subfailure damage in ligament: a structural and cellular...

Subfailure damage in ligament:a structural and cellular evaluation

PAOLO P. PROVENZANO,1 DENNIS HEISEY,2 KEI HAYASHI,3

RODERIC LAKES,4 AND RAY VANDERBY, JR.11Department of Biomedical Engineering and Department of Orthopedic Surgery and 2Department ofSurgery and Department of Biostatistics and Medical Informatics, 53792-3228; 3Department ofSurgical Sciences, School of Veterinary Medicine, 53706; and 4Department of Engineering Physicsand Department of Biomedical Engineering, University of Wisconsin, Madison, Wisconsin 53706-1687Received 17 July 2001; accepted in final form 14 August 2001

Provenzano, Paolo P., Dennis Heisey, Kei Hayashi,Roderic Lakes, and Ray Vanderby, Jr. Subfailure dam-age in ligament: a structural and cellular evaluation. J ApplPhysiol 92: 362–371, 2002.—Subfailure damage in ligamentswas evaluated macroscopically from a structural perspective(referring to the entire ligament as a structure) and micro-scopically from a cellular perspective. Freshly harvested ratmedial collateral ligaments (MCLs) were used as a model inex vivo experiments. Ligaments were preloaded with 0.1 N toestablish a consistent point of reference for length (andstrain) measurements. Ligament structural damage wascharacterized by nonrecoverable difference in tissue lengthafter a subfailure stretch. The tissue’s mechanical properties(via stress vs. strain curves measured from a preloaded state)after a single subfailure stretch were also evaluated (n 5 6pairs with a different stretch magnitude applied to eachstretched ligament). Regions containing necrotic cells wereused to characterize cellular damage after a single stretch. Itshould be noted that the number of damaged cells was notquantified and the difference between cellular area and areaof fluorescence is not known. Structural and cellular damagewere represented and compared as functions of subfailureMCL strains. Statistical analysis indicated that the onset ofstructural damage occurs at 5.14% strain (referenced from apreloaded length). Subfailure strains above the damagethreshold changed the shape of the MCL stress-strain curveby elongating the toe region (i.e., increasing laxity) as well asdecreasing the tangential modulus and ultimate stress. Cel-lular damage was induced at ligament strains significantlybelow the structural damage threshold. This cellular damageis likely to be part of the natural healing process in mildlysprained ligaments.

cell necrosis; fibroblast; medial collateral ligament; sprain;laxity

MICROTRAUMA OR SUBFAILURE INJURY in tendon and liga-ment may occur either as the result of overuse or as asingle traumatic event (5). Pathologies of the musculo-skeletal system, in which microtrauma is thought toplay a role, include tendinitis and tendinosis of theupper extremity (20), including tennis elbow (8) and

carpal tunnel syndrome (31), as well as lower limbpathologies such as Achilles tendinitis (8) and posteriortibial tendinitis (4), among others. Microtrauma dimin-ishes mechanical properties in knee ligaments (19) andis hypothesized to be a source of lower back painthrough microtrauma of either the iliolumbar ligament(35) or the intervertebral disk (42). Fruensgaard andJohannsen (10) reported that partial tears of the ante-rior cruciate ligament often lead to reduced levels ofactivity and performance. Although conservative treat-ment is often successful, tendon and ligament micro-trauma and partial tears may accumulate damage tothe point that load bearing is compromised and com-plete rupture or secondary damage occurs (12). Inaddition, ligament microtrauma may result in in-creased laxity, which in turn is associated with degen-erative joint disease and osteoarthritis (9). These stud-ies suggest the importance of studying subfailureinjury and intrinsic healing of ligaments.

A sprain is defined as an acute injury to a ligamentor joint capsule without dislocation. Sprains are clas-sified by severity on the basis of clinical examination orimaging. Grade I sprains are mild stretches with nodiscontinuity of the ligament and no clinically detect-able increase in joint laxity. Grade II sprains are mod-erate stretches in which some fibers are torn. Enoughfibers remain intact so that the damaged ligament hasnot failed. These grade II sprains produce detectableabnormal laxity at the joint compared with the unin-jured, contralateral side. Grade III sprains are severeand consist of a complete or nearly complete ligamentdisruption and result in significant joint laxity. Severesprains involving complete disruption of the ligamentand resulting in significant joint laxity (grade III) con-stitute ,15% of all ligament sprains (2). This leaves.85% of the sprains in which subfailure damage is thedominant issue.

Early studies of mechanical properties of tendonsand ligaments disclosed that irreversible mechanicalbehavior (and presumably damage) occurs at relatively

Address for reprint requests and other correspondence: R.Vanderby Jr., Dept. of Orthopedic Surgery, Orthopedic ResearchLaboratories, 600 Highland Ave., Univ. of Wisconsin, Madison, Mad-ison, WI 53792-3228 (E-mail: [email protected]).

The costs of publication of this article were defrayed in part by thepayment of page charges. The article must therefore be herebymarked ‘‘advertisement’’ in accordance with 18 U.S.C. Section 1734solely to indicate this fact.

J Appl Physiol92: 362–371, 2002.

