The ability of bone tissue to perceive and respond to the functional strain and stress milieu is critical to the structural success of the skeleton. As first proposed in Wolff's law39, the skeleton rapidly accommodates increased functional demands by site-specific increases in bone mass, while the removal of function promotes resorption of bone tissue, diminishing the metabolic burden of an overbuilt skeleton. This adaptive capacity enables the skeleton to withstand the extremes of functional load-bearing yet optimizes this structure by the strategic removal of so-called unnecessary bone tissue. However, bone adaptation can also have grave consequences, as demonstrated by fractures of the hip and spine that are potentiated by senile or postmenopausal osteopenia29, resorption of the calcar following prosthetic replacement of the femoral head8, stress fractures in athletes5 and military recruits21, and microgravity-induced bone loss in astronauts37. In order to curb these mechanically mediated abnormalities, it is imperative that we improve our understanding of the means by which biophysical stimuli regulate bone morphology.
There is little argument, in the orthopaedic and biomechanical community, that mechanical signals are critical to retaining an adequate skeletal structure. However, controversy arises when it becomes necessary to define the specific parameters of the functional milieu that are osteoregulatory and those that are irrelevant byproducts of loading. Considering the extreme complexity of the functional strain environment of bone18, the cell populations on and within the bone tissue (osteoblasts, osteoclasts, lining cells, and osteocytes) are subject to a wide range of stress and strain stimuli, such as normal, shear, and principal stresses and strains; strain energy density; strain rate; and gradients of strain. Clearly, a real difficulty in identifying the components that govern the modeling and remodeling process stems from the vast number of candidates in the stress-strain history of a bone.
It has been proposed, on the basis of several empirical and experimental models, that bone-remodeling is controlled by a specific singular parameter within the mechanical milieu. These proposed parameters include, but are not limited to, the strain magnitude16,32, strain distribution22,30, strain tensor12, strain frequency24, strain gradients18, strain rate25, number of load cycles31, strain history4,9,10,11, and peak strain energy density20. In the case of strain magnitude, for example, the mass and morphology of a skeletal element would be determined by the amplitude of the peak strains to which the bone tissue is subjected, regardless of the number of load cycles or the rate at which these loads are applied. While these models have contributed to the structural and biological understanding of Wolff's law, it is striking that none of the investigators has proposed that different parameters of the strain milieu play separate and distinct roles in defining morphology; either the proposed mechanical factor is considered to be relevant to bone-remodeling or it is not.
An alternative to these all-or-none hypotheses is that bone may have the ability to differentiate between specific components of the stress or strain tensor and that net adaptation is defined by the interaction of these competing signals. For example, it has been proposed10 that mineralization of the cartilage anlage is controlled by the relative degree of deviatoric stress (causing change in shape) and dilatational stress (causing change in volume). These components would each have essentially opposite roles in the control of mineralization processes, with the dilatational stimulus actively inhibiting mineralization and the deviatoric component promoting it. While this hypothesis was offered to address skeletal morphogenesis specifically and not bone-remodeling per se, it is unique in its proposal that two different components of a complex state of stress have two distinct regulatory responsibilities. From a biological perspective, this suggests that the cell population can distinguish changes in cell shape (deviatoric) from changes in cell volume (dilatational) and that these distinct, real-time changes in cell morphology subsequently activate discrete mechanistic pathways that ultimately orchestrate two distinct physiological responses.
In the experiments reported on here, we used the turkey-ulna model of disuse osteopenia to determine if adult cortical bone can also differentiate between components of a complex mechanical signal. This was done by comparing the surface modeling and intracortical remodeling stimulated by three distinct mechanical regimens: torsional loading, axial loading, and disuse.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding sources were National Institutes of Health Grants AR-39278 and AR-40411.
†Read in part at the Annual Meeting of the Orthopaedic Research Society, New Orleans, Louisiana, February 21, 1994.
‡Musculo-Skeletal Research Laboratory, Department of Orthopaedics, State University of New York at Stony Brook, Stony Brook, New York 11794-8181. E-mail address: clint@bone.ortho.sunysb.edu for Dr. Rubin.
