Growing bones and joints are dynamic structures, transforming in dimensions, mass, and physical properties in response to altered mechanical forces and/or loading environments, a process referred to as adaptive or plastic phenotypic response (Gotthard and Nylin, 1995). While the modeling and remodeling capabilities of long bones have been extensively investigated (e.g., Lanyon and Rubin, 1985; Biewener et al., 1986; Biewener and Bertram, 1993; Judex and Zernicke, 2000; Hamrick et al., 2006; Robling et al., 2006), the adaptive postnatal responses of cartilage, the most fundamental tissue involved in skeletal and joint formation, and its influence on skeletal morphology remains poorly studied (Frost, 1979, 1999; Hamrick, 1999; Plochocki et al., 2009).
Chondral modeling is the adaptive growth response of cartilage via changes in shape, size, and composition to create a phenotype that is presumably better suited to altered mechanical environments during ontogeny (Frost, 1979, 1999; Hamrick, 1999; Plochocki et al., 2009). To maintain the functionality of a skeletal element or joint system, chondral modeling must facilitate normal joint and bone movements as well as minimize potentially damaging tissue contact stresses (Hamrick, 1999; Plochocki et al., 2009). Putative chondral modeling sites include the articular surfaces, the physeal cartilage, and sites of fascial, ligamentous or tendonous insertion (Frost, 1979). The chondral modeling response is posited to include regional or widespread cartilage thickening, changes in cartilage cellular and extracellular matrix (ECM) composition and organization, and potential for increased calcification and ossification. Chondral modeling may result in differential mineralization and ossification of the deepest hyaline cartilage layers (i.e., the calcified layer of articular cartilage, hypertrophic zone of growth plate). As such, it may directly contribute to the form and proportions of bones via influences on subchondral and diaphyseal bone.
Frost (1979) outlined the basis of chondral modeling within cartilaginous tissues using several observations. Similar to the case for modeling and remodeling of bony elements (Lanyon and Rubin, 1985; Biewener and Bertram, 1993), Frost concluded that a physiological loading range must exist to maximally stimulate regional cartilage growth. Cartilage growth is generally reduced under routinely high compressive loads yet enhanced under moderate forces, although the specific magnitude and frequency of such loads was unclear. Under Frost's model, negative feedback from unequal mechanical loads is responsible for chondral modeling. For example, high load-bearing joint cartilage will cease growth yet compensatory growth will occur in adjacent areas to more equally distribute the load.
The changes associated with chondral modeling, including tissue thickness, content, organization, and production rates of extracellular matrix components, are regulated by chondrocytes (Kiviranta et al., 1992). Furthermore, chondrocyte proliferation and metabolism, as well as the morphological variables under their control (e.g., ECM synthesis, proteoglycan production) are known to be influenced by mechanical loading (Eggli et al., 1988; Kiviranta et al., 1988; Urban, 1994; Wu and Chen, 2000; Liu et al., 2001; Carter and Wong, 2003; Ravosa et al., 2007, 2008a, b). Hamrick (1999) reevaluated the chondral modeling theory, identified the optimum range and frequency of hydrostatic pressure that stimulates chondral modeling during ontogeny, and proposed specific ways that cartilage will adaptively respond to moderate levels of mechanical stimuli. In order to produce uniform hydrostatic pressure throughout the tissue, Hamrick (1999) predicted that cartilage should respond to altered mechanical loads through differential chondrocyte division and cartilage matrix synthesis.
Despite Hamrick's extensive review of the theory underlying chondral modeling, no published studies have explicitly tested the model's adaptive response mechanisms in vivo. Thus, we tested certain predictions of the chondral modeling theory via measures for altered matrix and cellularity in a group of exercised and sedentary pigs. Based on modern chondral modeling theory, we expected to find increased ECM, increased viscoelasticity through elevated proteoglycan content, increased cellularity, increased average cell size, larger femoral dimensions, and flatter joints in the exercised group (Frost, 1979; Paukkonen et al., 1985; Eggli et al., 1988; Urban, 1994; Hamrick, 1999; Plochocki et al., 2006, 2009). Increases in chondrocyte proliferation indicate the availability for differential mineralization, increased bone growth, and an increased ability to alter the morphological variables of cartilage (Hunziker and Schenk, 1989). Although cell size is expected to vary throughout the depth of cartilage itself, an increase in average cell size in high-load areas has been suggested to represent either enhanced physical properties or a metabolic functional adaptation to loading (Paukkonen et al., 1985; Eggli et al., 1988; Freeman et al., 1994). Holding cell size and cell number constant, increased ECM and changes in the composition (e.g., proteoglycan content) will physically alter the cartilage's thickness and viscoelasticity, and therefore alter its ability to withstand loading. Here we characterize the effects of endurance running on femoral head growth plate, articular cartilage, and bone in relation to the predictions of chondral modeling theory. Such information is critical to a more complete understanding of the process of chondral modeling and the role of ontogenetic variation in mechanical loading on intra- and interspecific variation in joint and limb proportions.
