Static Preload Inhibits Loading‐Induced Bone Formation

Nearly all exogenous loading models of bone adaptation apply dynamic loading superimposed upon a time invariant static preload (SPL) in order to ensure stable, reproducible loading of bone. Given that SPL may alter aspects of bone mechanotransduction (eg, interstitial fluid flow), we hypothesized that SPL inhibits bone formation induced by dynamic loading. As a first test of this hypothesis, we utilized a newly developed device that enables stable dynamic loading of the murine tibia with SPLs ≥ −0.01 N. We subjected the right tibias of BALB/c mice (4‐month‐old females) to dynamic loading (−3.8 N, 1 Hz, 50 cycles/day, 10 s rest) superimposed upon one of three SPLs: −1.5 N, −0.5 N, or −0.03 N. Mice underwent exogenous loading 3 days/week for 3 weeks. Metaphyseal trabecular bone adaptation (μCT) and midshaft cortical bone formation (dynamic histomorphometry) were assessed following euthanasia (day 22). Ipsilateral tibias of mice loaded with a −1.5‐N SPL demonstrated significantly less trabecular bone volume/total volume (BV/TV) than contralateral tibias (−12.9%). In contrast, the same dynamic loading superimposed on a −0.03‐N SPL significantly elevated BV/TV versus contralateral tibias (12.3%) and versus the ipsilateral tibias of the other SPL groups (−0.5 N: 46.3%, −1.5 N: 37.2%). At the midshaft, the periosteal bone formation rate (p.BFR) induced when dynamic loading was superimposed on −1.5‐N and −0.5‐N SPLs was significantly amplified in the −0.03‐N SPL group (>200%). These data demonstrate that bone anabolism induced by dynamic loading is markedly inhibited by SPL magnitudes commonly implemented in the literature (ie, −0.5 N, −1.5 N). The inhibitory impact of SPL has not been recognized in bone adaptation models and, as such, SPLs have been neither universally reported nor standardized. Our study therefore identifies a previously unrecognized, potent inhibitor of mechanoresponsiveness that has potentially confounded studies of bone adaptation and translation of insights from our field. © 2018 The Authors. JBMR Plus Published by Wiley Periodicals, Inc. on behalf of the American Society for Bone and Mineral Research.


Introduction
T he enhanced skeletal mass and morphology observed in athletes and other highly active individuals is thought to predominantly arise as a result of the skeleton's lifelong exposure to dynamic loading. (1)(2)(3)(4) Numerous facets of dynamic loading have been associated with amplifying the anabolic benefit of skeletal loading in preclinical studies (eg, magnitude, rate, cycle number, frequency, rest intervals, etc. (5)(6)(7)(8)(9) ). However, these insights have not led to effective exercise interventions that augment bone mass in adult humans. (10)(11)(12) Regardless of field, there are numerous challenges in translating preclinical observations to successful human interventions. One strategy to overcome this challenge is to minimize disparities between preclinical models and humans. (13,14) In considering potential barriers in our field, one evident challenge is that the majority of in vivo mouse bone adaptation models use exogenous loading protocols that result in tissue strains that are substantially larger than those induced by gait. For example, the murine tibia axial compression model that has been used to extensively explore bone adaptation in response to skeletal loading possesses numerous strengths that have led to its broad adoption (eg, ability to study both trabecular and cortical bone adaptation, ease of implementing a highly reproducible strain environment (15,16) ). However, finite element analysis (FEA) models suggest that the end loading magnitude predominantly implemented to induce an anabolic trabecular response (À9 N) causes peak normal strains ranging from À2800 to À3300 me at the tibia midshaft. (16,17) In contrast, locomotion in mice generates peak tibia midshaft normal strains that are less than 25% of the lower end of that range. (18) To begin to address this barrier, we developed an in vivo loading device that we anticipated would reduce the exogenous loading magnitude required to induce a given peak strain magnitude in the tibia. The device modestly achieved this goal via induction of off-axis compression, which caused increased tibia bending compared to the axial tibia compression device. Intriguingly, during calibration studies, we observed that our model required only an extremely small static preload (SPL) compared with previous in vivo models (Fig. 1). To our knowledge, nearly all in vivo exogenous skeletal loading models use SPL to ensure that each load cycle is reproducible during a loading intervention. Our review of tibia axial compression studies to date revealed that, with two exceptions, (15,19) SPL has varied from À0.5 N to over À2 N.
