INJURIOUS MECHANICAL COMPRESSION
OF BOVINE ARTICULAR CARTILAGE
INDUCES CHONDROCYTE APOPTOSIS
Andreas M. Loening, Ian E. James⋆, Marc E. Levenston∗,
Alison M. Badger⋆, Eliot H. Frank, Mark E. Nuttall⋆,
Alan J. Grodzinsky†, and Michael W. Lark⋆
Continuum Electromechanics Group
Center for Biomedical Engineering
Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology
⋆Department of Bone and Cartilage Biology
SmithKline Beecham Pharmaceuticals
King of Prussia, Pennsylvania
∗Department of Mechanical Engineering
Georgia Institute of Technology
Running Title:Injurious Compression Induced Chondrocyte Apoptosis
† to whom correspondence should be addressed:
M.I.T. Room 38-377
Cambridge, MA 02139
Tel: (617) 253-4969
Fax: (617) 258-5239
Key Words: apoptosis, mechanical injury, chondrocyte, cartilage, matrix
Subject Area: Proteoglycans and Extracellular Matrices
A bovine cartilage explant system was used to evaluate the effects of injurious compression on
chondrocyte apoptosis and matrix biochemical and biomechanical properties within intact carti-
lage. Disks of newborn bovine articular cartilage were compressed in vitro to various peak stress
levels and chondrocyte apoptotic cell death, tissue biomechanical properties, tissue swelling, gly-
ocosaminoglycan loss and nitrite levels were quantified. Chondrocyte apoptosis occurred at peak
stresses as low as 4.5 MPa, and increased with peak stress in a dose dependent manner. This in-
crease in apoptosis was maximal by 24 hours after the termination of the loading protocol. At
high peak stresses (>20 MPa), greater than 50% of cells appeared to apoptose. When measured in
uniaxial confined compression, the equilibrium and dynamic stiffness of explants decreased with
the severity of injurious load, though this trend was not significant until 24 MPa peak stress. In
contrast, the equilibrium and dynamic stiffness measured in radially unconfined compression de-
creased significantly after threshold injurious stresses of 12 and 7 MPa, respectively. Together,
these results suggested that injurious compression caused a degradation of the collagen fibril net-
work in the 7-12 MPa range. Consistent with this hypothesis, injurious compression caused a
dose-dependent increase in tissue swelling, significant by 13 MPa peak stress. Glycosaminogly-
cans were also released from the cartilage in a dose-dependent manner, significant by 6 MPa peak
stress. Nitrite levels were significantly increased above controls at 20 MPa peak stress. Together,
these data suggest that injurious compression can stimulate cell death along with a range of biome-
chanical and biochemical alterations to the matrix and, possibly, chondrocyte nitric oxide expres-
sion. Interestingly, chondrocyte programmed cell death appears to take place at threshold stresses
lower than those required to stimulate cartilage matrix degradation and biomechanical changes.
While chondrocyte apoptosis may therefore be one of the earliest responses to tissue injury, it is
currently unclear whether this initial cellular response subsequently drives cartilage matrix degra-
dation and changes in the biomechanical properties of the tissue.
Apoptosis is a normal physiological process involved in immune regulation, the removal
of potentially carcinogenic and damaged cells, and during development as evidenced by the apop-
tosis of hyaline cartilage chondrocytes during endochondral ossification . Aberrant apoptosis,
however, can be pathogenic and has been observed in diseases such as Alzheimer’s  and in
neuronal cell death following spinal cord injury . Recent studies of human osteoarthritic artic-
ular cartilage [4, 5, 6], along with evidence from an animal model of osteoarthritis (OA)  and
cartilage wounding experiments , have suggested that aberrant apoptosis may play a role in the
pathogenesis of OA.
Hypocellularity has been associated with osteoarthritic cartilage and apparently normal
cartilage in joints affected with OA . This hypocellularity is believed to be both a risk factor and
a contributor to the disease pathogenesis, and the finding of markedly increased levels of apop-
totic chondrocytes in diseased tissue has implicated apoptotic cell-loss as a possible cause of OA
hypocellularity . The precise role of apoptosis in the pathogenesis of OA is currently unknown,
but mechanisms involving calcium precipitation  and matrix degradation  by apoptotic bod-
ies have been proposed. Further evidence of a role for aberrant apoptosis in articular cartilage
comes from correlations of tissue age and percentages of apoptotic chondrocytes found in normal
tissue of adult animals . Both animal and human cartilage exhibit age-related decreases in
cellularity, and it has been suggested that the inability of hypocellular tissue to maintain and re-
pair itself may contribute to age related degeneration . While this decrease in cellularity with
age may be related to apoptosis, a correlation between apoptotic chondrocytes and age in normal
human cartilage has not been observed .
