Sost deficiency led to a greater cortical bone formation response to mechanical loading and altered gene expression

Bone adaptation optimizes mass and structure, but the mechano-response is already reduced at maturation. Downregulation of sclerostin was believed to be a mandatory step in mechano-adaptation, but in young mice it was shown that load-induced formation can occur independent of sclerostin, a product of the Sost gene. We hypothesized that the bone formation and resorption response to loading is not affected by Sost deficiency, but is age-specific. Our findings indicate that the anabolic response to in vivo tibial loading was reduced at maturation in Sost Knockout (KO) and littermate control (LC) mice. Age affected all anabolic and catabolic parameters and altered Sost and Wnt target gene expression. While load-induced cortical resorption was similar between genotypes, loading-induced gains in mineralizing surface was enhanced in Sost KO compared to LC mice. Loading led to a downregulation in expression of the Wnt inhibitor Dkk1. Expression of Dkk1 was greater in both control and loaded limbs of Sost KO compared to LC mice suggesting a compensatory role in the absence of Sost. These data suggest physical activity could enhance bone mass concurrently with sclerostin-neutralizing antibodies, but treatment strategies should consider the influence of age on ultimate load-induced bone mass gains.

and also promote osteocytic osteolysis via stimulation of carbonic anhydrase 2 expression 17 . No study has yet examined the resorptive response to mechanical loading under sclerostin deficiency, only effects of unloading on resorption in Sost deficient mice have been studied 7 .
Thus, it remains unclear whether mechanical loading will be effective in enhancing bone formation and reducing bone resorption further in adult patients treated with sclerostin inhibition and whether or how it is affecting Wnt signaling. We hypothesized that the bone formation and resorption response to mechanical loading is not affected by Sost deficiency, but is affected by skeletal maturation. We investigated changes in cortical bone morphology and the formation/ resorption response to two weeks of cyclic axial loading of the left tibiae compared to the right non-loaded tibiae in young growing 10 week old and skeletally mature adult 26 week old Sost KO mice and littermate control (LC) mice (Fig. 1). We performed in vivo microCT imaging at day 0, 5, 10, and 15 and conventional 2D histomorphometry, with labeling at day 3 and 12. In vivo time-lapsed 3D morphometry was performed on imaging data at day 0 and 15 to quantify the volume and surface area of bone formation and resorption over the 15 day interval. We also measured gene expression of several Wnt target genes and Wnt inhibitors at 3, 8 and 24 hours after a single loading session in an additional set of mice of both ages and genotypes.

Results
Altered load transmission in Sost deficient mice compared to littermate controls. Since we planned to conduct a strain-matched loading study, we first tested the relationship between the applied axial compressive load and the bone tissue deformation engendered at the left tibia. The following load levels were calculated for each group to attain peak strains of +900με at the strain gauge site on the medial surface of the cortical bone mid-diaphyseal: −12.9 N in 10 week old Sost KO mice, −14.5 N in 26 week old Sost KO mice, −7N in 10 week old LC mice and −7N in 26 week old LC mice. The slopes of the strain-load regressions 23 were the following: 10wk Sost KO mice: (−0.0143 ± 0.0021 N/με), 26wk Sost KO mice (−0.0161 ± 0.0016 N/με), 10wk LC mice: (−0.0077 ± 0.0016 N/με), and 26 wk old LC mice (−0.0079 ± 0.0006 N/με). The difference in the applied forces illustrates the strong divergence of the thickness and shape of the cortical bone in LC and Sost KO.
Load-induced increases in periosteal bone formation rates and newly mineralized surface area were significantly greater in Sost KO than LC mice. A significant anabolic response to loading was observed in static and dynamic microCT parameters as well as histomophometric measures of bone formation . Subanalyses of each genotype showed that at the beginning of the experiment (day 0), there were no difference in static microCT parameters (Ct.Th, Ct.Ar, Ct.Ar/T.Ar) between the left (loaded) and right (control) tibia, (inter-limb difference) for either Sost KO mice or LC mice of the same age (10 or 26 weeks old) ( Table 1). By day 15, static microCT parameters were greater in the loaded compared to control limb of both young (Ct.Ar: +8, Ct.Ar/T.Ar: +3%, Ct.Th: +7%) and adult (Ct.Ar: +5, Ct.Ar/T.Ar: +4%, Ct.Th: +9%) Sost KO mice (p < 0.014, Fig. 2, Table 1). In contrast, there were no significant differences in static microCT parameters measured in the loaded and control limbs of either young or adult LC mice (Fig. 2, Table 1).
