Oxidative metabolism is impaired by phosphate deficiency during fracture healing and is mechanistically related to BMP induced chondrocyte differentiation

Abstract

Prior studies of acute phosphate restriction during the endochondral phase of fracture healing showed delayed chondrocyte differentiation was mechanistically linked to decreased bone morphogenetic protein signaling. In the present study, transcriptomic analysis of fracture callus gene expression in three strains of mice was used to identify differentially expressed (FDR = q ≤ 0.05) genes in response to phosphate (Pi) restriction. Ontology and pathway analysis of these genes showed that independent of genetic background, a Pi-deficient diet downregulated (p = 3.16 × 10⁻²³) genes associated with mitochondrial oxidative phosphorylation pathways as well as multiple other pathways of intermediate metabolism. Temporal clustering was used to identify co-regulation of these specific pathways. This analysis showed that specific Ox/Phos, tricarboxylic acid cycle, pyruvate dehydrogenase. Arginine, proline metabolism genes, and prolyl 4-hydroxylase were all co-regulated in response to dietary Pi restriction. The murine C3H10T½ mesenchymal stem cell line was used to assess the functional relationships between BMP2-induced chondrogenic differentiation, oxidative metabolism and extracellular matrix formation. BMP2-induced chondrogenic differentiation of C3H10T½ was carried out in culture media in the absence or presence of ascorbic acid, the necessary co-factor for proly hydroxylation, and in media with normal and 25 % phosphate levels. BMP2 treatment led to decreased proliferation, increased protein accumulation and increased collagen and aggrecan gene expression. Across all conditions, BMP2 increased total oxidative activity and ATP synthesis. Under all conditions, the presence of ascorbate further increased total protein accumulation, proly-hydroxylation and aggrecan gene expression, oxidative capacity and ATP production. Lower phosphate levels only diminished aggrecan gene expression with no other effects of metabolic activity being observed. These data suggest that dietary phosphate restriction controls endochondral growth in vivo indirectly through BMP signaling, which upregulates oxidative activity that is linked to overall protein production and collagen hydroxylation.

Highlights

• •In vivo Pi-deficient diet downregulated genes associated with mitochondrial oxidative phosphorylation and intermediate metabolism during endochondral bone formation of fracture healing.
• •Reduced Phosphate in vitro during BMP2 induced mesenchymal stem cell differentiation did not directly affect oxidative metabolism.
• •BMP2 increases total oxidative activity and ATP synthesis during chondrogenic differentiation of mesenchymal stem cell.
• •The presence of ascorbate further increased total protein accumulation, prolyhydroxylation and oxidative activity and ATP production during chondrogenic differentiation.
• •Effects of in vivo Pi-deficient diet on cellular oxidative activity are indirectly mediated via effects on chondrogenic differentiation.

1. Introduction

Endochondral ossification is a process in which skeletal stem cells (SSC) condense and differentiate into a cartilage anlage modeling the structure of the bone into which it will develop. During this process, chondrogenic cells undergo a progression of differentiation in which the cells first produce a hyaline-like cartilage tissue. As the cells further differentiate, they begin to synthesize many specialized proteins that facilitate mineralization and eventual resorption (Blumer, 2021). Formation of the bones of the axial and appendicular skeleton during embryogenesis, postnatal epiphyseal bone growth, and long bone fracture repair all use this mechanism of bone formation. Endochondral ossification during postnatal epiphyseal bone growth has been extensively studied and used as the basic paradigm by which the cellular and molecular mechanisms of this process have been elucidated. Three separate cellular processes were shown to be associated with the overall longitudinal growth of epiphyses: the rate of cellular proliferation; the amount of matrix that is synthesized; and the hypertrophy of the chondrocytes (Hunziker, 1994; Wilsman et al., 1996). With the advent of genetic cloning, transgenic mice, and genetic analysis of congenital human skeletal pathologies, the major morphogenetic families of factors (BMPs, Wnt, Hh, and FGFs) have all been shown to regulate different aspects of this endochondral ossification (Kronenberg, 2003).

Dietary and congenital hypophosphatemia during development cause rachitic disorders that are associated with a failure in growth plate cartilage replacement with bone tissue, and an overall retardation in growth. Mechanistic studies have shown that the expansion of the zone of cells made up of hypertrophic chondrocytes within the growth plate is related to decreased numbers of cells undergoing mitochondria-mediated apoptosis (Sabbagh et al., 2005; Miedlich et al., 2010). A more recent study of acute phosphate restriction was shown both to increase marrow adiposity and to impair vascular tissue development within the marrow (Ko et al., 2016). Acute phosphate restriction also leads to delayed bone healing after fracture and recapitulates many cellular and molecular effects that are seen in rachitic growth plates. These studies also showed that under conditions of acute phosphate restriction, impairment of skeletogenic stem cell differentiation was related to decreased bone morphogenetic protein (BMP) signaling within these cells (Wigner et al., 2010). Our studies examining the time course of the fracture healing transcriptome under conditions of dietary phosphate restriction lead to both elongation and heightened expression of the molecular mediators of the circadian cycle (Noguchi et al., 2018) while PTH treatment shortened the circadian cycle within healing callus tissues (Okubo et al., 2015; Kunimoto et al., 2016). These results therefore suggest that mineral metabolism is regulated through aspects of circadian function. Comparative analysis of the genes in callus tissue that were differentially regulated by hypophosphatemia against published data for genes in bone that are diurnally regulated identified 1879 genes with overlapping differential regulation. Ontology assessment of these genes further showed that they were associated with oxidative metabolism and apoptosis, consistent with prior studies that identified TCA cycle and Ox/Phos metabolism as one of the primary downstream targets of circadian regulation (Panda, 2016).

