropriate gaseous anaesthesia using aseptic techniques. The external fixator system used in this protocol comprises two metal blocks of titanium alloy linked to two cylindrical stainless steel bars. Briefly, the fixator was applied to the craniolateral aspect of the femur using four threaded M1.2 stainless steel pins. Consistent positioning of the fixator pins was ensured using a drill locator template. After pin placement, a transverse osteotomy was created midway between the proximal and distal pins using an oscillating diamond bone saw, with saline irrigation throughout. The bone fragments were distracted to leave an osteotomy gap of 0.5 mm that was maintained by locking the fixator blocks on to the connecting bars. The rats were administered with 0.1 cc of Vetergesic (Alstoe Ltd, York, UK) for analgesia and 0.05 cc of cephalosporin (Sandoz Ltd, Camberley, UK), as a single dose to prevent infection, post-operatively and were returned to their cages. They were granted mobility immediately after regaining consciousness. Radiographs of the operation site were taken at 4 weeks post-fracture, the time where rats were euthanised under anaesthesia via the delivery of CO
2 into an inhalation chamber. Right tibiae were collected for micro-CT analysis of cortical and trabecular bone parameters while left osteotomised femora were collected for micro-CT analysis of fracture callus and histology.
Micro-CT analysis of mouse and rat tibiae
Right tibiae were harvested from 5-month-old OVX female C57BL/6-129Sv mice, fixed in 10 % neural-buffered formalin for 24–72 h and stored in 70 % ethanol at 4 °C. These tibia were then scanned with high-resolution (5 μm pixel size) micro-computed tomography (micro-CT, SkyScan 1172; SkyScan, Kontich, Belgium), as previously described [
7]. Right tibiae from the fracture study were dissected from rats, fixed and stored as above and scanned with a lower resolution of 14 μm pixel size due to the size of the bones. The whole tibiae were reconstructed using NRecon v.1.4.4.0 (SkyScan) and bone histomorphometric analyses in two and three dimensions (2D, 3D) were performed by SkyScan software (CT-Analyser v.1.5.1.3). For the analysis of trabecular bone, the cortical shell was excluded by operator-drawn regions of interest and 3D algorithms were used to determine the relevant parameters which included bone volume percentage (BV/TV), trabecular thickness, trabecular number, trabecular spacing, structure model index (SMI), trabecular pattern factor and degree of anisotropy. Analysis of cortical bone was performed using a 0.49-mm-long segment (or 100 tomograms) at 37 % of the tibias' length from its proximal end. For analysis of the cortical bone compartment, 2D computation was used and parameters were determined for each one of the 100 tomograms and then averaged. They included periosteal perimeter, endosteal perimeter and cortical thickness.
Assessment of fracture healing
X-ray analysis
Radiographs were taken at the study end point (8 weeks), prior to euthanasia. Both dorsal and ventral X-rays were performed to assess the extent of in situ healing and bridging of the fracture space. Fracture healing was scored from two dimensions, anterior–posterior and lateral–medial X-rays. The X-rays were scored using a three-point system, 1—no callus, 2—some callus formation and 3—significant callus, on all four cortices. The lowest score is thus 4, signifying no callus formation on all four cortices, and a highest of 12, significant callus growth in all four regions.
Micro-CT analysis of fracture healing
Left femora (fractured side) were scanned at 14 μm resolution using micro-CT (SkyScan 1172). A length of approximately 15 mm of the callus with the fracture in the centre was scanned. Histomorphometric analysis of fracture callus in 2D and 3D was performed by SkyScan software (v. 1.11.8.0). A ‘shrink-wrap’ algorithm was used to define the tissue perimeter as the volume of interest (VOI). Binarisation of the reconstructed datasets was by two methods that applied different thresholds since the fracture callus 4 weeks after fracture is heterogeneous and may contain low or highly mineralised woven bone; to automatically delineate the low mineralised callus, a specific threshold was applied that excluded the highly mineralised callus and cortical bone. After measurement, another thresholding was applied, which in contrast defined highly mineralised callus and cortical bone, excluding the very low mineralised callus. Two relevant parameters were therefore quantified, cortical and mineralised callus volume and low mineralised callus volume.
Histology
After micro-CT analysis, fracture calluses were decalcified in 0.34 M EDTA in PBS for 2 weeks at room temperature, bisected longitudinally and the lateral half embedded in paraffin as described previously [
25]. Sagittal sections (5 μM) were cut from the paraffin blocks using a microtome (HM360; Fisher Scientific UK Ltd, Loughborough, UK). Sections were stained with haematoxylin and eosin (H&E) for basic morphology and with Alcian blue and nuclear fast red for analysis of cartilage and bone.