MATERIALS AND METHODS
Procedures performed in this experiment were approved by the University of Missouri Animal Care and Use Committee under protocol 472–2. Fifteen castrated male juvenile miniature swine (Sus scrofa domesticus) were used in this study. The pigs were housed in contiguous plastic fence enclosures with concrete flooring, limiting physical but not visual and acoustic access to other pigs. The dimensions of the individual crates were 1.5 × 0.9 m (1.4 m2). The swine were supplied with water ad libitum and fed a high-fat diet provided once daily. All animals were fed the same amount of food, regardless of participation in the exercise regime or being sedentary. It should be noted that a high-fat diet has the potential to slow bone mineralization and cartilage regeneration (see Silberberg and Silberberg, 1950; Zernicke et al., 1995; Wohl et al., 1998), reducing the potential to induce and document major gross and histomorphometric changes.
Pigs are not skeletally mature until 5 to 6 years of age, with the femoral proximal growth plate remaining unfused until 3 years old (Barone, 1999; Dyce et al., 2002). The juvenile pigs began the protocol at 8 months of age and were sacrificed after seventeen weeks of participation in the experiment. The pigs were divided into two groups comprised of seven exercised and eight sedentary animals. The 15 pigs comprising the sample came from eight different litters, and brother pigs were divided between experimental groups as equally as possible. During the experimental period, exercised swine completed treadmill running to exertion limit 5 days a week, while the sedentary cohort was raised without exposure to exercise for the same seventeen week period.
The exercise training was done on electric motorized ClubTrack 3.0 PLUS treadmills (Quinton; Bothell, WA). Dynamic treadmill running consisted of four stages: warm-up (2.0–2.5 mph), a high-intensity sprint (4.0–7.0 mph), endurance running (3.0–5.0 mph), and cool-down (1.5–2.5 mph). A 5 min warm-up was followed by a 15 min sprint, a variable-length high-intensity endurance run, and was finished by a 5 min cool-down. The pigs were unable to maintain high-intensity speed and duration in the beginning of the experiment due to lower fitness levels and, thus, variation in duration in total work-out time and speed was necessary. Total endurance running times were variable, building from ∼45 min at the start of the protocol to 80 min in weeks 11–15. Average distance run built from 2 miles per day at the start of the protocol to over 5 miles per day by week 11 with a simultaneous increase in proportion of the run spent at high-intensity speeds (Table 1). It is believed that the pig joints experienced moderate to high levels of loading given the intensity of the experimental protocol.
Table 1. Mean exercise regime changes over 17 weeks for all 7 exercised pigs (± SD)
Mean sprint MPH
Mean sprint time (min.)
Mean endurance MPH
Mean endurance Time (min.)
Mean total distance (mi.)
The non-variable daily components of the exercise regime, warm up and cool down, averaged 2.4 mph/5 min and 2.3 mph/5 min.
Sacrifice week represents an incomplete exercise regime week.
All exercised pigs were exercised the day before sacrifice, and all pigs were weighed the day of sacrifice. Pigs were euthanized using a Telazol (5 mg/kg) and Xylazine (2.25 mg/kg) injection, followed by thiopental (25 mg/kg) to deep anesthesia. The secondary means of assuring death was through exsanguination and removal of the heart. Following sacrifice, the swine femora, including intact articular cartilage, were immediately dissected from the pelvis and surrounding tissues.
Thirteen femoral measurements (Table 2) were taken with digital calipers on the left femur (Fig. 1a), intact with cartilage, following Ruff (2002) (see footnote). Bones were fixed by immersion in 10% neutral buffered formalin. A DELTA band saw (DELTA; Jackson, TN) was used to remove the proximal end of the femur, which was then immersed in Surgipath Decalcifier II decalcification solution (Surgipath Medical Ind., Inc.; Richmond, IL) for 2 weeks. The left femoral head was divided coronally into 5 mm thick sections (Fig. 1b), photographed, dehydrated and paraffin imbedded for histological sectioning. Linear digital measures designed to quantify femoral head curvature were taken from images of the thin-sectioned femoral heads (Fig. 1c; Table 2). A ratio of (45° dorsal chord/45° ventral chord) captures the femoral head's departure from roundness, whereas the ratios (subchondral arc width/midpoint chord length) and (articular arc width/articular chord) show change influenced by epiphyseal bone shape and whole joint shape, respectively. Articular surface area was estimated based on Ruff's (2002) formula for a partial sphere 1.57*FHDP*(FHSI+FHAP). It should be noted that this surface area estimation is biased towards linear measures and not shape changes, which may influence articular area.