The biological and physical stimuli that may induce chondrocyte apoptosis in articular car-
tilage are not well understood. Retinoic acid  and antibodies to the CD95 (or Fas) receptor 
have both been reported to induce chondrocyte apoptosis in vitro. Additionally, IL-1 induced nitric
oxide in combination with oxygen scavengers has the capacity to induce apoptotic cell death of
chondrocytes . However, none of these studies have focused on chondrocytic cell death within
intact cartilage matrix. Proinflammatory cytokines such as TNF-α have also been reported to stim-
ulate cell death ; however, it is currently unclear if chondrocyte cell death is directly controlled
by these cytokines.
Studies of chondrocyte-mediated matrix turnover in a model for mechanical injury of car-
tilage have shown that mechanical load can produce non-viable cell populations exhibiting con-
densed nuclei , reminiscent of apoptosis. Additionally, secondary osteoarthritis is commonly
associated with mechanical injury of articulating joints . Together, these two observations
led us to the hypothesis that injurious mechanical compression of articular cartilage may cause of
chondrocyte apoptosis. In the present study, we examined the effects of graded levels of applied
injurious compression on the induction of chondrocyte apoptosis in cartilage explants in vitro. Our
objectives were to quantify threshold levels of mechanical stress that could induce chondrocyte
apoptosis as well as to quantify and compare the effects of these injurious compressive stresses on
biochemical, biomechanical, and compositional measures of cartilage degradation.
Explant and Culture
Articular cartilage disks (3 mm diameter by 1 mm thick, ∼8 mg wet weight at time of
explant) were obtained from 1-2 week old calves as previously described . Briefly, 9 mm
diameter cylindrical disks of cartilage and underlying bone were cored from the femoropatellar
grooveand inserted intoa sampleholderof a sledgemicrotome. The first 100-400 µm of tissuewas
then removed to provide a flat surface, and the next two 1 mm thick cartilage slices were obtained.
From each of these slices, four cartilage disks (3 mm diameter) were cored and maintained in
culture medium ( low glucose DMEM with 10% FBS, 10 mM HEPES, 1 mM sodium pyruvate,
0.1 mM nonessential amino acids, an additional 0.4 mM proline, 20 µg/ml ascorbate, and 25 µg/ml
gentamicin)in a 37◦C, 5% CO2environment. Conditioned media was frozen for laterbiochemical
After one to seven days in culture, cartilage disks were placed into individual wells of the
base of a polysulfone compression chamber  with 0.5 ml media per well. The disks were
held between impermeable platens in the chamber base and the overlying chamber lid in uniax-
ial, radially unconfined compression. The compression chamber (base and lid) was then placed
inside a custom built incubator-housed compression apparatus  for application of injurious
compression. Anatomically matched free-swelling disks served as controls. The injurious com-
pression protocol consisted of six repeated on-off cycles of a displacement controlled ramp (in
strain), applied at a strain rate of 1000 µm/s to a final strain level of 30%-50% and maintained at
this strain level for 5 minutes, followed by a release of compression for 25 minutes (Fig. 1). Load
measurements were recorded during all experiments, and the peak stresses produced during the
compression (between 4 and 25 MPa depending on the chosen strain level) were used for compar-
isons between different experiments. Following compression, injuriously compressed disks and
matched controls were returned to free swelling culture.
Injuriouslycompressed disks and matched controls to be used forthe detection of apoptotic
nuclei were frozen by direct immersion in liquid nitrogen four days post-compression. Serial
cryostat sections (8 µM) were taken through the entire thickness of the disk (∼125 sections/disk).
The sections were then immobilized on glass slides, air dried, fixed, and stained for the presence of
apoptotic nuclei according to the manufacturer’s protocol (ApopTag peroxidase in situ apoptosis
detection kit, Oncor, Gaithersburg, MD). The peroxidase enzyme label used yielded an insoluble
brown stain on positive nuclei.