Histomorphometry analysis of tibia
Tibia were fixed in 10 % neutral-buffered formalin for 24 h, dehydrated and embedded in methyl methacrylate (MMA) at low temperature to preserve enzymatic activity [
26]. Unstained 8-μm-thick sections were used for fluorescence microscopy to assess mineral apposition rate (MAR, μm/day). Mineralising surfaces were expressed as alizarin red-labelled surfaces per bone surfaces (MS/BS, %) and the bone formation rate was calculated as MS/BS × MAR (BFR/BS, μm
3/μm
2/day) [
27]. Alternatively, sections were stained for tartrate-resistant acid phosphatase (TRAP) (Leucognost® SP; Merck, Germany) and counterstained with Weigert haematoxylin solution. Histomorphometric parameters were measured on the trabecular bone of the metaphysis, on a region of interest consisting of 2 mm width below the growth plate. Measurements were performed using an image analysis software (Tablet’measure; Explora Nova, La Rochelle, France). Histomorphometric parameters were reported in accordance with the ASBMR Committee nomenclature [
28].
Protein extraction and western blot analysis
For the isolation of total proteins, right femora from 5-month-old female C57BL/6-129Sv mice were carefully dissected and all their surrounding musculature removed leaving the periosteum intact. We also dissected femora from wild-type C57BL/6 mice that were injected with metformin at 100 mg/kg/daily only for 3 days. The cartilaginous ends of the bones were separated and the remaining femoral shafts were flushed with PBS to remove the marrow. The femoral shafts were then snap-frozen and pulverised under liquid nitrogen using a mortar and pestle, and then lysed in cold denaturing lysis buffer (2 % SDS, 2 M urea, 8 % sucrose, 20 mM sodium glycerophosphate, 1 mM sodium fluoride and 5 mM sodium orthovanadate). Proteins were denatured by boiling for 10 min and concentrations determined by BCA protein assay. Twenty micrograms of proteins was size-fractionated using SDS–PAGE and electrotransferred onto Protran nitrocellulose membranes (Schliecher and Schuell, Dassel, Germany). Membranes were blocked for 1 h in 0.2 % (w/v) I-block (Topix, Bedford, MA, USA) before being incubated with primary antibodies. The blots were incubated overnight at 4 °C with antibodies against total AMPKα1/2 (tAMPK α1/2, rabbit), phospho-(Thr-172)-AMPKα1/2 (pAMPKα1/2, rabbit) (New England Biolabs, Hitchin, UK) and β-actin (goat) (Dako, Ely, UK), all added at a 1:1,000 dilution. The following secondary antibodies were used, goat anti-rabbit (New England Biolabs) against tAMPK and pAMPK1α1/2 and rabbit anti-goat (Dako) against β-actin antibody, both at 1:2,500 dilution at room temperature for 1 h. Proteins were visualised using the enhanced chemiluminescence detection system (ECL) (GE Healthcare UK Ltd, Little Chalfont, UK). The intensity of the specific bands was quantified by densitometry using Image J software.
RNA extraction and quantitative real-time PCR
Total RNA was isolated from left whole femora after removal of the bone marrow, as previously described [
7]. RNA from three femora in each treatment group was pooled and two separate extractions were performed. Total RNA was reverse-transcribed with Superscript II reverse transcriptase. Real-time QPCR was carried out as described earlier [
29] using QuantiTect SYBR green PCR kit and Opticon 2 LightCycler (MJ Research, Waltham, MA, USA). Primer sequences were obtained from Qiagen and are summarised in Table
1. The expression levels for
Osterix and
Runx2 were normalised to the reference gene 18s rRNA.