Operative Preparation of the Functionally Isolated Ulna
All operative and experimental procedures were reviewed and approved by the University Laboratory Animal User's Committee, and all met federal guidelines for the care and welfare of laboratory animals. Twenty-one adult male turkeys, ranging in age from thirteen to sixteen months and weighing between seventeen and twenty kilograms, were used in this study. These animals were selected because the epiphyses fuse at approximately nine months, the bone tissue normally undergoes haversian remodeling, and one ulna can be functionally isolated without compromising the daily activities of the turkey.
With use of aseptic conditions and halothane to induce general anesthesia, the left ulna of each animal was functionally isolated through transverse, parallel osteotomies at the proximal and distal metaphyseal-epiphyseal junctions33. Incisions, two centimeters in length, were made over the epiphyseal sites to expose the dorsal surface of the ulna. The osteotomies were performed with a u-shaped template clamped to the bone and, with use of an oscillating saw against the parallel planed surfaces of the template, a three-millimeter bone wafer was sectioned and removed. This procedure created a ten-centimeter diaphyseal preparation that was deprived of functional loads but that did not disturb the muscles or the nutrient or neural supplies of the diaphysis.
After the bone had been functionally isolated, two predrilled holes in the template, measuring ninety-two millimeters center to center, were used as guides to drill parallel dorsoventral holes through the bone. The template was removed, and stainless-steel caps were fit over the ends of the bone. When the predrilled holes of the caps were aligned with those just drilled in the bone, five milliliters of polymethylmethacrylate bone cement (Simplex P; Howmedica, Rutherford, New Jersey) was placed in each cap and the caps were pressed into position. Stainless-steel Steinmann pins, 100 by 4.7 millimeters, were introduced through the dorsal incision, piercing the caps, bone, and ventral muscles and emerging on the ventral surface of the wing. The wounds were sutured and dressed, and the ends of the pins were clamped with external fixators. The contralateral limb was left undisturbed to serve as a control for intra-animal comparison of modeling and remodeling activity. During the experimental period, all animals received food and water ad libitum.
Experimental Groups
The three treatment groups—axial loading, torsional loading, and disuse—were evaluated independently. In the first group, consisting of five animals, a sinusoidal eccentric axial load at 2.5 cycles per second was applied to the preparation to induce axial load and bending. Loads were applied to produce a peak compressive strain of 1000 microstrain, with a maximum strain rate of 15,000 microstrain per second. The turkey ulnae were loaded for 5000 cycles per day, five days per week, for a total of four weeks. The daily loading regimen required thirty-three minutes to complete. During the four-week period, each ulna was subjected to a total of 100,000 cycles of load.
Torsional load was applied to seven animals to produce a peak shear strain of 1000 microstrain. The number of loading cycles, the frequency at which they were applied, and the duration required to complete the loading were identical to those described for the group that had axial loading.
For the disuse model, the left ulna of six animals was functionally isolated and left unloaded for a four-week period.
Characterization of the Surface Strains Induced by the Axial and Torsional Loading Regimens
Both the axial and the torsional loading of the in vivo preparations was achieved with use of a specially designed, servohydraulic, closed-loop-feedback, computer-controlled actuator (Instron, Canton, Massachusetts). Three left ulnae, operatively prepared as described, were used to calibrate the loading regimen and to define the resultant stress and strain distributions. Three three-element rosette strain gauges (TML, Kenkyujo, Japan) with a length of two millimeters were attached subperiosteally at three surfaces on the plane of the middle of the ulnar shaft to allow beam-theory-based estimates of normal (tension or compression) longitudinal strain distributions within that cross section6,18. The nine channels of strain were conditioned through a bank of strain-gauge amplifiers (model 2120 system; Measurements Group, Instruments Division, Raleigh, North Carolina), sampled at 200 cycles per second through a twelve-bit analog-to-digital convertor (model DAS-16; Keithley Metrabyte, Taunton, Massachusetts), to a microcomputer (model 286; AST Research, Irvine, California). The strain-gauge-instrumented ulnae were placed in the axial and torsional frames for calibration of the load and torque signals.
For the axial loading condition, the two tines of one loading fork were connected to the proximal ulnar pins and, with the shaft of the fork acting as a 1:1 fulcrum, the handle of the loading fork was attached to the hydraulic actuator of the loading device. The tines of a second fork were fastened to the distal pins, with the handle again passing through a fulcrum and finally being connected to a load-cell, which supplied the axial load closed-loop-feedback error signal (Fig. 1).
