PHOSPHO1 is a skeletal regulator of insulin resistance and obesity

The classical functions of the skeleton encompass locomotion, protection and mineral homeostasis. However, cell-specific gene deletions in the mouse and human genetic studies have identified the skeleton as a key endocrine regulator of metabolism. The bone-specific phosphatase, Phosphatase, Orphan 1 (PHOSPHO1), which is indispensable for bone mineralisation, has been recently implicated in the regulation of energy metabolism in humans, but its role in systemic metabolism remains unclear. Here, we probe the mechanism underlying metabolic regulation by analysing Phospho1 mutant mice. Phospho1−/− mice exhibited improved basal glucose homeostasis and resisted high-fat-diet-induced weight gain and diabetes. The metabolic protection in Phospho1−/− mice was manifested in the absence of altered levels of osteocalcin. Osteoblasts isolated from Phospho1−/− mice were enriched for genes associated with energy metabolism and diabetes; Phospho1 both directly and indirectly interacted with genes associated with glucose transport and insulin receptor signalling. Canonical thermogenesis via brown adipose tissue did not underlie the metabolic protection observed in adult Phospho1−/− mice. However, the decreased serum choline levels in Phospho1−/− mice were normalised by feeding a 2% choline rich diet resulting in a normalisation in insulin sensitivity and fat mass. We show that mice lacking the bone mineralisation enzyme PHOSPHO1 exhibit improved basal glucose homeostasis and resist high-fat-diet-induced weight gain and diabetes. This study identifies PHOSPHO1 as a potential bone-derived therapeutic target for the treatment of obesity and diabetes.