All ∼125 sections/disk were scored blind for the presence of apoptotic nuclei (score from
-1 to 3). Sections with positive staining nuclei at their periphery were considered negative (score
of -1, 0, or 1), as these apoptotic cells were artifacts of the cutting process . Sections considered
positive had apoptotic nuclei away from their edges and were scored according to whether there
were small (score of 2) or large numbers (score of 3) of positive nuclei. The final percentage of
positive sections (score 2 or 3) from each disk is reported.
for a field (60-120 nuclei/field) in the central region of 3-5 sections on each disk using the x40 ob-
jective of an Olympus Vanox microscope. The apoptotic nuclei were expressed as a percentage of
the total number of nuclei counted.
Sulfated glycosaminoglycans (sGAG) were quantitated in the conditioned medium using
the dimethylmethyleneblue (DMMB) dye method , with shark chondroitin sulfate as the stan-
dard. Briefly, 20 µl of medium was mixed with 200 µl of DMMB dye, 20 µl of 70% ethanol was
added to remove bubbles, and the absorbance at 520 nm was measured with a microplate reader.
Nitric oxide production in the conditioned medium was measured using the Griess reaction with
sodium nitrite as the standard . Media were centrifuged at 16,000 g for 1 minute to remove
debris. 100 µl of the supernatant was then mixed with 50 µl 1% sulfanilamide in 5% H3PO4and
50 µl 0.1% naphthylethylenediamine dihydrochloride, with 20 µl of 70% ethanol added to remove
bubbles, incubated at room temperature for 20 minutes, and measured for optical absorbance at
560 nm in a microplate reader. Wet weights were measured by patting each disk with sterile gauze
and weighing inside a preweighed vial.
Biomechanical Characterization of Injury
Immediately after injurious compression, loaded and control disks were placed into PBS
containing proteinase inhibitors (leupeptin, pepstatin, PMSF and either E64 or pefablock). The
biomechanical properties of each disk were then measured first in uniaxial confined compres-
sion and then in radially unconfined compression using a Dynastat mechanical spectrometer, as
previously described [23, 19]. For both testing modes, disks were first subjected to sequential in-
crements of strain (to a final strain of ∼ 10% based on initial cut thickness) using ramp-and-hold
displacements under displacement control. After stress relaxation following each ramp-and-hold
displacement, the measured equilibrium load was normalized to the disk area to obtain the equilib-
rium stress; the equilibrium modulus was then calculated by fitting a quadratic to the equilibrium
stress and strain values. At ∼ 10% static offset strain, a series of 5 µm amplitude sinusoidal dis-
placements was then applied to each disk in the frequency range 0.01-1 Hz, also in displacement
control . The dynamic stiffness at each frequency was calculated as the measured load normal-
ized to the specimen area and dynamic strain amplitude. The confined and unconfined equilibrium
modulusand dynamicstiffnessofinjuriouslycompresseddiskswere normalizedto thatofmatched
Control and experimental groups were compared using either Student’s paired or unpaired
t test assuming equal variances, with significance at the level p < 0.05.
increase in the percentage of apoptotic nuclei as detected by TUNEL staining (Figs. 2 & 3A). The
data in Fig. 3A are reported as the percentage of apoptotic nuclei observed as a function of the
peak compressive stress achieved during loading (e.g., Fig. 1). As has been previously reported
regarding the cutting of articular cartilage , numerous cells at the cut edge in both the loaded
and unloaded disks stained positive for apoptosis. In contrast, the central region of the loaded
disks showed a dramatic increase in the number of apoptotic nuclei (Fig. 2) compared to unloaded
controls, reaching ∼ 50% apoptotic nuclei at 20 MPa peak stress. Upon evaluation of all of the
sections taken from a disk ( ∼125 sections/disk), a three-fold increase in the number of sections
judged positive for apoptotic chondrocytes (score of 2 or 3) was seen at a peak compressive stress
of just 4.5 MPa (Fig. 3B), although the total percentage of apoptotic cells was low.
The kinetics of the apoptotic response was determined by examining the percentage of
cells that were apoptotic in the tissue immediately after (i.e., time 0), 1 day after, and 2 days after
a 20 MPa peak compressive stress loading condition (Fig. 4). While there is evidence of some
apoptosis immediately after loading, peak levels of apoptosis are clearly observed by 24 hours
after loading (Fig. 4).