A separate load-frame was used to apply torsion. Linear displacement of the actuator was geared to rotate one fork, the tines of which were connected to the proximal set of ulnar pins. The second fork was fastened to the distal pins; the shaft of the fork was instrumented with rosette strain-gauges, the signal output of which served the torsional feedback error signal (Fig. 1).
Distribution of Normal and Shear Strains and Stresses as Determined with a Finite Element Model
A finite element model was used to determine the resultant stress-strain fields in the ulnae that were subjected to axial or torsional loading. Microradiographic images of 0.5-millimeter-thick sections, obtained every five millimeters from one of the three calibration ulnae, were used to build the geometry of the finite element model. The stainless-steel loading pins were also included in the three-dimensional finite element model. After performance of a convergence study, a meshing density was chosen that offered sufficient resolution to define small variations in the stress-strain field of the ulna28. To validate the finite element model, the strains measured directly from the bone surface were compared with those calculated in the model, and the boundary conditions were iteratively adjusted until the predicted strains matched those that were measured.
The diaphysis of the ulna was modeled as a homogeneous, orthotropic continuum13. The material properties were measured with use of the non-destructive ultrasound technique3, in which propagation velocities were measured for cortical-bone specimens from the ulna. Young moduli of 7.8, 11.7, and 17.0 gigapascals were used in the model for the radial, circumferential, and longitudinal directions, respectively. Shear moduli of 2.5 gigapascals (G12), 3.28 gigapascals (G13), and 4.88 gigapascals (G23) were used28, with Poisson ratios of 0.327 (?12), 0.306 (?13), and 0.246 (?23). The stainless-steel loading pins were modeled with isotropic material properties (a modulus of elasticity of 200 gigapascals and a Poisson ratio of 0.3). Rigid contact between the pins and the bone was assumed.
The axial loading condition was modeled by imposing longitudinal (three-directional) displacements to nodes on the distal loading pins that made contact with the forks while constraining the contact nodes on the proximal pins. For torsional loading, the proximal pins were similarly fixed and tangential displacements were applied to the distal pins. Stress and strain distributions were calculated with use of finite-element-analysis software (ABAQUS, version 5.2; Hibbitt, Karlsson, and Sorensen, Providence, Rhode Island) on an IBM RS/6000 computer workstation.
Six components—stress, strain, principal stresses and strains, strain energy density, maximum shear stress and strain, and longitudinal shear stress and strain—were calculated for each element at the mid-diaphysis of the ulna (Table I). In validating the finite-element-analysis model, both the longitudinal normal strain (e33) and the shear strain (e23) generated by the model were compared with the values for strain as measured at the locations of the three gauges.
Areal Properties and Intracortical Turnover
At the end of the loading protocol, the animals were killed with an intravenous injection of a saturated barbiturate. The left experimental and right control ulnae were immediately removed, dissected to the periosteum, transversely bisected at the middle of the diaphysis, and fixed for forty-eight hours in 70 per cent ethyl alcohol. After fixation, a low-speed diamond-wheel saw (model 650; South Bay Technology, Temple City, California) was used to obtain 250-micrometer transverse sections from the mid-part of the diaphysis of both the control and the experimental specimens. The bone sections were then ground to 150 micrometers, and high-resolution contact microradiographs (Faxitron; Hewlett-Packard, McMinnville, Oregon) were made with use of single-emulsion film (Industrex R; Eastman Kodak, Rochester, New York), exposed for two minutes at 22.5-kilovolt peak and 2.8 milliamperes.
To determine the over-all areal properties of each section, the microradiographs were photographically enlarged (ten times magnification) and were scanned at 300 dots per inch (2.54 centimeters) with a flatbed scanner (HSD, Mountain View, California); this resulted in a final resolution of approximately forty by forty micrometers (0.0016 square millimeter, or approximately twice the size of a lacuna) per pixel. The eight-bit (256-shade) gray-scale images were thresholded to eliminate background, and pixel-counting routines (PV Wave; Visual Numerics, Boulder, Colorado) were used to quantify areal properties with employment of custom-written software on an IBM RS/6000 computer workstation. To minimize any subjective human error, the measurement software routines for edge detection and pore-counting were automated. Repeatability of the measurements was maximized by setting a threshold level equal to the half-power of the maximum intensity (that is, v2/2 x 255 = 180).