A number of recent studies have implicated PHOS-PHO1 in the regulation of energy metabolism in humans [35][36][37][38]. Within the PHOSPHO1 gene, differential methylation sites have been identified as potentially useful biomarkers for clinical application in the early detection of type 2 diabetes [35] and significant associations between methylation at loci within the PHOSPHO1 gene and the future risk of type 2 diabetes exist [36,37]. Differential methylation in PHOSPHO1 was associated with three lipid traits (total cholesterol, high-density lipoprotein cholesterol, and triglycerides) [39,40]. Most recently, genetic variants of PHOSPHO1 in a bivariate twin study were found to be associated with body mass index and waist-hip ratio [38]. Taken together, these findings suggest that in addition to the established role of PHOSPHO1 in biomineralisation of the skeleton and dentition, Phospho1 ablation may result in improved glucose homeostasis and a reduction in metabolic disease susceptibility. We hypothesised that bone-derived choline may be an important regulator of global metabolism. To address this, we examined the metabolic phenotype of juvenile and adult Phospho1 −/− mice.
Glucose tolerance was improved in Phospho1 −/− mice after chronic HFD (Fig. 2b). Insulin secretion across the glucose tolerance test (GTT) was also lower in Phos-pho1 −/− mice on both CD and HFD, suggestive of insulin sensitisation rather than exaggerated β-cell insulin secretion as the major basis of the phenotype (Fig. 2c). This was confirmed with insulin tolerance tests (ITT) after chronic HFD, which revealed greater glucose disposal in Phospho1 −/− mice ( Fig. 2d-f).
Insulin sensitivity in Phospho1 −/− mice is independent of elevated adiponectin serum levels In an attempt to uncover the mechanism(s) responsible for the increased insulin sensitivity in Phospho1 −/− mice, serum levels of adiponectin and leptin were measured. Levels of high molecular weight adiponectin, a hormone linked to insulin-sensitisation [41], were decreased in Phospho1 −/− mice fed either a CD (2.62-fold) or a HFD (1.92-fold) (both p < 0.001) suggesting that insulin sensitivity and protection from obesity are independent of adiponectin (Fig. 4a). The observed decrease in circulating adiponectin in Phospho1 −/− mice was not due to decreased bone marrow adipose tissue, an endocrine organ that contributes significantly to serum adiponectin (Additional file 4: Fig. S4) [42].
Serum leptin levels in CD-fed mice were unaffected by Phospho1 deficiency, whereas in comparison to WT mice fed a HFD serum leptin levels were significantly decreased 1.31-fold (p < 0.05) in Phospho1 −/− mice fed a HFD (Fig. 4b), accordant with reduced fat mass [43]. Phospho1 −/− CD mice had increased circulating serum triglycerides compared to WT CD mice, but no change was observed in WT and Phospho1 −/− mice fed a HFD (Fig. 4c). Consistent with increased oxidative metabolism of carbohydrate and lipids in other peripheral tissues, mRNA levels of genes encoding key metabolic proteins were increased in adipose tissue (Lpl) muscle (Fndc5) and liver (Prkaa1, Pepck1 and Ppara) (Fig. 4d-f). The  Canonical thermogenesis does not underlie the metabolic protection observed in adult Phospho1 deficient mice Thirty-five-day-old Phospho1 −/− mice had decreased interscapular brown adipose tissue (BAT) mass compared to WT counterparts (WT, 5.11 ± 0.57 mg/g; Phos-pho1 −/− , 3.05 ± 0.40 mg/g; p < 0.01) (Fig. 5a). However this reduction in BAT mass did not persist to adulthood nor during high-fat feeding (Fig. 5b). Strikingly, adult and high-fat-fed Phospho1 −/− mice had smaller brown adipocytes compared to WT controls (Fig. 5c). In order to see if BAT activation and thermogenesis might be responsible for the observed phenotype, we measured key brown fat genes including uncoupling protein 1 (Ucp1) (Fig. 5d, e); no differences were observed in the mRNA and protein levels. Furthermore, there were no significant differences in respiratory exchange ratio (RER, indicative of metabolic substrate preference) or energy expenditure between WT or Phospho1 −/− mice fed either a chow or HFD housed at either room temperature or during cold exposure (4°C) ( Fig. 5f-i). These in vivo data show that increased canonical thermogenesis does not underlie the metabolic protection observed in the Phos-pho1-deficient mice so this line of investigation was not pursued further.
Diabetes mellitus-associated genes are enriched in Phospho1 −/− primary osteoblasts We next sought to unravel the genetic circuitry responsible for the improved glucose tolerance in Phospho1 −/− mice. To address this, a transcriptomic analysis of Phos-pho1-deficient osteoblasts was completed. There was a striking 20-fold upregulation of embryonic stem cell phosphatase (Esp) mRNA, the gene encoding the protein osteotesticular protein tyrosine phosphatase (OST-PTP) in Phospho1-deficient mice [7]. These data were validated in primary WT and Phospho1 −/− calvarial osteoblasts and Phospho1-deficient osteoblast overexpressing Phospho1 (Fig. 6a, b). Protein tyrosine phosphatases are recognised master regulators of insulin receptor signalling (INSR), negatively modifying osteoblast-insulin signalling and thereby controlling GLU13-OCN release [7,[44][45][46][47][48]. The identification of elevated Esp expression in Phospho1 −/− mice was strongly suggestive of a reciprocal regulation between OST-PTP and PHOSPHO1 in the control of glucose homeostasis. This increased Esp expression was however inconsistent with the improved glucose tolerance in the Phospho1 −/− mice. Therefore in an attempt to reconcile this anomaly, we measured circulating GLU-OCN and GLU13-OCN, which were found to be unchanged in juvenile and adult Phos-pho1 −/− mice (Fig. 6c). Only GLU-OCN and GLU13-OCN present in the serum have been shown to act as a hormone [5,49]. Therefore, these data implied that elevated serum levels of uncarboxylated or undercarboxylated OCN did not mediate the improved metabolic phenotype (Fig. 6c, d). The elevated levels of carboxylated and total OCN in Phospho1 deficiency was consistent with increased bone turnover in these mice as previously reported [30].
Phospho1 −/− calvarial osteoblasts show a greater metabolic capacity in utilising exogenous substrates compared to WT osteoblasts Primary calvarial osteoblast metabolism analysis revealed that differentiated Phospho1 −/− osteoblasts had elevated basal and FCCP (protonophoric uncoupler) induced oxygen consumption rates (indicative of oxidative phosphorylation) compared to WT osteoblasts when supplied exogenously with glucose, pyruvate and glutamine (Additional file 5: Fig. S5). There was a significant increase in the glycolytic rate of both the non-differentiated and differentiated Phospho1 −/− osteoblasts compared to WT osteoblasts suggesting that they have increased glucose metabolism. Based on these observations, we conclude that differentiating Phospho1 −/− calvarial osteoblasts show a greater metabolic capacity in utilising exogenous substrates compared to WT osteoblasts. To further clarify the genetic pathways underpinning these observations, 22 differentially expressed genes identified from the microarray analysis were found by Ingenuity Pathway Analysis (IPA) to be associated with glucose homeostasis (Additional file 6: Table. S6). Further predictive analysis of the differentially expressed genes of the original microarray data set identified > 40 genes to be associated with energy metabolism (p = 1.04 × 10 −6 ) ( Fig. 7a) of which 10 were found to be associated with both diabetes and bone, following an NCBI (PubMed) in silico search. Validation by RT-qPCR confirmed that 70% of the genes predicted by IPA (Vdr, Slc1a3, Adamts4, Cd68, Cfp, Fmod and Lum) were (See figure on previous page.) Fig. 4 Phospho1 −/− mice are insulin sensitive despite decreased adiponectin. a Adiponectin, b leptin and c triglyceride serum quantification. RT-qPCR analysis of tissue extracted from 120-day-old WT and Phospho1 −/− mice, d adipose tissue, e quadriceps femoris and f liver. RT-qPCR analysis of GLUT receptors from g adipose tissue, h quadriceps femoris and i liver. Data are represented as mean ± S.E.M. * p < 0.05. Different letters above the error bar show significant difference at p < 0.05 and the data were assessed as follows: a-c Two-way ANOVA with multiple comparisons, d-i two-tailed Mann-Whitney test for non-normally distributed data and two-tailed unpaired t test for normally distributed data differentially regulated in Phospho1-deficient osteoblasts (p < 0.05; Fig. 7b and Additional file 7: Table. S7). Furthermore, GeneMANIA network analysis predicted that Phospho1 both directly and indirectly interacts with 36 genes associated with glucose transport and metabolic processes and insulin receptor signalling (Additional file 8: Fig. S8). Of the output genes, Atf4, Foxo1 and Insr are recognised skeletal modulators of energy metabolism, suggesting crosstalk between Phospho1 and other metabolic regulatory genes [5,50,51].

