The Cortical Bone Metabolome of C57BL/6J Mice Is Sexually Dimorphic

ABSTRACT Cortical bone quality, which is sexually dimorphic, depends on bone turnover and therefore on the activities of remodeling bone cells. However, sex differences in cortical bone metabolism are not yet defined. Adding to the uncertainty about cortical bone metabolism, the metabolomes of whole bone, isolated cortical bone without marrow, and bone marrow have not been compared. We hypothesized that the metabolome of isolated cortical bone would be distinct from that of bone marrow and would reveal sex differences. Metabolite profiles from liquid chromatography–mass spectrometry (LC‐MS) of whole bone, isolated cortical bone, and bone marrow were generated from humeri from 20‐week‐old female C57Bl/6J mice. The cortical bone metabolomes were then compared for 20‐week‐old female and male C57Bl/6J mice. Femurs from male and female mice were evaluated for flexural material properties and were then categorized into bone strength groups. The metabolome of isolated cortical bone was distinct from both whole bone and bone marrow. We also found sex differences in the isolated cortical bone metabolome. Based on metabolite pathway analysis, females had higher lipid metabolism, and males had higher amino acid metabolism. High‐strength bones, regardless of sex, had greater tryptophan and purine metabolism. For males, high‐strength bones had upregulated nucleotide metabolism, whereas lower‐strength bones had greater pentose phosphate pathway metabolism. Because the higher‐strength groups (females compared with males, high‐strength males compared with lower‐strength males) had higher serum type I collagen cross‐linked C‐telopeptide (CTX1)/procollagen type 1 N propeptide (P1NP), we estimate that the metabolomic signature of bone strength in our study at least partially reflects differences in bone turnover. These data provide novel insight into bone bioenergetics and the sexual dimorphic nature of bone material properties in C57Bl/6 mice. © 2022 The Authors. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.

Metabolomics, the study of small molecule intermediates called metabolites, has the potential to provide novel insight into the connection between bone cell metabolism and bone quality. Bone cell metabolism is currently studied using several different approaches. The Seahorse assay is commonly applied to bone marrow to determine oxygen consumption rate. (3,(25)(26)(27)(28)(29)(30) However, cortical bone and its metabolism likely differs from bone marrow. Shum and colleagues (31) used an liquid chromatography-mass spectrometry (LC-MS) approach to study marrow-flushed cortical mouse bone. They found a glycolytic shift from 3 to 13 months of age in male C57Bl/6 mice. These results helped to contextualize changes to mitochondrial function with age. In another study, Zhao and colleagues (32) used LC-MS marrow-flushed mouse femurs to understand changes with ovariectomy (OVX) to cortical bone bioenergetics. They found that OVX mice had disordered lipid and amino acid metabolism. Collectively, these prior studies show that cortical bone metabolism changes in aging and disease models. However, it is still unknown how sex influences the cortical bone metabolome.
Another important question is whether, and how, cortical bone metabolism differs from bone marrow. Bone marrow contains mesenchymal stem cells, which give rise to numerous cell types including adipocytes, osteoblasts, osteocytes, as well as myeloid and immune cells. (33) Cortical bone is mostly cellularized by osteocytes (~90%), (34,35) but metabolites may persist in this tissue from bone marrow. This question can be addressed by comparing the metabolomes of isolated cortical bone, bone marrow, and whole bone using metabolomics.
The purpose of this study was to (i) evaluate differences in the metabolomes of whole bone (ie, cortical and trabecular bone and marrow), isolated cortical bone, and bone marrow; (ii) assess if the cortical bone metabolome is sexually dimorphic; and (iii) determine if metabolomic differences correspond with whole-bone strength, a sexually dimorphic cortical bone material property.