Image analysis of each bone section was performed without knowledge of either the experimental regimen or whether the section was from a control or an experimental ulna. The total bone area, total area of porosis, number of pores, and endosteal and periosteal envelopes were quantified. The size of the individual pores was also calculated. The cortical-bone area was calculated by subtracting the medullary area and the area of the intracortical pores from the periosteally enclosed area. Changes in the area during the experimental period were determined by comparing the areal properties measured in each experimental ulna with those observed in the intact, contralateral control bone. This comparison assumes that the areal properties of the right and left ulnae are similar at the beginning of the protocol, an assumption supported by the similarities observed in the mid-part of the diaphysis of the contralateral ulna in zero-time controls2,23.
Distribution of Normal Strain, Shear Strain, and Strain Energy Density Caused by Axial Loading Compared with That Caused by Torsional Loading
In the axial loading condition, differences of less than 6 per cent were evident between longitudinal normal strains measured in the ulna in vitro and those calculated with the finite element model at the locations of the three gauges (Table II). An axial load of 270 newtons in the finite-element-analysis model generated a nearly linear strain distribution with a peak longitudinal normal strain of approximately -920 microstrain in compression and 314 microstrain in tension (absolute mean, 337 microstrain across the mid-part of the diaphysis). The longitudinal shear strain around the cortex was low, ranging from -1.8 to sixty-two microstrain (mean, twenty-eight microstrain) (Table III). The maximum shear strain, located on planes approximately 45 degrees to the bone axis, was much higher, ranging from eleven to 741 microstrain (mean, 289 microstrain).
The finite-element-model calculations of shear strain magnitude at the gauge sites during torsion were within 3 per cent of the actual strains measured with the strain gauges. Although a finite element model was built from only one ulna, the pattern of strain (for example, the location of the neutral axis) and the strains at the gauge sites were very similar in the other two gauged ulnae, suggesting that the repeatability of the two loading regimens was very high. Under the calibration torsional loading of 2.4 newton-meters, the ulnae demonstrated a radially increasing shear strain distribution, with peak longitudinal shear strains along the periosteal surface ranging from 430 to 986 microstrain (mean, 662 microstrain across the cortex) (Table III). Because of the triangular morphology of the ulna, peak shear strain was located on the flat surfaces of the bone. Longitudinal normal strains in the mid-section were much lower, ranging from -71 to +72 microstrain, with an absolute mean of +21 microstrain.
Whereas the peak principal strains during axial and torsional loading were similar, the distributions and magnitudes of other components of the stress-strain environment were distinctly different (Table III). For example, under axial loading the distribution of strain energy density ranged from twelve to 6660 pascals; this reflected a more than 500-fold difference between the maximum and minimum values. The distribution of strain energy density in the ulnae that were subjected to torsion was more uniform, ranging from 1780 to 9110 pascals; this reflected a fivefold difference between the maximum and minimum values. The peak strain energy densities during axial and torsional loading differed by a factor of 1.4. The mean strain energy density was 1520 pascals under axial loading and 4300 pascals under torsion.
Modeling and Remodeling Response of the Ulna
The four-week protocol was completed for sixteen of the eighteen turkeys that had been used to monitor the remodeling. Two preparations in the group that had torsional loading loosened before four weeks, and these animals were not included in the analyses.
The areal analysis of the mid-part of the diaphysis focused on four parameters: the total bone area, the number of pores in the cortex that could be detected with the automated software, the total area of the pores in the cortex (in square millimeters), and the size of the individual pores (in square millimeters). Each of these four values was then compared with that measured in the intact, contralateral ulna. Quantification of the remodeling resulting from the loading regimens showed distinct responses for each condition (axial loading, torsional loading, and disuse) (Table IV).
As a baseline, the control ulnae from the sixteen turkeys for which the four-week loading protocol had been completed had a mean (and standard error) of 36 ± 8.5 pores; the mean area of bone lost because of intracortical porosis was 0.202 ± 0.062 square millimeters (Table IV and Fig. 2, top left). The mean size (and standard deviation) of the individual pores was 0.00561 ± 0.0029 square millimeter.
In the five animals that had axial loading (Fig. 2, top right), the increase in the total area compared with that in the controls averaged 6 ± 9.8 per cent; this difference was not significant, with the numbers available (Fig. 3-A). Both the mean number of pores (246 ± 40.5; Table IV and Fig. 3-B) and the mean area lost because of intracortical porosis (1.39 ± 0.252 square millimeters; Table IV and Fig. 3-C) increased significantly (by a factor of seven) compared with those measured in the controls (p < 0.05 for both). The mean size of the individual pores was 0.00565 ± 0.0019 square millimeter, almost identical to that of the controls (Fig. 3-D).