Subanalyses of the dynamic microCT measures of each genotype showed that the loaded limb had a significantly greater volume of newly formed bone than the control limb in young (MV/BV day0-15 : +208%) and adult (MV/BV day0-15 +88%) Sost KO mice over the 15 day period (p < 0.02, Fig. 3). Whereas, a significantly increased volume of newly formed bone in the loaded compared to the control limbs was also observed in the young (MV/ BV day0-15 : +143%, p = 0.018), but not in the adult (MV/BV day0-15 : +39%, p > 0.05) LC mice (Figs 3 and 4).
Surprisingly, the response to loading in terms of newly mineralized surface area (MS/BS day0-15 ) and nearly all static microCT and histomorphometric parameters was significantly greater in Sost KO mice compared to LC mice (Figs. 3-5 and Tables 1, 2; effect of (e) genotype & loading p < 0.05). Whereas the formation response to loading in terms of volume of newly mineralized bone (MV/BV day0-15 ) was not significantly different between the Sost KO and LC mice (Fig. 3, effect of (e) genotype & loading p = 0.625). A subanalysis examining the interlimb difference showed that the response to loading in young Sost KO animals was significantly greater than in LC mice, including the relative MS/BS day0-15 , nearly all relative day 15 static microCT parameters (rCt.Ar, rCt.Ar/T. Ar, rCt.Th, rCt.vTMD), all relative periosteal histomorphometric parameters, but none of the relative endocortical bone formation indices. The same analysis in the adult mice showed the Sost KO mice had a significantly greater response to loading measured in all relative day 15 static microCT parameters (except rT.Ar) and most relative histomorphometric parameters (rEc.dLS/BS, rEc.BFR/BS, Ps.dLS/BS, rPs.MS/BS, rPs.MARand rPs.BFR/ BS). Within both, young and adult mice, Sost KO mice had a greater load-induced total bone gain (Ct.Ar/T. Ar, Ct.Th) than LC mice (Fig. 2, Table 1). In adult mice this increased cortical area and thickness was achieved through a higher load-induced bone formation at both the periosteal and endocortical surface of Sost KO compared to LC mice (p < 0.001, Fig. 5). In young mice, the endocortical surface of both genotypes showed a similar load-induced increase in histomorphometric parameters, the periosteal surface showed a higher load-induced response to loading in young Sost KO compared to young LC mice (p < 0.001, Fig. 5).
The loaded limb had significantly decreased expression of the Wnt inhibitor Dkk1 compared to the nonloaded control limb of young Sost KO mice at 8 hrs. We also observed a tendency for lower Sost expression in the loaded limb compared to the nonloaded control limb of LC mice at 3 and 8 hours in 10 week old mice and at 8 Asterisk indicates a significant difference between loaded and control tibiae (Paired t-test; p < 0.05). and 24 hours in 26 weeks old (Fig. 6B). We found that loading had no effect on expression of the Wnt target gene Axin2, while the Wnt target gene Lef1 had significantly increased expression in loaded compared to control limbs of LC mice at 8 hrs. Loading led to significantly decreased expression of Lef1 in Sost KO mice at 24 hrs (Fig. 6B).
Comparisons of the loaded limbs from Sost KO versus LC using independent t-tests showed that the Wnt inhibitor Dkk1 was significantly upregulated in loaded limbs of Sost KO compared to LC mice (Fig. 6A). Also in the loaded limbs, there was a significantly lower expression of Lef1 in young Sost KO mice at 8 hrs and adult Sost KO mice at 8 hr and 24 hrs compared to LC mice (Fig. 6A).