Changes in Biomechanical Properties
a non-significant decrease with increasing injurious peak stress; there were no significant changes
versus controls for any loading condition (Fig. 5). The confined compression dynamic stiffness at
0.1 Hz also decreased with the severity of injurious load, though this trend was not significant until
24 MPa peak stress. In contrast, radially unconfined compression tests revealed significant reduc-
tions in the equilibrium modulus and dynamic stiffness of injuriously loaded tissue. Changes in
the equilibrium modulus became significant at 12 MPa and showed a 2.5 fold reduction at 24 MPa
peak applied stress. At 0.1 Hz, the loaded tissue exhibited a significantly reduced dynamic stiff-
ness by 7 MPa peak stress, with a nearly 7-fold reduction in stiffness by 24 MPa peak stress. The
dynamic stiffness computed at 1, 0.3, 0.03, and 0.01 Hz, showed trends similar to those reported
at 0.1 Hz for both confined and unconfined compression modes (data not shown). Together, these
results suggested that injurious compression caused a degradation of the collagen fibril network in
the 7-12 MPa range.
Changes in Biochemical Composition
Tissue wet weights were not significantly elevated compared to that of control for applied
stress levels at and below 8.5 MPa peak compressive stress (Fig. 6). At the 13 and 17 MPa com-
pression levels, the injured tissue swelled significantly compared to control tissue. This increase
arose entirely within the first day following the compression. The cumulativerelease of sGAG into
the medium after injurious compression displayed trends similar to the wet weight data (Fig. 7).
For the 13 and 17 MPa peak stress conditions, sGAG release rates were sharply increased during
thecompressionprotocol and remained significantlyelevatedfortwo tothree days before returning
to control values. In a separate experiment with a higher sample number, a small but significant
increase in the total sGAG released to the medium by day 4 was observed for a 6 MPa peak stress
loading condition (6 MPa: 10.1±.4, Control: 9.3±.4 µg sGAG/mg initial wet weight, n=12). Cu-
mulative nitric oxide released to the media was significantly increased for a 20 MPa peak stress
at four days post compression (20 MPa: .81±0.08, Control: .59 ±0.02 mM NO2/mg initial wet
weight, n=12), but 6 and 10 MPa peak stress conditions were unchanged compared to control
The results of the TUNEL staining demonstrate that mechanical compression can induce
articular cartilage chondrocytes to undergo apoptosis. A significant dose-related increase in apop-
tosis was observed at peak stresses as low as 4.5 MPa. While a small amount of apoptosis was
observed immediately after loading, there was a significant increase in apoptosis by 24 hours after
loading. This time lapse between the completion of the loading protocol and the emergence of an
apoptotic response suggests that there are biochemical changes taking place within the chondro-
cytes or tissue which eventually result in apoptotic cell death.
The biologicalpathway through which articularchondrocytes are induced to undergo apop-
tosis is currently unknown, but a variety of hypotheses have been proposed, including binding of
CD95 , elevated levels of NO [5, 15], and loss of extracellular matrix survival signals [6, 7].
This latter hypothesis is the most consistent with a mechanical origin for the initial apoptotic
signal, as a mechanical insult could possibly cause the chondrocytes to be separated from these
survival-promoting ECM signals.
Previous studies of bovine explants damaged by cutting  have shown that apoptotic
cells remained in the tissue and were detectable by TUNEL staining for at least 20 days following
injury. Additionally, there were very few empty lacunae in the sections evaluated for apoptotic
cells, suggesting that even after apoptosis the cell remnants remained within the tissue. Based on
these observations, the increase in TUNEL staining that we observed with increased peak stress
does not appear to be confounded by artifactual loss of apoptotic cells.
Interestingly, a significant increase of apoptotic cells was observed under loading condi-
tions (4.5 MPa peak applied stress) below or near the threshold stress levels required to produce
detectable changes in confined (24 MPa) and unconfined (7 MPa)biomechanical properties, sGAG
release (6-13 MPa), and tissue swelling in our system. These data suggest that a small percentage
of chondrocytes are sensitive to the low peak applied stress. As those cells begin to apoptose, they
may in turn degrade the surrounding matrix. This partially degraded matrix, in combination with
the increasing load, may further drive apoptosis and the dramatic matrix changes that are observed
at the high stresses examined in this study. Based on transgenic mouse studies, it has recently been
suggested that both MMP-3  and MMP-9  may play roles in controlling mesenchymal cell
apoptosis. In the loaded cartilage system it is possible that as the load is increased, MMP expres-
sion also increases , resulting in MMP-driven matrix degradation and further stimulation of
apoptosis. It will be interesting in this system to determine if MMP expression is in fact elevated
and if MMP inhibition can play any role in this process.