Identification of differentially expressed serum proteins in
To explore the secretome profile, quantitative SWATH (sequential window acquisition of all theoretical spectra) MS (mass spectrometry) proteomics [52] was conducted on serum from WT and Phospho1 −/− mice fed CD and HFD. Differentially expressed proteins (> 100) were identified in HFD Phospho1 −/− serum compared to HFD-fed WT mice. These proteins were highly associated with glycolysis, gluconeogenesis and 'metabolic pathways'  (Additional file 9: Table. S9). Pathway and network analysis predicted that the identified proteins interacted with miR-34a, a microRNA that is known to affect diverse parts of insulin signalling in the pancreas, liver, muscle and adipose tissue [53].

Bone-derived choline is involved in global energy regulation
Neutral sphingomyelinase 2 (nSMase2) catalyses the hydrolysis of sphingomyelin to form ceramide and PCho [54]. Furthermore, PCho is the preferred substrate for PHOSPHO1 yielding choline and Pi (Fig. 8a) [55]. As elevated levels of both ceramide and choline result in insulin resistance in mice [34,56] we targeted these intermediates to establish if alterations in their serum levels could explain the insulin sensitive phenotype in Phospho1 −/− mice [34,56]. The levels of various ceramide species were unchanged (Fig. 8b); however, Phos-pho1 −/− mice had a significant decrease in serum choline levels (WT, 0.152 ± 0.001 μg/ml; Phospho1 −/− , 0.128 ± 0.003 μg/ml; p < 0.01) (Fig. 8c) which were normalised upon choline supplementation. Supplementation of WT and Phospho1 −/− mice with a 2% choline diet (a welltolerated, palatable diet [57]), also normalised the insulin sensitivity measured in Phospho1 −/− , measured by GTT (Fig. 8d). However, unlike WT mice which when fed a 2% choline diet took longer to recover from the insulin challenge, Phospho1 −/− showed no metabolic change in response to insulin between the diets (Fig. 8e). Furthermore, choline supplementation normalised the lean phenotype observed in Phospho1 −/− mice (Fig. 8f, g). These results support the notion that Phospho1 deficiency improves the metabolic profile of mice in vivo and confers resistance to obesity and diabetes in part via the alteration of serum choline levels.