Table 2. Body size measurement means (±SD) of control (sedentary) and exercised pigs
Sedentary (N = 8)
Exercised (N = 7)
P-values were calculated from the sedentary and exercised individuals for each measure.
Significance P < 0.05. Percent difference is calculated as the absolute value of (exercised mean - sedentary mean) divided by the sedentary average. Less than 1% difference is considered approximately equal. Cartilage height, or thickness, is equal to layer area/300 μm. Cell area is approximated by the equation (0.5 h * 0.51 * π).
A Reichert-Jung 2040 Autocut microtome (Reichert-Jung, Inc.; West Germany) was used to obtain histological sections at 4 μm, which were then floated in a water bath, deparaffinized, and stained. Both groups were stained concurrently to reduce temporal variability in staining intensity. Hematoxylin and Eosin (H&E) staining was employed to identify cartilage zones for histomorphometric analyses. A second set of slides was stained with Safranin-O, a proteoglycan indicator which was used to qualitatively evaluate variation in proteoglycan content of the ECM. All things being equal, cartilage with higher amounts of proteoglycans (glycosaminoglycans - GAGs - bound to a protein core) typically has enhanced tissue stiffness and viscoelasticity (Jurvelin et al., 1986; Kiviranta et al., 1987; Tanaka et al., 2003).
Slides were analyzed with an Olympus BX41 microscope (Olympus Corp.; Tokyo, Japan). Eight sites on each H&E slide were assessed, four sites per physeal and four per articular cartilage (Fig. 1d). These sites were anatomically-determined and chosen by a single observer (ASH) for their repeatability on all slides. It should be noted that the epiphyseal plate exhibited stereotypical undulations (in coronal section) and the locations of measurement sites were selected based on equidistance between a characteristic inferior undulation (site VII in Fig. 1d) and growth plate articular cartilage. Digital images of all sites analyzed were saved in Tagged Image File Format (*.tiff) for histomorphometric analysis. All size values taken from the images were calibrated from a microscopy scale bar to pixels. A single observer (ASH) collected all measurements and was blinded to the experimental regime of each specimen during data collection.
NIH ImageJ 1.40 (National Institutes of Health; Bethesda, M.D.) was used to align the images so that the cartilage-bone interface was parallel to the horizontal axes. A standard longitudinal column 300 μm wide was selected from the center of each image. Within these standardized images, boundaries of the cartilage were delimited and defined as the superior cartilage border to the inferior edge of the “tidemark,” a boundary between the uncalcified and calcified cartilage that stains deeply with hematoxylin. Three layers (hypertrophic, proliferative, reserve) within the growth plate hyaline cartilage were manually delimited based on cell morphology (Niehoff et al., 2004). Cell morphology is an indicator for the behavior of the cells. The hypertrophic layer (HZ) was characterized by voluminous hypertrophying cells, where width approximated cell height, and included cells undergoing calcification and resorption. The proliferative layer (PZ) was defined by elongated, homogeneous cells whose width was typically twice cell height. The proliferative cells were further characterized by their distinct columnar configuration, oriented perpendicular to the horizontal axis of the growth plate. The reserve zone (RZ) was defined here by the start of the proliferative zone to the beginning of the epiphyseal bone and contained round stem-like chondrocytes. As in Fig. 2, the three layers of chondrocytes appear large and round in the hypertrophic layer, stacked and flat in the proliferative layer, and small and erratically-arranged in the reserve layer.
Unlike in the growth plate, chondrocyte morphology within the articular cartilage was not clearly identifiable. While raw thickness and cellularity were calculated, histomorphometric measures involving individual layers of the articular cartilage were not computed.
An average thickness was calculated for the articular cartilage and average thicknesses were calculated for each cartilage layer of the growth plate based on the formula “height” = area/300 μm. Cell counts were computed by individually numbering and counting all chondrocytes within the standard image frames. Cell counts were then scaled to the “height” of the cartilage (“cell count/height”). “Cell area” was calculated in the growth plate from cellular dimensions of 6 cells by the formula cell area = (0.5 h * 0.5 L * π). We did not use a randomizer to select cells in order to avoid oversampling cells from the same chondrocyte parent line. Instead, the cells used to compute cell area were selected throughout the image frame in order to ensure sampling from multiple cell lineages.
Between-group comparisons of linear metrics, ratios and cell counts were compared via a series of discriminant function analyses. Due to the small sample size, nonparametric ANOVA was used to assess variation in specific parameters (P < 0.05) between groups. The nonparametric ANOVAs also show directionality to facilitate the interpretation of discriminant function analyses between groups. Safranin-O staining intensity was evaluated qualitatively. Histological sample sites that were deemed unsuitable for inclusion in the analysis due to damage during histological preparation were not included in the statistical analyses. This is reflected in varying sample size values listed in variables used for histomorphometrics in Tables 3 and 4.