Prior studies of fracture healing comparing differing inbred strains of mice suggested that the coordinated development of vascular and skeletal tissues was regulated through a common set of signal-transduction pathways that as a group appear to integrate tissue response to systemic metabolic conditions with those that control local tissue morphogenesis (Grimes et al., 2011). In the present study, we extend upon initial identification of the effects of dietary phosphate restriction on intermediate metabolism with a detailed transcriptomic assessment of the regulatory effects of systemic phosphate restriction on oxidative gene expression and multiple pathways of intermediate metabolism within callus tissues from three strains of mice. We then assessed the relationships between BMP2 signaling and skeletal cell differentiation and examined those elements of cell function (proliferation vs protein accumulation) that showed association to oxidative energy consumption.

2. Materials and methods

2.1. Animal husbandry

Animal research was conducted in conformity with all federal and USDA guidelines, under an approved Boston University, Institutional Animal Care and Use Committee protocol. All studies were carried out using 8–10 week old male C57BL/B6J (B6), A/J (AJ), and C3H/HeJ (C3) mice. The choice of male mice was limited to what had been both proposed and funded by the NIH when this transcriptomic project was initiated 2013 and was based on comparisons to our prior transcriptomic studies performed only in male mice in 2009 and knowledge base of fracture healing at that time. All mice were obtained from Jackson Laboratories, Bar Harbor ME and were housed at Boston University Laboratory Animal Science Center under 12 h light dark cycle with free access to food and water. Mice were randomly divided into a normal diet group (controls) fed standard chow (Teklad 2018, Madison, WI, USA; 0.65 % phosphorus) or a phosphate-restricted diet group (Pi) (Teklad, PRD; Madison, WI, USA 0.06 % phosphorus). The Pi group was fed the low phosphate diet starting two days before fracture and continued for 16 days after which, a normal diet was re-introduced (14 days post fracture). Hypophosphatemia was confirmed by measuring serum phosphate levels: after 12 days, control mice had a mean serum phosphate concentration of 10.92 ± 0.85 mg/dl, whereas mice on phosphate-restricted diets were 6.91 ± 1.07 mg/dl (p < 0.001) (Wigner et al., 2010).

2.2. Closed fracture

Detailed methods for carrying out the fracture and fixation surgeries and various outcome methods are presented in Bragdon et al. (2015). The fracture device of Marturano et al. (2008) was used to generate fractures. Fig. 1A shows graphically, the general experimental design of the fracture studies.

Fig. 1.

Schematic summary of the experimental design. A) In Vivo design to feeding schedule and experimental time points for in vivo analysis. Three strains of male mice 8–12 week mice were subjected to blunt trauma to produce closed mid-diaphysis fracture which was fixed through the insertion of an intramedullary pin. Organ level assessments were made by plain film at all time points and CT measurements at selected times. B) Experimental design of the transcriptomic analysis. Differential gene expression was determined by ANCOVA at a significance of FDR q ≤ 0.01. That list was then subjected to ontology analysis using the DAVID bioinformatics tool and manual curation of all KEGG ontologies were used to identify using groups of common genes associated with specific aspects of intermediate metabolism. This subset of genes was then subjected to temporal clustering for individual strains followed by qualitative identification of individual clusters with overlapping temporal profiles.

2.3. Analysis of cartilage and bone by contrast enhanced (CE) μCT

Calluses were scanned at a resolution of 12 μm/voxel (μCT40, Scanco Medical, Brüttisellen, Switzerland) before and after 8 h of incubation in a cationic contrast agent, CA4+ (Joshi et al., 2009), which labels cartilaginous tissues (n = 10–12 per group). Cartilage volumes were determined as previously reported (Joshi et al., 2009; Morgan et al., 2008) while mineralized tissue, total callus, mineralized callus volume fraction, and tissue mineral density (considering only mineralized tissue) were quantified (Morgan et al., 2008). Preexisting cortex and intramedullary volume were excluded from these calculations.

2.4. RNA isolation

Callus tissues were dissected and cleaned of adherent soft tissues and RNAs were then extracted individually, from each callus using Qiazol lysis reagent (Qiagen, Valencia, USA) and homogenized using Qiagen Tissue Lyser II (n = 6 per group) as previously described (Wigner et al., 2010). RNA concentrations were determined based on the 260/280/310 nm ratios of their optical densities using a DU530 UV vis spectrophotometer (Beckman, Coulter, CA). Triplicate mRNA pools were made from a randomized pooling of two callus mRNA samples from the N = 6 callus samples. The final clean-up of the RNA prior to microarray analysis was made by passing 25 micrograms of each pooled RNA sample over an RNeasy micro kit column (Qiagen, Valencia, USA) as per the manufactures instructions.

2.5. Real time PCR

Real-time quantitative PCR was performed as previously described (Jepsen et al., 2008) using an Applied Biosystems 7300 PCR machine. The comparative CT method (ΔΔCT) (Schmittgen and Livak, 2008) was used for quantification with 18S RNA as the housekeeping reference. All probes were from Applied Biosystems.