Given that static stimuli may impair perception of primary signaling underlying bone mechanotransduction (eg, by reducing the potential for interstitial fluid flow (20) ), we speculated that SPL may inhibit the anabolic response of bone to dynamic loading. We therefore exploited the unique ability of our new loading model to stably load the tibia with minimal SPL, and assessed trabecular and cortical bone responses to a À3.8-N dynamic loading regimen superimposed upon a À1.5-N, À0.5-N, or À0.03-N SPL. The literature suggested that a À3.8-N dynamic loading regimen superimposed on a À1.5-N SPL would not be anabolic. (21)

Materials and Methods
Off-axis compression of the murine tibia To load the tibia in off-axis compression, the right foot is secured in a custom foot bed, with the knee and ankle joints aligned as at the time of peak ground reaction force (GRF) during free ambulation, (22) and the loading tine (oriented 60 degrees from vertical) positioned to contact the distal femur ( Fig. 2A,B). The loading tine is coupled to a linear force actuator (AFX10-200; Motran Industries, Ventura, CA, USA) in series with a force transducer (Model 31; Sensotec, Columbus, OH, USA), and openloop controlled via LabVIEW (National Instruments, Austin, TX, USA). Micrometer-driven stages (433 series; Newport Corp., Irvine, CA, USA) are used to fine-tune mouse and foot position relative to the loading tine in the x, y, and z directions. Once positioned, the loading tine axis is aligned to apply compressive force from proximal to anterior in the frontal plane and from medial to lateral in the sagittal plane, thereby achieving off-axis compressive loading of the murine tibia. To confirm stable positioning, the distal femur is first quasi-statically loaded to À0.5 N. If movement of the tibia is observed, positioning is adjusted and stability reconfirmed. Once stability is achieved, the loading tine is retracted (0 N) and the dynamic loading protocol is superimposed on the assigned SPL via LabVIEW software. The loading tine, foot bed, and mouse platform are each 3D printed from computer-aided design (CAD) drawings (provided on request).

Calibration of the loading-induced strain environment
We determined the strain distributions induced in the murine tibia by off-axis compression via combined strain gauging, mCT imaging, and FEA. (23) Briefly, a single element strain gauge (FLK-1-11; Tokyo Sokki Kenkyujo Co., Ltd., Tokyo, Japan) was attached to the medial periosteal midshaft surface (aligned with the tibia long axis) of female BALB/c mice (4 months old; n ¼ 3), and the mouse positioned in the off-axis loading device. We then assessed whether the coefficient of variation (CV) of peak gauge strains across loading cycles were influenced by the smallest SPL implemented in vivo in this study (À0.03 N), the most common SPL (À0.5 N), and an SPL that approximated the largest implemented in the axial compression model (À2.0 N). Each tibia was loaded over a range of loading magnitudes (À2 to À7 N, increments of À1 N) each superimposed upon each SPL (À0.03, À0.5, and À2.0 N). For each combination of variables, the applied forces and gauge strains were recorded for 10 consecutive, 1-Hz load cycles (gauge data sampled at 500 Hz).
To determine the induced strain environment, the tibias were then imaged by mCT (21-mm voxel resolution), and images processed to develop voxel-based finite element (FE) models ($60,000 eight-node brick elements). In order to determine the loading-induced strain distributions throughout the tibia-fibula structure, boundary conditions simulating the off-axis force orientations were applied, isotropic bone material properties implemented (E ¼ 19.7 GPa, n ¼ 0.3 (24) ), and the FE model analyzed using the open-source FE solver CalculiX (http:// www.calculix.de (22) ). The strains simulated at the FE model nodes representing the strain gauge location were averaged and used to validate an FE model against the strains recorded at the gauge site. Upon validation, boundary conditions were applied to animal-specific FE models (n ¼ 3) in order to determine the peak longitudinal normal strains in the metaphyseal trabecular bone compartment, and in the cortical bone compartment at the tibia midshaft across a range of applied loads in off-axis compression (À0.01 to À6 N).