Table 1
Quantitative real-time RT–PCR primer sequences (5′ → 3′)
Runx2
| GACTGTGGTTACCGTCATGGC | ACTTGGTTTTTCATAACAGCGGA | 474–557 | 84 | 225690525b1 |
Osterix
| ATGGCGTCCTCTCTGCTTG | TGAAAGGTCAGCGTATGGCTT | 1–156 | 156 | 18485518a1 |
18s | GTAACCCGTTGAACCCCATT | CCATCCAATCGGTAGTAGCG | 5231–5381 | 151 | NR_003286.2 |
Plasma quantification of metformin
Concentrations of basal metformin level in plasma were determined using a modified ultra high-pressure liquid chromatography (UHPLC) assay with UV DAD (diode array detector) as initially described [
30]. Liquid–liquid extraction of metformin was performed as follows: 200 μl of plasma sample was buffered by adding 200 μl of 8 M sodium hydroxide and spiked with 40 μl phenylbiguanide (internal standard). Then 2.6 ml of a mixture of 50:50 1-butanol/
n-hexane was added, the mixture centrifuged and 200 μl of 1 % acetic acid was added to the upper organic layer. The mixture was centrifuged, the upper organic layer discarded and 5 μl of the aqueous layer was then injected onto a Kinetex® Hilic column (100 × 4.6 mm ID, 2.6 μm) maintained at 40 °C. Flow rate was set 1 ml/min and compounds were detected at 234 nm on an Agilent DAD (1260 Infinity®). Retention times for phenylbiguanide and metformin were respectively 3.0 and 4.5 min. Lower limit of quantification was 15 ng/ml. Based on quality control samples, intraday and between-days precision and accuracy were less than 10 % over the entire range of quantification.
Statistics
The results were presented as mean values ± SD. Statistical analysis was performed using a two-tailed Mann–Whitney U test with GraphPad Prism software. P values less than 0.05 were considered to be statistically significant.
Discussion
With the increasing worldwide prevalence of T2DM which predisposes patients to osteoporosis and increased risk of fractures [
33,
34], there is an increasing need to evaluate the skeletal actions of anti-diabetic drugs and to examine their effects on healing of osteoporotic fractures. We show in this study that the anti-diabetic drug metformin is not ‘bone unfriendly’ but has no osteogenic action, as previously reported. In contrast, our data indicate that metformin reduces bone formation rate, has no major effect on bone mass in vivo in rodents and does not promote fracture healing.
We first used ovariectomised mice to examine the skeletal effect of metformin in conditions of low bone mass that are more representative of the frequent secondary osteoporosis observed in T2DM patients. Our results, which show no effect of metformin on bone architecture, contrast with two previous studies performed in ovariectomised rats [
12,
13], demonstrating that metformin inhibits the trabecular bone loss [
12] and the decrease in bone mineral density [
13] induced by OVX. In both studies, metformin was also administered to OVX rats by gavage at an identical concentration with the one used in our work. Although we did not perform a dose–response of metformin in our study, the concentration of metformin given orally has been extensively used in previous rodent studies [
35,
36]. Our metformin treatment was efficient since plasma levels of metformin were detected with a value of approximately 0.3 mg/l. In addition, metformin induced a small decrease in body weight in our study, a known effect of this anti-diabetic drug which promotes satiety, reducing the food intake [
37]. It is therefore difficult to reconcile our data with these previous rat studies, all the more since Gao’s study [
12] showed similar trabecular bone mass to ours after OVX and we previously showed that same-age OVX mice on this C57BL/6-129Sv background can experience large increases in trabecular bone volume when treated with intermittent PTH [
23]. The duration of metformin treatment is unlikely to explain those differences since we treated our mice with metformin for 1 month, but our rats for 2 months, similarly to the previous rat studies. The effects of metformin on bone may however vary depending on the rodent species and strain utilised, as previously demonstrated for the skeletal effect of rosiglitazone [
38,
39]. In our second study, we used non-OVX rats to examine the effect of metformin on basal bone mass. Rats were used as we wanted to utilise the non-fractured legs of our model of mid-diaphyseal, transverse osteotomy in the rat femur. Metformin was given this time in the drinking water as this mode of administration is less stressful than gavage for fracture experiments and also widely used. Similarly, we found no effect of metformin on bone architecture in contrast to a recent publication by Sedlinsky et al. [
14] showing by histology analysis that metformin increases trabecular area when administered to non-OVX adult rats for 2 weeks in the drinking water, at similar concentration, but in a different strain of rats. Although trabecular and cortical bone architectural parameters were not measured in this study using micro-CT, osteoblast numbers and resorption surfaces were quantified on paraffin sections and were both stimulated by metformin treatment, suggesting that metformin increases bone remodelling in favour of formation [
14]. In our mouse study, dynamic bone parameters measurements were performed in un-decalcified sections of tibiae, and we found that osteoclast surfaces were not affected by metformin treatment. In addition, we showed that the dynamic measure of bone formation, BFR, was significantly decreased in trabecular bone by metformin. This resulted from reduction of both MAR and MS/BS which reflects decreased osteoblast number and activity, although these two parameters of bone formation, when independent, were not decreased significantly with metformin treatment. The demonstration that metformin has no resulting effect on trabecular bone architecture, despite inducing a significant decrease in BFR in trabecular bone, could suggest other indirect effects of metformin, possibly affecting osteoblastogenesis. These results are in agreement with the demonstration that markers of osteoblast activity were reduced for women and men in the metformin group compared to the rosiglitazone one in T2DM patients from the ADOPT study [
21]. However, similarly to Wang’s study [
15], our preliminary results did not demonstrate changes in expression of osteoblast-specific transcription factors measured by quantitative RT–PCR in metformin-treated bones compared to control ones. The discrepancies between all these in vivo studies may therefore also arise from the fact that they measured diverse bone and cellular parameters.
