The guidelines for reporting animal research were followed (ARRIVE 2.0). The protocol (APAFIS#4197–2,015,091,514,435,707) was approved by the National Committee on Ethics in Animal Experimentation (CEEA 075) and legal requirements in France for the care and use of animals were followed. Thirty-three Sprague‒Dawley rats (Rattus norvegicus) aged 6 months and with homogeneous weights at baseline (Janvier Lab, Laval, France) [8] were studied. The rats were housed at the animal facility at the University of Lille (DHURE). The rats (n = 3 per cage) were housed in type 4 cages filled with Lignocel™ bedding and provided with horizontal tubes for climbing under controlled conditions at 22 ± 2 °C on a 12-h light/12-h dark cycle. Each animal was monitored daily and observed to determine any signs of poor adaptation to its environment. Assignments and surgeries were performed at 6 months by the provider to constitute a sham surgery group (SHAM, n = 11) and an ovariectomy group (OVX, n = 22). The assignment of the rats to groups was random. All experimenters were aware of the group allocation during performance of the experiment, result evaluation, and data analysis. The rats were sacrificed by exsanguination under anesthesia at 15 months of age (9 months post-OVX).

We previously performed and published our method for the collection and preparation of the tibia samples [8]. The right tibias were subjected to analysis of the bone microarchitecture using microcomputed tomography (µCT). The tibias were harvested, fixed for 48 h in 10% neutral buffered formalin (NBF), and then stored in phosphate-buffered saline (PBS) for a first µCT acquisition. Subsequently, they were prepared for analysis of BMAT via decalcification, osmium tetroxide staining, and storage in PBS, for a second µCT acquisition. The left tibias were used for the analysis of mineral parameters through quantitative backscattered electron imaging (qBEI) and for the analysis of bone composition through Raman microspectroscopy. The left tibias were fixed in 70% ethanol for 48 h before being embedded in polymethylmethacrylate (PMMA) resin. They were then sectioned and polished to a thickness of 100 µm.

The data and analysis protocol for bone microarchitecture and BMAT content investigation were previously published by our laboratory team [8]. The initial acquisitions were performed using a Skyscan 1172 µCT device (Bruker MicroCT, Kontich, Belgium) with the following parameters: isotropic voxels of 10 µm3, 80 kVp, 100 µA, an Al−Cu filter, an integration time of 2400 ms, and a rotation step of 0.5° over 180°. The data acquisition, reconstruction, analysis, and three-dimensional visualization were conducted using the following software: Nrecon™, Dataviewer™, CTAn™, and CTVox™ (Bruker µCT, Kontich, Belgium). Each bone sample was assessed before and after decalcification and osmium tetroxide staining. Osmium tetroxide binds to lipids stored in the cytoplasm of bone marrow adipocytes (BMAds), revealing BMAT in µCT. The region of interest is a cross-sectional slice of 2 mm thickness, located 1.5 mm below the growth plate. This protocol was carried out on the 33 tibias in this study. Bone density was represented by the ratio of bone volume to total volume (BV/TV, expressed as a percentage). BMAT was represented by the ratio of adipose volume to marrow volume (AdV/MaV, expressed as a percentage). The bone marrow was divided into 20 µm compartments relative to the surface of the trabecular bone (D1: 0–20 µm, D2: 20–40 µm, D3: 40–60 µm) to study the spatial distribution of BMAT.

The PMMA-embedded samples were analyzed using an environmental scanning electron microscope (Quanta 200™, FEI, Hillsboro, OR, USA) coupled with an energy-dispersive X-ray spectrometry detector (QuanTax™, Bruker). The parameters used included an accelerating voltage of 20 kV and a working distance of 10 mm. The bone mineral density distribution (BMDD) was evaluated using the method described by Roschger [15] and adapted by Olejnik [16]. The recorded images (Fig. 1a) are backscattered electron (BSE) images. The BSE images are displayed in grayscale. The BSE grayscale was calibrated using the “atomic number (Z) contrast" of reference materials, using PMMA resin (Z = 6), aluminum (Al, Z = 13), magnesium fluoride (MgF2, Zmean = 10), and tricalcium phosphate (β-TCP, Zmean = 14.4). The experimental gray-levels of PMMA and β-TCP were taken as 0 and 38.7% of weight of calcium, respectively. The BE gray-level was converted into weight concentration calcium. The intensity of the grayscale levels is dependent on the amount of calcium. The grayscale curve was extracted from the BSE images (Fig. 1b—upper x-axis). This curve corresponded to the BMDD. Through instrument calibration, the grayscale curve was converted into the mass percentage of calcium (Fig. 1b—bottom x-axis). The BMDD depicts the percentage of bone surface as a function of the percentage of calcium. Four parameters were calculated from the BMDD: (a) the full-width at half maximum (CaWidth), expressed as the mass percentage of calcium; (b) the most frequently occurring calcium concentration (CaPeak), expressed as the weight percentage of calcium; (c) the bone surface frequency of Capeak (FPeak), expressed as a percentage; and d) the average calcium concentration (CaMean), expressed as the mass percentage of calcium. Calculations were performed using MATLAB software (R2023a; MathWorks, Natick, MA, USA).