The biomechanicalcharacterization ofinjuriouslyloaded tissueshoweddistinctdifferences
tests emphasize the role of highly charged aggrecan molecules in resisting compression; uncon-
fined compression tests also emphasize the contribution of the collagen network tensile strength in
restraining tissue “bulging” that can occur at the disk periphery in the absence of radial confine-
ment. When measured in uniaxial confined compression, the equilibrium and dynamic stiffness
of explants decreased with the severity of injurious load, though this trend was not significant
until 24 MPa peak stress. In contrast, the equilibrium and dynamic stiffness measured in radi-
ally unconfined compression decreased significantly after threshold injurious stresses of 12 and
7 MPa, respectively. Together, these results suggest that injurious compression caused significant
degradationofthecollagennetwork at peak stress levelsin the7-12MParange. Since thesebiome-
chanical changes were detected immediately after injurious compression, it is possiblethat loading
caused direct damage to tissue matrix. However, it is not yet possible to rule out cell-mediated col-
lagenolytic activity, such as that associated with elevated MMP activity, based on these data alone.
Further studies are in progress to directly assess this possibility.
The results of tissueswelling and sGAG release measurements are also consistent with col-
lagen network damage arising from the more severe compression protocols. When tissue swelling
was observed, the increase in swelling relative to controls was most dramatic during the first day
following the compression. This swelling most likely occurred from the decreased ability of the
damaged collagen network to counteract proteoglycan-induced swelling pressure [24, 29]. The in-
creased rates of sGAG release, most dramatic during two to three days following the more severe
compressions, are also consistent with collagen network damage. A damaged collagen network
would be expected to have a greater effective pore size and hence increased proteoglycan diffusiv-
ity . Indeed, previous studies  suggest that this sustained release of sGAG after injurious
compressionin oursystemmaybe associatedwith increased release ofaggregatingspecies inaddi-
tion to a spectrum of degradation fragments found in controls. The additional role of cell-mediated
aggrecan degradation induced by injury is also the subject of further study in this system.
Our results indicate that threshold levels of mechanical stress are sufficient to induce chon-
drocyte apoptosis in articular cartilage, and suggest that injurious joint loading could cause chon-
drocyte death even in the absence of other observable biochemical or biomechanical changes to
the tissue. The increased biomechanical load could additionally stimulate chondroycte-mediated
matrix degradation and inhibit new matrix biosynthesis. Ultimately, apoptosis-mediated cell loss
could result in a significant reduction in cell number, leaving too few metabolically active cells to
repair the degraded matrix. Thus, a mechanical component may be included in hypotheses relat-
ing elevated chondrocyte apoptosis to the pathogenesis of OA. Conversely, OA-related fibrillation
of cartilage matrix may expose the chondrocytes to non-physiological levels of mechanical stress
and/or an unfavorably degraded pericellular matrix, again leading to apoptosis. In both cases, the
inductionofapoptosiscouldbe ascribed to mechanical compression, butin thelattercase apoptotic
chondrocytes would arise as a secondary result of OA. While chondrocyte apoptosis may therefore
be one of the earliest responses to tissue injury, it is currently unclear whether this initial cellu-
lar response subsequently drives cartilage matrix degradation and changes in the biomechanical
properties of the tissue.
This research was supported by a grant from SmithKline Beecham Pharmaceuticals, NIH
Grant AR33236, and an Arthritis Foundation Postdoctoral Fellowship (MEL). The authors thank
Elizabeth Lee-Rykaczewski, Catherine Healy, and Stephanie Soohoo for expert laboratory assis-
 Gibson, G. J., Kohler, W. J., and Schaffler, M. B. (1995) Dev. Dyn., 203, 468–476.
 Cotman, C. W. and J, A. A. (1995) Mol. Neurobio., 10, 19–45.
 Emery, E., Aldana, P., Bunge, M. B., Puckett, W., Srinivasan, A., Keane, R. W., Bethea, J.,
and Levi, A. D. (1998) J. Neurosurg., 89, 911–920.