Animals
Two sets of mice were utilized for this study (total n = 30). First, C57Bl/6J female mice (n = 10) were utilized to compare the metabolomes of whole bone, isolated cortical bone, and bone marrow. These mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and acclimated to the Montana State University animal facility for 3 weeks. Second, 20 C57Bl/6J mice (female, n = 10; male, n = 10) were utilized to evaluate sex differences in the cortical bone metabolome. These mice were born and raised at Montana State University.
All mice were housed in cages of three to five mice and fed a standard chow fed diet ad libitum (PicoLab Rodent Diet 20, 20% protein; LabDiet, St. Louis, MO, USA). All 30 mice were euthanized via cervical dislocation at age 20-21 weeks. All animal procedures were approved by the Institutional Animal Care and Use Committee at Montana State University. Investigators remained blinded to mouse sex during data collection and analyses.

Metabolomics
Experimental design to assess metabolic differences in bone and marrow We first evaluated differences in metabolomic profiles of isolated cortical bone, bone marrow, and whole bone for humeri from female mice (n = 10). To isolate bone marrow from cortical bone, the distal and proximal ends of right humeri were removed and the cortical shaft flushed with phosphate-buffered saline (PBS). Both cortical bone and marrow were preserved for analyses. The left humerus remained whole.
To extract bone metabolites, whole humeri and isolated cortical bone were placed in liquid nitrogen for 2 hours and pulverized to a powder using an autoclaved aluminum cold sink, rods, and a hammer to optimize extraction. Bone powder was extracted with 3:1 methanol:acetone. Samples were subjected to five cycles, each consisting of 1 minute of vortexing followed by macromolecule precipitation at À20 C for 4 minutes. Samples were stored overnight at À20 C to allow for precipitation of remaining macromolecules. The next day, samples were centrifuged to remove cells and debris, supernatant containing metabolites was collected, and dried down via vacuum concentration. Once the supernatant was isolated and dried, metabolites were resuspended with 1:1 acetonitrile:water (Fig. 1).
Flushed bone marrow underwent a similar extraction protocol. To extract bone marrow metabolites, 1 mL of 70:30 methanol:acetone was added and the mixture was subjected to five cycles of vortexing and À20 C precipitation. Samples were then stored overnight at À20 C to precipitate any remaining macromolecules. The next day, samples were centrifuged and supernatant containing metabolites was dried down via vacuum concentration. Dried metabolites were resuspended with 1:1 acetonitrile:water ( Fig. 1). All solvents used were high-performance liquid chromatography (HPLC) grade or higher.

Evaluation of sex differences in isolated cortical humerus metabolomic profiles
We investigated sex differences in metabolomic profiles of cortical humeri for C57Bl/6 mice (females, n = 10; males, n = 10). Left humeri were dissected, cleaned, and bone marrow flushed to isolate cortical bone tissue. Bones were stored in PBS-dampened gauze at À20 C after dissection until the metabolite extraction. For isolated bone metabolite extraction used the protocol described in section "Experimental design to assess metabolic differences in bone and marrow." Following metabolite extraction of whole bone, bone marrow, or isolated bone, all samples (n = 30) were analyzed by liquid chromatography-mass spectrometry (LC-MS) using an Agilent 1290 LC coupled to an Agilent 6538 Quadrupole-Time of Flight (Q-TOF) mass spectrometer in positive mode (resolution: 25,000 full width at half maximum [FWHM], accuracy: AE 5 parts per million [ppm]; Agilent Technologies, Santa Clara, CA, USA). A Cogent Diamond Hydride hydrophilic interaction liquid chromatography (HILIC) chromatography column was utilized (2.2μM, 120 Å, 150 mm Â 2.1 mm; MicroSolv, Leland, NC, USA). Five microliters (5 μL) of sample was injected and blank samples were analyzed every seven to 10 samples for quality control to prevent spectral drift and contamination. Agilent Masshunter Qualitative Analysis software was used to identify and export peak intensity values for m/z values in the experimental sample set. LC-MS data was then exported and converted for analysis using XCMS. Differential analysis of mass features was completed using Meta-boAnalyst. Kyoto Encyclopedia of Genes and Genomes (KEGG) was utilized to confirm retention time and exact mass of detected metabolite features.

