Male C57Bl6/J mice (Jackson Laboratories) were given a normal chow diet (ND) (13.5% calories from fat; LabDiet 5LOD) or 60% high fat diet (HFD) (Research Diets D12492) at 6 weeks of age for a duration of 12, 16, or 20 weeks. A third group of mice was put on HFD for 12 weeks and then on ND for 8 weeks [weight loss (WL) group]. Animals were housed in a specific pathogen-free facility with a 12-h light/12-h dark cycle at ~22°C and given free access to food and water. All animal use was in compliance with the Institute of Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and approved by the University Committee on Use and Care of Animals at the University of Michigan. The tibia was selected for our longitudinal analyses since it can be used to simultaneously monitor changes in rMAT (proximal tibia) and cMAT (distal tibia) within one sample (37). To compare the changes in bone within the tibia to those in the femur, as reported previously (12), we also analyzed the femurs in the 20-week groups.

Tibiae were fixed in formalin for 48-h and then placed in phosphate buffered saline (PBS). Specimens were embedded in 1% agarose and placed in a 19-mm diameter tube, and the length of the bone was scanned using a Micro Computed-Tomography (microCT) system (μCT100 Scanco Medical, Bassersdorf, Switzerland). Scan settings were: voxel size 12 μm, medium resolution, 70 kVp, 114 μA, 0.5 mm AL filter, and integration time 500 ms. Density measurements were calibrated to the manufacturer’s hydroxyapatite phantom. Analysis was performed using the manufacturer’s evaluation software.

Femurs were removed and frozen after wrapping in PBS-soaked gauze and then analyzed by microCT. Femora were scanned in water using cone beam computed tomography (explore Locus SP, GE Healthcare Pre-Clinical Imaging, London, ON, Canada). Scan parameters included a 0.5° increment angle, four frames averaged, an 80 kVp and 80 μA X-ray source with a 0.508 mm AI filter to reduce beam hardening artifacts, and a beam flattener around the specimen holder. All images were reconstructed and calibrated at an 18 μm isotropic voxel size to manufacturer-supplied phantom of air, water, and hydroxyapatite (38).

Following microCT scanning, femurs were loaded to failure in four-point bending using a servohydraulic testing machine (MTS 858 MiniBionix, Eden Prairie, MN, USA). All specimens were kept hydrated in lactated ringers solution-soaked gauze until mechanical testing. In the same mid-diaphyseal region analyzed by μCT, the femur was loaded in four-point bending with the posterior surface oriented under tension. The distance between the wide, upper supports was 6.26 mm, and the span between the narrow, lower supports was 2.085 mm. The vertical displacement rate of the four-point bending apparatus in the anterior–posterior direction was 0.5 mm/s. Force was recorded by a 50 lb load cell (Sensotec) and vertical displacement by an external linear variable differential transducer (LVDT, Lucas Schavitts, Hampton, VA, USA), both at 2000 Hz. A custom MATLAB script was used to analyze the raw force-displacement data and calculate all four-point bending parameters. Combining anterior–posterior bending moment of inertia data from μCT with mechanical stiffness from four point bending, the estimated elastic modulus was calculated using standard beam theory as previously described (38). The modulus of elasticity was derived based on previous methods with “L” set at 3.57 and “a” at 0.99 (39).

Tibia. Regions of interest (ROI) was located for both cortical and trabecular parameters. Analyses were performed with MicroCT software provided by Scanco Medical (Bassersdorf, Switzerland). A mid-diaphyseal cortical ROI was defined as ending at 70% of the distance between the growth plate and the tibia/fibula junction. A ROI spanning 360 μm (30-slices) proximal to this region was analyzed with standard plugins using a threshold of 280. The trabecular ROI was defined as starting 60 μm (5-slices) distal to the growth plate and ending after 600 μm total (50-slices). Trabecular analyses were performed with standard Scanco plugins with a threshold of 180.

Femur. ROI was located for both cortical and trabecular parameters. A diaphyseal cortical ROI spanning 18% of total femur length was located midway between the distal growth plate and third trochanter. Cortical bone was isolated with a fixed threshold of 2000 Hounsfield Units for all experimental groups. Parameters including cortical thickness, endosteal and periosteal perimeter, cross sectional area, marrow area, total area, anterior–posterior bending moment of inertia, and tissue mineral density (TMD) were quantified with commercially available software (MicroView v2.2 Advanced Bone Analysis Application, GE Healthcare Pre-Clinical Imaging, London, ON, Canada). A trabecular ROI 10% of total femur length was located immediately proximal to the distal femoral growth plate and defined along the inner cortical surface with a splining algorithm. Trabecular metaphyseal bone was isolated with a fixed threshold of 1200 Hounsfield Units.

