Bone 50 (2012) 1152–1161

Contents lists available at SciVerse ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r .com/ locate /bone

Original Full Length Article

Decreased bone tissue mineralization can partly explain subchondral sclerosisobserved in osteoarthritis

L.G.E. Cox a, C.C. van Donkelaar a,⁎, B. van Rietbergen a, P.J. Emans b, K. Ito a

a Department of Biomedical Engineering, Eindhoven University of Technology, PO Box 513, 5600 MB, Eindhoven, The Netherlandsb Department of Orthopaedic Surgery, University Hospital Maastricht, PO Box 5700, 6202 AZ, Maastricht, The Netherlands

⁎ Corresponding author at: Department of BiomeUniversity of Technology, PO Box 513, GEM-Z 4.101, 5600MFax: +31 40 2443744.

E-mail address: c.c.v.donkelaar@tue.nl (C.C. van Don

8756-3282/$ – see front matter © 2012 Elsevier Inc. Alldoi:10.1016/j.bone.2012.01.024

a b s t r a c t

a r t i c l e i n f o

Article history:Received 21 September 2011Revised 25 January 2012Accepted 27 January 2012Available online 9 February 2012

Edited by: David Fyhrie

Keywords:OsteoarthritisBone adaptationMineralizationCartilage degenerationSclerosis

For many years, pharmaceutical therapies for osteoarthritis (OA) were focused on cartilage. However, it has beentheorized that bone changes such as increased bone volume fraction and decreased bone matrix mineralizationmay play an important role in the initiation and pathogenesis of OA as well. The mechanisms behind the bonechanges are subject of debate, and a better understanding may help in the development of bone-targeting OAtherapies. In the literature, the increase in bone volume fraction has been hypothesized to result from mechano-regulated bone adaptation in response to decreased mineralization. Furthermore, both changes in bone volumefraction and mineralization have been reported to be highest close to the cartilage, and bone volume fractionhas been reported to be correlated with cartilage degeneration. These data indicate that cartilage degeneration,bone volume fraction, and bone matrix mineralization may be related in OA. In the current study, we aimed to in-vestigate the relationships between cartilage degeneration, bone matrix mineralization and bone volume fractionat a local level.WithmicroCT,we determined bonematrixmineralization and bone volume fraction as a function ofdistance from the cartilage in osteochondral plugs from human OA tibia plateaus with varying degrees of cartilagedegeneration. In addition, we evaluated whether mechanoregulated bone adaptation in response to decreasedbone matrix mineralization may be responsible for the increase in bone volume fraction observed in OA. For thispurpose, we used the experimentally obtained mineralization data as input for bone adaptation simulations. Wesimulated the effect of mechanoregulated bone adaptation in response to different degrees of mineralization,and compared the simulation results to the experimental data. We found that local changes in subchondralbone mineralization and bone volume fraction only occurred underneath severely degenerated cartilage,indicating that bonemineralization and volume fraction are related to cartilage degeneration at a local level. In ad-dition, both the experimental data and the simulations indicated that a depth-dependent increase in bone volumefraction could be caused by decreased bone matrix mineralization. However, a quantitative comparison showedthat decreased mineralization can only explain part of the subchondral sclerosis observed in OA.

© 2012 Elsevier Inc. All rights reserved.

Introduction

Osteoarthritis (OA) is a complex multi-factorial joint disease,characterized by degeneration of cartilage and modification of thestructural and material properties of subchondral bone [1]. Formany years, pharmaceutical therapies have been focused on cartilage,because the bone changes were thought to occur secondary to carti-lage degeneration, and not to play a major role in the disease process.However, currently the bone is being reconsidered as a therapeutictarget [2,3]. The bone can be a source of pain in OA, and in additionit has been theorized that bone changes may play an important rolein the initiation and pathogenesis of OA [2-5].

dical Engineering, EindhovenB, Eindhoven, The Netherlands.

kelaar).

rights reserved.

The most frequently discussed bone change observed in OA isprobably subchondral sclerosis. Subchondral sclerosis has been de-scribed in many clinical [6-10] and animal studies [11-14]. Radin etal. [15] were the first to suggest that bone sclerosis may play animportant role in cartilage degeneration. They hypothesized thatsclerosis would lead to a stiffer subchondral structure with a de-creased shock-absorbing capability, thereby increasing the stress inthe overlying cartilage [15]. Alternatively, the sclerotic bone may in-fluence cartilage degeneration via the release of biochemical signals.Due to the increased bone metabolism associated with the sclerosis[4,16-21], various cytokines, including insulin-like growth factors,transforming growth factor ß, and bone morphogenic proteins,which are known to modulate chondrocyte proliferation and differ-entiation, and matrix synthesis, are released [5,22,23]. In addition tobone volume fraction, bone material properties are altered in OA.Bone matrix mineralization is lower [6,18,24,25] and as a result thematerial stiffness of OA bone is decreased [18,24,26]. In contrastwith the theory of Radin et al., it has been suggested that this

1153L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

decrease in bone material stiffness may increase cartilage deforma-tion and induce cartilage degeneration upon joint loading [19,27].