8750-7587/02 $5.00 Copyright © 2002 the American Physiological Society http://www.jap.org362

low levels of strain [2.5–4.5% beyond preloaded refer-ence levels (32, 38)]. In one study of rabbit anteriorcruciate ligaments, subfailure injury (;80% of failurestretch) altered the shape of loading curves andthereby increased joint laxity (28). In another study,dramatic reductions in peak force were observed inhuman cadaveric inferior glenohumeral ligament afterrepeated subfailure loadings (39). Laws and Walton(23) used a sheep model to investigate a grade II injuryto the medial collateral ligament (MCL). They reportedan initial 13% reduction in the tensile strength and anincrease in laxity but full recovery at 6 wk. They alsoreported progressive healing by infiltrating fibroblastswith less of the typical inflammatory response seen incomplete failure models. The first time interval forhistological investigation was after 1 wk, so the earlyhealing response was not described. The injury was notwell controlled, and the molecular responses in theMCL were not described. Although changes in liga-ments have been observed at subfailure loadings, themechanisms and effects of damage accumulation arestill not known.

Cells in the body are subjected to complex mechani-cal loadings, consisting of tension, compression, shear,or combinations of the three types of load and defor-mation. Researchers have studied the mechanicalstimulation and behavior of cells in vitro (1, 11, 13, 14,40). Hsieh et al. (14) studied human fibroblasts fromthe anterior cruciate and medial collateral ligamentsin vitro under equibiaxial strains of 5 and 7.5%, show-ing that cell strain induced expression of types I and IIIcollagen. They suggested that remodeling of ligamenttissue may take place by a continuous microhealingprocess whereby scar tissue (predominately type IIIcollagen) is formed and later matures into remodeledtissue. Arnozczky et al. (3) showed that cell deforma-tion increased with tissue strain in situ but that whenthe relationship is nonlinear and wide, a variation wasobserved. None of these studies examined cell necrosisin situ or the mechanism by which mechanically in-duced cell necrosis occurs in soft tissue.

This study examines structural and cellular damagein ligaments after a subfailure damage injury. Sprainsare not currently well quantified in this subfailureregion. This study 1) quantifies the onset of structuraldamage in the medial collateral ligament as character-ized by nonrecoverable change in tissue length (frompreload) after a subfailure stretch, and 2) quantifiesregions of cellular damage (not the number of necroticcells) in ligament as a function of subfailure ligamentstrain. The tissue’s mechanical properties after a sub-failure stretch were also evaluated. It should be notedthat for this study structural damage is considerednonrecoverable changes in the entire ligament struc-ture and does not reflect examination of the extracel-lular matrix (ECM) (i.e., collagen fiber or fibril integ-rity) at the microstructural level.

METHODS

Experimental studies. This study was approved by theinstitutional animal use and care committee and meets Na-

tional Institutes of Health (NIH) guidelines for animal wel-fare. Sprague-Dawley male rats (weight 250 6 25 g) wereused as our animal model. Immediately after death, all ex-traneous tissue was carefully dissected to expose each MCL.Intact MCLs, including femoral and tibial bone segments,were carefully harvested with care taken not to disturb theligament insertion sites. The tissues were kept hydrated inHanks’ physiological solution (pH 5 7.4). Structural damage(n 5 25 ligaments) and cellular damage (n 5 24 ligaments)were evaluated in this study. Twelve additional ligaments(n 5 6 pairs) were used to evaluate changes in the ligaments’mechanical properties after a subfailure stretch.

To initiate a subfailure stretch in the MCL, tissues wereplaced into a custom-designed load frame with special struc-tures to hold the femur and tibial end sections of the samplein an anatomic position that loads the fibers as uniformly aspossible. Engineering strain (e) was measured by placinggraphite-impregnated silicon grease markers on the speci-men and using video dimensional analysis to measure dis-placement. The experimental system (load frame, camera,and image processor) has a resolution of at least 10 mm andrepeated measurements at a fixed length of optical markersare consistently reproducible to at least 10 mm. Force wasmeasured with a transducer having a resolution ;0.0025% ofthe maximum force. Data were acquired on a personal com-puter with Labtech Notebook (Laboratory Technologies, Wil-mington, MA) and recorded on video to synchronize force anddisplacement data. NIH Image (25) was used to capture andanalyze individual frames of the videotaped tests. The x-ycoordinate center (centroid coordinates) of each marker wasused to calculate the distance between the silicon markersand thus ligament displacement after loading. For the struc-tural damage group (n 5 25), the tissue was preconditioned(10 cycles at 1% strain) and allowed to recover for 10 min.After preconditioning and recovery (to reduce a viscoelasticcreep-recovery response), the gauge length (LO) was estab-lished after loading of the specimens with 0.1 N. Precondi-tioning and preload were included in the experimental pro-tocol to obtain a uniform zero point and to allow the sampleto settle into the testing apparatus. After preload, the tissuewas subjected to a subfailure stretch in displacement controlat a strain rate of 10%/s. Tissues that were subjected to zerostretch (0% controls) were harvested and hydrated as statedabove and placed in the load frame with markers in the samepreloaded fashion as the subfailure stretched ligaments withlength measured at each time point. The level of subfailurestretch not resulting in grade III ligament failure applied toeach ligament varied between 0 (no stretch) and stretch;12% based on preliminary studies in this laboratory. If agrade III ligament injury occurred (tissue failure displayingan abrupt drop in force), the ligament was excluded from thestudy. It should be noted that some grade III failures oc-curred at slightly lower strains than 12% whereas somefailed at larger strain; thus ;12% strain was the highestsubfailure stretch we were able to attain. After the subfailurestretch, the tissue was unloaded and allowed to recover for 10min [.3003 the length of loading during the test; at least103 the length of loading is recommended by Turner (37)], inan attempt to ensure that changes in length were not theresult of a viscoelastic creep-recovery response. Viscoelasticexperiments on rat MCLs subjected to low loads or displace-ments below 5% tissue strain performed in this laboratoryhave shown that a recovery time equal to 103 the test timeallows the rat MCL to return to the initial LO at preload (30).After tissue recovery, the length (LS) of the MCL was mea-sured at the preload level of 0.1 N. The difference in nonre-coverable length (laxity) is defined as structural damage for

