Experimental Design

Sample size was calculated a priori to detect a minimally significant effect of swimming on femur volumetric BMD (vBMD) between swimming and control experimental animals groups based on 1.18 effect size, considering α = 0.05 and 80% statistical power [22]. Twenty male Wistar Han rats (eight weeks; 323 ± 18 g; Charles River laboratories; France, Les Oncins) were housed individually in 800 cm2 floor area eurostandard type III H cages (Tecniplast, Buguggiate, Italy) in a 24 ºC temperature and 60% humidity-controlled animal facility with a 12/12 h inverted dark–light cycle. Ad libitum standard diet (Diete Standard 4RF21; Mucedola, S.r.l., Lombardia, Italy) and water were provided. After two weeks of quarantine, animals were randomly assigned to a control (CON, n = 10) or swimming (SW, n = 10) group through the “randombetween” function in Microsoft Excel.

SW group underwent a swimming exercise protocol for eight consecutive months. CON group was allowed to perform voluntary physical activity as in their natural environment by being housed in cages with an activity wheel (Tecniplast, Buguggiate, Italy) [23]. All animals were daily inspected for signs of stress, injury or illness that could dictate premature cessation of the protocol in accordance with the pre-defined humane endpoints. Food consumption and weight were recorded weekly. The protocol was approved by the local ethics committee (CEFADE 06 2021). All the experimental procedures and animals housing were in accordance with the Guide for the Care and Use of Laboratory Animals from the Institute for Laboratory Animal Research (ILAR, NIH Pub. No. 85–23, Revised 2011).

Swimming Protocol

Swimming was maintained during eight months and followed an established protocol [24]. Each animal swam inside a 20-cm diameter cylinder to prevent interference between animals with water at 30–32 ºC. First week consisted on acclimatization, where animals were gradually introduced to the water environment. Afterwards, duration increased progressively 10 min each 2 days up to 2h/d of swimming, 5 d/week, during eight months (supplementary Fig. 1). A 3% body weight load was attached to each animal tail to prevent fluctuation, as recommended elsewhere [24]. Whenever the animal submerged > 5s, it was removed from the water and allowed to recover for 1 min. One of the animals was excluded from the experiment due to refusal to collaborate with the protocol. After each swimming session, animals were dried and returned to their cages.

Animals Sacrifice and Data Collection

Weight, food intake and running wheel activity were recorded throughout the experiment. After eight months, all animals were sacrificed by an overdose of ketamine (Nimatek, Dechra; 100 mg/ml) and xylanize (Rompun, Bayer; 20 mg/ml). Immediately after sacrifice, 5 mL of blood were collected from the inferior vena cava to EDTA-coated tubes (Fisher Scientific, 15,452,520), centrifuged at 2100 G, 4 ºC for 10 min and the plasma stored at −80 °C. Gastrocnemius and soleus muscles, lumbar vertebrae (L4 and L5) and both femurs and tibias were collected. Right femur, left tibia and L4 vertebra were cleared of surrounding soft tissue and wrapped in saline-soaked gauze pads to prevent dehydration, sealed in plastic bags and stored at −80 ºC for µ-CT analysis. Left femur and L5 lumbar vertebrae were immediately fixed in 4.0% w/v buffered formaldehyde (PanReact AppliChem, 252,931.1315) for histological analysis.

Measurement of Bone Growth, Mass, Geometry and Microarchitecture

Left femur (from the greater trochanter to the lateral condyle) and right tibia (from the intercondylar eminence to the medial malleus) lengths were assessed with a 0.01-mm resolution digital calliper (Powerfix, Germany) to determine bone growth. Lumbar vertebrae growth, bone mass [cortical (Ct.vBMD) and trabecular volumetric bone mineral density (Tb.vBMD)], geometry [cortical bone volume (Ct.Volume), cortical thickness (Ct.Th), trabecular bone volume (Tb.Volume)] and trabecular microarchitecture [bone volume per tissue volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), trabecular separation (Tb.Sp) and trabeculae connectivity density (Conn.D)] of the L4 vertebra, right femur and left tibia were analysed by µ-CT.

Right femur and L4 vertebra were scanned with the SkyScan 1276 µ-CT (Bruker, Kontich, Belgium), with the following parameters: 85-kV voltage, 200-μA current [1-mm aluminium filter], and 7.0-µm pixel size. Images were reconstructed using the NRecon software (Version 1.7.5.2, SkyScan, Bruker, Kontich, Belgium). Images were transversally oriented and saved in the transaxial plane using the DataViewer software (Version 1.5.6.3, SkyScan, Bruker, Kontich, Belgium). CTVox (Version 3.3.0 r1412, Bruker, Kontich, Belgium) was used to visualize the 3D structures in cross-section. Two- and 3D models and the quantification of bone morphometry variables were obtained with CTAn (Version 1.20.3, SkyScan, Bruker, Kontich, Belgium). The left femur midshaft diaphysis was selected for cortical bone analysis. A region of interest (ROI) was defined in the femur midshaft and 100 slices were analysed (50 slices above and 50 below midshaft), representing a ROI of 0.7 mm length. Femur trabecular bone was analysed in the distal metaphysis. The selected ROI started 100 slices proximally to the growth plate (0.7 mm) and 400 slices were analysed, representing a 2.8-mm ROI. A threshold of 80–255 grayscale index was used to identify cortical bone at the femur midshaft and distal metaphysis, whilst a threshold between 70 and 255 grayscale indices was used for femur distal metaphysis trabecular bone. The L4 vertebra trabecular bone was assessed at the centre of the vertebral body in a ROI of 400 slices representing 2.8 mm length that excluded the cranial and caudal endplate regions. In the L4 vertebra, 2-D µ-CT images of the centre of the ROI in the axial plane were used to assess L4 vertebral geometry, namely total antero-posterior and vertebral body antero-posterior lengths (Supplementary Fig. 2), using ImageJ (version 1.53K; National Institutes of Health, USA). Trabecular bone at the L4 vertebral body was selected based on a threshold between 80 and 255 grayscale index.