The cortical bone resorption response to loading is similar between Sost KO and LC mice. Since young Sost KO and LC mice had minimal resorption over the 15 day period, as observed in the control limb, a reduction in resorption with loading was hardly possible (Fig. 3). Although a significant effect of loading was observed in both EV/BV day0-15 and ES/BS day0-15 (Fig. 3), sub-analyses of only adult mice showed that for both genotypes, there was a non-significant reduction in the volume of resorbed bone (EV/BV day0-15 ), but a significant reduction in the resorbed surface area (ES/BS day0-15 ) over the 15 day period in loaded compared to control limbs (p≤ 0.05). The percent difference in the volume and surface area of resorbed bone for the loaded compared to control limbs was similar for the adult Sost KO mice (EV/BV day0-15 : −28%, ES/BS day0-15 : −25%) and LC mice (EV/ BV day0-15 : −25%, ES/BS day0-15 : −27%). Genotype did not significantly influence the EV/BV day0-15 , although Sost KO mice had a significantly lower ES/BS day0-15 than LC mice (Fig. 3). The resorption response to loading (EV/BV day0-15 and ES/BS day0-15 ) was not significantly different between the Sost KO and LC mice (Fig. 3, effect of (e) genotype & loading p > 0.358). A subanalysis examining the interlimb difference showed that within both ages, the relative EV/BV day0-15 and relative ES/BS day0-15 were similar between the Sost KO and LC mice. Skeletal maturation diminished the bone formation response to loading independent of sclerostin. Both genotypes were less mechanoresponsive (in terms of volume and area of newly mineralized bone in response to loading) in adult compared to young mice (Fig. 3, effect of (f) age & loading p < 0.002). For example, the load-induced percent increase in MV/BV day0-15 was greater in young (+143%) compared to adult (+39%) LC as well as for young (+207%) compared to adult (+87%) SostKO mice (Fig. 3). A subanalysis examining the interlimb difference showed that within both genotypes, the relative MV/BV day0-15 and relative MS/BS day0-15 were significantly greater in young compared to adult. In Sost KO mice, maturation diminished the bone formation response to loading in the periosteal cortical bone (Ps.BFR/BS: −34%, Ps.MAR: −35%) (p < 0.046, Fig. 5) and therefore the overall bone formation response (MV/BV day0-15 : −75% and MS/BS day0-15 : −46%) (p < 0.011, Fig. 3). Interestingly there was no difference in mechanoresponse to loading at the endocortical surface between young and adult Sost KO mice; relative (interlimb) endocortical bone formation histomorphometric indices were similar. Whereas the relative Ec.BFR/BS was (−62%) significantly lower in adult compared to young LC mice (Fig. 5). Age significantly affected Sost expression in LC mice as well as Wnt target genes Lef1 and Axin2 expression in Sost KO and LC mice (Fig. 6C). Interestingly, the expression of Lef1 at the 24 hr time point was significantly lower in both the loaded and control limbs of adult compared to young Sost KO mice. At the 3 and 8 hr time points, Axin2 expression was significantly greater in both the loaded and control limbs of adult compared to young Sost KO mice. We also observed a significantly greater Dkk1 expression at 24 hrs in the control limbs of young compared to adult Sost KO mice. Lastly, we measured a significantly lower Sost expression at 3 hrs in the control limb of adult compared to young LC mice.
Control limbs of young and adult LC mice had a similar volume of resorption to that of age-matched Sost KO mice (EV/ BV day0-15 (mm²/mm²) LC 10wk: 0.002 ± 0.001; Sost KO 10wk: 0.004 ± 0.003; LC 26wk 0.024 ± 0.012; Sost KO 26wk 0.018 ± 0.007). However, adult LC mice had a significantly greater bone resorption surface than adult Sost KO mice (ES/BS day0-15 : +103%, p = 0.001), while young mice of both genotypes exhibited near negligible levels. Therefore in the control limbs, the difference in bone resorption surface between young and adult mice was much greater in LC compared to Sost KO mice.