363SUBFAILURE DAMAGE IN LIGAMENT

J Appl Physiol • VOL 92 • JANUARY 2002 • www.jap.org

this study. The length difference was normalized by theoriginal length (LO) and is expressed as a percentage: DS 5100 z [(LS 2 LO)/LO], where DS is a measure of structuraldamage. It should be noted that the symbol DS in this studydiffers from the use of the symbol D as the damage variablein continuum damage mechanics (6, 21).

To demonstrate the effect of a damage-inducing subfailurestretch on the mechanical properties of the ligament, stress-strain curves for six additional pairs of ligaments (n 5 12ligaments) were examined. Paired ligaments from each ani-mal were used with one MCL as a control, i.e., not receivinga subfailure stretch before being pulled to failure, and thecontralateral ligament receiving a subfailure stretch beforebeing pulled to failure. The assumption of symmetry in con-tralateral knee ligaments, used herein, has been validated(7). The cross-sectional area was calculated by measuring thewidth and thickness (10-mm resolution) of the ligament andassuming an elliptical cross section. The ligaments werepreconditioned at 10 cycles of 1% strain, allowed to recover,and stretched in displacement control at 10%/s. Each subfail-ure-stretched ligament received a different magnitude ofsubfailure stretch, ranging from 0 to 9%. The six ligamentsreceiving a subfailure stretch were loaded, allowed time torecover viscoelastic deformations as above, and then pulledto failure at 10%/s. The mechanical properties of the liga-ments that received subfailure stretches were compared withthe mechanical properties of the contralateral control liga-ment.

Tissues evaluated for cellular damage (n 5 24) werestretched in the same fashion as the structural damagegroup. Confocal microscopy (n 5 22 ligaments) with a cellviability assay detected live and necrotic cells (mainly fibro-blasts in ligament). Tissues were prepared for confocal mi-croscopy immediately after tissue stretch, and the time fromstretch to viewing of the tissues was ;1 h, during whichstaining was performed. By use of a technique similar to thatused in Ohlendorf et al. (27), in situ staining was done withcalcein and ethidium homodimer, whereby live cells metab-olize calcein and show green fluorescence and membranes ofnecrotic cells are penetrated by ethidium homodimer, whichresults in red fluorescent staining of the nuclei. Image scansthrough the depth of the tissue were compiled and overlaid.Yellow areas on the images are a summation of both live andnecrotic cells and were counted as regions of cellular damage.Black sections indicated that no stained cells were present inthat region. Tissues that were subjected to zero stretch wereharvested from the rat, as stated above, and placed in theload frame with markers in the same fashion as the subfail-ure stretched ligaments, preloaded, and then processed as0% strain baselines. Images of the ligaments from confocalmicroscopy were stored digitally, and regions containing ne-crotic cells were quantified with NIH Image (25). The numberof necrotic cells was not counted and is not quantified in thisstudy because the relationship between fluorescence area(present in the images) and cell area is not known and scansare overlaid. The colors from red to yellow to green that arepresent in the ligament were ordered and assigned an index.A constant threshold was set as the boundary between yellow(regions with some necrotic cells) and green (regions wherecells had intact membranes). The number of pixels of greenwas used to quantify the regions of the ligament with viablecells, and the yellow and red pixels were used to quantify theregions of the ligament with necrotic cells. The backgroundindex (black in color) of the image was identified and pixels ofthat index were eliminated, so that the nonred, nonyellow,and nongreen pixels were not counted. The area of the liga-ment was measured in units of pixels and used to normalize

the regions containing necrotic cells. Hence, the measure ofcellular damage (DC) is a nondimensional unit representedas a percentage of the form DC 5 100 z (regions containingnecrotic cells/tissue area) (pixels/pixels).

In addition, transmission electron microscopy (TEM) (n 52 ligaments) was performed on control (e 5 0%) and stretched(e 5 3.2%) ligament to view the cells. The specimens werefixed in modified Karnovsky’s solution, postfixed in 1% os-mium tetroxide, and stained with 1% uranyl. After sequen-tial dehydration in ethanol and infiltration in Epon-Aralditeand propylene oxide, specimens were embedded in 100%Epon-Araldite and polymerized at 60°C. Ultrathin (70 nm)sections were cut, placed on grids, stained with lead citrate,and viewed by using a transmission electron microscope.