2.6. Microarray analysis

All procedures were performed at the Boston University Microarray and Sequencing Resource. 200 ng of RNA from each mRNA sample was labeled with biotin using the Ambion WT Expression Kit (Life Technologies, Grand Island, NY) followed by use of the GeneChip WT Terminal Labeling and Controls Kit (Thermo Fisher, Waltham, MA) as per the manufacturer's protocol. Hybridization washing and chip scanning was as previously described (Noguchi et al., 2018). A total of 239 samples were processed and scanned separately in five batches, which were balanced for all experimental variables, and included a set of 17 samples that were present in all batches as an internal control. All 239 CEL files were normalized together to produce gene-level expression values using the implementation of the Robust Multi-Array Analysis (RMA) algorithm (Irizarry et al., 2003) in the affy R package (version 1.36.1) and an Entrez Gene-specific probe set mapping (version 17.0.0) from the Molecular and Behavioral Neuroscience Institute (Brainarray) at the University of Michigan. Array quality was assessed by computing Relative Log Expression (RLE) and Normalized Unscaled Standard Error (NUSE) values using the affyPLM R package (version 1.34.0), and by re-normalizing the CEL files using Expression Console (build 1.3.0.187) to generate the Area Under the [Receiver Operating Characteristics] Curve (AUC) metric using positive and negative control probes. All samples had similar distributions of RLE and NUSE values, and AUC values >0.85, indicating that all samples were of suitable quality for analysis. Expression values were processed using the implementation of ComBat in the sva R package (version 3.4.0) to adjust for any technical effects with respect to microarray processing, and the ComBat-adjusted expression. The data has been submitted to NCBI and is available as Gene Expression Omnibus Series GSE 99580.

2.7. Microarray statistical and informatics analysis

Analysis of covariance (ANCOVA) (Storey, 2002) that included time and strain as covariates was used to conduct preliminary assessment of diet -related differential gene expression in batch-corrected expression values of the 21,187 genes from the microarray data (SAS 9.4, Inc., Cary, NC). A false discovery rate q-value (FDR q-value < 0.01) was used to select the subset of 2162 differentially expressed genes based on treatment (Supplemental Table 1). Ingenuity Pathway Analysis (IPA) software (http://www.ingenuity.com/) and DAVID (Huang et al., 2009) were then used for all biological and functional assessments of 2162 genes identified in our initial analysis. Only ontology groupings with modified Fisher exact test p < 0.05 were considered significantly overrepresented for these assessments. The Kyoto Encyclopedia of Genes and Genomes (KEGG) ontology groups identified in DAVID were used as the reference database to manually curate a comprehensive list of genes that were related to intermediate metabolism and energy production under five main branches of intermediate metabolism: amino acid metabolism, carbohydrate metabolism, energy metabolism (oxidative phosphorylation), lipid metabolism. We further selected three signaling pathways (PI3k-Akt, AMPK and mTOR) that have been associated with regulation of metabolic functions to further investigate in the context of hypophosphatemia. This group contained 577 genes (Supplemental Table 2). To account for genes with multiple functions being assigned to more than one of the categories, each gene was assigned a weighted average to each of the categories under which it participated. If a gene participated in multiple pathways, it was assigned a value as a percentage of its contribution across the multiple pathways. Fig. 1B summarizes the flow chart of the gene selection protocol that was used for our subsequent analysis and the overall parentage distribution of the various metabolic functions showing significant differences in expression across the three strains.

2.8. Temporal clustering assessments

Using the subset of metabolism significant genes a normal mixtures approach was used to cluster the genes together that had similar temporal ratios of expression (Pi/Ctrl), These analysis were carried out using JMP Pro (version 12.1.0 by SAS Inc., Cary, NC) for all three strains together and each strain separately and in pairwise comparisons. Non-normalized ratios were used for this analysis and the following settings were applied: the clustering had a convergence criterion of 1 × 10⁻⁸ with the number of independent restarts set to 100 to protect against finding a local solution and the EM algorithm was allowed a maximum of 1000 iterations to reach the convergence criterion. To determine the optimal number of clusters that each strain should be grouped, the genes were iteratively clustered from 2 to 25 clusters. For all three strains, the clustering results were qualitatively evaluated from the Bayesian information criterion (BIC) minimum to the Akaike information criterion (AIC) minimum to determine which metric yielded best resolution.

2.9. Cell culture

C3H10T½ murine stem cell line was used for all studies (Taylor and Jones, 1979). Cells were grown as previously described in Shea et al. (2003). For the experiments in which biochemical measurements were made cells were seeded on 12 well cell culture plates (Corning Inc., Corning, NY) at a density of 1.5 × 10⁴ cells per well. For RNA studies cells were seeded on 6 well cell culture plates (Corning Inc., Corning, NY) at a density of 4.0 × 10⁴ cells per well. For all experiments, after the cells were seeded on their respective cell culture plates, they were incubated for 2–3 days in pre-differentiation growth medium to reach approximately 60 % confluence. When this level of confluence was reached, the cell culture medium was changed to one of three types: control growth medium, 100 % Pi medium, and 25 % Pi medium. Experimental groups containing 100 % (1 mM) Pi and 25 % (0.25 mM) Pi medias were induced to differentiate by supplementation with 1× insulin-transferrin-selenium (Lonza Walkersville Inc., Walkersville, MD). To examine the effects of ascorbic acid and BMP-2, groups were treated with ±200 ng/ml BMP-2 (Research and Diagnostic Systems, Inc., Minneapolis, MN) and ±0.2 mM l-ascorbic acid (Sigma-Aldrich Corporation, St. Louis, MO). In total, eight experimental groups were established with varying combinations of BMP-2, l-ascorbic acid, and phosphate levels. The cells were fed with media every two days, with BMP-2, l-ascorbic acid, and 1× insulin-transferrin-selenium being prepared immediately before the media change. At day 7, cell materials were extracted and collected and the assays were performed. All experiments were repeated twice.

