Increased mechanical loading through controlled swimming exercise induces bone formation and mineralization in adult zebrafish

Exercise promotes gain in bone mass through adaptive responses of the vertebrate skeleton. This mechanism counteracts age- and disease-related skeletal degradation, but remains to be fully understood. In life sciences, zebrafish emerged as a vertebrate model that can provide new insights into the complex mechanisms governing bone quality. To test the hypothesis that musculoskeletal exercise induces bone adaptation in adult zebrafish and to characterize bone reorganization, animals were subjected to increased physical exercise for four weeks in a swim tunnel experiment. Cellular, structural and compositional changes of loaded vertebrae were quantified using integrated high-resolution analyses. Exercise triggered rapid bone adaptation with substantial increases in bone-forming osteoblasts, bone volume and mineralization. Clearly, modeling processes in zebrafish bone resemble processes in human bone. This study highlights how exercise experiments in adult zebrafish foster in-depth insight into aging-related bone diseases and can thus catalyze the search for appropriate prevention and new treatment options.

Bone microstructural changes. Micro-computed tomography revealed morphological differences between zebrafish vertebrae subjected to exercise and non-exercised controls, even though there were no significant macroscopic differences regarding standard length (Fig. 2a,b) and total body weight between the individuals from the two groups (Fig. 2c).
Bone modeling characteristics. Calcein double labeling allowed the detection of areas that showed increased bone formation. De novo bone formation occurred in vertebral body end plates (Fig. 4a). The investigated end plates showed more regions with overlapping, non-distinguishable double labels in the non-exercised group. In contrast, the exercised group displayed double labels with larger interlabel distances, indicative of increased bone formation (Fig. 4b). Increased bone formation was most notably observed in posterior caudal vertebrae. The quantification of dynamic histomorphometric parameters yielded a significantly higher mineralizing surface per bone surface (MS/BS, percent of bone surface that displays a calcein label and reflecting active mineralization) in the exercised group than in the control group (30.38 ± 3.86% vs. 23.63 ± 4.07%, p = 0.04) as depicted in Fig. 4c. Although there was no significant difference regarding mineral apposition rate (MAR, rate of new bone deposition) between the groups (Fig. 4d), the bone formation rate (BFR, amount of new bone formed in unit time per unit of bone surface) was 46% higher in the exercise group compared to the control group (25.88 ± 9.65 µm 3 /µm 2 /y vs. 11.93 ± 4.07 µm 3 /µm 2 /y, p = 0.038, Fig. 4e). Zebrafish belonging to the exercise group were placed inside the blue chamber during exercise units; white arrows indicate the direction of the water current. Magnified image on the right displays the exercise group during an exercise session. (b) Zebrafish belonging to the exercise group were trained for four weeks at step-wise increasing hours per day (4 h/day in week one, 5 h/day in week two and three, 6 h/day in week four) while control zebrafish did not experience any exercise. Both groups were labeled using fluorescent chromophore calcein submersions before and after the experiment. (c) Schematic depiction of a zebrafish vertebra. Lateral view (left) and transverse cross section (right).   . Bone microstructural changes due to musculoskeletal exercise. (a) Contact X-ray image displays the volume of interest (VOI) in the caudal spine of zebrafish. All parameters were extracted from caudal vertebrae. (b) Micro-CT images display the typical hourglass shape of the autocentrum that represents the core of the zebrafish vertebral body. The volume parameters extracted to quantify skeletal changes due to the swimming exercise are displayed on the sagittal cross-sections of the vertebrae: tissue volume (TV), bone volume (BV), and vertebral length (VL). The neural arch (NA) and the hemal arch (HA) were not included in the quantification of structural parameters via micro-CT. (c) Tissue volume was significantly higher in the exercise group. (d) Bone volume of the vertebral body was significantly greater in the exercise group. (e) Vertebral length increased substantially in the exercise group. The larger bone anatomy of the vertebral bodies of the zebrafish exposed to exercise indicates an increase in bone mass due to increased swimming activity. Osteocyte characteristics. 3D X-ray microscopy (3D XRM) was used to depict the zebrafish vertebrae in 3D and at high resolution and to quantify osteocyte characteristics at an adequate resolution (0.7 µm) (Supplemental Video 1). In both groups, the analysis showed that the ellipsoidal osteocyte lacunae had a distinct orientation within the zebrafish vertebrae (Fig. 6a). In the vertebral body end plates, the long axes of lacunae were orientated tangential to the circumference, while in the central region of the autocentrum the long axis of the lacunae was aligned with the long axis of the vertebral body. Vertebrae of the exercise and the control group were similar with regard to the lacunar sphericity (0.66 ± 0.03 vs. 0.64 ± 0.01, Fig. 6b) and the mean lacunar volume (83.6 ± 21.8 µm³ vs. 77.6 ± 10.3 µm³, Fig. 6c).