Statistical analysis. Statistical analysis was performed todetermine the strain at which the onset of damage occursfrom structural and cellular standpoints and to determinewhether the two were different from one another. An obser-vation of cellular damage (cellular damage per ligament) wasdefined as Dci. The general expression Dci ; N(mci, sc

2) indi-cates that Dci was expected to have the mean value mci, anormal distribution about this mean value with variance sc

2.The mean value was modeled as a function of strain eci as

mci 5 bc~εci 2 uc!Ic@εci# (1)

with

Ic~εci! 5 0 for εci , uc

Ici~εci! 5 1 for εci $ uc

where uc was the strain threshold below which cellular dam-age is zero and above which cellular damage increases lin-early with strain with slope bc. The kink at uc was obtainedby the indicator function Ic(eci), which equals 0 when eci , uc

and equals 1 otherwise. A similar development was appliedfor structural damage Dsi, modeling the mean value (msi) as afunction of strain (esi), hence

msi 5 bs~εsi 2 us!Is~εsi! (2)

with

Is@εsi# 5 0 for εsi , us

Is@εsi# 5 1 for εsi $ us

To obtain the “best” estimates of the parameters bc, uc, bs,and us (estimates of the nuisance parameters sc and ss werealso needed), maximum likelihood estimates (MLEs) wereused (e.g., Ref. 24). MLEs are those values of the parametersthat maximize the probability of observed data. Thus, toobtain MLEs, the probability density function of the observeddata was constructed; this probability density function isreferred to as the likelihood equation. For mathematicalconvenience, rather than maximizing the likelihood directly,the log likelihood was maximized (the position of maxima areinvariant to log transformation).

For the nc cellular damage samples, the observations wereassumed to be independent and normally distributed. Hence,the log likelihood for the cellular data was expressed as

Lc 5 2~nc / 2! ln~2psc2! 2 Sc

2/2sc2) (3)

where Sc2 represented the cellular damage sum of squares.

This defined the sum of squared errors about the meanvalues, or

Sc2 5 (

i 5 1

nc

~yci 2 mc!2

364 SUBFAILURE DAMAGE IN LIGAMENT


or more specifically,

Sc2 5 (

i 5 1

nc

~yci 2 bc~xci 2 uc!Ic@xci#!2 (4)

A similar expression for the log likelihood of structural dam-age (Ls) was written. Because of the additive nature of loglikelihoods, the log likelihood for all of the observed data,cellular and structural combined, was defined as L 5 Lc 1 Ls.Then L was maximized with respect to the six parameters(bc, uc, bs, us, sc, ss). This was done by taking the first partialderivatives, setting them to 0, and solving for the parame-ters. This was conveniently done in two steps. The first stepwas to maximize L just with respect to sc and ss. Because thesolution to ]L/]sc 5 0 is sc

2 5 Sc2/nc, sc was eliminated from Lc

by substituting s*c2, yielding

L*c 5 2~nc / 2!@ln~2ps*c2! 2 1# (5)

the so-called cellular profile log likelihood. The cellular pro-file log likelihood (i.e., the cellular log likelihood maximizedwith respect to sc) became a function of bc and uc. Thusmaximizing the total profile log likelihood L* 5 Lc

* 1 Ls* with

respect to just the four parameters (bc, uc, bs, us) was equiv-alent to maximizing the total log likelihood with respect to allsix. For any fixed uc and us, L* was maximized by individuallyfinding the least-squares estimates of bc and bs, which indi-vidually maximized the components Lc

* and Ls*.

A goal of this study was to determine whether the thresh-old for cellular damage was different from that for structuraldamage. That is, a difference (i.e., rejecting the null hypoth-esis H0: uc 5 us in favor of the alternative hypothesis HA: ucÞus) was sought with a likelihood ratio test. The maximum loglikelihood obtained under the alternative model was definedas LA; this was simply the maximum of the log likelihooddescribed above with all six parameters free to vary. Themaximum log likelihood obtained under the constrained nullhypothesis model where uc 5 us was denoted as L0. Then,x2 5 2(LA 2 L0) had a x2 distribution with one degree offreedom under the null distribution.

A line search was used to find the value of u that maxi-mized the null hypothesis log likelihood, L0. To start, u wasset to 0. Then, conditional on this u, the values of bc and bs

that maximize L* were found by least squares. Then u wasincremented by a small amount, and the process was re-peated. L0 was the largest value of L* observed, and theMLEs under the null hypothesis were the correspondingestimates for bc, bs, u, sc, and ss. A similar process was usedto maximize the alternative hypothesis log likelihood to ob-tain LA. Because of the additive nature of the log likelihood,Lc

* and Ls* were maximized separately by individual line

searches.Confidence intervals about the parameter estimates were

constructed with the profile likelihood method. Let Lp be L*maximized for all parameters but us, a so-called profile like-lihood. Ninety-five percent profile likelihood confidence inter-vals for us were constructed by finding the values of us thatsatisfy the equality 3.84 5 2(LA 2 Lp), where 3.84 is the 95thpercentile of the x2 distribution.

RESULTS

Results indicate that tissue stretch-induced struc-tural damage and cellular damage are fundamentallydifferent in their behaviors when analyzed as a func-tion of applied strain. Qualitatively, regions of cellulardamage can be seen to increase with strain by exami-

nation of the confocal microscopic images (Fig. 1). Inaddition, the strain produced extensive damage to in-dividual cells as seen in TEM images at 3.2% ligamentstrain (Fig. 2). These images show a typical cell forwhich the membrane is intact and the cellular contents

Fig. 1. Rat medial collateral ligaments at strains of 0 (A), 6 (B), and11% (C). Red and yellow regions indicate cell necrosis, and greenregions are healthy live cells. Note the extent of necrosis in thestrained ligaments and the increase in necrosis with higher strains.



are undisturbed in a ligament with 0% strain. A typicalmechanically damaged cell in the strained ligamentdisplays a cell membrane that has been ruptured andthe cellular contents that have been released into the