In the five animals that had torsional loading, the modeling response was essentially abolished and the remodeling response remained at control levels. The mean bone area did not increase substantially (a change of 1 ± 2.1 per cent) compared with that in the controls (Table IV and Fig. 2, bottom left). Furthermore, neither the mean number of pores (67 ± 22.6) nor the mean area lost to porosis (0.352 ± 0.114 square millimeter) was significantly different from those in the controls (Fig. 3-C). Finally, the mean size of the individual pores was 0.00525 ± 0.0035 square millimeter, almost identical to that of the controls.
In the six ulnae that were subjected to disuse, the mean area lost to porosis increased fivefold compared with that in the controls (1.05 ± 0.35 square millimeters), although, with the numbers available, the number of pores did not increase significantly (59 ± 22.4; Table IV and Fig. 2, bottom right). The mean size of the pores also increased significantly (p < 0.05) to 0.0178 ± 0.0126 square millimeter; this was three times the mean size in the control, torsional, and axial loading groups (Fig. 3-D). Most of the bone area was lost as a result of thinning of the cortex through endosteal resorption, decreasing 12.3 ± 1.9 per cent compared with that in the controls (p < 0.05). The ulnae that had been subjected to disuse showed the only significant changes in total areal properties.
In these experiments, a carefully controlled, highly characterized mechanical stimulus was applied to turkey ulnae that had been operatively isolated from function. The goal was to quantify the potential of three distinct regimens (axial loading, torsional loading, and disuse) to promote new-bone formation, inhibit resorption, and mediate intracortical turnover. In essence, these experiments are evaluations of the relative capacity of shear and normal strain to substitute effectively for a complex distribution of functionally induced mechanical stimuli that had, before their removal, maintained a remodeling equilibrium in the bone tissue. By examining the relative abilities of distinct mechanical stimuli (or their absence) to mediate modeling and remodeling activity, some indication of the mechanisms responsible for the control of bone turnover may be provided.
Axial loading of 1000 microstrain at one cycle per second for 100 cycles per day has been shown to maintain bone mass32. As it has been reported that torsional loading was not as osteogenic as axial loading27, the number of loading cycles per day in the current experiments was increased to 5000, to minimize concern that the cycle-number threshold under torsional loading may differ from that under axial loading. It was hoped that this would prevent the possibility that 100 cycles of torsion per day would not be sufficient to saturate the end-response. Furthermore, in vitro efforts to apply torque of more than 1000 microstrain (for example, 1500 microstrain) to the ulna caused the preparation to fail. Finally, while in most experiments in which this preparation was used the modeling and remodeling response was evaluated after eight weeks of loading (or disuse), under torsional loading the pin-cap portion of the preparation began to loosen at or beyond four weeks. Animals that had evidence of loosening of the pins were not included in the current study. We believe that this four-week protocol of 5000-cycle-per-day, 1000-microstrain torsional loading was at the extreme limit of the capacity of this preparation.
These experiments were performed on turkeys, in which the ulnae are subject to a minimum number of peak strain events in a given day2. It is important to consider the sensitivity of the bone tissue to the induced strain signals (or the lack of them) in light of the minimum strain to which this bone is normally exposed. Although the normal strain is small, removal of the bone causes disuse osteopenia. It is possible that bones that have a more vigorous functional role (that is, more structural responsibility) demonstrate a remodeling response that differs from that observed in the current study. However, when functional strain is considered during a full twenty-four-hour period and is characterized over its full spectral content, the strain history of the turkey ulna is remarkably similar to that of bones considered to have a more active functional role15. Bone tissue that is in remodeling equilibrium should be considered to have adapted to the confounding stimuli of the hormonal and mechanical milieus to which it is exposed. Because there was so little bone turnover in the control ulnae in the current study, and because this bone normally undergoes haversian remodeling, we believe that our data can be extrapolated to the adult human skeleton.
While the magnitudes of the axial load and torque inputs were carefully controlled (peak principal strains of approximately 1000 microstrain), the complex morphology of the bone precluded uniformity of any given component of the mechanical stimulus about the cortex. Therefore, these results can be viewed only as indicative of the relative role that each distinct mechanical factor plays. Furthermore, neither axial nor torsional loading creates a situation that is devoid of a given strain parameter. The axial loading condition, because of induced bending and the Poisson effect, does not extinguish shear strain, and the torsional loading condition still elicits normal strain. However, the distributions of normal and shear strains in the ulnae were quite different for the two loading protocols (Table II).