Bone mineral density distribution. The bone mineral density distribution (BMDD) analysis by quantitative backscattered electron imaging (qBEI) is a compositional analysis enabling the measurement of the degree of bone mineralization based on the average atomic number of the material 16,[38][39][40] . In addition to a higher bone mass in zebrafish in the exercised group, as evidenced by micro-CT and dynamic histomorphometry, the swimming exercise essentially resulted in a higher degree of bone mineralization (Fig. 7a). The histograms that show the calcium weight percentages in bone display a shift from lower mineralized bone in the control group to higher mineralized bone in the exercise group. This change is expressed by significantly higher mean calcium content (CaMean) in the exercise group compared to controls (26.74 ± 0.83 wt% vs. 24.82 ± 1.13 wt%, p = 0.03, Fig. 7b). While the amount of lower mineralized bone (CaLow) did not differ between the groups (Fig. 7c), the amount of highly mineralized bone (CaHigh) was larger in the exercise group (21.97 ± 11.29% vs. 6.96 ± 2.53%, p = 0.04, Fig. 7d).

Discussion
Utilizing high-resolution multi-scale analyses enabled the visualization and quantification of processes of bone adaptation in response to musculoskeletal exercise in zebrafish and allowed to assess their potential as an attractive animal model for bone research.
The data provides important insight into the impact of swimming exercise on zebrafish bone. In particular, it is shown that exercise triggers bone adaptation in response to loading. Interestingly, musculoskeletal exercise did not only induce a higher rate of bone formation, but also resulted in a higher degree of bone mineralization. These findings are comparable with results obtained from exercise experiments on larger fish species, which showed a higher mineral content in exercised rainbow trout (Oncorhynchus mykiss) and Atlantic salmon (Salmo salar) 41,42 . Increased swimming activity in zebrafish has previously been associated with muscle growth and muscle growth marker gene expression 43,44 as well as with improvement of swimming performance over a certain timespan 25 . In addition to confirming the initial hypothesis, the results of our study show that zebrafish vertebral bodies, similar to human vertebral bodies, have a comparable capability of adapting to increased mechanical load despite their anatomical and developmental differences.
In humans, bone formation and mineralization relies on the interaction between muscle forces and mechanosensitive osteocytes 13 . Osteocytes detect mechanical strain in the bone tissue, which is induced by muscle forces. Our experimental setup allowed the investigation of this mechanism in zebrafish by exposing the animals to musculoskeletal exercise that triggered new bone formation and increased mineralization as shown in detail in Fig. 8. It has been observed in humans that physical activity such as weight-lifting or practicing a wide range of sports has a beneficial influence on bone quality 24 , particularly on bone mineral density (BMD) 45,46 . The level of bone adaptation in humans obviously depends on the type of loading and appears to be limited to the regions that are exposed to the highest stresses 9,45,47 . Due to ethical considerations exercise studies in humans rely primarily on BMD measurements by dual-energy X-ray absorptiometry (DXA) and blood serum levels 48 . BMD measurements provide, however, only limited and not differentiated information about bone composition and bone structure in age-related deteriorations of bone quality [49][50][51] . The insufficient data on the influence of mechanical loading on human bone quality and fracture risk highlights the need for animal models suited for comprehensive exercise experiments.