ECM. Data for structural damage (DS) and cellulardamage (DC) seen in Figs. 3 and 4, respectively, clearlyshow the different strain-dependent behaviors forstructural and cellular damage. Statistically, the onset

Fig. 2. Fibroblasts in a normal unstrained liga-ment (left) and fibroblasts in a strained (e 5 3.2%)ligament (right). Normal ligaments have intactcellular membranes in a longitudinal section at24,0003 (top left), transverse section at 9,1003(middle left), and longitudinal section at 24,0003(bottom left). Ligament that has been strained to3.2% (right) shows cellular damage in the form ofruptured cell membranes and the release of cellu-lar contents in the extracellular matrix in a longi-tudinal section at 24,0003 (top right), transversesection at 9,1003 (middle right), and longitudinalsection at 24,0003 (bottom right). Note interdigi-tation of collagen fibrils with normal ligaments inleft column.



of structural damage (us) was found to be at a strain of5.14% from preload, and this level of strain can be seento be well within the linear region of the stress-straincurve for the rat ligaments used in this study (Figs. 5).The 95% profile likelihood confidence intervals forstructural damage are 4.50–5.69. The onset of changesin cellular damage (uc) was found to be 0 (e 5 0%). Thatis, statistically, changes in cellular damage in ratMCLs begin with the application of ligament strainfrom preload, and structural damage occurs at strains.5.14% from preload. Some regions of cellular damagecan be seen at 0% strain (i.e., preloaded treatment), yetstatistical analysis could not reveal any other onsetbecause of scatter in the data. No consistent localiza-tion of regions of cellular damage could be identified inthe ligaments. The slope values (b) for structural andcellular damage are 1.42 and 4.00, respectively. Fromthese values of u and b, Eqs. 1 and 2 are

mci 5 4.00~εci 2 0.00!Ic~εci! (1)

msi 5 1.42~εsi 2 5.14!Is~εsi! (2)

with the indicator function I(ei) equal to 0 when ei , uand equal to 1 otherwise (Fig. 3). The null hypothesis ofidentical thresholds has H0: uc 5 us, and the alternativehypothesis is HA: ucÞ us. Performing a line search findsthe maximum likelihood under the null hypothesis tobe L0 5 2127.152 (u estimated to be 4.79). Performinga grid search yields LA 5 2109.5 (uc estimated at 0, usestimated at 5.14). The equation x2 5 2(LA 2 L0) wouldhave a x2 distribution with one degree of freedomunder the null distribution. With x2 5 35.3, the nullhypothesis is soundly rejected (P , 0.0001) and thealternate hypothesis accepted, indicating that the on-set of structural and cellular damage occur at signifi-cantly different levels of strain. Stress-strain curvesdemonstrate no change in shape when the subfailurestretch is below the damage threshold (Fig. 5, A–C). Asubfailure stretch above the damage threshold elon-gates the toe region of the stress-strain curve and

decreases tissue stiffness and ultimate stress (Fig. 5,D–F).

DISCUSSION

Subfailure damage has been shown to alter the me-chanical properties of the anterior cruciate ligamentand lengthen the toe region of the force-displacementcurve, which results in increased joint laxity (28). Re-sults from this study indicate that rat medial collateralligaments strained above 5.14% from preload do notregain their original length after significant recoverytime (3003 the time of test). This recovery is consider-ably longer than other studies showing that ligamentsstretched below 5% completely recover in ,103 thetime of test (30). Hence, we conclude that ligamentsstretched beyond this threshold remain “stretched.”These findings are valuable to researchers performingmultiple tests on the same ligament specimen becauseno change in length or properties is evident in thetissue below ;5% when testing under the methodsdescribed in this study for rat.

We speculate that the increase in elongation after;5% strain and change in mechanical properties arethe result of fiber damage arising from two possiblemechanisms. One mechanism would be torn or plasti-cally deformed fibers. Torn fibers would be consistentwith the fiber failure mode of Hurschler’s microme-chanical model for ligament behavior (16), and plasti-cally deformed fibers could be supported by experimen-tal observations by Sasaki et al. (33, 34) and Kukretiand Belkoff (22) who observed that collagen fibrils,which make up collagen fibers, elongate during tendonloading. In addition, Yahia et al. (41), using scanningelectron microscopy, reported damage to collagen fibersin subfailure strained ligaments. Another possiblemechanism for the observed ligament laxity could bebiochemical degradation of the ECM from proteaserelease associated with the observed cellular necrosis.Regardless of the mechanism, the resulting increase intissue length represents tissue laxity and can be hy-pothesized to increase joint laxity.

Fig. 4. Cellular damage (DC) vs. strain (e) in the rat medial collateralligament. Necrotic area is normalized by the area of the ligament.Equation 1 demonstrates that cellular damage statistically beginswith the onset of loading (from a preloaded state) and occurs at lowerstrain levels than structural damage.

Fig. 3. Structural damage (DS) vs. strain (e) in rat medial collateralligament. Structural damage represents tissue laxity. The onset ofdamage is at 5.14% strain, as can be seen by Eq. 2. Note that thislevel of strain is well into the linear region of the stress-strain curvefor this tissue, as can be seen from Fig. 5.