Comparisons of the control limbs from Sost KO versus LC showed that the Wnt inhibitor Dkk1 was significantly upregulated in control limbs of Sost KO compared to LC mice (Fig. 6A). We also measured significantly greater expression of Lef1 at 24 hrs in the control limb of young Sost KO compared to LC mice.

Discussion
In this study, we examined whether two weeks of in vivo tibial loading could enhance bone formation and reduce bone resorption in young and adult female Sost KO and LC mice. We hypothesized that the bone formation and resorption response to mechanical loading is not affected by Sost deficiency, but is affected by skeletal maturation. We investigated changes in cortical bone morphology and the adaptive (re)modeling response using in vivo microCT imaging, registered microCT data as a 4D imaging biomarker of bone formation and resorption, and conventional 2D histomorphometry. We also examined the gene expression in Wnt target genes (Axin2 and Lef1) and Wnt inhibitors (Dkk1 and Sost) at 3, 8, and 24 hours after a single loading session. We observed that Sost deficiency not only led to increased formation, but also decreased resorption in the control nonloaded limbs. Our results show that the bone (re)modeling response and the expression of Wnt target genes is dependent on skeletal maturation in Sost KO mice (Fig. 6C). Sost KO mice had significantly higher Dkk1 expression compared to LC mice, independent of loading at nearly all time points and ages. We demonstrated that the cortical bone resorption response to loading was similar in Sost KO and LC mice, while the cortical adaptive response in terms of formation was enhanced in Sost KO compared to LC mice. Additionally, recent finite element analyses from our group showed that compressive and tensile strains were lower in the Sost KO than in the LC mice, for both young and adult animals 24 . Therefore, the greater cortical bone formation response to loading in Sost KO mice compared to LC mice was not due to higher mechanical strains engendered during loading.
The cortical adaptive response in terms of formation was surprisingly enhanced in Sost KO compared to LC mice (Figs 2-5). Although the formation response to loading in terms of volume of newly mineralized bone (MV/BV day0- 15) was not significantly different between the Sost KO and LC mice, Sost KO mice had significantly greater load-induced increases in the absolute and relative newly mineralized surface area (MS/BS day0-15 ), most absolute and relative static microCT parameters and histomorphometric parameters compared to LC mice (Figs 2-5 and Tables 1, 2; effect of (e) genotype & loading p < 0.05 and t-tests of interlimb difference between Sost KO and LC mice within each age p < 0.05). A recent study by Robling et al. 25  control mice. However, regional analyses showed differences in terms of strain-dependent distribution of newly formed bone in response to loading. They measured greater rBFR/BS in regions of high strain (medial and lateral region) compared to low strain (caudal and cranial region) in wild-type mice. Whereas, Sost KO mice had lower rBFR/BS relative to wild-type mice in high strain regions, but greater rBFR/BS relative to wild-type mice in low strain regions. The only other study to examine mechanoresponse during Sost deficiency 7 , showed region-specific varied responses to tibial loading in 10 wk old female Sost KO mice compared to wild-type controls. In their strain-matched results, they reported a greater response to loading in Sost KO mice compared to wild-type at the metaphyseal region (Ct.Th: 23% in Sost KO, 8% in wild-type), but a similar response to loading in Sost KO mice compared to wild-type at the 37% mid-diaphyseal region (Ct.Th: 15% in Sost KO, 15% in wild-type). At the same region as in our study, 50% mid-diaphyseal region, Morse et al. reported a slightly lower response to loading in Sost KO compared to wild-type mice (Ct.Th: 11% in Sost KO, 14% in wild-type).We recently showed that the mouse tibial metaphyseal and mid-diaphyseal regions have different set points (or threshold above which an anabolic response to loading occurs) and slopes of the relationship between engendered strains and remodeling response 26 , which may explain the region-specific