Left tibia was scanned with a µ-CT40 (Scanco Medical) using the following parameters: 55-kVp voltage, 145-μA current, medium resolution and 12-μm isotropic voxel size. Scanco Eval v.6.0 was used for measuring bone volume. Tibias were scanned from the proximal end to the distal tibiofibular joint and the distal 10 slices were used for measurement of cortical indices, with a threshold of 200 mg/cm3. Trabecular bone was evaluated in 100 slices in the metaphyseal region starting 100 slices distal to the growth plate with a threshold of 220 mg/cm3. Tb.vBMD (mg/cm3), Tb.Volume (mm3), BV/TV (%), Tb.Th (μm), Tb.N (mm−1), Tb.Sp (mm) and Conn.D (mm−3) were obtained from the L4 vertebra, femur distal metaphysis and tibia proximal metaphysis. Ct.vBMD (mg/cm3), Ct.Volume (mm3) and Ct.Th (mm) were also assessed at the femur and tibia midshaft diaphysis. All bone morphometric measurements and nomenclature are in accordance with recommendations of the American Society for Bone and Mineral Research. Representative µCT images depicting trabecular bone ROI assessment are provided in supplementary material (Supplementary Fig. 3).

Bone and Skeletal Muscle Histological Analyses

Histological analysis of the left gastrocnemius and soleus, left femur and L5 vertebra were performed to determine muscle fibre cross-sectional area and osteocyte and empty lacunae density, respectively. After 48 h of fixation on 4% formaldehyde w/v, the left femur and L5 vertebra were decalcified in 10% w/v ethylenediaminetetraacetic acid (EDTA; Fisher Scientific, D/0450/50) in phosphate-buffered saline (PBS; Fisher Scientific, BP2944100) at room temperature for ~ 30 days under constant agitation in a platform rocker (Stuart Scientific, STR6; United Kingdom). Complete decalcification was assessed by the ammonium oxalate test. Then, tissues were routinely processed for histological analysis by dehydration through graded ethanol solutions (Fisher Chemical, E/0650DF/C17), clearing through xylene (Thermo Scientific, 6615) and embedding in paraffin (Thermo Scientific, 8336). Five-µm-thick sections were cut with a rotary microtome (Leica RM2125 RT, Leica Microsystems; Nussloch, Deutschland), collected with silane-coated slides, stained with haematoxylin (Thermo Scientific, 72,704) and eosin (Thermo Scientific, 71,204), and visualized in a light microscope coupled with a digital camera (Axio Imager A1, Carl Zeiss; Germany). About 35 representative images of the gastrocnemius and soleus, distributed through the entire tissue section, were analysed with ImageJ (version 1.53K; National Institutes of Health, USA) to measure individual fibre cross-sectional area (µm2). In each animal, fibre CSA was determined as the average of 350 individual muscle fibres analysed in the soleus and gastrocnemius. The multifidus muscle fibre cross-sectional area surrounding L5 vertebrae was also measured, as the average area of 150 fibres.

Osteocyte density, determined as the number of osteocytes identified per cortical bone area (N.Ot/Ct.Ar; N/mm2), empty lacunae density (N.Lc/Ct.Ar; N/mm2), and osteocyte lacunae occupancy rate, determined as the filled lacunae with osteocyte per total lacunae number [(filled lacunae with osteocyte/ total lacunae number)*100] [25], were assessed in histological sections of the femur distal diaphysis and L5 vertebrae (transverse and frontal plane, respectively). Eight images from the femur and six images from the L5 vertebrae were analysed per animal to determine osteocyte and empty lacunae density. For L5 and femur distal diaphysis, the intra-observer coefficient of variation was 3 and 5% for N.Ot, respectively, and 10% for N.Lc (Supplementary Fig. 4 highlights the regions selected for this assessment).

Statistical Analysis

Statistical analyses were conducted in SPSS statistical software (version 29.0) and data were presented as mean ± standard deviation (SD) with a significance level set at p < 0.05. Outliers were identified and removed from the analysis when significantly affected the results, as was the case of femur Conn.D. For this purpose, outliers were identified as data points that are (i) deviated ≥ 3 standard deviations (SD) from the mean, (ii) clearly isolated from the rest of the dataset, and (iii) that after removal did not lead to the appearance of other outliers. Variables normality was verified by the Shapiro–Wilk test and adequate transformation was performed to obtain normality whenever required. If the transformation was ineffective, statistical analysis was conducted with bootstrapping using 1000 bootstrap samples, specifically in tibia midshaft Ct.vBMD. Independent samples t tests were performed to compare all variables between CON and SW groups and Cohen’s d was reported as a measure of effect size when p < 0.05.