In contrast to humans 14,15 , zebrafish display a reduced lacuno-canalicular network 52 . In tetrapods this network is believed to be essential for the osteocyte-driven bone remodeling. The results of this study indicate that despite the lower osteocyte connectivity observed in zebrafish, increased bone formation was achieved as a result of loading. When osteocyte processes are scarce or absent, Witten & Hall suggest that osteoblasts and bone lining cells could function as alternative receptors for mechanical load 26 . Indeed, various experiments present evidence that morphology, structure and patterns of gene expression change in the jaw bones of advanced teleost species that have no osteocytes in response to altered mechanical load 53,54 . The question of how load is specifically translated into a tissue response in advanced teleosts with anosteocytic bone remains unanswered. The question   Effects of exercise on bone quality in zebrafish. Schematic depiction of the sum of effects induced by musculoskeletal loading of zebrafish following the swim tunnel experiment. Zebrafish from the exercise group (bottom) were subjected to higher muscular forces that translated into a higher load on zebrafish bone. Higher forces acted as a stimulant for osteoblast activity, which led to an increase of bone formation and mineralization. As a result, exercised zebrafish displayed higher bone mass with a higher degree of bone mineralization when compared to the control group. In humans, increased physical activity 45,46 and higher bone mineralization is positively correlated with improved bone quality 16 . The results obtained from zebrafish exercise experiments provide therefore deeper insight into the complex mechanisms governing mechanical sensitivity of bone and the pathways that control bone formation and mineralization, thus affecting bone quality. is especially relevant in the context of early developmental stages as all teleost including zebrafish skeletons are initially devoid of osteocytes 55 . In this context, Fiaz and collagues suggested the possibility of the notochord as a mechanosensor 56 . Some teleost species with anosteocytic bone display cells on the bone surface that extend cytoplasmic cell processes into the matrix similar to mechanosensing osteocytes. Meunier describes this skeletal tissue as primary canaliculated bone in the Spangled emperor fish Lethrinus nebulosus as a representative of this type of anosteocytic bone 57 . In mammals, Vatsa et al. showed that osteoblastic and osteocytic cells in vitro are sensitive to mechanical load 58 . Likewise, mammalian odontoblasts are considered to have sensory capacity 59 . Zebrafish, which display a less developed lacuno-canalicular system, likely rely on similar complementary sensory pathways that do not involve osteocytes. The involvement of osteocytes regarding this aspect can be tested by comparing our results with adaptions to biomechanical load in the species medaka (i.e. advanced teleost with similar morphology but anosteocytic bone).
As hypothesized, bone formation in the exercise group occurred most prominently in the vertebral body end plates. Here, the higher number of osteoblasts and their larger size indicate greater osteoblastic activity 5 and offer an explanation for the observed increase in bone formation. Zebrafish from the control group displayed in the same region fewer osteoblasts, indicative of less bone formation 5 . The musculoskeletal exercise of zebrafish appears to have enhanced the forces acting on the vertebrae and induced bone formation. Regions that experienced the highest bone formation activity, in this case the end plates of the vertebral bodies, most likely indicate regions that experienced the highest stress concentration. In humans, a similar behavior can be observed as regions of bone that experience higher stress also display higher bone formation and higher osteoblast activity 10,45 .
Key aspects of bone modeling such as resorption lacunae and multinucleated osteoclasts can be observed in the zebrafish skull 33,55 . Different from teleost hemal and neural arches that require bone resorption of the endosteal bone surface for extension of the lumina, the vertebral body endplates grow and extend by periosteal bone apposition only; osteoclastic resorption is not required for normal development 60 . In adult zebrafish vertebrae, we were able to observe how the orientation of osteocyte lacunae changes in different regions of the vertebral bodies. In the central region, osteocyte lacunae are oriented parallel to the long axis of the vertebrae while in the region of the vertebral body end plates, osteocyte lacunae are oriented circumferentially within the vertebra. This specific arrangement of osteocyte lacunae found in the zebrafish vertebrae does not only indicate the orientation of the collagen 61 but also signifies the direction of the main forces experienced by the vertebrae 62 . Osteocyte lacunae represent also volumes with a low modulus inside a material and thus are expected to produce stress concentrations around them 62 . This fact explains the tendency of osteocyte lacunae to be oriented parallel to the forces acting on bone in order to reduce the stress concentrations surrounding osteocyte lacunae 62 . The orientation of the long axis of the osteocyte lacunae therefore likely represents the direction in which zebrafish vertebrae experience the highest stress. During early developmental stages, the stresses acting upon the vertebrae may explain the change in orientation of osteocyte lacunae as the vertebral diameter increases. Initially, the main forces acting upon the vertebrae should possess an axial orientation contributing therefore to the axial orientation of osteocyte lacunae. As the vertebrae grow in diameter, more complex stress arrangements must be accounted for, leading to a circumferential arrangement of osteocyte lacunae.