2.10. Biochemical assays

DNA and protein were extracted from individual wells using 250 μL of an extraction buffer consisting of 4 M guanidine HCl, with 1 % Triton X-100 (Sigma-Aldrich Inc., Saint Louis, MO) in 1× Tris EDTA buffer (Life Technologies Inc., Beverly, MA). Samples were transferred to 1.5 ml microcentrifuge tubes, and vortexed after which they were diluted with an additional 250 μL DNAse-free water at stored at -80 °C. DNA, protein, and hydroxyproline assays were carried out consecutively from aliquots of the same extraction solutions. DNA content was quantified using a Quant-iT™ PicoGreen dsDNA Assay Kit (Life Technologies Inc., Beverly, MA) using 50 μL of the total sample. Total protein was determined using 25 μL of the extraction solution using the MicroBCA protein assay kit (Thermo Fisher Scientific, Waltham, MA). The hydroxyproline analysis was carried out on the remaining sample using the hydroxyproline assay kit from (Sigma-Aldrich, Saint Louis, MO). For these studies the remaining extraction solutions were centrifuged at 1400 rpm for 5 min to isolate the insoluble material. The resulting supernatant containing soluble proteins was added to new 1.5 ml Eppendorf tubes. An amount of 75 μL of the supernatant containing soluble materials was added to the Eppendorf tubes with the insoluble substance. The entirety of the insoluble material and 75 μL of the soluble materials were transferred into pressure-tight vials with PTFE-lined caps and 100 μL of concentrated (12 N) hydrochloric acid was added to each tube. Acid hydrolysis was carried out for 4 h and samples were transferred to a vacuum oven where they were dried at 60 °C. All assays were run against standard curves and read at the recommended wavelengths using a BioTek Synergy 2 Multi-mode Microplate Reader.

2.11. Oxygen consumption rate quantification

7.5 × 10³ cells per well were plated and within specialty 24 well tissue plates designed for the SeaHorse assay system. The same growth and differentiation conditions were used as in the biochemical and RNA studies. All assays were carried out at the 8 day endpoint. Oxygen consumption rates and mitochondrial function were assessed using the Seahorse XF24 mitochondrial stress test. All medium and assay material, except where noted, were from Seahorse Bioscience Inc. (North Billerica, MA). Assay medium was supplemented with 20 mM d-glucose, 4 mM l-glutamine and 5 mM sodium pyruvate (Life Technologies, Beverly, MA), and the pH was adjusted to 7.4 at 37 °C. Prior to the initiation of the Sea horse Assay XF24 sensor cartridges were hydrated with XF Calibrant Solution and incubated in a non-CO₂ incubator at 37 °C for a minimum of 4 h. At the time of assay, the growth medias were aspirated from the wells and each well was washed with supplemented XF Assay Medium before being filled to a final volume of 450 μL. After changing to the supplemented XF Assay Medium, the cells were incubated in a non-CO₂ incubator at 37 °C for a minimum of 20 min. The oxygen consumption assay was carried out using three inhibitors of oxidative phosphorylation: 3 μM oligomycin (EMD Millipore, Billerica, MA), 2 μM, carbonyl cyanide-p-(trifluoromethoxy) phenylhydrazone (FCCP) (Sigma-Aldrich Inc., Saint Louis, MO), and 3 μM antimycin A (Enzo Life Sciences Inc., Farmingdale, NY). Inhibitor concentrations were chosen based on prior titration studies for optimal assay performance. The inhibitors were loaded into the sensor cartridge at a volume of 50 μL. Basal OCR was measured 4× for 1.5 min with each assay period separated by a 5-minute mix and a 30 second wait. Following the basal measurement, 3 μM oligomycin was injected into all wells in order to determine the oxygen consumption accounted for by proton leak. Proton leak was measured from 3 cycles of 3.5 min with a mix of 30 second wait, and 2 minute measure. Two μM FCCP was injected into all wells after the proton leak measurements, to assess the cells' maximal respiration rate. Maximal respiration was measured for three cycles of 5.5 minute mix, and 1.5 minute measure. Finally, 3 μM antimycin A was injected into all wells to assess oxygen consumption accounted for by non-mitochondrial respiration. Non-mitochondrial respiration was measured for three cycles of 4 minute mix, 30 second wait, and 2.5 minute measure. All wells were imaged before and after the stress test using an Olympus Stereo Fluorescence Microscope (Olympus Corporation of the Americas, Center Valley, PA), and any wells showing significant separation of the cell layer from the surface of the well or damage to the cell layer as a result of the stress test were excluded from the final OCR quantification. Measures of oxygen consumption were calculated according to the Seahorse Bioscience XF Stress Test Report Generator User Guide and all values were normalized to well DNA content.

3. Results

3.1. Genetic variability in the effect of hypophosphatemia on the progression of endochondral bone development during fracture healing

Cartilage contrast-enhanced (CE) μCT analysis was used to assess the variations in the progression of cartilage tissue formation and replacement during fracture healing. This analysis was carried out at 14 days after fracture, based on our prior studies that had shown this time-point would best capture delays in the replacement of the cartilage with bone (Jepsen et al., 2008). Differences in both structure and tissue composition of the calluses were observed among the three strains (Fig. 2A). Analysis of the amount of cartilage showed that all three strains had larger amounts of cartilage in the callus tissues in the diet-restricted mice; however, due to considerable variability for these measurements, only the AJ strain was statistically significant (Fig. 2B). As previously reported, the B6 strain had the largest calluses followed by the C3 and AJ stains (Jepsen et al., 2008); however, the total size of the callus seen for the Pi-restricted callus was not different among strains. Subsequent analysis assessed mineral accumulation over the 35-day time course of healing (Fig. 2C–D). Callus BV/TV was lower with diet restriction for both the AJ and B6 strains at 14 days, consistent with an increase in cartilage volumes at this time; however, at 14 days after return to a normal diet, no difference in callus BV/TV were observed between the two diets. TMD was lower in the diet-restricted mice of all three strains.