In vivo experiment
We used BALB/c mice as their substantial basal trabecular bone volume/total volume (BV/TV) (eg, compared to C57BL/6 mice) enhanced our experimental resolution to detect varied extents of either anabolic or catabolic bone adaptation. For in vivo loading, female BALB/c mice (16 weeks; The Jackson Laboratory, Bar Harbor, ME, USA) were anesthetized (2% isoflurane) and were positioned in the off-axis loader as described in the section titled "Off-axis compression of the murine tibie". The right tibias were then exogenously loaded with a À3.8-N, 1-Hz, 50-cycle/day (10-s rest between each load cycle) regimen 3 days/week (Monday, Wednesday, Friday) for 3 weeks (ie, 9 loading bouts/ mouse). Each load cycle was implemented as a trapezoidal waveform with 0.1-s linear rise to peak (0.025/s strain rate), 0.8-s dwell at peak total load, 0.1-s linear decline to baseline, and a 10s rest interval at baseline (ie, the specified SPL). Based on a power analysis of preliminary data, experiments were designed with n ¼ 12 mice per group. Mice were randomly assigned to receive the À3.8-N dynamic loading superimposed on an SPL of À1.5 N, À0.5 N, or À0.03 N. Dynamic load and SPL magnitudes were recorded throughout each loading bout. However, during the second week of the experiment, four mice in the À0.5-N SPL Fig. 1. Schematic of DL superimposed on SPL. All in vivo exogenous bone loading models require SPL to ensure reproducible DL (where TL ¼ DL þ SPL). In this example, three load cycles of a À3.8-N DL waveform (0.1 s to peak load, 0.8 s dwell at peak load, 0.1 s to baseline) with 10-s rest inserted between each load cycle is superimposed on a À1.5-N SPL. DL ¼ dynamic loading; SPL ¼ static preload; TL ¼ total load. group and two mice in the À0.03-N SPL group accidently received an incorrect SPL (À1.5 N). These mice were euthanized and removed from the study when the error was recognized prior to the subsequent loading bout. This error resulted in final group sizes of n ¼ 12, 8, and 10 mice for the À1.5-N, À0.5-N, and À0.03-N SPL groups, respectively. All mice were calcein labeled (day 10, day 19; 10 mg/kg i.p.) and euthanized on day 22 (CO 2 inhalation).
Using separate mice, we also assessed whether gait-induced GRFs were altered following exogenous off-axial loading of the tibia (female BALB/c, 16 weeks old, n ¼ 4 for each SPL). GRFs were recorded on day 0 (prior to initiation of in vivo loading experiments) and on day 22 following nine bouts of dynamic loading (À3.8 N) superimposed upon one of three SPLs (ie, À1.5 N, À0.5 N, or À0.03 N). GRFs were assessed during free ambulation as each mouse repeatedly walked along an enclosed Plexiglas walkway (50 mm wide by 750 mm long) equipped with a miniature force plate (AMTI HE6X; Advanced Mechanical Technology, Inc., Watertown, MA, USA), centered in the walkway transit. Following acclimation to the testing apparatus, GRFs were collected from force plate contacts with the right hindpaw (ie, the experimental limb), and verified via a synchronized video capture system (Prosilica EC 650C high speed digital camera; Allied Vision, Exton, PA, USA). Trials were repeated until the right hindpaw made clear force plate contact three times per mouse. Force data were recorded at 200 Hz and video was captured at 30 Hz via customized LabVIEW software.
All mice were provided food and water ad libitum and housed five mice per cage in the specific-pathogen-free vivarium within the Harborview Research & Training Building (University of Washington [UW], Seattle, WA, USA). All experimental procedures were approved by the UW IACUC. No adverse events were noted in the experiments. All analyses were performed blinded to group identity.