It is interesting to note, that new bone formation is exclusively induced in regions of higher stress concentration, while at the same time bone mineral density is increasing throughout the entire vertebral body. In humans, newly formed bone typically displays a lower mineral density than old bone. Following primary mineralization, secondary mineralization leads to further increases in bone mineral content, which correlates with tissue age 63 . The difference in bone mineral density distribution measured between groups indicates that the musculoskeletal exercise induced substantial increases in calcium content of old and newly formed bone in a relatively short period of time. Additionally, the higher degree of mineralization (i.e. calcium content) of the entire zebrafish vertebra is not restricted to the vertebral bone tissue but also affects the mineralized notochord sheath. This supports the notion that increases in mineral content can be achieved without the contribution of osteocytes as the notochord sheath is completely devoid of mechanosensing osteocytes. Bone mineralization in humans is clearly correlated with fracture resistance 50,64,65 , which implies that the higher bone mineralization observed in the zebrafish exercise group likely represents an improvement of the mechanical properties of bone.
In conclusion, the results of this study demonstrate that musculoskeletal exercise of zebrafish is a novel investigative approach that can promote new research in the fields of bone regeneration and age-related bone diseases. Utilizing zebrafish as model has a few limitations: Zebrafish have no osteons as it is known from human load-bearing cortical bone and zebrafish have a swim bladder, which changes the experienced mechanical loading during swimming in comparison to tetrapods. Further on, zebrafish live in environmental conditions that differ substantially from humans (i.e., aqueous ionic concentrations, temperature of the water (≈28 °C)).
The here reported exercise regimen can be applied to genetically modified zebrafish strains with skeletal phenotypes, which resemble common metabolic bone pathologies including osteoporosis 66 or rare genetic disorders including osteogenesis imperfecta 67 . Musculoskeletal exercise of zebrafish not only offers the chance to investigate the impact of exercise on the progression of these diseases, but also the chance to determine the efficacy of new or existing treatments when combined with musculoskeletal exercise. Interestingly, zebrafish show impaired swimming performance with advanced age 42 , indicating that the zebrafish is an extremely useful candidate to study aging-related bone health in combination with musculoskeletal exercise. Specifically, with regard to the similarities in essential bone reorganization mechanisms between humans and teleosts, zebrafish qualify as a valuable and promising model for future mechanistic research in the field of osteology.

Methods
To test the increased skeletal loading on bone in zebrafish, we designed a musculoskeletal exercise regimen. After completion of the exercise regimen, the caudal vertebrae were compared using high-resolution imaging methods in order to detect exercise-induced bone changes. All bone histomorphometric measures were determined considering the ASBMR nomenclature 68 .
Animal models. We acquired 20 zebrafish (Danio rerio) of four and half months of age and divided them into a control group and an exercise group (n = 10 per group). Each group was placed in a 60 l water tank with identical water conditions regarding temperature (26 °C), pH (7.4), water hardness (6-12°dH), nitrite (<0.3 mg/L) and nitrate (25-100 mg/L) levels. All zebrafish were fed flake food twice a day until satiation and were kept on a 14/10 hours light/dark photoperiod. The experiment gained prior approval by the institutional review board (IRB) of the University Medical Center Hamburg-Eppendorf (UKE) and the ethics committee of the city of Hamburg (No. 42/16), and the methods were performed in accordance with the approved guidelines.
Experimental setup and sample preparation. A third 70 l water tank containing a custom-built swim tunnel was used to exercise the zebrafish (Fig. 1a). The swim tunnel was adapted with a stack of flow tubes to maintain a continuous laminar flow. The swimming exercise regimen consisted of five days of exercise per week for a period of four weeks. Zebrafish from the exercise group were coerced to swim against the flow and moved inside the swim tunnel in a natural motion pattern throughout the exercise as shown in Fig. 1a and Supplementary Video 2. Flow velocity in the swim tunnel was kept constant at 12 cm/s while the number of exercise hours increased from four hours per day during week one, to six hours per day during week four. The control group remained in a 60 l water tank for the duration of the experiment and was not subjected to exercise.