Fig. 2.

MicroCT analysis of cartilage and bone composition with callus tissues. A) Representative microCT reconstructions of cartilage contrast enhanced callus tissues from control and phosphate restricted mice. Contrast agent binding used for quantitative assessments of cartilage contents is denoted in red. Bone is denoted in pale yellow to white. B) Quantitation of total volume and cartilage volumes of 14 day post fracture callus tissues (N = 10–12 mice). C) Representative microCT images of mid sagittal slices of femoral callus tissue at 35 days post fracture. D) Quantitation of BV/TV and TMD of 35 day post fracture callus tissues.

The expression of three mRNAs for both the cartilage and bone lineages that were representative of early lineage commitment, the middle and late stages of each lineage's differentiation were next assessed. The mRNA expression seen in the phosphate restricted callus tissues for all three strains showed a greater and more prolonged expression of Sox9 with lower and slightly later expression of the other two lineage makers of cartilage differentiation, consistent with delayed progression of chondrogenic differentiation. Similarly, the progression of osteogenic differentiation was also delayed as inferred from the prolonged expression of Sp7. Unlike what was seen for the cartilage gene expression, later lineage stage markers of osteogenic differentiation were not decreased suggesting that once the phosphate levels were restored after 14 days that the osteogenic cells underwent an elevated progression of differentiation (Fig. 3).

Fig. 3.

Effects of hypophosphatemia on the progression of cartilage and bone cell differentiation. Expression of cartilage and bone-related genes are as denoted in the figure: Error bars represent ±1 standard error of the mean of the 2–3 repeat analyses performed on six samples per group. *A significant t-test p-value comparing Pi and Control expression levels at each time point (p < 0.05). ^#A trend in t-test p-value comparing Pi and Control expression levels at each time point (0.05 < p < 0.07).

3.2. Effects of phosphate restriction on transcriptomic expression of genes regulating oxidative and intermediate metabolism

The group of differentially expressed gene across all three strains (FDR q-value < 0.01) is summarized in Supplemental Table 1A, and was analyzed for those canonical and disease and biological function ontologies that had the most highly represented number of genes (Fig. 4A). Both canonical, disease, and bio-function groupings identified oxidative phosphorylation, mitochondrial disorders, and metabolic disorders in the top three ontology groups at very significant levels. A more narrow focus on just those gene ontologies related to intermediate metabolism and oxidative function identified 577 diet significant metabolism genes (Supplemental Table 1B). These genes were composed of signal transduction genes related to mTOrR, PI3k-Akt, and AMPK pathways, which represented 24.18 %, and carbohydrate metabolism 19.84 %. The remaining categories were relatively equally represented for oxidative phosphorylation (16.55 %), amino acid metabolism (14.21 %), nucleotide metabolism (13.60 %) and lipid metabolism (11.62 %).

Fig. 4.

Alterations in intermediate and oxidative metabolism gene expression within callus tissue is the major transcriptomic response to hypophosphatemia A) Ontology assessment of the top ten associated canonical pathways and disease and bio function related groupings using IPA based on the DE genes group with an FDR q = ≤ 0.05. B) Gene distribution between metabolic functions corresponding to genes that grouped together in the same temporal cluster across multiple strains of mice. The total number of genes in all cluster groups is shown at the top. Individual distributions of gene groups from the three pairwise comparisons are each presented. C) Graphical presentation of the down regulated genes in the mitochondrial electron transport pathway is denoted in green. (Graph as rendered from DAVID.) D) Estimates of gene expression separately for each strains of mice and for the mean expression total of oxidative phosphorylation molecules showing significant interaction between diet and strain. All of the genes were consistently down regulated, and on average the down-regulation was most pronounced in the B6 mice compared to AJ and C3 mice. E) Selected time profiles of three of the expressed genes showing both a decrease and delay in the peak expression until after mice were returned to their normal diet.

The 577 diet significant metabolism genes found in the AJ, B6, and C3 transcriptomes were separately clustered into 6, 8, or 7 unique temporal groups respectively. The genes showing common association to the various metabolic pathways that clustered together in two or more strains were then grouped together as a set. One hundred and two unique genes tracked together among all the pairwise comparisons while 74 genes clustered together across all strains. The compositional break down of all these genes based upon their associated biological function and the percentage of the total biological functions that tracked in a pairwise manner within metabolism is shown in Fig. 4B.

The primary function that tracked throughout all three of these comparisons was oxidative phosphorylation containing ~68 % (65 out of 96) of all the oxidative phosphorylation genes in the original data set. Seventeen of these genes tracked across all three strains while the remainder of these genes tracked in pairwise comparisons between each of the strains. The extent to which the genes encoding the core mediators of oxidative metabolism in all five of the electron transport complexes are down-regulated in callus tissue derived from mice fed a Pi restricted diet is shown schematically in Fig. 4C. It is interesting to note that the majority of genes in every complex of the electron transport system were downregulated. The interaction of the genes associated with oxidative function in the electron transport chain between diet and strain and their ratio of differential expression are next shown in Fig. 4D. Out of a total of 58 genes associated with the electron transport chain, 25 molecules had a significant interaction between diet and the mouse strain (FDR q-value ≤ 0.01) showing that restriction in dietary Pi has a global down-regulatory effect on mitochondrial oxidative function. This down-regulation was the most pronounced in the B6 mice compared to AJ and C3 mice. The temporal profile of three of the mitochondrial three genes showing significant difference in their expression is shown in Fig. 4E, which is presented to demonstrate the relationship of peak metabolic activity to the time course of fracture healing.