Immediately following euthanasia, the metaphyseal trabecular compartment of the loaded ipsilateral and contralateral tibias were imaged by mCT (Scanco vivaCT40, 10.5 mm voxel, 55 kVp, 145 mA; Scanco Medical AG, Br€ uttisellen, Switzerland). The imaged region encompassed 2.23 mm of the proximal tibia, with image analysis performed on the 0.85 mm immediately distal to the growth plate. (25) Raw mCT image data were preprocessed using a Gaussian Filter algorithm to remove image noise (Sigma ¼ 1.2, Support ¼ 2.0) followed by trabecular bone segmentation using a semiautomated contouring algorithm in which the endocortical shell of the proximal metaphysis was used to guide the total analysis volume. (26) A mCT threshold of 490.5 mg hydroxyapatite (HA)/cm 3 was used to quantify trabecular bone and provided optimal comparisons between 2D binarized images and the original grayscale images. (27) We then determined total volume (TV), bone volume (BV), ratio (BV/ TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular spacing (Tb.Sp) within the metaphyseal trabecular compartment per standard algorithms. As is standard, TV represented the endocortical volume within the metaphyseal region. The periosteal volume (Ps.Vol) and cortical bone volume (Ct.Vol) for the metaphyseal region were also quantified.
After mCT imaging, the right and left tibias were dissected, soft tissue removed, and cross-sections (700 mm thick) spanning the midshaft were obtained. The sections were ground to 90 mm thickness and imaged (Zeiss Apotome.2 epifluorescent microscope; Zeiss, Inc., Thornwood, NY, USA). Images were blinded Off-axis loading is achieved as the compressive end loads are applied via a loading tine that is offset from posterior to anterior in the frontal plane and medial to lateral in the sagittal plane (ie, the loading tine is not aligned with the long axis of the tibia in neither anatomical plane). Pr ¼ proximal A ¼ anterior. 2) for À2.0-N, À0.5-N, and À0.03-N SPL groups, respectively, across the 10 load cycles. The coefficient of variation of peak strain across load cycles of dynamic loading were 1.45%, 2.08%, and 2.83% for the À2.0-N, À0.5-N, and À0.03-N SPL groups and were not significantly altered by the SPL applied (p ¼ 0.15). and analyzed using custom ImageJ-based software (NIH, Bethesda, MD, USA; https://imagej.nih.gov/ij/) to quantify mineralizing surface (MS/BS), mineral apposition rate (MAR), and bone formation rates (BFRs) at the endocortical (e.) and periosteal (p.) surfaces. (28) Statistical analysis All statistical analysis was performed in R (R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project. org/). For calibration data, Feltz and Miller's asymptotic test (29) was implemented to test for the equivalence of the coefficients of variation of induced strains across load cycles in order to access whether the stability of dynamic loading was influenced by SPL. In the in vivo experiments, ANOVA (with Tukey honest significant difference [HSD] post hoc tests) was used to assess whether SPLs and/or dynamic loads were significantly different across groups. Follow-up equivalence testing was performed to confirm that the applied dynamic loads were equivalent across in vivo groups. (30) Because in vivo outcomes data residuals (eg, BV/TV, BFR) were not normally distributed (via a Kolmogorov-Smirnov test), we implemented nonparametric statistics for assessment of bone adaptation. As such, we used Kruskal-Wallis tests with Mann-Whitney follow-ups to determine whether outcome measures (eg, BV/TV, periosteal BFR [p.BFR]) were significantly increased across groups and used Wilcoxon normal strains were calibrated against strain gauge data (mouse #1) and the derived BCs were applied to animal-specific FE models to predict strain gauge data in remaining specimens (mouse #2 and mouse #3). Upon calibration (rms error ¼ 16.9%), the animal-specific FE models were used to determine the peak tension (e T ) and compression (e C ) longitudinal normal strains induced at the cortical midshaft and for metaphyseal trabecular bone (dotted lines indicate 95% confidence intervals). (C) Exemplary strain distributions induced in the tibia-fibula structure, at the metaphyseal trabecular and midshaft cortical bone sections by À3.8 N dynamic loading. BC ¼ boundary condition; rms ¼ root mean squared. rank-sum tests to contrast outcomes in loaded versus contralateral bones. GRF alterations across time (day 0, day 22) and interactions with applied SPL were assessed via a nonparametric equivalent of the repeated measures AN-OVA. (31,32) For all tests, p 0.05 was considered statistically significant.