Bone marrow adiposity
Proximal tibias were dissected and fixed for 18 hours in 10% neutral buffered formalin. Tibias were then decalcified with ethylenediamine tetraacetic acid (EDTA), dehydrated in a graded ethanol series, and embedded in paraffin. Five-micrometer (5-μm) sections were cut longitudinally onto glass slides and stained with hematoxylin and eosin (H&E) per standard protocols.
Sections were imaged using a Nikon Eclipse E-800 (Nikon, Tokyo, Japan) and Universal Imaging Corporation's MetaVue software (version 7.4.6; Bedford Hills, NY, USA). Images were taken with a 4Â objective (909 pixels/1 mm). A central section proximal tibia was selected for analyses. Adipocytes were manually segmented, counted, and measured using the iPad app YouDoodle and custom MATLAB (MathWorks, Natick, MA, USA) code. Measurements included mean adipocyte area (mm 2 ), marrow cavity area (mm 2 ), and adipocyte number density (number of adipocytes per marrow cavity area).
To confirm that the technique utilized to flush marrow from the humerus, additional humeri were assessed. These bones were dissected, harvested, and sections were prepared, imaged, and analyzed consistent with the description for tibias. Microscopy confirmed that marrow was almost entirely removed from the flushing step (Fig. S1).

Trabecular microarchitecture and cortical geometry
After dissection, left femurs were stored at À20 C wrapped in PBS-soaked gauze. These were thawed for micro-computed tomography (μCT) (Scanco uCT40; SCANCO Medical AG, Brüttisellen, Switzerland) analysis and afterward refrozen as before until flexural testing. Scans were acquired using a 10-μm 3 isotropic voxel size, 70 kVp, 114 μA, 200 ms integration time, and were subjected to Gaussian filtration and segmentation. Image acquisition and analysis adhered to the JBMR guidelines. (36) Trabecular bone microarchitecture of the distal femur metaphysis was evaluated in a region beginning 200 μm superior to the top of the distal growth plate and extending proximally 1500 μm proximally. The trabecular region was identified by manually contouring the endocortical region of the bone. Trabecular bone was segmented from soft tissue using a threshold of 375 mg hydroxyapatite (HA)/cm 3   (500-μm-long region) at the femoral mid-diaphysis and the region included the entire outer most edge of the cortex. Cortical bone was segmented using a fixed threshold of 700 mg HA/cm 3 . Measurements included cortical tissue mineral density (Ct.TMD, mg HA/cm 3 ), cortical bone area (Ct.Ar, mm 2 ), polar and minimum moment of inertia (pMOI, I min , mm 4 ), medullary area (Ma.Area, mm 2 ), total cross-sectional area (bone + medullary area) (Tt.Ar, mm 2 ), and bone area fraction (Ct.Ar/Tt.Ar, %).
Femoral whole-bone mechanical and material properties Left femurs were frozen and thawed once before three-point bending. The test was performed to failure at a rate of 5 mm/min on a custom fixture with an 8-mm span, such that femurs were loaded with the posterior side facing down (1 kN load cell; Instron, Inc., Grove City, PA, USA; Instron 5543). Femurs were hydrated before testing using PBS. Using load-displacement data and the I min and C min (the distance between the center of mass and bone surface in the anterior-posterior direction) values from μCT, modulus, yield strength, maximum strength, and toughness were calculated using standard equations for the mouse femur. (37) Right femurs were assessed for notched fracture toughness following methods in alignment with prior description. (38) Thawed femurs were hydrated with PBS and notched on the posterior side to a target of 1/3 of the anterior-posterior width on the posterior surface using a custom precision saw. Notched femurs were then tested with the notched side down in three-point bending at a rate of 0.001 mm/s using an 8-mm span until failure. Fractured femurs were dried overnight at room temperature and imaged using field emission scanning electron microscopy (Zeiss SUPRA 55VP; Carl Zeiss Microscopy, Inc., Dublin, CA, USA) in variable pressure mode (VPSE, 20 Pa, 15 kV) and analyzed using a custom MATLAB code to quantify cortical geometry and the initial notch angle. Fracture toughness (K c ) was calculated from the initial notch angle and the maximum load. (38) To assess how differences in bone strength associate with metabolic profiles, we defined high-strength and low-strength groups for additional comparisons. Males demonstrated higher variability of strength than females. "Higher-strength" and "lower strength" groups were defined by a threshold of AE7% away from the mean strength for males. For females, highstrength bones were those >5% greater than the female group mean. Low-strength bones were not selected for females because there was not a cluster of distinctly low-strength bones for this sex.