Discussion
The fundamental observations presented in this study further strengthen the concept that the skeleton acts as an endocrine organ and provides empirical evidence for a critical role for PHOSPHO1 in energy metabolism. Phospho1 deficiency results in decreased blood glucose levels, improved insulin sensitivity, glucose tolerance and conferred protection from diet-induced obesity and diabetes in mice despite a 60-fold upregulation of Esp expression by Phospho1 −/− osteoblasts. Mice lacking Esp in osteoblasts present with severe hypoglycaemia and hyperinsulinemia resulting in postnatal lethality in the first 2 weeks of life [7]. Conversely, mice overexpressing Esp exclusively in osteoblasts were glucose intolerant and insulin resistant [7,[44][45][46][47][48]. Intriguingly, the increased insulin sensitivity in Phospho1 −/− mice was not associated with the expected rise in serum GLU13-OCN levels suggesting that PHOSPHO1-regulated energy metabolism is via OCN-independent mechanisms. This notion has previously been observed when partial genetic ablation of osteoblasts profoundly affected energy expenditure, gonadal fat weight and insulin sensitivity which were not restored by the administration of OCN [20,21]. Nevertheless, it is possible that the increased insulin sensitivity noted in Phospho1 −/− mice may be primed by an initial rise in GLU13-OCN levels, which is eventually normalised in a compensatory manner by the observed increase in Esp expression. This being the case we would predict that the loss of Esp on a Phospho1 −/− background would exacerbate the insulin sensitivity due to increased GLU13-OCN serum levels. These data strengthen the concept that a novel pathway exists between osteoblasts and glucose homeostasis; however, it does highlight the potential cross-talk between OCNdependent and OCN-independent mechanisms of glucose metabolism. The role of PHOSPHO1 in controlling bone mineralisation has been extensively investigated through the use of both in vitro and in vivo mouse models. Crucial for the initiation of mineralisation within matrix vesicles, PHOSPHO1 hydrolyses membrane lipid derivatives, primarily PCho to produce Pi (utilised in hydroxyapatite formation) and choline [55]. Phosphocholine can be formed from choline via choline kinase activity or phosphatidylcholine via PLA2 and ENPP6 as well as from the hydrolysis of sphingomyelin, via nSMase2 to form PCho and ceramide [58]. Mindful of this, it has been reported that elevated levels of both ceramide and choline result in insulin resistance in mice [34,56]. We saw no change in ceramide species in Phospho1 −/− mice; however, there was a significant decrease in serum choline levels in Phospho1 −/− mice, which was normalised in Phospho1 −/− mice fed a 2% choline rich diet resulting in a normalisation in insulin sensitivity and fat mass. This study (See figure on previous page.) Fig. 7 Ingenuity Pathways Analysis network summary predictions. a Ingenuity Pathways Analysis was used to predict further genes associated with glucose homeostasis based upon the 22 genes found to be differentially expressed in the microarray. b Predicated genes were analysed by RT-qPCR in WT and Phospho1 −/− primary calvarial osteoblast's. Results were normalised to the Atp5b housekeeping gene. Data are represented as mean ± S.E.M (n = 3 replicates). *P < 0.05, **P < 0.01, ***P < 0.001. Red = Upregulated. Green = Downregulated (the darker the shade of green and red colour indicates a more extreme up/down regulation, conversely the paler the shade indicates a more subtle up/down regulation. Dashed line = indirect interaction (blue = inhibition, yellow = findings underlying the relationship are inconsistent with the state of the downstream node, grey = Ingenuity Pathways Analysis prediction) and the data were assessed in (b) using a two-tailed unpaired t test The regulation of global energy metabolism by the skeleton is a complex, multifactorial process, and it is therefore likely that presently undefined bone secreted factors may have a significant role in PHOSPHO1's ability to regulate global energy. We identified > 100 secreted proteins that may contribute to the energy regulation via the skeleton. These unique proteins were highly associated with glycolysis, gluconeogenesis and 'metabolic pathways' and association with miR-34a, a microRNA that affects diverse parts of insulin signalling in the pancreas, liver, muscle and adipose tissue [53,59]. Furthermore, lumican, a proteoglycan secreted by differentiating osteoblasts and a constituent of the bone matrix was found to be enriched in serum of Phos-pho1 −/− mice by both proteomic and microarray analysis. Interestingly, lumican has also been observed in the decidua of diabetic patients [60,61]. Further investigations of these and other candidates may uncover new skeletal regulators of energy metabolism that act via osteocalcin independent mechanisms.
The metabolic phenotype observed in the Phospho1deficient mice was not due to increased BAT activation despite striking histological differences between WT and Phospho1 −/− tissue. BAT is a thermogenic organ that increases energy expenditure to generate heat, maintaining body temperature in a cold environment [62]. When activated by cold exposure or through activation of β3adrenergic receptors, BAT improves insulin sensitivity and lipid clearance, highlighting its key role in metabolic health [63]. PHOSPHO1 has been previously suggested to have a role in murine BAT function. Increased expression of Phospho1 in both isolated brown and white adipocytes was observed following treatment with the βadrenergic agonist (CL316,243) compared to placebotreated white adipocytes [64]. Mice with an adiposespecific defect in fatty acid oxidation (Cpt2 A−/− ) showed a loss of β3-adrenergic induced Phospho1 expression in BAT [64] and expression of Phospho1 was shown to be elevated in Ucp1-deficient animals; however, WT mice had very low Phospho1 expression in BAT and no detectable PHOSPHO1 was observed in WT iWAT and BAT [65]. No difference in Phospho1 gene expression was observed between human supraclavicular and subcutaneous adipose progenitor cells (GEO Dataset: GDS5171/8016540) and murine WAT had elevated Phospho1 gene expression compared to BAT (GEO Dataset: GDS2813/1452485_at). In our hands, we have noted high expression of Phospho1 in murine BAT but no protein expression was detected (Additional file 10: Fig. S10). Taken together, these data suggest that Phos-pho1 is likely to play a role in BAT function; however, the BAT phenotype observed in Phospho1 deficient mice does not appear to underlay the metabolic protection we see in these animals. Further studies are necessary to unravel the role of PHOSPHO1 in BAT using a BAT conditional knock-out model of Phospho1.