Studies that have investigated the in vitro effects of metformin on bone have also shown discrepancies. While the majority of studies reported osteogenic effects of metformin in vitro [
4‐
9,
40], there are reports indicating that metformin has no osteogenic effect [
10] or inhibits osteoblast differentiation [
11]. Metformin was also shown to inhibit osteoclast differentiation in vivo and in vitro by stimulating osteoprotegerin and inhibiting RANKL expressions [
13,
41], although Bak et al. [
40] showed no effect of metformin on osteoclast formation. Few clinical studies in diabetic patients have assessed the effect of metformin as a monotherapy on fracture risk, and they show overall poor evidence that it has major anabolic effects on bone. No direct links between metformin and falls [
42] were demonstrated, and data regarding the association of metformin with fracture risk are unclear [
16,
43,
44]. Borges et al. [
45] have recently shown that 80 weeks of metformin treatment in drug-naïve T2DM patients induces very modest increases in lumbar spine and total hip BMD. However, metformin treatment was recently shown to decrease circulating sclerostin levels in men with T2DM [
46], suggesting that it could improve skeletal fragility in those patients. More clinical studies have compared the effects of combined TZDs and metformin therapies to TZDs alone and have more consistently shown that metformin decreases fracture risk compared to TZDs [
17‐
20].
Metformin is an AMPK agonist [
32,
47], and our previous work has established that AMPK is important for bone mass in vivo [
7,
23]. The contribution of AMPK to the skeletal action of metformin is unknown. Our results demonstrate that both 3-day and 1-month treatments with metformin did not stimulate AMPK phosphorylation in bone in WT and OVX mice, respectively. The absence of association between metformin treatment and AMPK activation in bone in vivo may suggest that metformin’s effect on bone could be more relevant in the context of diabetes and primarily indirect by reducing the inflammatory state, the accumulation of advanced glycation end-products (AGEs) and the formation of reactive oxygen species (ROS).
We show for the first time that metformin, at the dose given, has no effect on fracture healing in a model of mid-diaphyseal transverse osteotomy in rats. We evaluated the effect of metformin 4 weeks after fracture to examine the endochondral ossification process, and our data show no effect of metformin on callus size or on the speed of the healing process. Diabetes mellitus has been associated with impaired fracture healing, mainly due to suppressed osteoblastogenesis caused by low expression of genes that control osteoblast differentiation [
48‐
53]. Both intramembranous and endochondral ossification are impaired and diabetic bone shows delayed bone regeneration [
53]. The effects of anti-diabetic drugs on fracture healing have not been extensively studied. Molinuevo et al. [
9] have found that metformin treatment stimulates bone lesion regeneration in a defect model in parietal bone in control and diabetic rats. Similarly, Sedlinsky et al. [
14] have shown, in a similar minimal lesion defect in rats, that metformin treatment increases the reossification of this small lesion while rosiglitazone impaired it. Interestingly, metformin increased TRAP activity in these parietal bone lesions, a marker of osteoclast activity. Our data suggest that metformin does not affect the endochondral ossification process, but we cannot exclude that metformin could have an effect on the remodelling of the fracture callus and its mechanical strength.
Our study has several limitations, including the use of a single dose of metformin and the fact that we did not investigate the impact of T2DM on the skeletal effect of metformin. Nevertheless, it strongly indicates that metformin does not promote bone formation or fracture repair in non-diabetic rodent models, in contrast to the increased osteogenesis shown in several in vitro and in vivo studies. This suggests that, similarly to what was shown for TZDs, the skeletal effects of metformin are not always observed and could vary depending on factors such as the strain/sub-strain of rodents, gender, age, dose and duration of treatment as well as the hormonal and the inflammatory states.