The statistical threshold for structural damage iswell into the linear region of the stress-strain curve, aregion unlikely to be regularly reached in normal ac-tivity. The upper half of the linear region of stress-

strain curve has been associated with partial tearing ofanterior cruciate ligaments (26). A microstructuralstudy of the rat MCL in situ using scanning electronmicroscopy has shown that during knee flexion normal

Fig. 5. Stress-strain curves for a control rat medial collateral ligament (F), subfailure stretch (h), and stretchedligament (■) that received a subfailure stretch. The subfailure stretch reached a peak strain of 0% (A), 4.7% (B),5.1% (C), 6% (D), 7% (E), and 9% (F). Subfailure strains (h) below 5.14% (A–C) did not change the toe region ortangential modulus on a subsequent stretch (■), and the toe region, tangential modulus, and ultimate stress of theligament (■) were not changed compared with the contralateral normal ligament (F). Subfailure strains (h) above5.14% (D–F) elongated the toe region and lowered the tangential modulus (■) compared with the initial stretch (h).The toe region was elongated and the tangential modulus and ultimate stress decreased (■) compared with thecontralateral normal ligament (F). Note the relative symmetry in contralateral ligaments in 0% subfailurestrain (A).



ligament has a crimped pattern in knee flexion that isonly removed near full extension (17). Our preloadremoves some crimping and hence results in our pre-loaded reference length being longer than typical ref-erence lengths in vivo.

The statistical threshold of cellular damage wasfound to be at 0% strain from preload in the rat MCL.That is, statistically, cellular damage begins with theapplication of tissue strain. It should be noted thatphysically one would not expect an increase in cellulardamage at small strains as our statistical analysisimplies. However, necrotic cells are present in thecontrol tissues (e 5 0, i.e., preload) and are presentafter very small strains (above reference preload). Thisbehavior did not allow the authors to identify anystatistical threshold other than the preloaded value.One would not expect high levels of cell necrosis duringnormal activity; therefore, as mentioned in the previ-ous paragraph, the strains represented when thismethod is used are likely to be higher than thosetypically seen in the rat in vivo. The ratio of the area offluorescence to the actual cell dimensions is not known,and therefore the number of necrotic cells is notcounted nor is the percent area of necrotic cells in thetissue quantified. In addition, cellular changes maytake place during the processing of the tissues becauseof processing time and handling, elevating the magni-tude of cellular damage. However, all tissues wereprepared in the same controlled manner, and changesin cellular damage were seen between preloaded zerostrain controls and stretched ligaments, with increas-ing cellular damage with increasing strain indicatingchanges relative to the reference values. This differ-ence in damage properties indicates fundamentallydifferent behaviors between cellular and structuraldamage.

This study quantifies regions of cellular damage inthe ligament as a function of strain. As stated in theprevious paragraph, this work does not count the num-ber of necrotic cells in the ligament. To the authors’knowledge, no other study has quantified cellular dam-age in ligament as a function of strain. However, me-chanical stimulation of cells in vitro has been studied(1, 11, 13, 14, 40). Strain in fibroblasts during in vitroequibiaxial testing on membranes are often higherthan the tissue strains at which we are reporting celldamage (e . 2% from a preloaded state). However, therelationship between reference (initial) strains used inthese studies compared with the complex cell loadingin an in vivo state is unknown. Furthermore, differ-ences between the reference strain in our study andpreviously published in vitro cell deformation studiesis also unknown.

We propose that microstructural irregularities inECM organization create local distortions in fibro-blasts during ligament strain that result in the cellulardamage reported in this study. Fibroblasts in connec-tive tissue are found to be in columns between fibrilbundles (18, 36), and fibrils cross and interweave (29).Using confocal microscopy to track cells as a measureof microstructural strain during tendon tissue elonga-

tion, Hurschler et al. (15) reported that microstruc-tural strains in ligament can be both larger andsmaller than macroscopic ligament strain (6 ;2% ofthe ligament strain) at different locations in the sametissue and that, at the microstructural level, even neg-ative strains can exist. Because fibroblast cells adhereto ECM collagen fibrils, cell deformation is related tofibril displacement and deformation. Squier andBausch (36) reported junctional attachments betweenprocesses of different and same fibroblasts in rat tailtendon. In addition, Squier and Bausch (36) showedthat fibroblasts exhibited invaginations of single orgroups of collagen fibrils and stated that “this arrange-ment may be indicative of fibril elongation or serve totransmit tension between the fibroblast and the colla-gen fibrils.” This result is supported by the TEM im-ages in Fig. 2, which show fibrils interdigitating withthe cell membrane. Because fibril orientation aroundthe cell is not completely uniform, the fibrils willstretch independently, creating nonlinear cell strainsand nonuniform cellular distortions as seen in Arnozc-zky et al. (3). The nonuniform fibril strain patternsseen must produce local cellular distortions far beyondwhat occurs on monolayers of cultured cells and mustcontribute to mechanically induced cell death reportedherein. These large local distortions from nonuniformfibril loading, fibrils slipping past one another, andcomplex cell-matrix interactions help explain the re-sults of this study where changes in cellular damageoccur before the onset of structural damage. In contrastto the complex loadings and large distortions of fibro-blast in vivo, fibroblasts in vitro on stretched mem-branes undergo less complex and less distortional load-ings. Although in vitro stretched membrane celldeformation studies provide extremely valuable infor-mation about biological mechanisms, they may notaccurately simulate in situ cells during ligament defor-mation.