Comparisons of the temporal expression patterns found for individual clusters of genes and their relationship to intermediate metabolic function was next carried out. The profiles representative of those temporal clusters that showed the greatest similarity in their temporal expression profiles are presented in Fig. 5. These comparisons show the slight differences that was seen across strains while identify those genes that showed highest conservation in their expression across strains that were responsive to restriction in dietary phosphate. It is interesting that almost all of the functions that were identified to track together across all three strains, including glycolysis, elements of the TCA cycle, Complex I and IV of the electron transport chain and arginine and proline metabolism are directly affected by the Hif1-α transcription factor. Comparing gene sets that comprise functions regulated by HIF1-α (glycolysis – Hunziker, 1994; Wilsman et al., 1996; Ko et al., 2016, TCA - Blumer, 2021; Sabbagh et al., 2005, Complex I - Wilsman et al., 1996; Sabbagh et al., 2005, Complex IV – [6,x,2], and arginine and proline metabolism – Blumer, 2021; Miedlich et al., 2010; Wigner et al., 2010), we can see that all of the functions are regulated similarly over time in both the Pi and Control groups (Fig. 5). Glycolysis and prolyl 4-hydroxylase peaked earlier than genes in the electron transport chain. The expression levels of oxidative phosphorylation in the Pi groups were attenuated compared to the Ctrl expression and peak expression was shifted from post-operative day 14 to 18. It is also of importance that these peaks all occur after the re-introduction of phosphate at post-operative day 14. Another difference to note was that the B6 strain expression differs from AJ and C3 with regard to ETC (Complex I and IV) expression and the amplitude for these functions is attenuated through the entire timeline and does not seem to fully recover in the Pi groups.

Fig. 5.

Averaged temporal genes expression pattern in strain specific temporal clusters for four spate ontologies associated with oxidative and intermediate metabolism that are commonly regulated by Hif-1a. Each ontology group and the genes associated in the set temporal clusters for the set are denoted in the figure. Strain for the temporal clusters, time after fracture and the treatment groups for each line is denoted in the figure.

3.3. Relationship between skeletal cell differentiation, collagen protein hydroxylation and oxidative metabolism

The murine mesenchymal stem cell line C3H10T½ was used to define the relationship between phosphate levels, oxidative metabolism and differentiation. The goal of these experiments was to assess the interrelationships between phosphate levels, oxidative metabolism, cellular differentiation, proliferation and matrix formation. In order to assess these various interactions the cells were grown in their normal growth media containing 100 % and 25 % its normal phosphate content. BMP2 was used to promote differentiation, while the addition of ascorbate was used to examine the role of collagen hydroxylation in energy consumption. The phosphate concentration of the media made no difference to any of the biochemical parameters, under any of the differing growth conditions. On the other hand, BMP2 induced differentiation under all growth conditions lead to ~50 % lower DNA accumulation, and ~three fold higher total protein accumulation. In those cultures in which BMP2 was not added, levels of hydroxylproline were also all statistically lower, while the absence of ascorbate in the media led to a further reduction of ~20 % to ~50 % of hydroxylproline compared to those cultures in which ascorbate was added dependent on the other culture conditions. When looking at the expression of both collagen genes (Col2a1 and Col10a1) that are associated with early and late stages of chondrocyte differentiation, the only thing that effected these gene's expression was the presence of BMP2 in the media. Interesting while BMP2 treatment similarly effected aggrecan expression, both ascorbate and lower phosphate levels in the media led to further decreases in the levels of this gene's expression (Fig. 6).

Fig. 6.

Biochemical and molecular responses of C3H10T½ cells to phosphate levels BMP2 differentiation and ascorbate acid treatment. All assays were performed at 7 days after the cultures had reached confluence. Values represent the mean values of n = 24 wells averaged from two replicate experiments. All conditions are indicated at the bottom of the figure. Individual outcome functions that are measured for each graph are indicated for the graph.

In the final experiment that is presented, the oxidative metabolism was measured under the various culture conditions described above. Under all growth conditions, BMP2 treatment increased oxidative activity and under all conditions, the absence of ascorbate led to lesser activity. In contrast, the amount of phosphate in the media only led to higher levels of ATP production and an increased basal capacity with no effects on either maximal of spare capacity while these changes in oxidative metabolism were only seen in cultures that where grown in ascorbate and treated with BMP2. While other differences were seen between individual aspects of oxidative capacity these differences were specific for separate combinations in culture conditions; however none of these differences were consistent across a whole treatment group (Fig. 7).

Fig. 7.

Oxidative function of C3H10T1/2 cells to phosphate levels BMP2 differentiation and ascorbate acid treatment. All assays were performed at 7 days after the cultures had reached confluence. Values represent the mean values of n = 24 wells averaged from two replicate experiments. All values were normalized to the DNA content of the well in which the values were determined. All conditions are indicated at the bottom of the figure. Individual outcome functions that are measured for each graph are indicated for the graph.