Conclusions
Collectively, the results of this study add further credibility to the concept that GLU13-OCN is not the sole mediator of the endocrine function of the skeleton [20]. We suggest, as others have, that further undefined bonederived proteins/lipids work in partnership with OCN to regulate the metabolic function of the skeleton and affect other metabolic organs such as muscle and liver [20]. Indeed, this study has identified other potential protein mediators and raised the possibility that bonederived choline may contribute to the regulation of the development of the metabolic syndrome. Furthermore, our results suggest that Esp may act as a fine controller of insulin sensitivity in mice, offering protection from severe hypoglycaemia and dyslipidaemia.
Several previous reports have suggested an association between PHOSPHO1 expression in disorders of altered energy metabolism such as obesity and diabetes [35][36][37][38][39]. The data from this present study is both supportive of such an association but also provides insight into the mechanisms by which PHOSPHO1 may contribute to the regulation of energy metabolism, inclusive of insulin sensitivity, glucose tolerance and fat metabolism. Also, inhibitors of PHOSPHO1 activity such as the proton pump inhibitor lansoprazole, commonly prescribed to control and prevent symptoms of gastroesophageal reflux disease and dyspepsia, have been associated with improved glycaemic control in diabetic patients [28,[66][67][68]. Taken together, the identification of PHOSPHO1 in the role of energy metabolism in both the human and g Quantification of epididymal fat adipocyte diameter and representative histology. Data are represented as mean ± S.E.M (c-g) n = 3-5 per group. Different letters above the error bar show significant difference at p < 0.05 and the data were assessed as follows: c, f, g Two-way ANOVA with multiple comparisons. b One-way ANOVA with Dunnet's or Tukey's tests for multiple comparisons. d, e repeated measures two-way ANOVA with Sidak's multiple comparisons test mouse offers the potential to manipulate key targets of the PHOSPHO1 pathway to improve metabolic health [36,37].