Results
The CV of induced strains did not significantly differ across the implemented SPLs (Fig. 3). The normal strains recorded at the gauge site, and peak tensile and compressive strains estimated in the trabecular and cortical compartment (via FEA) revealed a heterogenous normal strain environment (Fig. 4A, B). For the À3.8-N load implemented in vivo, peak longitudinal normal strains of À2490 AE 260 me were induced at the tibia midshaft and À1600 AE 110 me in metaphyseal trabecular bone (Fig. 4C). As quantified during in vivo loading, and as designed, SPL magnitude was significantly different across experimental groups (p < 0.01), whereas dynamic loading magnitude was statistically equivalent across experimental groups (p 0.01; Table 1). GRFs recorded at day 0 and day 22 in the experimental limb during gait were not differentially influenced across the SPL groups. Specifically, although GRF significantly declined between day 0 and day 22 (p < 0.05), the main effect of SPL and the interaction between days and SPL were not significantly different (p ! 0.5; Supporting Table 1).
Metaphyseal trabecular morphology of contralateral bones did not differ across SPL groups (p ! 0.11). Metaphyseal endocortical TV was significantly expanded in the dynamically loaded tibias of the À1.5-N SPL group versus contralateral tibias (4.6%, p < 0.05), but was not altered in other SPL groups (Fig. 5A). Trabecular BV in the À1.5-N SPL group was reduced in the dynamically loaded versus contralateral tibias (À9.2%, p < 0.05). In contrast, BV was significantly elevated when the same loading regimen was superimposed on a À0.03-N SPL (15.4% versus contralateral, 40.9% versus À0.5-N SPL, both p < 0.05; Fig. 5B). As a result, the significant reduction of BV/TV observed in tibias exposed to dynamic loading superimposed on the À1.5-N SPL group (À12.9% versus contralateral; p < 0.05) was reversed to an anabolic response when SPL was minimized to À0.03 N (12.3% versus contralateral, 46.3% versus À0.5-N SPL and 37.2% versus À1.5-N SPL-loaded tibias, all p < 0.05, Fig. 5C). The enhanced BV/ TV in the À0.3-N SPL group (versus contralateral) was associated with increased Tb.Th versus both larger SPLs (Table 2). In the metaphyseal region, total periosteal volume of the cortical shell was not altered in experimental versus contralateral bones, whereas cortical bone volume was significantly decreased in experimental versus contralateral bones in the À1.5-N SPL group (Table 2).

Discussion
Consistent with the literature, a À3.8-N loading regimen generating peak dynamic trabecular normal strains of 1600 me did not enhance trabecular bone morphology in the presence of either a À1.5-N or À0.5-N SPL. However, superimposing the identical loading regimen on a À0.03-N SPL significantly elevated trabecular BV/TV compared to contralateral tibias. As well, periosteal bone formation at the midshaft by the loading regimen superimposed on À1.5-N or À0.5-N SPLs was significantly amplified when SPL was reduced to À0.03 N. To our knowledge, there are no previous reports identifying SPL as a potential confounding variable in exploring how dynamic mechanical loading influences bone cell function, yet it is present in nearly all in vivo models of long-bone exogenous loading and is most often not even reported. Thus, our data identify a previously unrecognized inhibitor of dynamic loadinginduced bone anabolism that, given its near ubiquitous presence, is likely to have unintentionally confounded interpretation of preclinical skeletal adaptation studies.
We view our off-axis tibia compression model as an incremental advance in the in vivo application of loading to the mouse skeleton. (15,33,34) Its design draws directly upon the axial tibia compression device. (15,18) Contrasting our calibration data with strains induced by axial tibia compression indicates that the required exogenous load magnitude to achieve a given cortical peak strain is reduced by 37% in our model, whereas induced trabecular peak strains are 3% greater in axial compression than off-axis compression for a given load magnitude. (16,35) A potentially unique feature of our device is that it is capable of applying high-fidelity dynamic loading interventions using SPL magnitudes as small as À0.01 N. We believe that this minimal SPL (approximately 10% of peak GRFs during free ambulation) could be further reduced if our device were to be coupled with a highfidelity closed-loop controller. In contrast, À0.2 N (ie, 2Â gaitinduced GRF and 20Â our device minimum) is the smallest SPL ever reported for the axial tibia compression device. (15,19) Our lowest-magnitude SPL (À0.03 N) was only 6% of the most commonly implemented SPL in axial tibia compression studies (À0.5 N (36)(37)(38) ). Qualitatively, we believe that our device achieves stable dynamic loading with minimal SPL because it places (and loads) the lower limb in an anatomical orientation as during gait. In contrast, the axial tibia compression device requires increased knee flexion (<35 degrees (15,18) ) versus our device ($60 degrees). Although increased knee flexion appears to have required larger SPL to maintain limb stability, device modifications (eg, deeper fixtures) may also permit the axial compression device to stably load the tibia with smaller SPL. Finally, although we report calibration data showing that stable dynamic loading is possible with minimal SPL in our device, we have since confirmed that this observation is independent of waveform shape (eg, haversine, triangular) and/or frequency (data not shown).