Absence of a Modeling Response
Neither the axial nor the torsional loading condition stimulated sufficient new-bone formation to result in a significant difference from the control values. Only disuse caused sufficient endosteal resorption and intracortical porosis to result in a significant difference in the total cross-sectional area. These results also indicate that relatively low amplitudes of peak mechanical stimuli (1000 microstrain), when induced at relatively low frequencies (2.5 cycles per second), were not sufficient to stimulate new-bone formation, suggesting that the mechanical signals did not pass a threshold required to initiate modeling (new-bone formation at the surfaces). Another interpretation of these data is that the torsional and axial loading conditions were equally successful in maintaining bone mass. The idea that torsion was as influential as axial loading in this respect conflicts with previous results, which suggested that shear strain was less effective than normal strain as a determinant of bone morphology27.
Although the maximum shear and principal strains were on the same order of magnitude for the two loading conditions, when the rather large differences in the distribution of the stress and strain components are considered it is striking that these distinct conditions were both effective substitutes for the background remodeling stimulus. The non-uniform distribution of strain signals did not change through the course of the experiment; for example, there was a fivefold difference in the distribution of strain energy density in the axial condition. This implies that bone homeostasis, to some degree, is dependent on spatial integration of information, perhaps achieved through intercellular communication as mediated by gap-junction interconnectivity35.
Differential Intracortical Response
While both the axial and the torsional loading condition caused bone mass to be retained, the means by which this was achieved were quite different. Axial loading caused intracortical resorption that was different from that in the controls, while torsional loading failed to increase either the area of intracortical porosis or the number of pores as compared with those parameters in the controls. It appears that 5000 cycles per day of this relatively low torsional loading regimen is sufficient to retain the status quo, both at the surface and within the cortex. The finding that axial loading allowed a substantial increase in the number of intracortical pores while torsional loading did not suggests that the cell population responsible for the control of remodeling (intracortical turnover) can differentiate between these two types of stimuli.
The peak magnitudes of the maximum shear strain and principal strains within the cortex were similar for axial and torsional loading, while normal longitudinal strain was high in axial loading and essentially nonexistent in torsion. This implies that intracortical turnover was actually stimulated by the presence of strain along the long axis of the ulna. Perhaps some parameter that is dependent on axial strains (for example, changes in cell volume), or on gradients in axial strains (for example, strain-induced fluid flow, fluid pressure, or related factors such as streaming potentials or flow-induced shear stresses), serves as a signal to enhance cell activity. Correspondingly, if a modeling or remodeling inhibitory signal is provided by shear strain, then even though fluid flow and strain gradients are minimum this mechanism is dependent on changes in cell shape rather than cell size. The fact that fluid flow and strain gradients were minimum in disuse as well but remodeling was not inhibited emphasizes that matrix strain per se, either normal or shear in nature (as opposed to strain and fluid flow), may be all that is necessary for the maintenance of bone mass. Such a hypothesis could be tested by causing fluid flow in bone in the absence of matrix strain (for example, by increasing blood pressure) and determining if both the number of pores and the net resorption of the endosteal surface increased. Furthermore, while substantial intracortical activity occurred in the axial loading condition, no significant surface formation (modeling) occurred under either axial or torsional loading. This suggests that the signals that control modeling (surface activity) and remodeling (intracortical activity and resorption) are distinct and not necessarily coupled.
Strain-Regulated Size and Number of Pores
The mean area of the intracortical pores was essentially identical in the control, torsional, and axial loading groups, but it tripled in the ulnae that were subject to disuse. The number of intracortical pores increased significantly (by a factor of seven; p < 0.05) in the axial loading condition, while neither disuse nor torsional loading caused any significant change from the controls. These data again suggest that the axial loading condition actively increased cell kinetics and that the axial strain component was responsible for the increase. This is in contrast to the condition of disuse, which passively increased the net loss of bone area; there was no axial strain component and no increase in cortical turnover. The fact that the area of intracortical porosis increased with disuse while the number of pores remained the same suggests that the removal of a mechanical stimulus caused the sites already involved in turnover to continue simply to resorb but not to couple to formative (bone-replacement) activity. Conversely, in the axial condition, even though the mean size of the pores was identical to that in the ulnae that had torsional loading and to that in the controls, the number of pores increased by a factor of seven. This finding implies that axial loading caused a dramatic increase in turnover activity while the formation and resorption mechanisms remained coupled. Thus, while axial strain increased turnover, shear strain was necessary to retain coupling. The absolute test of this hypothesis would be to load a bone such that pure normal strain was induced, to determine if coupling was retained. However, this goal is beyond the capacity of our in vivo model and, because more than 85 per cent of functionally induced strain is due to bending18,30, it is also not physiologically realistic.