Table 3. Articular cartilage measurement means (±SD) of sedentary and exercised pigs
Number of individuals is indicated by superscript numbers. 1 = 8 individuals, 2 = 7 individuals, 3 = 6 individuals, 4 = 4 individuals, 5 = 3 individuals.
Cartilage thickness (‘height’)
I Total height raw (μm)
II Total height raw (μm)
III Total height raw (μm)
IV Total height raw (μm)
Cell count scaled to cartilage thickness
I Total cell count: total height
II Total cell count: total height
III Total cell count: total height
IV Total cell count: total height
Table 4. Growth plate measurement means (±SD) of sedentary and exercised pigs
Number of individuals is indicated by superscript numbers. 1 = 8 individuals, 2 = 7 individuals, 3 = 6 individuals, 4 = 4 individuals, 5 = 3 individuals.
Cartilage layer height
V height reserve zone (μm)
VI height reserve zone (μm)
VII height reserve zone (μm)
VIII height reserve zone (μm)
V height proliferative zone (μm)
VI height proliferative zone (μm)
VII height proliferative zone (μm)
VIII height proliferative zone (μm)
V height hypertrophic zone (μm)
VI height hypertrophic zone (μm)
VII height hypertrophic zone (μm)
VIII height hvpertroohic zone (μm)
Cell count scaled to layer height
V reserve cell count: height reserve zone
VI reserve cell count: height reserve zone
VII reserve cell count: height reserve zone
VIII reserve cell count: height reserve zone
V proliferative cell count: height proliferative zone
VI proliferative cell count: height proliferative zone
VII proliferative cell count: height proliferative zone
VIII proliferative cell count: height proliferative zone
V hypertrophic cell count: height hypertrophic zone
VI hypertrophic cell count: height hypertrophic zone
VII hypertrophic cell count: height hypertrophic zone
VIII hypertrophic cell count: height hypertrophic zone
Average cell area
V reserve cell area (μm2)
VI reserve cell area (μm2)
VII reserve cell area (μm2)
VIII reserve cell area (μm2)
V proliferative cell area (μm2)
VI proliferative cell area (μm2)
VII proliferative cell area (μm2)
VIII proliferative cell area (μm2)
V hypertrophic cell area (μm2)
VI hypertrophic cell area (μm2)
VII hypertrophic cell area (μm2)
VIII hypertrophic cell area (μm2)
No significant difference in body mass was found between the exercised and sedentary control groups at the start or conclusion of the experiment (Table 2). There was more variability in body mass in the sedentary sample, with both the largest and smallest weight values observed in the sedentary group. Nonetheless, due to the lack of mean differences, between-group variation in joint and limb form is more readily attributable to the exercise regimen than body mass.
Data for articular cartilage and growth plate cartilage are summarized in Tables 3 and 4, respectively. Tables 5 and 6 summarize discriminant function classification matrices. The articular cartilage zone thickness was not highly distinct between groups with only 67% classifying correctly in a discriminant function analysis (Table 3, 5). The exercised group demonstrated thinner growth plate cartilage zones, with 93% of the sample classifying correctly in a discriminant function analysis (Table 4, 6). This was confirmed by the univariate analyses, where the growth plate of exercised swine tended to be thinner, with reductions in average thickness localized in the proliferative zone.
Table 5. Articular cartilage discriminant function classification matrices
The canonical correlations are 0.638 for heights raw and 0.320 for total cell count scaled to total height.
Number of individuals is indicated by superscript numbers.
Total height (Sites I–IV)
Sedentary (N = 6)
Exercised (N = 3)
Total cell count scaled to total height (Sites I-IV)
Sedentary (N = 6)
Exercised (N = 3)
Table 6. Growth plate discriminant function classification matrices
The canonical correlations are 0.793 for zone heights raw, 0.819 for cell counts scaled by section height, and 0.996 for average cell area.
Number of individuals is indicated by superscript numbers.
Zone heights raw (Sites V–VIII; reserve, proliferative, hypertrophic)
Average cell area (Sites V–VIII; reserve, proliferative, hypertrophic)
Sedentary (N = 8)
Exercised (N = 6)
Cartilage ECM Composition
Micrographs of articular and physeal cartilage in the two groups are shown in Figs. 2a–d. High levels of safranin staining were observed in the physeal cartilage in the exercised and sedentary groups, with both groups appearing very similar in overall staining characteristics and matrix composition (Figs. 2c,d). Maximum positive staining occurred in the hypertrophic region in both groups, with moderate staining in all other regions. Negligible Safranin-O staining was present in the articular cartilage for both treatment groups.
Like articular cartilage thickness, articular cellularity is a poor discriminator as well, with only 44% classifying correctly (Table 3, 5). Articular cartilage cellularity tended to be higher in the exercised group, however, with all dorsally sampled areas (e.g., sites I-III, the lunate surface contact site) displaying an increased cellularity signal.