differences. In addition, some possible reasons for the contrasting results in these studies 7, 25 compared to our data may be due to differences in loading parameters, anatomical region loaded, or use of wild-types as controls rather than littermates. Also, neither study 7, 25 included baseline controls which would confirm if the contralateral limb was a suitable control. We included baseline controls in the form of in vivo microCT measurements at day 0, prior to loading, where we observed similar interlimb differences between Sost KO and LC mice. Also, Morse et al. reported a woven bone response due to loading, which was absent in our study; only lamellar bone was formed in response to loading. There were also stark differences in the reported strain-load relationship of the Sost KO mice between the two studies. Morse et al. 7 reported engendering 1200 με in 10 week old Sost KO mice by using −12.5 N, while we engendered 900µɛ in the tibial mid-shaft of young mice using −7N (LC) and −12.9N (Sost KO). Originally we loaded a group of 10 week old Sost KO mice at a load level (−17N), which through in vivo strain gauging we determined engendered 1200µɛ at the mid-diaphysis. However, we had to discontinue the experiment prematurely since this load/strain level led to ankle swelling and limping in the mice (see Method's section). The difference in the load-strain relationship between the two studies remains unclear, since the Sost KO mice in both studies were from the same source. As expected, since the formation response to loading is dependent on the maximum strain levels used 27 , we measured a lower bone formation response to loading with 900 με in the LC mice in the current study compared to what we reported previously using 1200με in C57Bl/6 wild-type mice 28 . Lastly, previous studies have suggested that load-induced bone formation is inversely proportional to sclerostin abundance 29,30 , which may have contributed to the different results between the studies. Our study is the first to examine the effect of mechanical loading on Wnt signaling under sclerostin ablation. Although, Lin et al. 31 showed that unloading via tail suspension led to upregulation of Sost and downregulation of Lef1 expression in 7 and 17 week old female wild-type mice. They measured no alteration in Lef1 in Sost KO mice of either age with unloading. We observed significantly decreased expression of Lef1 (loaded vs control limb) in Sost KO mice at 24 hrs (Fig. 6B). Similar to the significant increased expression of Lef1 after ulnar loading in young wild-type mice reported by Tu et al. 6 , tibial loading in our study led to significantly increased expression (loaded vs control limb) in LC mice at 8 hrs. Our data are consistent with Holguin et al. 32 who reported that loading led to a significantly decreased expression of the Wnt inhibitor Dkk1 in wild-type mice of various ages. We measured no changes in the expression of the Wnt target gene Axin2 with loading, which is in line with data of Holguin et al. 32 who report that loading had no effect on expression of Wnt target gene Axin2 after a single loading bout in wild-type mice. In contrast, they also reported no effect of loading on Lef1 expression, which we observed in adult LC mice at 8 hrs (Fig. 6B). We could not detect expression of Wnt ligands (Wnt1 and Wnt7b).
Our data also showed that ablation of the Sost gene in mice resulted in an increased expression of another Wnt inhibitor, Dkk1, possibly due to a feedback mechanism intended to compensate for the loss of Sost. However, the bone anabolic response was even higher in Sost KO mice, suggesting that Dkk1 is not sufficient to compensate for the loss of Sost. A significantly lower expression of Lef1 was measured in the loaded limbs of young Sost KO mice at 8 hrs and adult Sost KO mice at 8 hr and 24 hrs compared to LC mice (Fig. 6A). We also measured significantly greater expression of Lef1 at 24 hrs in the control limb of young Sost KO compared to LC mice. Lin et al. 31 also observed this interaction between genotype and age, as they also reported that young Sost KO mice had greater Lef1 expression than wild-type mice, while there was no difference in Lef1 expression in adult Sost KO compared to wild-type mice.