The Osteocyte's Perception of the Strain Environment
These experiments suggest two critically discrete characteristics of bone-cell activity: that modeling (surface) and remodeling (intracortical) activity are not necessarily coupled, and that the osteocyte syncytium can differentiate between normal and shear strains and stresses within the tissue. The ability to distinguish between normal and shear signals may be due to several distinct mechanisms. For example, cells may be able to respond selectively to changes in their shape caused by shear strains as compared with changes in their volume caused by normal strains10. This hypothesis is supported by in vitro findings that bone cells can perceive both hydrostatic and shear stresses but respond differently to these mechanical signals7,14. Alternatively, other biophysical factors related to and derived from the strain environment may be responsible for cellular signaling. Dynamic loading of bone tissue not only results in a dynamic stress-strain environment but is also associated with other matrix-related events related to the fluid content within the porous space of bone tissue24,36. These interstitial flow phenomena are driven by gradients in tissue deformation17, which are significantly different under conditions of bending as compared with torsion. Thus, the observed differences between axial and torsional loading may be due to differences in fluid pressurization40, flow-induced shear stresses38, or strain-generated electric potentials26 in the microenvironment of the cells. Regardless of the origin of the stimulus, it is impressive that these cells, embedded in the matrix, are so sensitive to the mechanical milieu they not only can differentiate between strain components but also can regulate their activity on the basis of this stimulus.
In an attempt to summarize the results and consider them in light of the specific mechanical conditions that may control bone-remodeling, it is necessary to simplify what is undoubtedly a complex interaction between biophysical stimuli and the bone-cell population. In the case of disuse, it appears that the absence of matrix strain resulted in a decrease in bone mass, achieved through uncoupled resorption and formation. Resorptive activity per se did not increase (the number of pores was the same), but porotic infilling was suppressed. Torsional loading at a level to cause a principal strain of 994 microstrain and a peak strain energy density of 9110 pascals was sufficient to inhibit intracortical turnover and maintain bone mass. The matrix strain and cell deformation caused by torsion is predominantly distortional, minimizing interstitial fluid flow36 and changes in cell volume19. This suggests that shear strain, in the absence of normal strain, may have the capacity to inhibit cell activity, perhaps through the changes in cell shape that it may cause. Conversely, axial loading sufficient to cause 1000 microstrain and a peak strain energy density of 6660 pascals maintained bone mass, despite an increase in intracortical turnover. This loading condition resulted in significant axial and shear components of the loading environment, implying significant matrix strains as well as strain-induced fluid flow and gradients in interstitial fluid pressure. This suggests that axial strains increase cell activity, perhaps through the changes in cell volume that they may cause. Whether the retained coupling between formative and resorptive processes is also regulated by axial strain or whether it occurs because shear strain is also present could not be determined in these experiments.
In conclusion, the turkey-ulna model provides a unique means with which specific mechanical components can be evaluated. The results of the current study imply that shear strains inhibit bone resorption and retain coupling, axial strains promote cell activity, and disuse uncouples formation and resorption. The finding that axial loading conditions caused substantial intracortical porosis while torsional loading did not supports the hypothesis that cell populations responsible for the control of remodeling (intracortical turnover) are indeed able to differentiate between various (normal and shear) components of the strain environment. However, because of the distinct differences between the remodeling and modeling responses to axial loading and disuse, we believe that both of these components of the strain milieu play an active role in determining bone morphology. In the case of adult bone, we propose that axial strains increase activity while shear strains inhibit it and that both influence bone mass equally but in different ways. The delineation of a cellular mechanism that can realistically differentiate the surface from the cortex as well as normal from shear stresses will greatly increase the understanding of the interaction between mechanical loads and bone adaptation.