Aim
Examine glucose metabolism and the energy status of Phospho1 −/− mice.

Reagents
All chemicals, tissue culture medium and buffers were from Sigma-Aldrich (Dorset, UK) and Invitrogen (Paisley, UK) unless otherwise stated. PCR oligonucleotides were purchased from MWG Eurofins (Ebersberg, Germany) and Primer Design (Southampton, UK). PHOSPHO1 HuCAL Fab bivalent antibody was purchased from AbD Serotech (Kidlington, UK). All antibodies were diluted 1: 1000 unless otherwise noted.

Animals
Phospho1 null mice were generated as previously described [31]. Offspring carrying the mutant Phospho1 gene were identified by genotyping (F: 5′-TCCTCCTC ACCTTCGACTTC-3′, R: 5′-TCCTCCTCACCTTCGA CTTC-3′). All in vivo studies were conducted at 120 days of age unless otherwise stated. Male mice were fed a high fat diet consisting of 58% of calories from fat (DBM Scotland, Broxburn, UK) or control diet (6.2% calories from fat; Harlan Laboratories, Indianapolis, IN, USA) starting at 4 weeks of age. Male mice were fed a 2% supplemented choline diet (Harlan Laboratories) or control diet (Harlan Laboratories) for 5 weeks prior to cull at 120 days. Ad libitum food consumption was monitored for 6 days and basal nocturnal activity was quantified using an AM524 Single Layer X, Y IR activity monitor and associated Amonlite software (Linton Instrumentations, Norfolk, UK). Juvenile metabolic activity was measured using indirect calorimetry (Oxymax Lab Animal Monitoring System: CLAMS (Columbus Instruments, OH USA). Adult metabolic rate was measured using indirect calorimetry (TSE PhenoMaster 1.0, with software version 6.1.9). Cold exposed mice were first housed in these cages for 3 days at room temperature (RT) for acclimation and baseline measurements. Mice were then house for 72 h at 4°C. All experiments were conducted blind to the operator. Animals were maintained under conventional housing conditions with a 12h light/dark cycle with free access to food and water (except when food was restricted during fasting). All animal experiments were approved by The Roslin Institute's Animal Users Committee, and the animals were maintained in accordance with UK Home Office guidelines for the care and use of laboratory animals.

Metabolic studies
Male juvenile and adult (35 and 120 days old respectively) were weighed and fasted for 4 h between 9 am and 1 pm. Prior to the start of the tests, a basal blood sample was collected by venesection into EDTA powder-coated capillary tubes (Starstedt, Leicester, UK). Basal glucose levels were measured using a glucose monitoring system (Accu-Chek® Aviva, Roche, Leicester, UK). Two milligrammes of D-glucose (Sigma, Poole, UK) per g of body weight was administered by gavage or 0.5 mU of insulin (Actrapid, NovoNordisk, Bagsvaerd, Denmark) per g/ body weight was administered intraperitoneally (i.p.). At precisely 15, 30, 60 and 120 min following administration, blood glucose was measured with an Accu-Chek® Aviva glucose meter (Roche Diagnostics Ltd., Lewes, UK) and insulin was measured by ELISA (ChrystalChem, Chicago, IL, USA). Animals were allowed to recover for 2 weeks prior to euthanasia. Tissues were collected for protein, gene and histological analysis.

Primary osteoblast isolation and culture
Under sterile conditions, calvaria were isolated from 2to 4-day-old new-born WT and Phospho1 −/− mice as previously described [69]. Osteoblasts were expanded in flasks in growth medium consisting of α-MEM supplemented with 10% FBS and 1% gentamicin in a humidified atmosphere of 95% air/5% CO 2 and maintained at 37°C. When the cells reached 80-90% confluency, they were seeded at a density of 2.5 × 10 4 /cm 2 in multi-well plates. Conditioned medium was collected upon plate confluency, centrifuged to remove particulates and frozen at − 80°C until required. For overexpression studies, primary osteoblasts were transfected with empty (EV) or overexpressing (OE) vectors as previously described [70].