Marrow adipose tissue volume within the tibia was assessed as described previously (37, 40). After the initial microCT scan, bones were decalcified in 14% EDTA solution, pH 7.4 for 14 days at 4°C. Decalcified bones were stained with 1% osmium tetroxide solution in Sorensen’s phosphate buffer pH 7.4 at room temperature for 48 h. Osmium-stained bones were re-scanned using the Scanco microCT settings described above. For analysis of MAT within the tibia, four regions were defined as follows: (1) the proximal epiphysis between the proximal end of the tibia and the growth plate, (2) the proximal metaphysis, beginning 60 μm (5-slices) distal to the growth plate and ending after 600 μm total (50-slices), (3) the growth plate to the tibia/fibula junction (GP to T/F J), and the distal tibia between the tibia/fibula junction and the distal end of the bone. MAT volume analyses were performed with standard Scanco plugins with a threshold of 500.

Statistical comparisons were performed in GraphPad Prism (GraphPad Software, Inc., La Jolla, CA, USA). The following planned comparisons were performed on the graphs in Figures 1, 2, 3 and 5: 12-week ND vs. HFD (two-tailed t-test); 16-week ND vs. HFD (two-tailed t-test); 20-week ND vs. HFD vs. WL (1-way ANOVA); 12-, 16-, 20-week ND (1-way ANOVA); 12-, 16-, 20-week HFD (1-way ANOVA); 12-week HFD vs. 20-week WL (two-tailed t-test). These results were corrected for multiple comparisons using the Benjamini-Hochberg procedure as described previously (41). For comparisons in Figures 4, 6 and 7, a one-way ANOVA with Tukey’s correction was applied. In Figure 8, linear regression was applied to test the significance of the correlations. Raw data for the skeletal morphology, marrow fat quantification, and biomechanical testing is available in Data Sets 1–3 in Supplementary Material.

Mice were fed normal chow diet (ND) or 60% high-fat diet (HFD), starting at 6 weeks of age, for 12, 16, or 20 weeks. A separate group of mice received 12 weeks of HFD, followed by 8 weeks of ND to mimic weight loss (Figure 1A). Comparison of the 12-week HFD group to the WL group was used to determine if weight loss reversed changes that were already present at 12 weeks, or, rather, prevented further deterioration induced by continued HFD.

Increases in body mass relative to ND control were apparent after 12-weeks of HFD (Figure 1B). Increases in body mass relative to ND control persisted at the 16- and 20-week time points (Figure 1B). Over time, body mass continued to increase from 12 to 16 weeks of HFD but stabilized between 16 and 20 weeks (Figure 1B). Body mass returned to normal after weight loss (Figure 1B). Weight gain was due, at least in part, to increases in liver, inguinal and gonadal white adipose tissue (WAT) mass after 20-weeks of HFD (Figure 1C). With weight loss, liver mass returned to normal; however, WAT mass was only partially rescued (Figure 1C).

Within the tibia, MAT expansion became significant, relative to ND control, after 16 weeks of HFD (Figures 2A,B). With HFD, changes in the proximal tibial epiphysis mimicked what was observed between the growth plate and tibia/fibula junction, with a 5.5- and 4.3-fold increase at 16 weeks, relative to 12 weeks, respectively (Figure 2B). The distal tibia was similar, though there was only a 2.1-fold increase between 12 and 16 weeks of HFD, likely owing to the higher baseline MAT in this region (Figure 2B). No additional MAT accrual in any region of the tibia was observed between 16 and 20 weeks of HFD (Figure 2B).

In the weight loss group, the MAT in the regions of the proximal epiphysis and GP to T/F J was indistinguishable from that of the 20-week ND group (Figure 2B). However, it was also similar in magnitude to the 12-week HFD group, suggesting that switching to ND was sufficient to block HFD-induced MAT expansion – rather than reversing MAT accrual that had already occurred.

In the distal tibia, age-associated increases in MAT were noted from 12- to 16-weeks of age in the ND group. Prior to correction for multiple comparisons, MAT within the distal tibia in the WL group was higher than chow (p = 0.019) but less than HFD (p = 0.050). This suggests that weight loss blunted, but did not entirely prevent, HFD-induced MAT expansion in the distal tibia at 20-weeks (Figure 2B). Raw data for these comparisons is available as Data Set 1 in Supplementary Material.