Considering the organization of the fibroblast-ECMinteraction, one may expect to see cellular damage inlongitudinal lines along the axis of the ligament. How-ever, it is reasonable that cellular damage is not seenin longitudinal lines in this study. By using the abovetechniques, slices through the tissue are compiled andoverlaid, and because all fibers and fibrils are notperfectly aligned and do sit skewed from one anotherthrough the depth of the tissue one would not expect tosee clear longitudinal distributions. By overlayingslices, the slight mismatching of collagen fibers orfibrils, size effects, and fibril interweaving and overlaplimit the ability to see longitudinal columns because allthe slices are “compressed” together. It is entirely pos-sible that many semilongitudinal scans are adjacent,skewed, or slightly angled as a result of complex fibrilorganization, resulting in the distributions seen in Fig.1. It should be noted that, because only regions ofnecrotic cells are being quantified (overlaid yellow re-gions are counted as damaged) and green (viable) re-gions are not being quantified, the overlaying processdoes not induce an error in the cellular damage mea-



sure used in this study because cellular damage onlyconsiders necrotic cells normalized by tissue area.

Limitations exist in this study and should be noted.Structural damage was measured at only one rate ofloading from a preloaded state, and the effect of mul-tiple loadings was not investigated. In addition, theligament microstructure was not examined for collagenfiber or fibril integrity (tearing or plastic deformation),and the mechanical properties were evaluated withonly one tissue per applied subfailure stretch. Cellulardamage was examined only at one time point and,therefore, only early cellular damage is quantified. Inaddition, processing times may have increased themagnitude of cell death in the tissue, yet changes incellular damage are seen to increase with strain intissues processed in the same controlled manner. Thetime-dependent aspects of cellular damage and thepossibility of further necrosis resulting from apoptosisremain unexplored. In addition, in vivo loads in ratMCLs are not known and, hence, this behavior must bestudied in a better model with more elucidation intothe in vivo reference states regarding tissue micro-structure and deformation during normal activity. Therat model used herein leaves questions open regardingeffects of ligament size and species. Further studies inhuman ligaments are recommended.

In summary, structural and cellular damage occur atdifferent levels of tissue strain in a rat MCL. Subfail-ure strain above the damage threshold changed themechanical properties of the ligament. Further inves-tigations into loading rate, multiple loadings, ligamentmicrostructural displacements and deformations, celldeformation, and cell biology need to be performed tounderstand the subfailure behavior of ligament andthe role of cell death in the healing process.

We thank Dr. Yan Lu for assistance with confocal microscopy, Dr.Tom Warner for evaluation of transmission electron microscopy, andGlen Leverson for computational assistance.

This work was thankfully funded in part by the National ScienceFoundation (Grant no. CMS-9907977), National Aeronautics andSpace Administration (Grant no. NAG9-1152), and the University ofWisconsin-Madison Graduate School.

REFERENCES

1. Almekinders LC, Banes AJ, and Ballenger CA. Effects ofrepetitive motion on human fibroblasts. Med Sci Sports Exerc 25:603–607, 1993.

2. Andriachi T, Sabiston P, De Haven K, Dahners L, Woo S,Frank C, Oakes B, Brand R, and Lewis J. Ligament: injuryand repair. In: Injury and Repair of the Musculoskeletal SoftTissues, edited by Woo SL-Y and Buckwalter JA. Park Ridge, IL:AAOS, 1987.

3. Arnoczky SP, Hoonjan A, Whallon B, and Cloutier B. Celldeformation in tendons under tensile load: a morphological anal-ysis using confocal laser microscopy. Trans Orthop Res Soc 40:495, 1994.

4. Blake RL, Anderson K, and Ferguson H. Posterior tibialtendinitis. A literature review with case reports. J Am PodiatrMed Assoc 84: 141–149, 1994.

5. Buckwalter JA and Woo SLY. Effects of repetitive loading andmotion on the musculoskeletal tissues. In: Orthopaedic SportsMedicine: Principles and Practice. Philadelphia, PA: Saunders,1994.

6. Burr A, Hild F, and Leckie FA. Micro-mechanics and contin-uum damage mechanics. Arch Appl Mech 65: 437–456, 1995.

7. Chelikani S and Panjabi MM. Biomechanical symmetry of therabbit ACL. Ann Mtg Am Soc Biomech Stanford, CA, 1995.

8. Curwin S and Stanish WD. Tendinitis: Its Etiology and Treat-ment. Lexington, MA: Collamore, 1984.

9. Daniel DM, Stone ML, Dobson BE, Fithian DC, RossmanDJ, and Kaufman KR. Fate of the ACL-injured patient. Aprospective outcome study. Am J Sports Med 22: 632–644, 1994.

10. Fruensgaard S and Johannsen HV. Incomplete ruptures ofthe anterior cruciate ligament. J Bone Joint Surg Br 71: 526–530, 1989.

11. Gilbert JA, Weinhold PS, Banes AJ, Link GW, and JonesGL. Strain profiles for circular cell culture plates containingflexible surfaces employed to mechanically deform cells in vitro.J Biomech 27: 1169–1177, 1994.

12. Gomez MA. The biomechanical properties of normal and heal-ing ligaments. In: The Anterior Cruciate Ligament: Current andFuture Concepts, edited by Arnoczky SP, Frank CB, Woo SL-Y,and Simon TM. New York: Raven, 1993, p. 227–233.

13. Gudi SR, Lee AA, Clark CB, and Frangos JA. Equibiaxialstrain and strain rate stimulate early activation of G proteins incardiac fibroblasts. Am J Physiol Cell Physiol 274: C1424–C1428, 1998.