Statistical analysis
The effect of sex on all bone outcomes except metabolomic measures was tested using two-sample t tests (Minitab, v.19; Minitab, Inc., State College, PA, USA). Data were checked for normality. Non-normal data were evaluated using the nonparametric Mann-Whitney test. Significance was set a priori to <0.05.
A concise overview of the statistical methods used in this study are discussed in detail in Appendix S1. In brief, HCA and PCA are unsupervised multivariate statistical analyses that assess differences in metabolomic profiles between experimental groups. HCA identifies sub-groups of samples and determines differences between groups. PCA is an unsupervised technique that finds components (PCs) of the dataset that align with the overall variation to examine the underlying structure of the data. PLS-DA is a supervised analysis that partitions variation within the data base on a priori knowledge of experimental groups. PLS-DA further calculates a variable importance in projection (VIP) score to quantify how much each metabolite feature contributes to discriminating between cohorts. Taken together, these three approaches, HCA, PCA, and PLS-DA, provide a global view of the thousands of metabolites comprising the metabolome. This view describes how experimental groups are both similar and different. These complementary analyses provide insight into interactions occurring at the metabolic level.
Finally, volcano plot and fold change analyses are utilized to identify differentially regulated metabolite features between two groups. Employing these two tests allows identification of metabolites that differ in intensity between groups. These differentially expressed metabolites are then subjected to pathway analysis. Pathways are determined using MetaboAnalyst's MS Peaks to Pathways feature with the mummichog algorithm. This allows for metabolite compounds identified from the statistical tests described to be associated with biological pathways and for prediction of networks of functional cellular activity. Significance was defined a priori as 0.05. In the case of multiple comparisons, p values were corrected for false discovery rate (FDR) to maintain family-wise error at 0.05.

Results
The metabolome is distinct for whole bone, isolated cortical bone, and bone marrow We first tested whether metabolomic profiles are distinct for whole bone (ie, cortical and trabecular bone and marrow), isolated cortical bone, bone marrow from female mice (Table S1). Unsupervised (HCA and PCA) and supervised (PLS-DA) multivariate statistical analyses were utilized to compare the metabolome of these three tissue groups. The global metabolomic profiles of bone marrow, isolated cortical bone tissue, and whole bone are substantially different ( Fig. 2A-C). HCA showed perfect clustering of samples within their respective groups ( Fig. 2A). PCA and PLS-DA also showed clear separation between the metabolomes of the three groups (Fig. 2B,C). There were 2178 metabolite features that were significantly different between tissues groups using analysis of variance (ANOVA) with FDRcorrected p FDR <0.05. Further, 1120 metabolite features were detected that differed between the three experimental groups with p value <0.0001 (Fig. 2D). In-depth analysis revealed metabolite features unique to isolated cortical bone and bone marrow. Unsupervised clustering of the median intensity for each metabolite within cohort summarizes the major differences between cohorts (Fig. 2E).