Measures of cellularity within each growth plate cartilage layer were accurate indicators of exercise treatment group (Table 4). Cellularity measures classified the samples correctly 86% of the time (Table 6). The proliferative and hypertrophic layers showed the most consistency, with three of four sample sites each having an average increase in cell density in the exercised group. The proliferative chondrocytes had a larger average cell area in the exercised group as well (also see Figs. 2e, f), with analyses of chondrocyte size correctly classifying all members (100%) of each locomotor treatment group.
Femoral Size and Shape
All members of the exercised and sedentary locomotor groups were correctly classified in a discriminant function analysis using the 13 femoral linear measurements (Table 7). In univariate comparisons, the exercised treatment group tended to exhibit larger femoral dimensions than the sedentary group in 8 of 13 measures (Table 2). While joint size did not vary between groups, the exercised cohort had relatively taller epiphyses which created more expansive dorsal subchondral and articular surfaces. The measure of epiphyseal curvature indicated by the ratio (45° dorsal chord/45° ventral chord) was significantly different between groups, with the smaller ratio indicative of dorsal flattening found in the exercised treatment group.
Table 7. Classification matrix for 13 femoral linear measurements
The canonical correlation value for the femoral measures is 0.895.
Number of individuals is indicated by superscript numbers.
Sedentary (N = 8)
Exercised (N = 7)
The primary function of chondral modeling is to maintain a morphology that maximizes the ability of bone and cartilage to resist dynamic mechanical loads while ensuring the overall functional integrity and congruence of the structures or joint system (Frost, 1979, 1999; Hamrick, 1999). Chondral modeling has been proposed to occur through differential chondrocyte mitosis and synthesis of the ECM, and this signal should be evidenced by (1) increased cartilage thickness and differences in extracellular composition, (2) increased chondrocyte proliferation and average cell size, and (3) differences in gross limb dimensions and shape. We only found support for differential physeal chondrocyte proliferation and altered morphology as well as bone growth, suggesting that chondral modeling theory may have understated key implications for adaptive chondrogenesis in bone growth. These findings and their considerations for postcranial bone and joint form are considered below.
Both indicators for ECM activity, cartilage thickness (“height”) and proteoglycan content via Safranin-O staining were not clearly different between groups in the articular cartilage. When considering the sample sites most directly affected by loading, however, we may see evidence of chondral modeling. As Frost (1979) predicted, articular cartilage thickness decreased in highly loaded dorsal sample sites (I-II) with a simultaneous increase in thickness adjacent to the high-load areas (III-IV). It is unclear whether these results indicate an adaptive ECM response to loading or that articular cartilage is a poor indicator for modeling. Other possibilities include that the loading regime was not within the optimum threshold for adaptive chondrogenesis or that our articular cartilage sample was comprised largely of adult chondrocytes less capable of ECM synthesis.
The exercised and sedentary articular cartilage had similar matrix composition (e.g., proteoglycan content), with a minimal Safranin-O staining intensity. Articular cartilage is known to display a reduction in proteoglycans under high intensity exercise regimes (Kivranta et al., 1992; Ravosa et al., 2007) or from connective tissue pathology (Archer, 1994; Ostergaard et al., 1999; LeRoux et al., 2001). These do not explain the poor proteoglycan content in the articular cartilage of this sample, however, as both the exercised and sedentary treatment groups displayed equally low GAG levels. We are investigating alternative theories for lack of proteoglycan content in the articular cartilage, including an examination of the collagen fiber orientation and other matrix constituents. Nonetheless, it is worth noting that Safranin-O histological pilot data from exercised and sedentary juvenile hypercholesterolemic miniature pigs has yielded identical results.
In the growth plate, the similar Safranin-O staining between groups qualitatively indicates similar matrix composition and presumably tissue viscoelasticity levels for both treatment groups. There was consistently high safranin staining throughout the growth plate in both groups, especially in the hypertrophic zone ECM. Strong hypertrophic staining does not indicate zonal enhanced viscoelasticity, however, and is a result of proteoglycan concentration in the hypertrophic zone due to the reduction in matrix volume and calcification of structures that accompanies normal cell hypertrophy (Alini et al., 1992). Further exploration of other matrix constituents (e.g., water, collagen, noncollagenous proteins) may implicate other tissue microstructures and/or their organization as part of the phenotypic response.