14. Hsieh AH, Tsai CM, Ma QJ, Lin T, Banes AJ, Villarreal FJ,Akeson WH, and Sung KL. Time-dependent increases in type-III collagen gene expression in medical collateral ligament fibro-blasts under cyclic strains. J Orthop Res 18: 220–227, 2000.

15. Hurschler C. Collagen Matrix in Normal and Healing Liga-ments: Microstructural Behavior, Biological Adaptation and aStructural Mechanical Model (PhD thesis). Madison, WI: Uni-versity of Wisconsin, 1998.

16. Hurschler C, Loitz-Ramage B, and Vanderby R Jr. A struc-turally based stress-stretch relationship for tendon and liga-ment. J Biomech Eng 119: 392–399, 1997.

17. Hurschler C, Provenzano PP, and Vanderby R Jr. Scan-ning electron microscopic investigation of healing and normalrat medial collateral ligaments fixed under slack and loadedconditions. Trans Orthop Res Soc 23: 1032, 1998.

18. Kastelic J, Galeski A, and Baer E. The multicomposite struc-ture of tendon. Connect Tissue Res 6: 11–23, 1978.

19. Kennedy JC, Hawkins RJ, Willis RB, and Danylchuk KD.Tension studies of human knee ligaments. Yield point, ultimatefailure, and disruption of the cruciate and tibial collateral liga-ments. J Bone Joint Surg Am 58: 350–355, 1976.

20. Kiefhaber TR and Stern PJ. Upper extremity tendinitis andoveruse syndromes in the athlete. Clin Sports Med 11: 39–55,1992.

21. Krajcinovic D. Damage mechanics. Mech Matls 8: 117–197,1989.

22. Kukreti U and Belkoff SM. Collagen fibril D-period maychange as a function of strain and location in ligament. J Bio-mech 33: 1569–1574, 2000.

23. Laws G and Walton M. Fibroblastic healing of grade II liga-ment injuries. Histological and mechanical studies in the sheep.J Bone Joint Surg Br 70: 390–396, 1988.

24. Lindgren BW. Statistical Theory. New York: Macmillan, 1976.25. National Institutes of Health. NIH Image. http://rsb.info.nih.

gov/nih-image, 2000.26. Noyes FR, Butler DL, Grood ES, Zernicke RF, and Hefzy

MS. Biomechanical analysis of human ligament grafts used inknee-ligament repairs and reconstructions. J Bone Joint SurgAm 66: 344–352, 1984.

27. Ohlendorf C, Tomford WW, and Mankin HJ. Chondrocytesurvival in cryopreserved osteochondral articular cartilage.J Orthop Res 14: 413–416, 1996.

28. Panjabi MM, Yoldas E, Oxland TR, and Crisco JJ 3rd.Subfailure injury of the rabbit anterior cruciate ligament. J Or-thop Res 14: 216–222, 1996.

29. Provenzano PP, Hurschler C, and Vanderby R Jr. Micro-structural morphology in the transition region between scar andintact residual segments of a healing rat medial collateral liga-ment. Connect Tissue Res 42: In Press, 2001.

30. Provenzano PP, Lakes RS, Keenan T, and Vanderby RJr. Non-linear ligament viscoelasticity. Ann Biomed Eng 29:908–914, 2001.



31. Raudino F. Radial sensory nerve entrapment in carpal tun-nel syndrome. Electromyogr Clin Neurophysiol 35: 201–205,1995.

32. Rigby BJ, Hirai N, Spikes JD, and Eyring H. The mechan-ical properties of rat tail tendon. J Gen Physiol 43: 265–289,1959.

33. Sasaki N and Odajima W. Elongation mechanism of collagenfibrils and force-strain relations of tendon at each level of struc-tural hierarchy. J Biomech 29: 1131–1136, 1996.

34. Sasaki N, Shukunami N, Matsushima N, and Izumi Y.Time-resolved X-ray diffraction from tendon collagen duringcreep using synchrotron radiation. J Biomech 32: 285–292, 1999.

35. Sims JA and Moorman SJ. The role of the iliolumbar ligamentin low back pain. Med Hypotheses 45: 511–515, 1996.

36. Squier CA and Bausch WH. Three-dimensional organizationof fibroblasts and collagen fibrils in rat tail tendon. Cell TissueRes 238: 319–327, 1984.

37. Turner S. Creep in glassy polymers. In: The Physics of GlassyPolymers, edited by RH Howard. New York: Wiley, 1973.

38. Viidik A. Simultaneous mechanical and light microscopic studiesof collagen fibers. Z Anat Entwicklungsgesch 136: 204–212, 1972.

39. Wang VM, Bucchieri JS, Pollock RG, Flatow EL, PawlukRJ, Bigliani LU, and Mow VC. The effects of cyclic loads onthe subfailure properties of the inferior glenohumeral ligament.ASME Bioeng Conf 35: 137–138, 1997.

40. Winston FK, Macarak EJ, Gorfien SF, and Thibault LE. Asystem to reproduce and quantify the biomechanical environ-ment of the cell. J Appl Physiol 67: 397–405, 1989.

41. Yahia L, Brunet J, Labelle S, and Rivard CH. A scanningelectron microscopic study of rabbit ligaments under strain.Matrix 10: 58–64, 1990.

42. Yoganandan N, Maiman DJ, Pintar F, Ray G, MyklebustJB, Sances A Jr, and Larson SJ. Microtrauma in the lumbarspine: a cause of low back pain. Neurosurgery 23: 162–168.



subfailure damage in ligament: a structural and cellular...

Documents