The cortical bone metabolome differs by sex
We then compared the isolated cortical bone metabolome for female and male mice. A total of 2129 distinct metabolite features were detected. These features were further analyzed (Fig. 3). HCA showed that males and females cluster separately except for one male mouse (Fig. 3A). We then utilized PCA to analyze the overall variation in the dataset between males and females. Principal component 1 accounted for 24% of the variation in the dataset (ie, if the metabolomes were the same, principal components would account for 1/[2129 metabolite features Â 20 samples] = 0.00002% of the variation in the dataset).This result strongly suggests that the metabolome for isolated cortical bone from males and females is different (Fig. 3B).
We used PLS-DA to further examine the variation between the male and female cortical bone metabolome. This supervised analysis produced a complete separation between the metabolomes of male and female mice (Fig. 3C). VIP scores from PLS-DA analyses were used to identify specific metabolite features that contribute the most to differences between male and females. Of the top 300 metabolite features that contributed to the distinction between groups, the majority (75%) were upregulated in male mice. Significant pathways that contributed to the separation of groups and are primarily associated with male isolated cortical bone included amino acid metabolism (cysteine, methionine, arginine, proline) and central energy metabolism (glycolysis, tricarboxylic acid [TCA] cycle).
We then utilized t test, fold change, and volcano plot analysis to identify specific metabolite features that differ between male and female cortical bone (Fig. 3). Based on a priori significant level of 0.05, volcano plot analysis revealed 126 metabolite features were upregulated among female mice and 225 were upregulated among male mice (Fig. 3D). Fold change analysis yielded 487 metabolite features that were differentially regulated between groups (Fig. 3E). Student's t test identified 318 metabolite features (Fig. 3F). Numerous metabolic pathways were found for male and female cortical bone ( Table 1). The main metabolic theme upregulated for males compared with females was amino acid metabolism. Upregulated metabolites included cysteine, methionine, alanine, aspartate, and glutamate. Central energy metabolism, including glycolysis, the TCA cycle, and purine metabolism, were also upregulated in males. The main metabolic theme upregulated for females was lipid metabolism. This includes sphingolipid metabolism, GPI-anchor biosynthesis, glycerophospholipid metabolism, and the fatty acid pathways of biosynthesis, degradation, and elongation.

Females have increased bone marrow adiposity and serum biomarkers of resorption
Because lipid metabolism was prominent in the female bone metabolome, we sought to investigate if metabolic upregulation of lipid metabolism corresponds to an upregulation in adiposity in females compared to males. Adipocyte count and number of adipocytes per marrow cavity area were 81% (p < 0.0001) and 83% (p < 0.0001) higher in females compared to males (Fig. 4A-E). Thus, metabolic and histological data agree that females had both increased bone marrow adiposity as well as an increased metabolic signature of lipid metabolism.  We also assessed serum biomarkers to estimate global bone resorption and formation activities. CTX1, a biomarker of global bone resorption, was higher in females compared to males (+51.4%, p = 0.011) (Fig. 5A). P1NP, a biomarker of global bone formation, was lower in females compared to males (À22.8%, p = 0.027) (Fig. 5B). The ratio of CTX1 to P1NP was higher in females compared to males (+ 87.1%, p = 0.001) ( Fig. 5C; Table S1).
Sex differences in the cortical bone metabolome are associated with sex differences in bone strength Females in this study had higher femur strength than males. Although females did not have much spread in strength (coefficient of variation: 4.7%), males showed more variation (coefficient of variation: 7.6%). Thus, we evaluated whether the cortical humerus metabolome differed for males with the highest strength and highest lowest strength femurs. We also evaluated whether the cortical humerus metabolome differed between males and femurs with the highest strength femurs of their sex.