In both Sost KO and LC mice, skeletal maturation led to increased cortical thickness, cortical area fraction and reduced bone formation in the control bones (Figs 2-5 and Tables 1, 2). The increase in bone elastic modulus 33 and cross-sectional moment of inertia with skeletal maturation would result in a decreased strain in the older mice for a given force, which we did not observe at the medial midshaft. This apparent contradiction can be explained by changes in whole bone geometry with aging. We and others have previously shown that the bones of 26 wk old mice are more curved than those of 10 wk old C57Bl/6 mice 11, 23, 28 , which will increase their bending stresses under a given load and counteract age-related increases in mineral and geometric properties. Additionally, a recent study from our group 24 showed that although finite element models predicted approximately 900 µε at the strain gauge position, lower strains are predicted in 26 week old mice compared to 10 wk old LC mice at other positions along the tibial length, suggesting that strain gauging one position is inadequate to characterize the strain distribution. The increase in absolute and relative cortical area fraction and cortical thickness due to skeletal maturation in the control bones was higher in Sost KO compared to LC mice (Fig. 2). However in both LC and Sost KO mice, there was a lower anabolic response to loading in adult mice compared to young mice. The load-induced percent difference in MV/BV day0-15 was greater in young (+143%) compared to adult (+39%) LC as well as for young (+207%) compared to adult (+87%) Sost KO mice. An age-specific gene expression response was present in both Wnt signaling target genes examined (Lef1 and Axin2), as well as in Sost expression in the LC mice. Axin2 expression was higher in skeletally mature animals, while Lef1 was higher in young mice, thereby indicating that different Wnt effectors might be involved in bone metabolism, depending on the skeletal development stage.
We showed that with Sost deficiency the load-induced cortical resorption response, in terms of volumes and surface area of resorbed bone in response to loading, was similar to that of LC mice (Fig. 4). As in our previous studies of female C57Bl/6 mice 16, 32 , we observed only a minimal amount of bone resorption occurring in the cortical diaphysis of the tibia from young LC and Sost KO mice. There was, however, a small but significant decrease in the eroded surface of the loaded compared to control tibiae of adult LC mice. Similarly, Morse et al. 7 showed no changes with unloading in osteoclast number, osteoclast surface and the fraction of bone surface with osteoclasts, in both young C57BL/6 and Sost KO mice. Also, Tian et al. 34 reported that resorption (Ec.Er.S/BS) was similarly reduced at the tibial midshaft of middle-aged rats (10 months old) after 4wks of sclerostin neutralizing antibody treatment in control and unloaded (immobilized) bones. Interestingly, there was a greater eroded surface in the control limbs of adult LC than Sost KO mice (Fig. 3). Similarly, it has been shown that short-term sclerostin inhibition leads to reduced bone resorption in ovariectomized rats and cynomolgus monkey vertebrae and femoral endocortex 19,35,36 . Moreover, recent studies by Atkins and colleagues show that sclerostin promotes osteoclastic bone resorption through regulation of RANKL and also promotes osteocytic osteolysis via stimulation of carbonic anhydrase 2 expression in human primary osteocyte-like cells and mouse MLO-Y4 cells 17,22 . Our data suggests that long-term sclerostin deficiency may have a protective influence against age-related increases in cortical bone resorption.
A limitation in our study is that we did not investigate the mechanoresponsiveness of elderly Sost KO mice. There were a few reasons for this. Firstly, we learned from 10 and 26wk old Sost KO mice that the load levels necessary to achieve comparable strains in elderly Sost KO mice would have caused damage to the joints. Secondly, we observed an increased mortality in Sost KO mice in our breeding program and therefore, aging a population of SCIeNTIfIC RepoRts | 7: 9435 | DOI:10.1038/s41598-017-09653-9 Sost KO mice would have required an unjustifiable number of mice. Previous studies in wild-type mice from our group and others have shown that most of the reduced formation response to loading occurs already at skeletal maturation and therefore a great deal of information can be gained from examining young and skeletally mature mice 37,38 . Lastly, in order to reliably mimic the effects of anti-sclerostin treatment it would be necessary to conditionally knock out the Sost gene after the growth phase, since treatment is targeting the elderly population with normal bone mass accrual during growth.
Our model aims to investigate structural adaptation of the mouse tibia through surface modeling and remodeling processes to additional mechanical loading. Since these processes include both formation and resorption, we analyze both to understand the structural response of the bone to loading. Models of disuse may be more appropriate for other investigations focused more on resorptive processes. One must keep in mind that the mechanisms responsible for bone formation to additional loading and maintaining bone mass with normal load, may not be the same as those for recovering bone mass after unloading 39 .