Growth plate cartilage layer thickness, an indicator of differential ECM synthesis, was negatively responsive to the loading regime. Contrary to expectations of chondral modeling theory (Hamrick, 1999), our study does not demonstrate an increase in ECM via cartilage thickness in loaded animals and, in fact, shows the opposite signal. It is possible that our exercise regime exceeded the optimum loading threshold to elicit ECM synthesis associated with chondral modeling and in fact inhibited the production of ECM, an avenue that should be explored with additional experimentation in mild to moderate exercise tasks. ECM synthesis was considered an indicator of chondral modeling for a variety of reasons, including known ECM decreases associated with cartilage degradation and/or chondrocyte deformation as a result of loading (Hamrick, 1999). Furthermore, chondrocytes increase their matrix production in the proliferating stage, thus it was expected that we would see increased ECM production downstream from this stage (e.g., the proliferative and hypertrophic zones). The sedentary group tended to be thicker and was unambiguously thicker in proliferative height, however. This suggests that ECM production in this region was either not mechanoresponsive or displayed cartilage degradation without repair. Either explanation questions the predicted role of ECM synthesis in chondral modeling.
Both the exercised articular cartilage and the growth plate cartilage displayed elevated cellularity levels when scaled to cartilage layer thickness, confirming that chondrocyte proliferation plays a key role in joint mechanobiology. The growth plate in the exercised group exhibited relative increases chondrocyte numbers in the proliferative and hypertrophic zones. Previous research has shown altered mechanical loading to increase proliferative chondrocytes (Wu and Chen, 2000), as well as increases in hypertrophic chondrocytes related to increased subchondral mineralization (Ravosa et al., 2007). This directly supports predictions of the chondral modeling theory regarding chondrocyte mitosis (Frost, 1999; Hamrick, 1999), although it should be noted that the research herein does not directly document an increase in chondrocyte mitosis. As the proliferative and hypertrophic layers of growth plate cartilage are responsible for mitosis and calcification of chondrocytes, respectively, increases in chondrocytes in these regions indicates an elevated number of cells available for mineralization. Cellularity increases in the growth plate, especially increased hypertrophic chondrocyte density, are typically reflected in limb elongation (Hunziker and Schenk, 1989).
An additional factor to consider is increased cell size in the proliferative zone of the growth plate. The relationship between cell size and altered loading, while poorly studied in physeal cartilage, has been examined thoroughly in articular cartilage (Paukkonen et al., 1985; Eggli et al., 1988; Freeman et al., 1994; Stokes et al., 2006). Chondrocytes and matrix regularly undergo deformation as a result of loading, particularly in the upper levels of articular cartilage (e.g., the superficial zone - Guilak, 2000; Carter and Wong, 2003; Grodzinsky et al., 2006). Our experimental model shows that the proliferative chondrocytes, in addition to increasing in density, are relatively larger in the exercised treatment group. The significance of this plasticity response in proliferative cell size is unclear. However, increases in articular chondrocyte size during loading have been attributed to altered physiological states, changes in intracellular composition, and changes in viscoelasticity, osmotic and hydrostatic pressure (Paukkonen et al., 1985; Eggli et al., 1988; Freeman et al., 1994; Guilak, 2000; Stokes et al., 2006). As the hypertrophic chondrocyte stage follows the proliferative stage, it may be logical to assume that the exercised hypertrophic cells would also be increased in size compared with the sedentary control cells. In fact, such hypertrophic chondrocytes were smaller than those in the sedentary group, which leaves one to speculate that this is due to an increased turnover rate in the hypertrophic cells of the exercise group.
Bone Dimensions and Joint Shape
Our results demonstrate that relative length and shape of postcranial elements in growing mammals is indeed differentially influenced by postnatal variation in loading behavior. This has been linked to an increase in physeal cellular proliferation and hypertrophy, which initiates a cascade of cellular and molecular events that are crucial for bone growth (e.g., apoptosis, resorption, ossification). Interestingly, while there is an apparent relationship between cartilage cellularity and bone measures, our findings show that the correspondence among loading, bone growth, and growth plate thickness are not necessarily complementary. Niehoff et al., (2004), who studied the effects of varying levels of exercise on distal femoral growth plates in rats, observed that growth plate height and proliferative zone height were lower in association with exercise yet the femur lacked length changes. Robling et al., (2001) found longitudinal bone growth and growth plate cartilage thickness uncoupled as well, although differing results for physeal cartilage thickness response to altered loading. Our results demonstrate that long bone growth and growth plate thickness do not necessarily reflect increased loading (Hunziker and Schenk, 1989; Robling et al., 2001; Niehoff et al., 2004; but see also Seinsheimer and Sledge, 1981). That is, elevated loading does not result in a correspondingly larger growth plate, and a larger growth plate is not essential for greater longitudinal bone growth.