The highest-strength (n = 4) and lowest-strength femurs (n = 4) from males and only the strongest female femurs (n = 3) were grouped (males = AE7%, females = +5% difference from the mean values for each sex). There was not a distinct lower-strength group for females, so only higher-strength femurs were grouped. We employed HCA, PCA, and PLS-DA to assess if metabolomic profiles were associated with each strength group (Fig. 6). High-strength females, high-strength males, and low-strength males had partial separation in metabolomic profiles, as seen from HCA (Fig. 6A). PCA found differences between samples within their respective cohorts. PC2 and PC3 were analyzed and together accounted for almost 30% of the overall variation in the dataset (Fig. 6B,C). PLS-DA showed that each of the three strength groups are metabolically distinct from each other (Fig. 6D).
Unique pathways and features for each strength group were identified using HCA and heat maps (Fig. 6E). The data suggest that strong female femurs are associated with terpenoid backbone biosynthesis compared to high-and lowstrength males (Table 1). Metabolic pathways shared by males and females with stronger femurs included tryptophan and porphyrin metabolism. Metabolic themes unique to strong male femurs included purine and pyrimidine metabolism, beta-alanine metabolism, and aminoacyl transfer RNA (tRNA) biosynthesis. The metabolome of males with weaker femurs included the PPP, pantothenate and coenzyme A (CoA) biosynthesis, and phosphatidylinositol signaling system.
The metabolomes of cortical bone, bone marrow, and whole bone are distinct We found that metabolomic profiles of isolated cortical bone, bone marrow, and whole bone of the humerus are distinct. Pathways upregulated in bone marrow included steroid metabolism, vitamin metabolism, and purine metabolism. A common upregulated pathway shared by all three tissues was amino acid  Data are presented as mean AE standard deviation from the mean.
JBMR ® Plus metabolism. However, different amino acids were upregulated in each tissue. Phenylalanine, tryptophan, tyrosine, and lysine metabolism were upregulated in bone marrow. These aromatic amino acids are commonly grouped together because of their aromaticity and because they feed into, or can be the products of, glycolysis and the TCA cycle. (57) These findings are consistent with the paradigm of bone marrow as an active endocrine organ, with many cells that serve as a long-term energy reserve. (58) It is important to note that bone marrow is composed of numerous cell types and is influenced by sex, age, and endocrine factors. (33) Therefore, metabolic signatures of flushed humeri-derived bone marrow may be driven by many unique factors. Pathways upregulated in isolated cortical bone and whole bone, but not marrow, included cysteine, methionine, histidine, and beta-alanine metabolism. Methionine is an essential amino acid that plays a large role in several metabolic processes including nucleotide synthesis. Methionine restriction prolongs lifespan in mice. However, methionine restriction also has negative impacts on bone density, overall bone structure, and the innate immune system. (59) Through the conversion of methionine to homocysteine, a common downstream product of methionine metabolism is cysteine. (60) Cysteine, along with other glucogenic amino acids, can be used as an energy source by conversion into glucose. In postmenopausal women, low cysteine has been associated with low BMD and increased bone turnover. (61) Dysregulation of cysteine, and upstream methionine, can lead to bone fracture and loss. (62) Elevated levels of histidine in orthopedic tissues, including synovial fluid and cartilage, is found both in response to mechanical stimulation and osteoarthritis. (63,64) Beta-alanine was elevated in synovial fluid, serum, and subchondral bone in osteoarthritis animal models. (44,52,65) Pathways unique to cortical bone include mannose type-O glycan biosynthesis, linoleic metabolism, glycosphingolipid biosynthesis, and quinone biosynthesis (ubiquinone, terpenoid-quinone). Isolated cortical bone and whole bone differ not only from marrow flushing but also because whole bone includes trabecular bone. Therefore, pathways that are different between cortical bone and whole bone that do not align with those of bone marrow may also reflect differences apparent between cortical and trabecular compartments.