Mito stress test
An XF24 Analyser (Agilent Technologies, Santa Clara, CA, USA) was used to measure the respiratory function of primary osteoblasts. Osteoblasts were plated at a density of 50,000 cells per well and transferred to a 37°C CO 2 incubator until and calvarial osteoblasts were differentiated using osteogenic differentiation media containing 8 mM β-glycerophosphate and 50 μg/ml ascorbic acid for 3 days. On the day of the assay, cells were washed in XF Assay Media supplemented with 25 mM glucose and 10 mM pyruvate and placed in a non-CO2 incubator at 37°C for 1 h prior to start of assay. Reagents were prepared for the assay (injection volume of 75 μL for each reagent per well) from 2.5 mM Seahorse stock solutions, Oligomycin (1.2 μM). Following equilibration, the Seahorse plate was placed in the Seahorse XF24 Analyser for sample analysis. The raw data was normalised to protein content in each well at the end of the assay [71].

Gene expression analyses and immunoblotting
RNA extractions from tissues and cells were performed using the RNeasy Lipid Tissue Kit (Qiagen). The Super-Script First Strand Synthesis System (Invitrogen) was used for reverse transcription. Real-time PCR amplification with the 2 x precision master mix (Primer design, Southampton, UK) using the Stratagene Mx3000P realtime QPCR system (Agilent Technologies, Santa Clara, CA, USA). Each sample was tested in triplicate and compared to a housekeeping gene (Atp5B in osteoblasts and bone tissue, Lrp10 in adipose tissue and Gapdh in all other tissues) using MxPro software (Cheshire, UK), and the relative expression of the analysed genes was calculated using the ΔΔCT method [72]. All primer sequences are described in Table 1. Primers sequences are available in Table 1. For protein extraction, cells were scraped and tissues homogenised in an appropriate volume of radio-immunoprecipitation assay (RIPA) buffer containing 15% of complete mini protease inhibitor cocktail (Roche, Burgess Hill, West Sussex, UK). Protein concentration was determined by the Bio-Rad DC protein assay (Bio-Rad, Hertfordshire, UK). Immunoblotting was conducted with specific antibodies and protein bands were visualised using the enhanced chemiluminescence (ECL) Western Blotting Detection System (GE Healthcare, Chalfont St Giles, UK) or the Odyssey infrared detection system (LICOR). All antibodies are described in Table 2.

Tissue histology
Tissue was fixed in 4% PFA and embedded in paraffin wax. Five-micrometre sections were stained with haematoxylin and eosin (H&E) using the Leica Autostainer and mounted in DePeX (VWR, Lutterworth, UK). Adipocyte diameter and pancreatic β-cell islet number and size were quantified using ImageJ software as previously described [7,73].

Osmium staining
Mouse tibiae were fixed in 10% neutral-buffered formalin and decalcified in 14% EDTA, pH 7.4. Mouse bones were stained according to [74]. Briefly, bones were stained with a 1% osmium tetroxide solution for 48 h at room temperature. Bones were washed in Sorensen's buffer and embedded in 1% agarose prior to μCT scanning (μCT100 Scanco Medical, Bassersdorf, Switzerland) − 12 μm, medium resolution, 70 kVp, 114 μA, 0.5 mm AL filter and integration time 500 ms. Analysis was performed using the manufacturer's software.

Proteomic analysis
Proteins from serum was extracted and prepared as previously described [75][76][77], and the extracted peptides was analysed using a RSLC 3000 nanoscale capillary LC followed by qTOF mass spectrometry (5600 Triple-TOF, Sciex). Sequential window acquisition of all theoretical spectra (SWATH) was used to profile all proteins in each sample using a data-independent acquisition method [78]. ProteinPilot™ was used for protein identification and quantitation, as well as visualising peptideprotein associations and relationships.