The human femoral head is slightly nonspherical to maximize contact with the acetabulum during high loading and reduce vertical resultant forces (Radin, 1980; Afoke et al., 1984; Adams, 2006). Ratios reflective of changes in epiphyseal height and flattening of the dorsal loading surface characterized the exercised treatment group, which seemingly corresponds with predictions that joint surfaces should become flatter with increased loading (Latimer and Lovejoy, 1989; Plochocki et al., 2006, 2009). The shape changes in the exercised femoral heads are a direct result of taller epiphyses (i.e., the mean distance between growth plate and articular surface is larger) and a less-spherical femoral head (45° dorsal chord/45° ventral chord), likely due to a combination of bone modeling and adaptive chondrogenic activity. A taller epiphysis may create a more expansive dorsal loading surface or an increased range of motion (Eckstein et al., 1994, 1997; Steppacher et al., 2008). As the exercised pigs maintained their normal adducted and extended femoral position during loading, the relatively taller epiphyses may be functioning to create a larger loading surface rather than enhance mobility. If the pig femora mirror the biomechanical constraints of humans, less-spherical femoral heads will lessen forces transmitted through the hip by distributing joint loads normally across a larger joint surface during high loading (Afoke et al., 1984; Eckstein et al., 1994). Comparison of changes in the complementary lunate surface would be necessary to further evaluate this hypothesis. This suggestion would likewise benefit from experimental data regarding femoral head position during peak ground reaction forces.
Joints where flattening has been hypothesized to occur have largely been hinge joints, such as the tibiotalar joint or the knee joint, not the highly integrated rotational ball-and-socket type joint (e.g., Latimer and Lovejoy, 1989; Plochocki et al., 2009; but see Plochocki et al., 2006). Major flattening of the femoral head is unlikely to be phenotypically adaptive and is more commonly associated with pathologies such as femoracetabular impingement and hip dysplasia (Lequesne et al., 2004; Steppacher et al., 2008). In the condyles of the human distal femur, however, a flatter surface increases joint contact area and creates a larger surface for loads to pass normal to the joint during bipedal locomotion than a highly grooved or curved surface (Heiple and Lovejoy, 1971; Latimer et al., 1987; Organ and Ward, 2006; Sylvester and Organ, 2010). Interestingly, the femoral condyles appear to become both mediolaterally wider and superoinferiorly taller in the exercised pig group, showing changes in joint form in the distal femoral articular surface as well (Table 2). Thus, the postnatal plasticity response of joints due to altered loading may depend on joint type and mobility requirements, and include adaptive shape changes rather than global increases in size. Indeed, while we investigate our findings vis-à-vis chondral modeling theory, generalized bone plasticity responses should not be overlooked as a contributing factor to changes in skeletal morphology.
High adaptive plasticity responses in bones and cartilage, including altered joint shapes, can be achieved if stimulated during early growth (Frost, 1979, 1999; Robling et al., 2001; Niehoff et al., 2004; Ravosa et al., 2007, 2008a, b). The altered macro- and microanatomical variables produced here in a larger experimental model animal with an extended limb posture loaded with moderate to high loading largely corresponds with work done in small animals with habitually flexed hips. This suggests that there may be a pattern of adaptive changes in mammalian joint form despite inherent anatomical or postural differences (Robling et al., 2001; Niehoff et al., 2004; Plochocki et al., 2006). It would be interesting to conduct an interspecific comparison of effects of loading on joint shape, growth and, potentially, altered locomotion. Our experimental model, while unable to show clear altered locomotor adaptations from chondral modeling, did result in differing joint morphologies for a cohort of male pigs engaged in differing intensities of their normal locomotor activity. These results appear to suggest that adaptive chondrogenesis and bone plasticity during ontogeny is likely involved with intra- and interspecific variation in joint and bone dimensions in fossil mammals as well.
Despite the overall increase towards larger dimensions in the exercised group, joint size remained similar and a slightly smaller average articular surface area was found in the exercised group. Our results correspond well with the earlier findings of Lieberman et al., (2001), who found articular surface area to be conserved regardless of loading regime. As we failed to demonstrate a plastic phenotypic increase in hip joint size or articular surface area associated with loading, this has provocative implications for unproven associations between endurance exercise, its supposed anatomical correlates, and the evolution of Homo (see Bramble and Lieberman, 2004). If enlarged joint size or area is an anatomical correlate for this major behavioral adaptation, then it is more likely to have been a result of directional genetic changes, rather than phenotypically plastic response to altered behaviors during postnatal development.
Chondral modeling has been theorized to maintain joint congruence in altered loading environments by increasing cellularity and cartilage ECM production (Hamrick 1999; also Frost, 1979, 1999). Despite different joint shapes in the two experimental groups, ECM synthesis and cartilage viscoelasticity do not appear to increase in response to an exercise loading regime, showing that extracellular matrix synthesis and ECM proteoglycans may not be fundamental in chondral modeling processes. The articular cartilage demonstrated a poor response to mechanical loading, with minimal differences between treatment groups, and it appears that the articular cartilage itself can vary greatly in thickness and cell counts even within an experimental group. It is surprising that the joint morphology was altered with small histomorphometric differences in the articular cartilage, although it should be noted that the increased articular histomorphometric measures occurred in low load areas as Frost (1979) predicted. Given the lack of a perichondrium on the articular surface, bony articular surface shape changes are primarily due to articular cartilage modeling activities, although the bony remodeling activities of subchondral bone should also be considered during postnatal loading and adaptive chondral responses (see Rubin and Lanyon, 1984; Murray et al., 2001; Robling et al., 2006).