Our key findings in the current study are: 1) the load-induced cortical bone formation response was significantly enhanced in Sost KO compared to LC mice 2) Dkk1 expression was significantly greater in Sost KO compared to LC mice, and was significantly downregulated with loading (loaded versus control nonloaded limb) 3) the load-induced cortical resorption response was similar in Sost KO and LC mice, 4), the anabolic response to loading was reduced at maturation in both LC and Sost KO mice coincident with age-dependent expression of Wnt target genes in Sost KO and LC mice as well as Sost gene expression in LC mice. Although we were not able to obtain data from elderly Sost KO mice as explained above, our findings suggest that future treatment strategies which include long-term sclerostin inhibition should consider the age-dependent effectiveness in reducing resorption or increasing formation as well as the role of additional physical activity regimens. Continued understanding of how long-term Sost/sclerostin deficiency influences (re)modeling processes is critical to effectively treat patients with low bone mass.

Materials and Methods
Animals and genotyping. The sperm of four male Sost−/− mice (provided by Novartis) was pooled into two groups and intracytoplasmic sperm injection with the oocytes from female C57Bl/6 mice was performed at the Charité medical university animal facility. The first heterozygous generation was mated among themselves. In following generations, homozygous Sost KO and littermate control (LC) mice were identified using a Multiplex PCR with mice tail cuts, according to a protocol provided by Novartis. Three primers (Sost-specific endogen: 5′ TCC ACA ACC AGT CGG AGC TCA AGG 3′, Sost-specific endogen and target: 5′ ACT CCA CAC GGT CTG GAA AGT TTG G 3′ and Neo target: 5′ GGG TGG GAT TAG ATA AAT GCC TGC TCT 3′, acquired from TIB MOLBIOL, Berlin, Germany) were used to detect the homozygous LC (Sost-specific endogen -Sost -specific endogen and target) and Sost KO (Neo target -Sost -specific endogen and target) mice. All animal experiments described were carried out according to the policies and procedures approved by the local legal research animal welfare representative (LaGeSo Berlin, G0021/11).
In vivo strain-load calibration. Single element strain gauges (EA-06-015LA-120, Micromeasurements, USA) were prepared and attached to the medial surface of the tibial mid-shaft aligned with the bone's long axis of both limbs of 28 mice (n = 7/age/genotype) as explained previously 28,40 . A range of axial dynamic compressive loads (peak loads ranging from −2 to −14 N in LC and −2 to −20 N in Sost KO mice) was applied between the flexed knee and ankle using an in vivo loading device (Testbench Electro Force LM1, TA Instruments, USA), while strain measurements were recorded simultaneously using WinTest software to determine the relationship between the applied compressive loads and bone tissue deformation.
In vivo mechanical loading. The left tibiae of 10 week (young) and 26 (adult) week old Sost KO and LC mice (n = 7/age/genotype) underwent 2 weeks (5 days/week, Monday-Friday) in vivo cyclic compressive loading (216 cycles applied daily at 4 Hz, peak strains at strain-gauge position of +900 με) (Fig. 1) while under anesthesia. The triangle waveform included 0.15 sec symmetric active loading/unloading, with a constant strain rate of 0.016 ε/sec maintained during both the loading and unloading ramp of the waveform in mice of both ages. The waveform also included a 0.1 sec rest phase (−1N) between load cycles and a 5 sec rest inserted between every four cycles. The right tibia served as internal control as described before 28,41 .
Three days after the last loading session, the mice were sacrificed (day 15), while under anesthesia (ketamine 60 mg/kg and medetomidine 0.3 mg/kg) through an overdose of potassium chloride. The weight was measured before loading and daily throughout the experiment. No mice had to be excluded from the experiment because of weight loss. There were higher weight variations in the Sost KO mice (young: 21 g ± 1.5 g; adult: 25.6 g ± 1.5 g) than in the LC mice (young: 19 g ± 0.8 g; adult: 22.9 g ± 0.4 g) at day 15. Additionally, no mice exhibited any complications or signs of limping throughout the experiment.