Choline extraction
Serum samples were analysed using tandem mass spectrometry (LC-MS/MS) and a multiple-reaction monitoring (MRM) methodology. Five microliters of serum was extracted with 90 μL of an organic solution (10% methanol and 90% acetonitrile), containing the deuterium-labelled internal standard (IS, D9-Cho at 10 μg/mL). This resulted in precipitation of proteins, which were removed by filtration with a Millex 0.45-μm filter followed by centrifugation for 2 min at 6000g. The mass transitions used to measure the analytes are choline (mass transition m/z 104 → 60) and D9-choline (mass transition m/z 113 → 69). A QTRAP 5500 triple-quadrupole mass spectrometer (AB Sciex, Warrington, Cheshire, UK) with ESI ion source was used for data acquisition. Separation of analytes was performed in an Acquity UPLC-MS/MS (Waters, Hertfordshire, UK), with a binary pump system at a flow rate of 0.3 mL/min, connected to the mass spectrometer. The injection volume was 10 μL. Samples were separated using a Cogent 100 mm × 2.1 mm, 4 μm Diamond Hydride silica column (Microsolv Technologies, NJ, USA) and a linear gradient from 65% buffer B (0.1% formic acid in Acetronitrile) and 35% buffer A (0.1% formic acid in water) to 35% buffer B over 7 min. Analyst software (AB SCIEX) was used for HPLC system control, data acquisition and data processing.

Statistics
The data were analysed using various statistical models. All data were analysed for normal distribution within each experimental group using the Shapiro-Wilk normality test. Linear regression and correlation analysis based on Excel (Microsoft Office 10) built-in functions with interval of confidence and testing of the correlation coefficients were performed according to standard procedures. The SAS software was used to fit the generalised linear model (Microsoft Office 10). Normally distributed data were analysed by ANOVA or t tests, as appropriate. Where data were not normally distributed, non-parametric tests were used. When appropriate, P values were adjusted for multiple comparisons. Data are presented as histograms (means ± standard error) or box and whisker plots (boxes indicate the 25th and 75th percentiles; whiskers display the range; and horizontal lines in each box represent the median). Regression and correlation coefficients are given with the intervals of confidence (p = 0.05). Statistical analyses were performed using Sigma Plot software (v 11.0) (Systat Software Inc., London, UK) and Prism software (GraphPad, USA). P value < 0.05 was considered statistically significant.
Additional file 3: Fig. S3. μMRI adipose quantification from WT and Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) using the Seahorse X-24 analyser in WT and Phospho1 −/− primary calvarial osteoblasts cultured in growth media and growth media supplemented with osteogenic differentiation media for 3 days following a Mito Stress Test. Data represented as means ±S.E.M from average of two independent seahorse runs, with n = 5 wells per group. **p < 0.01, ***p < 0.001.
Additional file 6: Table S6. Osteoblast microarray candidates involved associated with glucose homeostasis. 21 genes from the WT and Phospho1 −/− osteoblast microarray were identified by Ingenuity Pathway Analysis to be associated with glucose homeostasis p = 1.04 × 10 − 6 .
Additional file 7: Table S7. In Silico analysis of Ingenuity pathways predictions In Silico analysis of genes predicated to be associated with bone and diabetes mellitus.
Additional file 8: Fig. S8 GeneMANIA network summary predictions GeneMANIA network generated using Ingenuity Pathways Analysis gene predictions. The network highlights potential interactions between Phospho1 and related osteoblast genes involved in the glucose metabolic process, encompassing; glucose transport, insulin receptor signalling, response to insulin and cellular response to insulin stimulus. Query genes (black) with the exception of Spic and Runx2 which were inputted manually, other genes (grey) were generated by the programme using a large set of inbuilt functional association data. Node size are based on GO terms. Network line colour corresponds to interaction: purple = coexpression, pink = physical interactions, blue = co-localisation, green = shared protein domains orange = predicted, grey = other.
Additional file 9: Table S9. In Silico analysis top diseases associated with unique WT HFD and Phospho1 −/− HFD proteins. (CSV 3 kb) Additional file 10: Fig. S10. Gene and protein expression of Phospho1 mRNA and PHOSPHO1 protein in murine tissue (a) RT-qPCR of Phospho1 in murine tissues, high expression was seen in the gonad and brown adipose tissue (BAT) (b) Protein expression of PHOSPHO1 was detectable by western blot in the calvaria and bone. Non-specific binding of the PHOS-PHO1 antibody was observed in the pancreas, seen in both WT and Phos-pho1 −/− pancreatic tissue.
Additional file 11. Raw data.