To allow the detection of dynamic bone formation, zebrafish were labeled using chromophore calcein (Calcein, Sigma-Aldrich Chemie GmbH, Munich Germany) three days before starting the exercise experiment. This procedure was repeated on the last day of the exercise experiment. Calcein labeling was performed using the submersion protocol described by Du et al. 69 . Briefly, immersion solutions (0.2%) were prepared by dissolving 2 g of calcein powder in 1 liter of deionized water, subsequently immersing zebrafish for 6 min. All zebrafish were sacrificed three days later using an overdose of tricaine methane sulfonate (MS222, 250 mg/l, Sigma-Aldrich Chemie GmbH, Munich Germany). All zebrafish were measured after the experiment to quantify possible macroscopic body changes induced by the musculoskeletal exercise. Body length was measured as standard length, excluding the length of the caudal fin. Three zebrafish spines per group were dissected for subsequent scanning using micro-CT. The remaining zebrafish were fixed in 3.7% formaldehyde for three days, dehydrated with an increasing alcohol series and embedded in methyl-methacrylate blocks. Samples used for static histomorphometric analysis were cut into 4 µm thick sections while dynamic histomorphometry was performed on 12 µm thick sections. Masson-Goldner trichrome and von Kossa/van Gieson protocols were used for the analysis of bone cells and mineralized hard tissue, respectively 70,71 . Micro-computed tomography. Vertebral microstructural differences were assessed using micro-computed tomography (Skyscan 1272, Bruker, Kontich, Belgium) at a resolution of 1 µm. We scanned nine vertebrae per group with 45 kV and 200 µA without an X-ray filter. Ring artifact and beam hardening corrections were kept constant for all samples during reconstruction with NRecon (Bruker, Kontich, Belgium). After applying a fixed threshold for all samples, 3D evaluation was conducted using CTAn (Bruker, Kontich, Belgium). Neural and hemal arches were excluded from the volume analysis. The tissue volume (TV) was defined as the whole volume of a vertebra including the inner open space, whereas the bone volume (BV) was defined as the vertebral body without the inner space. Volume and length (VL) of the vertebral bodies was determined. 3D X-Ray microscopy. The osteocyte network in zebrafish vertebral bodies was quantified using a 3D X-ray microscope (ZEISS Xradia 520 Versa, Carl Zeiss X-ray Microscopy, Pleasanton, CA, USA). Zebrafish vertebrae were scanned at a resolution of 0.7 µm and vertebrae were analyzed using 3D evaluation software (Visual SI Advanced, Object Research Systems Inc, Montreal, Canada; ImageJ2 v1.5, https://imagej.nih.gov/ij). Orientation of the osteocyte lacunae in relation to the long and short axis of the vertebral bodies, sphericity (Sph, 0-1) and mean lacunar volume (Lc.V/N.Lc, μm 3 ) parameters were imaged with 3D X-ray Microscopy (3DXRM) in a total of six caudal vertebrae from two animals per study group.
Dynamic and static histomorphometry. To detect and quantify bone formation, calcein double labels were assessed in fluorescent light microscopy. The following parameters were evaluated using OsteoMeasure (OsteoMetrics, Atlanta, GA, USA): The mineralizing surface per bone surface (MS/BS,%), representing active mineralization, calculated as percent of bone surface that displays a calcein label; the mineral apposition rate (MAR, µm/d), representing the rate of new bone deposition, calculated as distance between consecutively deposited double labels divided by the time interval between them; and the bone formation rate (BFR, µm 3 /µm 2 /y), representing the amount of new bone formed, calculated by multiplying mineralizing surface with the mineral apposition rate.
Osteoblast-and osteoid-related parameters were also quantified using the Detector, Type 202; K.E. Developments Ltd., Cambridge, England). Bone mineral density distribution (BMDD) analysis was performed as previously described 16 to assess the degree of mineralization of the vertebral bone (N = 4 per group). A carbon-aluminum standard serves to calibrate the system and consequently allows a quantification of mineral as a calcium weight percentage 49,72 . Recorded parameters were mean calcium content CaMean (wt%), amount of low mineralized bone (CaLow in % bone area, percentage of bone area that is mineralized below the 5th percentile of a defined reference group) and amount of highly mineralized bone (CaHigh in % bone area, percentage of bone area containing Ca concentration above the 95th percentile). Statistical Analysis. Statistical analysis was carried using SPSS (Version 24, IBM, Armonk, NY, USA) and using a significance level of 5%. Normal distribution was tested using Shapiro-Wilk tests and homogeneity of variance using Levene's test. All parameter comparisons between groups were carried out using independent t-tests.
Data availability statement. The data that support the findings of this study are available from the corresponding author upon reasonable request.