4. Discussion

4.1. Role of phosphate in metabolism

Our prior studies showed dietary phosphate restriction led to longer and reduced circadian gene expression and function in fracture callus tissues (Noguchi et al., 2018). The transcriptomic data shown here are consistent with the linkage between phosphate metabolism and intermediate metabolism since the regulation of TCA cycle and oxidative phosphorylation metabolism are known downstream targets of circadian regulation (Panda, 2016). Our observations also suggest that the two known primary systemic environmental factors, starvation (Farnum et al., 2003) and phosphate restriction (Liu et al., 2014), that lead to either cessation or slower endochondral bone formation in epiphyseal growth plates might be mechanistically linked through changes in circadian function. The fracture healing data presented here also provide a clear link to how other systemic disturbances in metabolism including diabetic condition and obesity could delay fracture healing. In this regard, it is noteworthy that congenital hypophosphatemia and decreased systemic phosphate levels lead to a diabetic condition (Zelenchuk et al., 2014), while congenital and dietary hypophosphatemic rickets disorders in both human and animal models also show changes in insulin sensitivity (Paula et al., 1988), glucose levels and the metabolism of fat (Xie et al., 2002; Hess et al., 2007; Hruska et al., 1995). It also is of interest to note that systemic phosphate levels correlate with changes in glucose production, energy metabolism and oxygen consumption (Xie et al., 2000; Ditzel and Lervang, 2010; Haap et al., 2006). Conversely, primary diabetic acidic ketosis leads to hypophosphatemia (Ditzel and Lervang, 2010) and hyperparathyroidism is associated with impaired glucose-tolerance, hyperglycemia and reduced insulin sensitivity (Corbetta et al., 2018). Such data suggest that essential nutrient factors, whether they are various organic nutrient factors or inorganic elements such as phosphate might share common mechanisms of controlling metabolism and bone growth. Finally it should be noted that while many of these changes are most likely mediated systemically through actions on kidney function, the current data suggest that they also act locally at the level of skeletal tissue cellular function.

4.2. Mesenchymal differentiation and the progression of chondrocytes and osteoblast development and oxidative phosphorylation metabolism

Prior studies of phosphate restriction in both post-natal growth plate and fracture callus showed that both ends of growth plate differentiation were affected, leading to both increased numbers of stem/progenitor cells (Wigner et al., 2010) and decreased hypertrophic cells showing apoptosis (Sabbagh et al., 2005). The overall gene expression for oxidative and intermediate metabolic activity observed in this study is consistent with increased numbers of MSCs (Shum et al., 2016) and delayed chondrocyte maturation, since both MSCs and chondrocytes at earlier stages of their development show lower oxidative phosphorylation (Bentovim et al., 2012a; Wang et al., 2021). Also consistent with our observations are studies showing that growth cartilage has a higher aerobic oxidative phosphorylation than articular cartilage and that deeper chondrocyte zones within articular cartilage closer to the underlying marrow have higher oxidative phosphorylation metabolism (Heywood et al., 2010). Other studies performed within the C3H10T½/cell line induced to undergo osteogenic differentiation through canonical Wnt signaling also showed increased oxidative phosphorylation metabolism (An et al., 2010). This observation though does not appear to be unique to this cell line, since increased mitochondrial oxidative phosphorylation (OxPhos) appears to be a common feature of osteogenic differentiation in both primary human and mouse marrow stromal stem cells (Shares et al., 2020; Smith and Eliseev, 2021), the murine MC3T3 osteogenic cell line (Shares et al., 2018), or primary cultures of mouse calvaria cells (Guntur et al., n.d.).

4.3. Relationship of BMP to oxidative function

The studies accessing the direct role of phosphate in the media of C3H10½ cells did not show any effect on oxidative metabolism, suggesting that the systemic effects of phosphate on oxidative metabolism were indirectly mediated. Our prior studies of in vivo fracture healing and stem cell differentiation under conditions of acute phosphate restriction however, had shown that the impairment of skeletogenic stem cell differentiation was related mechanistically to decreased bone morphogenetic protein signaling (Wigner et al., 2010). In other studies, we had also shown that chondrogenic differentiation of C3H10½ cells cell line was promoted by BMP2 (Shea et al., 2003), leading us to assess if BMP2 would also increase oxidative metabolism of this cell line. We found that BMP2 treatment led to a ~2-fold increase in oxidative metabolism. Consistent with our studies examining BMP2 effects in these cells, studies examining the role of Wnt signaling in C3H10½ cells showed similar increases in oxidative aerobic metabolic activity associated with osteogenic differentiation. Interestingly, BMP2 and insulin treatment used to promote adipogenic differentiation in this cell line also increase oxidative metabolism (An et al., 2010). Further, BMP2 signaling is essential for glucose metabolism during embryonic skeletal development (Lee et al., 2018), although how this requirement is related to oxidative metabolism is unclear. Our findings would suggest that progression of differentiation alone, independent of the signaling processes that promote cell differentiation, was associated with elevated aerobic oxidative phosphorylation. Again, such findings are generally consistent with many with other studies showing that differentiation in general favored pyruvate oxidation and OxPhos function (Dahan et al., 2019). Finally, it is interesting to note that Wnt and BMP signaling can mediate their actions through many other intracellular mechanisms. Some of these will either directly act on mitochondrial function or through other signal transduction process such as mTOR (Karner et al., 2017; Huang et al., 2021) and HIF1α, or through their actions with Hippo signal transduction which integrates BMP and Wnt signaling processes and is known to regulate intermediate metabolism (Koo and Kun, 2018).