Males had elevated P1NP, lower CTX1, and upregulated amino acid metabolism. Metabolic pathways previously associated with P1NP include energy metabolism (eg, TCA cycle), amino acid metabolism, and pyrimidine metabolism. (66) Energy and amino acid metabolism were also upregulated in male isolated cortical bone. Upregulated amino acids included alanine, aspartate, glutamate, cysteine, methionine, proline, and arginine. Alanine is generated from pyruvate and is therefore closely linked to glycolysis and the TCA cycle. Alanine has been shown to aid in the recycling of carbon backbones in skeletal muscle and the liver. (67) Glutamate is an excitatory neurotransmitter, and there is a growing body of evidence of a regulatory role for glutamate in osteoblast and osteoclast differentiation, as well as bone homeostasis. (68)(69)(70)(71)(72) Our metabolomic and serum biomarker analyses suggest that the elevated P1NP concentration in males is associated with the upregulation of energy and amino acid metabolism (Figs. 5 and 6, Table 1; Table S2).
Female mice had greater bone marrow adiposity and upregulated lipid metabolism ( Table 1). The metabolism of lipids is essential for energy regulation, membrane dynamics, and signaling. (73) Fatty acids and other lipids are transported systemically via chylomicrons to be cleared by the liver and bone. After the liver, the femur diaphysis is the second most active organ for chylomicron reuptake, supporting the idea that normal osteoblast proliferation is fueled by serum lipoproteins. (74,75) This is consistent with the canonical role of bone's importance in fatty acid clearance for energy purposes. (76,77) When lipids are limited, osteogenesis is negatively influenced because osteoblasts depend on fatty acid oxidation. (78) Upregulation of fatty acid metabolism for females may demonstrate potential usage of fat as a substrate. Further research investigating the relationship between osteocyte metabolism, marrow adiposity, and energyrelated substrates is required because adenosine triphosphate (ATP) levels were not measured in this study.
It is possible that this increased lipid metabolism for females has a functional role for bone resorption. We observed that females had greater biomarkers of bone resorption (CTX1 and CTX/P1NP) than males. In a study investigating healthy young adult serum using metabolomics and biomarker analysis, CTX1 was positively correlated with lipid metabolism (fatty acid biosynthesis), beta oxidation, and carbohydrate metabolism in humans. (66) Previous studies found that osteoclasts' high energy demand can be fueled by lipids, (66,(79)(80)(81) but it is uncertain if osteoblasts and osteocytes are fueled by lipids in the same way. Our study utilized marrow-flushed cortical bone. It is possible that metabolic products of osteoclasts may still be present within cortical tissue.
Bone metabolism differs in association with bone strength and bone turnover Because bone turnover strongly influences bone strength (82) and requires cellular energy production and utilization, (23,24) we hypothesized that the cortical bone metabolome would differ between groups of different bone strength. The cortical bone metabolomes of high-strength males had upregulated nucleotide metabolism, whereas high-strength females had upregulated terpenoid backbone biosynthesis and low-strength males had an upregulation of the PPP (Table 1).
Both high-strength male and female mice had upregulated tryptophan metabolism. Tryptophan is a precursor to serotonin, melatonin, and kynurenine. (83)(84)(85) In humans, downstream metabolites of tryptophan metabolism, such as kynurenic acid, have been shown to influence bone remodeling by inhibiting glutamate receptors. (70,84,86) Other mechanisms of tryptophan prominently affecting bone remodeling is through stimulating proliferation and differentiation of osteoblasts and bone marrow mesenchymal stem cells. (87) In both humans and mice, tryptophan has been associated with osteoclast activity and was positively associated with CTX1 levels. (66) In our study, there was a moderate correlation between bone strength and CTX1/P1NP for males. The increased bone strength in females, despite an increase in CTX1/P1NP, is likely influenced by the higher cortical thickness of female femurs (Table S2). Overall, the detection of upregulated tryptophan metabolism for high-strength females and males is consistent with the increased global bone resorption observed for the same mice.