Consistent with previous reports (42, 43), we observed an age-related decrease in trabecular bone volume fraction (BVF) and trabecular number, with a corresponding increase in spacing, in the proximal tibial metaphysis of ND control mice (Figures 3A,C,E,G).

Relative to ND controls, mice fed a HFD had a significant decrease in trabecular BVF, bone mineral content (BMC), and number after 12-weeks of diet (Figures 3C,D,E). Structure model index and trabecular spacing were reciprocally increased (Figures 3F,G). Trabecular thickness remained unchanged (Data Set 1 in Supplementary Material). Loss of trabecular BVF and BMC with HFD, relative to control ND, persisted at the 16- and 20-week time points (Figures 3C–G). Weight loss partially rescued decreases in trabecular BVF, BMC, and number (Figures 3C–E) and increases in spacing (Figure 3G).

Unlike loss of trabecular bone at 12-weeks, MAT volume was not significantly increased relative to ND control until 16- and 20-weeks of HFD (Figure 3B). Over time, MAT increased by 3.5-fold from 12- to 16-weeks of age in the HFD group (Figure 3B). No further increases were present from 16- to 20-weeks of age. Weight loss completely prevented HFD-induced MAT accumulation from 12- to 20-weeks (Figure 3B).

Relative to ND control, HFD caused MAT expansion within the femur (Figure 4A). MAT expansion was absent in the WL group (Figure 4A). The absolute amount of MAT in the distal femoral metaphysis was ~90% less than the proximal tibia (p < 0.0001, t-test) (Figures 3B and 4A). The magnitude of the loss of BVF with HFD at 20-weeks was comparable between femur and tibia (50 vs. 45%) (Figure 4B; Data Sets 1 and 2 in Supplementary Material). There was also a significant, HFD-induced decrease in trabecular number and increase in trabecular spacing (Figures 4D,F). Trabecular thickness and tissue mineral density remained unchanged (Figures 4C,E). With weight loss, there were no statistically significant differences in trabecular morphology, relative to HFD, in the femur (Figures 4B–F). There was a non-significant trend toward a decrease in trabecular spacing in the WL group relative to HFD (Figure 4F).

Within the tibia, there were no statistically significant changes in mid-diaphyseal cortical morphology after 12, 16, or 20 weeks of HFD or after weight loss (Figures 5A–F). However, slight differences may have been missed after statistical correction for multiple comparisons. For example, with standard one-way ANOVA at the 20-week time point only, there was a slight decrease in cortical thickness in the 20-week HFD group relative to ND control (p = 0.045) (Figure 5D). Raw data are available in Data Set 1 in Supplementary Material.

In the femur, 20-week HFD caused a significant decrease in cortical thickness relative to ND control (Figure 6D). Cortical tissue mineral density was also decreased with HFD (Figure 6F). With weight loss, cortical thickness and total mineral density improved relative to ND (Figures 6D,F). No differences in total area, marrow area, cortical area, or Iyy were noted (Figures 6A–F). Raw data are available in Data Set 2 in Supplementary Material.

In the control ND group, pooled over all ages (12, 16, and 20 weeks of diet), there was a significant inverse correlation between MAT volume in the proximal metaphysis and measures of trabecular morphology including BVF, trabecular thickness, and BMC – but not with trabecular number (Figures 7A–D). By contrast, in the HFD group, MAT volume was negatively correlated with trabecular BVF, thickness, and BMC in addition to trabecular number (Figures 7A–D). Cortical thickness was significantly negatively correlated with GP to T/F J MAT volume in the HFD group only (Figure 7E). Cortical TMD did not correlate with MAT volume in either group (Figure 7F).

Four-point bending was performed to assess the biomechanical integrity of HFD and WL femurs. The femurs from the 20-week HFD and WL groups broke under a reduced maximum load relative to ND, trending toward less total work to induce fracture (Figures 8A,B). This indicates that despite recovery of cortical thickness and mineral density with WL (Figure 6), the bone quality remains impaired, leading to decreased fracture resistance and poor post-yield behavior. The yield load and post-yield work trended toward a decrease relative to ND in the HFD and WL groups, respectively (p < 0.1) (Figures 8D,E). The stiffness, post-yield displacement, and modulus of elasticity (39) were not significantly different between groups (Figures 8C,F) (Data Set 3 in Supplementary Material).