 Kouri, J. B., Rosales-Encine, J. L., Chaudhuri, P. P., Luna, J., and Mena, R. (1997) Med. Sci.
Res., 25, 245–248.
 Blanco, F. J., Guitian, R., Vazquez-Martul, E., de Toro, F. J., and Galdo, F. (1998) Arthritis
Rheum., 41, 284–289.
 Hashimoto, S., Ochs, R. L., Komiya, S., and Lotz, M. (1998) Arthritis Rheum., 41, 1632–
 Hasimoto, S., Takahashi, K., Amiel, D., Coutts, R. D., and Lotz, M. (1998) Arthritis Rheum.,
 Walker, E. A. and Archer, C. W. (1998) Trans. Orthop. Res. Soc., 23, 502.
 Mankin, H. J., Dorman, H., Lippiello, L., and Zarins, A. (1971) J. Bone Joint Surg., 52,
 Hashimoto, S., Ochs, R. L., Rosen, F., Quach, F., McCabe, G., Solan, J., Seegmiller, J. E.,
Terkeltaub, R., and Lotz, M. (1998) Proc. Natl. Acad. Sci., 95, 3094–3099.
 Adams, C. S. and Horton, W. E. (1998) Anat. Rec., 250, 418–425.
 Meachim, G. and Collins, D. H. (1962) Ann. Rheum. Dis., 21, 45–50.
 Dietz, U. H. and Sandell, L. J. (1996) J. Biol. Chem., 271, 3311–3316.
 Hashimoto, S., Setareh, M., Ochs, R., and Lotz, M. (1997) Arthritis Rheum., 40, 1749–1755.
 Blanco, F. J., Ochs, R. L., Schwarz, H., and Lotz, M. (1995) Am. J. Pathol., 146, 75–85.
 Jilka, R. L., Weinstein, R. S., Bellido, T., Parfitt, A. M., and Manolagas, S. C. (1998) J. Bone
Min. Res., 13, 793–802.
 Quinn, T. M., Grodzinsky, A. J., Hunziker, E. B., and Sandy, J. D. (1998) J. Orthop. Res.,
 Davis, M. A., Ettinger, W. H., Neuhaus, J. M., Cho, S. A., and Hauck, W. W. (1989) Am.
J. Epidemiol., 130, 278–288.
 Sah, R. L., Kim, Y.-J., Doong, J. H., Grodzinsky, A. J., Plaas, A. H. K., and Sandy, J. D.
(1989) J. Orthop. Res., 7, 619–636.
 Frank, E. H., Jin, M., Loening, A. M., Levenston, M. E., and Grodzinsky, A. J. A versatile
shear and compression apparatus for mechanical stimulation of tissue culture explants. In
 Farndale, R. W., Buttle, D. J., and Barrett, A. J. (1986) Biochim. Biophys. Acta, 883, 173–
 Green, L. C., Wagner, D. A., Glogowski, J., Skipper, P. L., Wishnok, J. S., and Tannenbaum,
S. R. (1982) Anal. Biochem., 126, 131–138.
 Frank, E. H., Grodzinsky, A. J., Koob, T. J., and Eyre, D. R. (1987) J. Orthop. Res., 5,
 Bonassar, L. J., Frank, E. H., Murray, J. C., Paguio, C. G., Moore, V. L., Lark, M. W., Sandy,
J. D., Wu, J.-J., Eyre, D. R., and Grodzinsky, A. J. (1995) Arthritis Rheum., 38, 173–183.
 Tew, S. R., Kwan, A. P. L., Hann, A. C., Poole, A. R., Thomson, B., and Archer, C. W. (1998)
in Trans 2nd Symposium of the ICRS.
 Vu, T. H., Shipley, J. M., Bergers, G., Berger, J. E., Helms, J. A., Hanahan, D., Shapiro, S. D.,
Senior, R. M., and Werb, Z. (1998) Cell, 93, 411–422.
 Alexander, C. M., Howard, E. W., Bissell, M. J., and Werb, Z. (1996) J. Cell Biol., 135,
 Martin, J., Heiner, A., Brown, K., Schroder, A., Band, R., and Buckwalter, J. (1999) Trans. Download full-text
Orthop. Res. Soc., 24, 624.
 Maroudas, A. (1976) Nature, 260, 808–809.
 Jeffrey, J. E., Thomson, L. A., and Aspden, R. M. (1997) Biochim. Biophys. Acta, 1334,