We performed a combined strain gauge and FEA calibration to characterize the exogenous normal strain distribution induced by the off-axis device. The midshaft normal strain environment confirmed that the tibia was placed in compression and bending by the device. The orientation of the neutral axis ran from the anterior to posterior cortex and was rotated approximately 25 degrees from that induced by axial tibia compression, (16) and nearly 40 degrees from that estimated for running activity in BALB/c mice (Supporting Fig. 2). This normal strain distribution arose as the complex curvature of the tibia was subject to bending about multiple axes. Metaphyseal trabeculae were exposed primarily to compressive boundary conditions and were therefore exposed to smaller normal strain magnitudes compared to the cortical midshaft. Our calibration of the midshaft normal strain distribution emphasized that strains on the medial periosteal surface alone (the primary site of calibration strain gauge attachment (15,(39)(40)(41) ) vary profoundly (for our device at À3.8 N dynamic load, between 2000 me and À100 me). Experimentally, peak strain magnitude estimates are used as a means of equilibrating loading-induced stimuli across variables that would alter tissue deformations (eg, age, sex, genotype). However, given the profound magnitude gradient in this region, our data reinforce that attempting to equilibrate a given loading intervention across study variables via single strain gauge recordings is inaccurate and likely to substantially underestimate the maximum normal strains induced by the loading intervention. Despite recent identification of this concern, (35) review of the axial tibia compression literature demonstrates that this potential problem continues to be overlooked in many studies. For example, a recent study described using a À9-N loading regimen to induce 1200 me normal strain on the tibia midshaft of female 10-week-old C57 mice. (42) Although this strain value reflects tension on the portion of the medial midshaft cortex where the strain gauge is attached, it vastly underestimates peak strains induced by the protocol, which FEA models predict would reach À3300 me. (16,17) Relevant to our study, we have estimated that a À9-N dynamic load would induce peak tibia midshaft compressive strains >4000 me in 16-week-old female BALB/c mice, whereas caged BALB/c mice are unlikely to generate normal strains that exceed À660 me (Supporting Fig. 2).
The axial tibia compression literature provides a rich set of data to consider the impact of varied SPLs given the numerous combinations of loading magnitude and SPL magnitude that have been implemented. In general, the greater the magnitude of reported SPL, the greater the magnitude of loading that has been implemented in studies designed to induce an anabolic trabecular response. For example, the two studies prior to 2008 that did implement smaller SPL (À0.2 N) required less dynamic load magnitude to induce a profound anabolic trabecular response (>þ10% BV/TV) than later studies using larger SPLs. (15,19) Placed in the context of our data, the inhibitory influence of large SPLs (eg, À2 N) may underlie why very large dynamic loading magnitudes (approximately À16 N) have been used to induce substantial trabecular bone anabolism. (43,44) The rationale for using SPLs much larger than those used in the initial axial compression studies has not been addressed.
Given the near ubiquitous presence of SPL in exogenous skeletal loading models, (6,33,34,(45)(46)(47) but lack of reporting and/or standardization (including by our own group), we believe this variable may have unintentionally confounded experimental outcomes and interpretation in the bone adaptation literature. This type of unrecognized bias has been highlighted as problematic by NIH's own rigor and reproducibility initiative (https://www.nih.gov/research-training/rigor-reproducibility). We therefore believe that the potential impact of our observation extends beyond a technical issue with a particular in vivo model. First, the presence of an unintended inhibitor of bone mechanoresponsiveness in studies exploring bone adaptation to skeletal loading is likely to have confounded both broad interpretations (eg, the effect of aging on bone's response to loading) and to have biased assessment of the mechanisms underlying any perceived response (eg, mitigation of expected anabolism in a KO mouse). Second, to an unknown extent, SPL of the human lower limb occurs in a gravitational environment (eg, standing). Although speculative at this time, it is therefore possible that SPL has also confounded attempts to use exercise as a means of enhancing bone mass and morphology in humans.