4.4. Protein and collagen hydroxylation and not proliferation are major consumers of oxidative metabolism

The examination of BMP2 and ascorbate effects on oxidative metabolism in the C3H101/2 cells relative to basic aspects of the growth and differentiation of these cells was informative to how oxidative metabolism is used in these processes. In the absence BMP2 and ascorbate these cells had much lower oxidative metabolism while showing a 2- to 3-fold increase in their total DNA relative to those cultured treated with BMP2 and ascorbate. These results are consistent with the hypothesis that cell proliferation is primarily supported aerobic glycolysis (Vander Heiden et al., 2009). In contrast BMP2 induced differentiation or ascorbate treatment alone enhanced protein accumulation ~2-fold and together increased it by 3-fold while at the same time producing much higher levels of oxidative metabolism. The relationship between increased oxidative phosphorylation and protein synthesis has been extensively documented, relative to how both mitochondria and gene expression associated with the biosynthesis of these organelles are co-regulated with mitochondrial and cytoplasmic translation and associated with greater levels of oxidative metabolisms (Ott et al., 2016). It is less clear how this relationship is related to the overall protein synthetic activities associated with differentiated functions; however, the data that are presented here suggest that this relationship is more generally applicable within cell types whose differentiated function is related to high levels of protein synthesis.

As a further assessment of what aspects of cellular function were associated with increasing oxidative metabolism, we tested if collagen hydroxylation would have an impact on oxidative phosphorylation since this enzymatic function has a requirement for ascorbate as a cofactor and needs oxygen iron, and α-ketoglutarate as part of the enzymatic process (Stegen et al., 2019). Multiple aspects of the oxidative pathway are also needed for the production of collagen including the production of α-ketoglutarate for the use in hydroxylation of collagen as well as in the synthesis of proline which is biochemically derived from the conversion to glutamate and in urea cycle (Albaugh et al., 2017) which is coupled to energy metabolism via OxPhos. It is interesting to note however, that embryonic chondrocyte differentiation is dependent on active Hif-1a (Bentovim et al., 2012b) and that the transcription of prolyl 4-hydroxylase, which catalyzes collagen hydroxylation was dependent on Hif-1a. Since this enzyme is normally associated with active hypoxia and aerobic glycolytic activity this observation is contrary to the observations reported here. One possibility is that these prior results were based on results within embryonic growth plates and primary cells derived from embryonic tissues have differing mechanism of metabolic regulation than observed in the C3H10T/½ cell line. Conversely, studies have shown that mechanisms of self-renewal and clonal expansion during embryogenesis and post-natal bone growth are very different within the epiphysis. Using the PthrP-creER promoter for lineage tracking, studies (Mizuhashi et al., 2018) showed that individual SSC could clonally give rise to full growth plate columns containing both proliferative and hypertrophic chondrocytes in both embryonic mice and early post birth mice for up to seven days. Interesting, prolonged chase times after birth gave rise to subpopulations of SSC that further expressed osteogneic genes suggesting a new population of cells arise postnatally. Other lineage tracking studies using the Col2a1 showed that there was an initial depletion of clonal cells capable of self-renewal within the growth plate in the first month of post-natal life. Subsequently, there was a renewal of epiphyseal SSPCs that once again were able to show clonal expansion and full formation of chondrogenic growth plate columns (Newton et al., 2019). Consistent with the idea that skeletal development uses different mechanisms of embryonic and postnatal control, was the observation that BMP2 is not necessary for embryonic endochondral tissue differentiation but it is required for post-natal development and fracture healing (Tsuji et al., 2006). These data therefore suggest that SSC have widely different growth capacities, mechanisms of maintenance, and functional abilities for regeneration of skeletal tissues that is different between the embryonic and postnatal states (reviewed in Kurenkova et al. 2020; Lui, 2020; Chagin and Newton, 2020). Such differences are reminiscent of variations in the hematopoietic tissue development that are dependent on oxygen tension involving a switch from embryonic to postnatal forms of hemoglobin expression. Similarly, such developmental differences might be dependent on the need for new mechanisms of growth control related to the varying availability of metabolites, which are relatively constant during embryonic growth but can be highly variable during in the post-natal environment.

In summary, these studies show that in vivo hypophosphatemia lead to delayed endochondral stem cell differentiation, and that these cells have reduced gene expression for mitochondrial oxidative metabolic function. These data further suggest that the in vivo effects of dietary phosphate restriction of endochondral growth are indirectly mediated through BMP signaling, which upregulates oxidative activity that is linked to overall protein production and collagen hydroxylation.

The following are the supplementary data related to this article.

Funding

Supported with grants from the National Institute of Arthritis and Musculoskeletal and Skin Diseases RO1AR05974 to LCG. Institutional support was provided by the Department of Orthopaedic Surgery Boston University School of Medicine and Boston University School of Medicine.

CRediT authorship contribution statement

Amira I. Hussein: Formal analysis, Data curation, Writing – review & editing, Visualization, Supervision, Project administration. Deven Carroll: Methodology, Data curation, Software, Formal analysis, Validation, Investigation, Writing – original draft. Mathew Bui: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Alex Wolff: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Heather Matheny: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Brenna Hogue: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Kyle Lybrand: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Margaret Cooke: Methodology, Data curation, Formal analysis, Validation, Investigation, Writing – original draft. Beth Bragdon: Writing – review & editing, Supervision. Elise Morgan: Writing – review & editing, Supervision. Serkalem Demissie: Conceptualization, Writing – review & editing, Supervision. Louis Gerstenfeld: Conceptualization, Formal analysis, Data curation, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Louis C Gerstenfeld reports financial support was provided by National Institutes of Health.

Data availability

Data will be made available on request.