The growth plate was mechanoresponsive and showed that chondrocyte hypertrophy and proliferation are important processes in adaptive chondrogenesis and, potentially, in bone plasticity and growth. Overall, the growth plate appears more responsive to exercise-induced loading than articular cartilage, due likely to higher metabolic activities, increased vascular supply as well as differentially greater involvement in limb elongation. These findings may reflect the inherent nature of these two forms of hyaline cartilage (primarily limb development vs. joint function) as well as their innate responsiveness to mechanical stimuli (sensitive vs. conservative). A greater understanding of how the hierarchically organized structures of the proximal femur behave under different loading regimes and ultimately contribute to morphological variation may provide a better interpretation of locomotor behavior in living and fossil species.
While “chondral modeling” may be an appropriate description of the adaptive plasticity of cartilage related to altered joint function, it is unclear if the specific mechanisms, tissue/cellular responses, signal pathways, etc. previously linked to this hypothesis indeed apply equally to all types of cartilage, joint configurations, species, ages, and types of loading (Paukkonen et al., 1985; Kiviranta et al., 1987; Eggli et al., 1988; Kiviranta et al. 1992; Urban, 1994; Sibonga et al., 2000; LeRoux et al., 2001; Robling et al., 2001; Niehoff et al., 2004; Plochocki et al., 2006; Ravosa et al., 2007, 2008a, b). Experimental models have shown variable plasticity responses for cartilage thickness, cellularity, chondrocyte size, proteoglycan content, and skeletal correlates, although one thing remains clear: cartilage and the morphological parameters influenced by cartilage are in turn modulated by mechanical loading. Therefore, it may be more appropriate to consider chondral modeling a form of adaptive chondrogenesis owing to the role of postnatal chondral modeling in both adult cartilage and skeletal morphology.
Given the disparity between our findings and certain predictions, one should examine a variety of joint types from different species so as to better gauge the broader applicability of chondral modeling to notions about cartilage plasticity. Arguably, a long-term integrative perspective should be employed so as to more fully characterize the coordinated series of changes at the gross, cellular and molecular level that facilitate the adaptive process of chondral modeling (Ravosa et al., 2007, 2008a, b). Moreover, as plasticity responses decrease with age in a wide range of organisms (Hinton and McNamara, 1984; Meyer, 1987; Bouvier, 1988; Rubin et al., 1992; Ravosa et al., 2008b), appropriate controls should be employed in comparing developmental data across taxa. It would also be informative to examine chondrogenic response at sites of tendonous and ligamentous insertion and how it may contribute to skeletal morphology. The roles of subchondral modeling and remodeling are also integral to the shape changes associated with chondral modeling, yet subchondral and chondral modeling and remodeling have only been examined independently and without a unifying synthesis. Moreover, despite its known involvement in the formation of subchondral bone and articular cartilage, the role of adaptive chondrogenesis in osteoarthritis remains poorly studied (Arokoski et al., 2000; Aspden 2008).
The chondral modeling response, perhaps more accurately termed adaptive chondrogenesis, appears to be complex, site-specific, and highly variable even within hyaline cartilage, hinting at the importance of intrinsic cellular mechanisms that underlie the process. This study suggests that greater insight into adaptive chondrogenesis would profit considerably from research directed at understanding the nature of joint loads in vivo. In sum, our experimental analyses regarding joint plasticity in a high endurance environment have implications for the mechanobiology of limb growth and form, specifically in terms of identifying which connective tissue components are most responsive to exercise stimuli and thus potentially linked to normal and abnormal phenotypes.
Jason Organ, Valerie DeLeon, Timothy Smith, and Qian Wang are thanked for inviting us to contribute to their volume on experimental approaches to morphology. Sincere gratitude to the Laughlin Lab for providing the experimental specimens, especially Dave Harah and Dr. Harold M. Laughlin. Specimens were acquired with the assistance of NIH grant PO1-HL52490 to HML. Stephanie Child and Ian George are thanked as well as two anonymous reviewers. The Veterinary Medical Diagnostic Lab (VDML) assisted with certain histological methods. ASH was supported by a MU Life Sciences Fellowship and a Life Sciences Travel Award.
1FHAP is a measure of the anteroposterior femoral head maximum width, when the observer orients the bone vertically from a supporting surface. FHSI is a measure of the superoinferior femoral head maximum width, when the observer orients the bone vertically from a supporting surface. FHDP is a measure taken perpendicular to FHSI through the center of the head to its intersection with the lateral border of the articular surface, with the anterior (or cranial) surface facing the observer. All linear measure definitions from Ruff (2002).