Grayscale BSE image of the trabecular and cortical bone in a left tibia from the SHAM group (a). The grayscale curve ranging from 85 to 255 (upper x-axis), with grayscale levels converted to mass percentages of calcium ranging from 0 to 38.7% (lower x-axis) (b). This curve is designated the BMDD and is characterized by the parameters CaWidth, CaPeak, FPeak, and CaMean. Only trabecular bones were chosen for the calculation of the BMDD

BSE images enable the localization of bone surface areas based on their calcium content. The images were segmented into three natures of packets: PMMA resin (black), newly formed bone (new bone), and mature bone (old bone). New and old bone packets were differentiated according to the calcium content: new bone was represented by a calcium content ranging from 0 to 90% of the maximum (primary mineralization), while old bone was represented by a calcium content ranging from 90 to 100% (secondary mineralization) [17]. Standardizing the areas of new bone and old bone based on calcium content allows for the identification of relevant analysis regions for Raman microspectroscopy measurements. Only BSE images containing sufficient trabecular bone were used for this analysis. Cortical bone was not considered in this study.

Raman analyses were conducted using a LabRAM HR800 microspectrometer (HORIBA, Jobin Yvon, Villeneuve d’Ascq, France) equipped with a diode laser (785 nm, 100 mW), a CCD detector (1024 by 256 pixels), and a × 100 objective (NA = 0.80; Olympus, France). The spectral acquisition window was set between 300 and 1800 cm–1. The lateral resolution was 1 µm. A scrambler was used to minimize polarization effects. The acquisition time consisted of two acquisitions of 30 s each. Four bone areas were identified from the BSE images (see Fig. 2): new bone–trabecular surface (NB-Tb.S), new bone–trabecular core (NB-Tb.C), old bone–trabecular surface (OB-Tb.S), and old bone–trabecular core (OB-Tb.C). For each sample and each bone area, ten Raman spectra were acquired. All Raman spectra were processed using LabSpec 6 software (HORIBA, Jobin Yvon, France). The physicochemical parameters were calculated using MATLAB software (R2023a; MathWorks, USA) as follows.

The grayscale BSE image was colored into black (grayscale 0–99), pink (grayscale 100–193, accounting for 90% of mineralization), and purple (grayscale 194–255) to locate new bone and old bone prior to the area selection on the Raman microspectrometer (a). A magnified view of a trabecula (b). Simplified image of the four categories, showing position (surface or core) and mineralization (new bone or old bone) (c). * new bone–trabecular surface; † new bone–trabecular core; ‡ old bone, trabecular surface; ¶ old bone, trabecular core

The mineral-to-matrix ratio (MMR) describes the amount of mineral relative to the amount of organic matrix in bone [13]. It is the ratio of the area under the ν1PO4 peak (range 900–990 cm–1) to that of the δ(CH2) peak of collagen (range 1434–1490 cm–1). Crystallinity (CRYST) refers to the size and perfection of mineral crystals, which gradually increase with the formation of apatite [13]. This peak is the inverse of the full-width at half maximum of the ν1PO4 peak. The relative content of carbonate B (CARB) measures the amount of type B carbonates in the bone mineral [13]. It is the ratio of the area under the CO32− type B peak (range 1052–1092 cm–1) to that of the ν1PO4 peak. The hydroxyproline/proline ratio (HPP) provides information about the posttranslational modifications of collagen [18]. It is calculated by the ratio of the intensity of the proline peak (range 828–898 cm–1) to that of the hydroxyproline peak (range 828–898 cm–1). Collagen maturity (COLL) corresponds to collagen crosslinking modifications [13, 19]. It is calculated by the ratio of the intensity between two successive amide I peaks (1660 and 1690 cm–1). The relative content of glycosaminoglycans (GAGs) is indicative of noncollagenous organic constituents [13]. Glycosaminoglycans play a role in assembling the organic content, modulating mineralization and remodeling, and preserving the nonmineralized organic matrix of osteocytes and canaliculi. It is calculated by the ratio of the area under the glycosaminoglycan peak (range 1365–1390 cm–1) to that of the amide III peak (range 1243–1269 cm–1). These six parameters were measured at four different bone locations: NB-Tb.S, OB-Tb.C, NB-Tb.C, and OB-Tb.S (Fig. 2).

Qualitative variables are described in terms of frequencies and percentages. Quantitative variables are described in terms of the mean and standard deviation. The normality of distributions was assessed visually and using the Shapiro‒Wilk test.

Comparisons between two elements were conducted using the Student’s t test for normally distributed data and the Mann‒Whitney test for nonnormally distributed data. Multiple comparisons were performed using one-way ANOVA for normally distributed data and the Kruskal‒Wallis test for nonnormally distributed data. When a significant difference was found, a post hoc Dunn’s test was applied to characterize the difference. The relationship between the parameters and adiposity was assessed using the Pearson correlation coefficient and its associated test. Each correlation was assessed for the whole sample and for separate groups. For parameters significantly associated with adiposity at a threshold of 0.10, we investigated, using the same method, the relationship between the parameter and adiposity in each of the three compartments (D1, D2, and D3). Statistical comparison analyses were conducted using GraphPad Prism v7.0 software (GraphPad Software, San Diego, CA, USA). Statistical correlation analyses were performed using SAS software (SAS Institute version 9.4). The significance level was set at 5%.