High-strength males had upregulated nucleotide metabolism (ie, purine and pyrimidine metabolism) compared to high-strength females and low-strength males. In low-strength males, the PPP was upregulated. Both purine and pyrimidine metabolism are derived from the PPP. Purine and pyrimidine metabolism are important pathways for DNA and amino acid synthesis, (88) which are consistent with the relatively increased bone formation relative to resorption (eg, lower CTX1/P1NP) for low-strength males. This finding may represent a higher cellular demand for nucleotides. However, additional studies are needed to determine the mechanistic details and relevance of nucleotide metabolism to bone strength.
Overall, the metabolic differences between high-strength and low-strength bone were different than those found between the broader groups of all female and male mice. For example, when comparing all males and females, lipid metabolism was upregulated for females and amino acid metabolism was upregulated in males. By contrast, when comparing high-strength males and females, terpenoid backbone biosynthesis was upregulated in high-strength females and purine and pyrimidine were upregulated in high-strength males. Our data suggest that metabolic differences between male and female strength groups likely reflect sex differences in bone turnover, because females have higher CTX1/P1NP (Fig. S2).

Limitations
There are important limitations to this study. First, the femur was used for mechanical testing while humeri were used for metabolomic assessments. Therefore, the metabolism and material and mechanical properties of other bones may differ. Second, marrow-flushing can leave trace amounts of marrow, although from our histological assessments we estimate that marrow remaining was very minimal (Fig. S1). Third, although female and male C57Bl/6J mice were for experiments involving bone quality characterization, only female C57Bl/6J mice were used to assess metabolic differences between whole bone, isolated bone, and bone marrow. Finally, humeri were not weighed, therefore, it is unknown if all bones had the same amount of cellular content.

Summary
Bone quality is sexually dimorphic for C57Bl/6 mice. (1,2,(4)(5)(6)(7)9,10,(12)(13)(14)16) Our data demonstrate that the cortical bone metabolome is distinct from marrow and is also sexually dimorphic. We found that males and females rely on different metabolic pools and pathways to generate ATP and meet energy demands. For example, female mice predominantly utilized lipid metabolism to meet energy demands, whereas males utilized amino acid metabolism. Males and females with higher values of bone strength had upregulated tryptophan metabolism. Because the stronger groups (eg, high-strength males versus low-strength males; high-strength females versus high-strength or lowstrength males) had higher CTX1/P1NP, we estimate that the metabolomic signature of bone strength in our study least partially reflects differences in bone turnover. The sex differences evident in the cortical bone metabolome may help to connect sex differences in bone cell health and behavior with tissue-level differences in bone quality.

Acknowledgments
Funding was provided by the National Science Foundation (CMMI 1554708, CMMI 2120239) and the National Institutes of Health (NIAMS R01AR073964, NIGMS P20GM103474, NIH R03AG068680). We thank the Montana State University Mass Spectrometry Facility and Dr. Katie Steward for assisting in LC-MS analysis. Funding for the Proteomics, Metabolomics and Mass Spectrometry Facility used in this publication was made possible in part by the MJ Murdock Charitable Trust, the National Institute of General Medical Sciences of the National Institutes of Health under Award Numbers P20GM103474 and S10OD28650, and the MSU Office of Research, Economic Development and Graduate Education. Additionally, assistance from Maria Jerome at the Montana State University Histology Core Facility, and Dr. Heidi Smith and Dr. Markus Dieser at the Center for Biofilm Engineering at Montana State University in histological preparation, imaging, and analysis is gratefully acknowledged. We acknowledge the Center for Advanced Orthopaedic Research for μCT analyses. Finally, we thank Maya Moody, Leah Davidson, Kenna Brown, and Priyanka Brahmachary for assisting with tissue harvests.
Authors' roles: HDW performed dissections, extracted metabolites, analyzed data, and drafted the manuscript. GV performed dissections, bone flexural testing, and analyzed data. STW supplied study mice. BB assisted in analyzing data. SAM provided serum biomarker analysis materials and assisted with analysis. CMH designed experiments and analyzed data. RKJ designed experiments and analyzed data. All authors have read and revised the manuscript.