We do not currently have experimental data that directly support a specific biophysical mechanism by which SPL inhibits bone anabolism. However, the literature suggests a number of candidates. A recent study demonstrated that a À2-N SPL, by itself, only modestly alters bone cell function (no cortical changes, no change in trabecular BV/TV (48) ), which suggests that the inhibitory mechanism of SPL is more likely associated with impairing early mechanotransduction. In this context, it is likely that SPL creates a pressurized state that minimizes the transient fluid flow that is normally induced by dynamic loading in the absence of SPL. (20,(49)(50)(51) Further, secondary aspects of altered fluid flow, such as hypoxia, are also known to mediate bone cell function. (52)(53)(54) Alternatively (or additionally), application of SPL prior to the initiation of dynamic loading (as occurs during in vivo studies) could lead to inadvertent cellular stimulation (eg, Ca 2þ oscillations), which may cause desensitization of and impair critical aspects of cellular signaling (eg, repeated Ca 2þ oscillations) to subsequent superimposed dynamic loading. (55)(56)(57) We believe that in vivo studies hold potential to clarify which of these pathways are most promising. For example, the literature suggests that redistribution of fluids following SPL onset (and reestablishment of the naive bone fluid state upon removal of SPL) would occur at faster timescales compared to the time course of cellular desensitization and subsequent recovery of sensitivity. (55,(58)(59)(60) If so, experiments designed to vary when SPL is initiated and removed prior to imposition of dynamic load could provide initial insights into whether physical (ie, fluid flow) and/or biochemical mechanisms (eg, Ca 2þ fluctuations) are promising candidates. At that point, complimentary experimental (eg, tracer transport, in situ Ca 2þ oscillations (61,62) ) and computational modeling studies (eg, of fluid flow dynamics (63) ) could be utilized to further clarify the mechanisms underlying the inhibitory impacts of SPL on dynamic loading-induced bone anabolism.
Our observations should be placed in context of the study limitations and the current lack of clarity regarding the mechanism(s) by which SPL inhibits the anabolic response of bone to loading. Although these results represent the first published data with our device, the inclusion of a À3.8-N dynamic loading regimen superimposed on À1.5-N SPL was intended as a negative control based on the axial tibia compression literature, which suggested that this intervention would not be anabolic for trabecular bone. (18,21,64) At this time, we do not know whether loads smaller than À3.8 N will be osteogenic in this model. When we designed this study, we considered a variety of potential strategies, but chose to control dynamic loading magnitude across protocols primarily because dynamic loading has long been considered a primary stimulus for bone anabolism. This choice did result in varying total load across groups (ie, maintaining constant dynamic loading [DL] amplitude while increasing the magnitude of SPL). However, if total load [TL] influenced adaptation in our study, we would anticipate observing the largest anabolic response with the greatest SPL (and thus the greatest TL). Instead, we found that the group with the greatest TL showed a reduced response to loading (including a catabolic trabecular response). Further, our data have not yet delineated a critical SPL magnitude threshold where inhibition begins, whether that threshold is dependent upon the magnitude/type of dynamic loading, or whether that relation is linear or nonlinear. Finally, we did not directly assess trabecular osteoblast function in this study because the label intervals were selected for cortical bone, not trabecular bone dynamics.
In conclusion, data from our study demonstrate that when SPL is sufficiently minimized, it is possible to enhance cortical and trabecular bone formation with a dynamic loading intervention that is otherwise ineffective when superimposed on SPLs that are currently implemented with the field standard trabecular adaptation model. This observation was serendipitously enabled by our new off-axis tibia loading device that is capable of applying stable dynamic loading to the tibia with SPLs as small as À0.01 N. We anticipate that future studies exploring the mechanism underlying SPL inhibition of dynamic loading-induced bone formation may hold potential to identify new strategies that may transform otherwise ineffective skeletal loading into a potent stimulus for enhanced bone morphology.

Disclosures
All authors state that they have no conflicts of interest.