Longitudinal In vivo micro-computed tomography of cortical mid-diaphysis. Longitudinal in vivo
micro-computed tomography (microCT) with a voxel size of 10.5μm (vivaCT 40, Scanco Medical, Brüttisellen, Switzerland; 55kVp source voltage, 145 μA source current, 300 ms integration time, no frame averaging, range of 180 degrees) was performed at day 0, 5, 10, and 15 to assess the cortical bone compartment in both the right and left tibiae. Average scan time was 48 minutes. For each imaging session, mice were anesthetized (ketamine 60 mg/kg and medetomidine 0.3 mg/kg). To prevent motion artifacts during microCT scanning, anaesthetized mice were constrained in a custom-made plastic mouse bed. Scans were reconstructed and analyzed using standard filtered backprojection implemented using software from the microCT. The microCT was calibrated weekly against a hydroxyapatite mineral phantom for determining in-plane spatial resolution. An error occurred during imaging Dynamic histomorphometry. Calcein (20 mg/kg) was administered to the LC mice via intraperitoneal injection at day 3 and 12. Sost KO mice were labeled with calcein at day 3 (20 mg/kg) and alizarin (30 mg/kg) at day 12. Although the labeling dyes were different between the genotypes, Sun et al. 44 showed that calcein and alizarin staining provide comparable results. After dissection of the tibiae from the surrounding soft tissues, their lengths were measured using digital calipers to be the following: The tibiae were dehydrated in ascending grades of ethanol to absolute, cleared in xylene, infiltrated and finally embedded in methyl-methacrylate. The blocks were sectioned transversal to the bones long axis at the cortical midshaft. The slices were ground and polished to an approximate thickness of 60 μm and viewed at a magnification of 200 × under a mercury lamp microscope (KS400 3.0, Zeiss, Oberkochen, Germany) for evidence of fluorochrome labels (Figs 1, 5). Images were acquired using commercially available software (Axiovision, Zeiss, Oberkochen, Germany). The analyzed region of interest for the cortical bone included endo-and periosteal surface. The single-and double-labeled surface per bone surface (sLS/BS, dLS/BS), mineralizing surface (MS/BS), mineral apposition rate (MAR), and bone-formation rate (BFR/BS), were analyzed using ImageJ. MS/BS was calculated as 0.5x sLS/BS + dLS/BS. When a specimen had no double-labeled surface (dLS/BS = 0), it was labeled as "no data" for MAR and BFR/BS 45 . The amount of newly mineralized bone per day was calculated using the averaged double label distances divided by the 9-day labeling interval, and expressed as the MAR in units of microns per day. For determining MAR, the entire endocortical (Ec) and periosteal (Ps) surfaces were analyzed (Fig. 5). Except for the young Sost KO control limb (n = 6), 7 specimen per group were analyzed ( Table 2). qPCR analysis. The left tibiae of an additional 72 female mice (n = 6/genotype/age/time point) was loaded for a single loading session. The loading procedure has been described above (see section In vivo mechanical loading). Mice were sacrificed at either 3, 8, or 24 hours after the single loading session. Bone marrow was removed and RNA was extracted using TRIzol ® reagent (Thermo-Fisher), followed by purification with RNEasy kit (Qiagen).
Statistical analysis for microCT parameters,histomorphometry and qPCR. The within-subject effect of loading (loaded, control limbs) and between-subject effects of age (10 week old, 26 week old) and genotype (Sost KO mice and LCs) as well as interactions between these terms was assessed using a repeated measures ANOVA (SAS 9.4, Cary, USA) for absolute values. A separate ANOVA was used to assess between-subject age (10 week old, 26 week old) and genotype (Sost KO mice and LCs) and interaction effects for relative values, the interlimb differences (∆ interlimb = loaded limb − control limb). Furthermore, subanalyses were performed on absolute values and relative values (interlimb difference) of all outcome measures using paired or independent t-tests. The percent difference is presented as %Δ = ((loaded limb -control limb)/control limb) × 100%). Statistical analyses of the qPCR analysis were performed on the ΔCt values (Fig. 6). A p value ≤0.05 was considered significant.