Sclerosis and bone matrix mineralization may be closely related, asthe increase in bone volume fraction has been hypothesized to resultfrom bone adaptation in response to decreased mineralization [26,28].In a previous simulation studywe showed that bone adaptationmay in-deed increase bone volume fraction to counteract an increase in bonetissue strain resulting from decreased bone material stiffness [29].There are also indications that bone volume fraction andmineralizationare related to cartilage degeneration. Both bone matrix mineralization[25,30], and bone structure [7] have been reported to change mostmarkedly directly underneath the cartilage, and bone structuralchanges seemed correlatedwith the degree of degeneration of the over-lying cartilage [7]. A possible explanation for these findings could bethat biochemical signaling from degenerating cartilage may influencebone mineralization and/ or bone volume fraction.

In the current study, we aimed to investigate the relationships be-tween cartilage degeneration, bone matrix mineralization, and bonevolume fraction at a local level. For this purpose, we determined bonematrix mineralization (tissue mineralization, TMD) and bone volumefraction (BV/TV) using microCT, at different depth levels in humanosteochondral plugs with varying degrees of cartilage degeneration.These plugs were harvested from OA tibia plateaus obtained aftertotal knee replacement. In addition, we used the experimentallyobtained mineralization data as input for bone adaptation simulations,to predict the effect of bone mineralization on bone volume fraction.With this combined experimental–mathematical approach, we aimedto evaluate whether the differences in bone volume fraction observedfor different OA grades could be explained by bone adaptation in re-sponse to changes in matrix mineralization.

Methods

Sample preparation

We examined 32 human OA tibia plateaus, obtained after totalknee replacement (University hospital Maastricht, Maastricht, TheNetherlands). With a diamond coated bit, we cored cylindrical speci-mens perpendicular to the articular surface, with a diameter of 8 mmand variable depth, dependent on the height of the tibia plateaus.Preferably, one specimen was analyzed for each side (medial and lat-eral) of the plateau (Fig. 1a). However, for 11 plateaus, specimensfrom one side could not be used for further analysis, due to either lim-ited height of the plateaus as a result of the location of the surgicalcutting plane or the presence of cysts, leading to a total number of53 bone–cartilage specimens (25 medial, 28 lateral).

sample1

sample2

a)

Fig. 1. Sample preparation. (a) Schematic representation of a tibia plateau indicating possible2 mm slice used to determine mineralization and bone volume fraction values at 1 mm dep

MicroCT

Each specimen was fixated in formalin and scanned with a microCTscanner (vivaCT40, Scanco Medical AG) at an isotropic voxel size of21 μm (70 kVp, 114 μA, 500 projections per 180 degrees, 300 ms inte-gration time). A beam-hardening correction algorithm was applied toall scans. The computed linear attenuation coefficient of the X-raybeam in each voxel, which can be considered to be proportional to thelocal degree of mineralization, was stored in an attenuation map andrepresented by a gray value in the reconstruction.

Image processing included Gaussian filtering (sigma=0.8, sup-port=1) and segmentationwith a constant threshold value to discrim-inate between bone and soft tissue. The segmentation was visuallychecked. The degree of mineralization for a specific volume of interestof the segmented images was quantified by comparing the attenuationcoefficients of the tissue quantified as bone with reference measure-ments of a phantom containing hydroxyapatite of different densities.For these measurements, the outermost voxel layer characterized asbone was removed, as it is likely to be corrupted by partial volumeeffects.

Bone volume fraction (BV/TV) and bone matrix mineralization(TMD) were determined for 2 mm thick bone slices parallel to the car-tilage using the Scanco Medical AG software. Analyzing 2 mm thickslices is necessary for accurate measurement of structural parameters.To detect depth-dependent changes in bone volume fraction and bonematrix mineralization, we evaluated slices at 5 different distancesfrom the articular cartilage, which will be referred to according totheir mean distance as 1, 1.5, 2, 2.5, and 3 mm depth. This means thatthe bone volume fraction and bone matrix mineralization that we de-termined at 1 mm depth, are the average values determined for thebone slice between 0 mm and 2 mm distance from the cartilage(Fig. 1b). It also implies that there was a 1.5 mm overlap between con-secutive slices. Due to the limited height of some of the tibia plateaus,we could not obtain measurements at all five depth levels for all 53specimens, leading to a decreasing number of measurements with in-creasing depth level.

Cartilage degeneration

All specimens were dehydrated in graded ethanol solutions (70% to100%) and embedded in polymethylmethacrylate (PMMA). The embed-ded specimens were bisected perpendicular to the cartilage using abandsaw, and then sectioned using a microtome (Leica RM2165, LeicaMicrosystems), thereby exposing the trabecular microstructure at thesurface of the specimens. These surfaces were stained with toluidineblue for 1.5 h and then rinsed in running tap water. Images were

cartilage

trabecularbone

subchondralplate

1mmdepth

2mm

b)

sample locations. (b) Schematic representation of an osteochondral plug indicating theth.

Table 1ICRS cartilage degeneration scoring system.

ICRS grade Description

0 Normal intact cartilage1 Chondral softening and blistering, superficial lesions,

fissures and cracks, soft indentation2 Fraying, lesions and fissures extending down to b50% of

cartilage depth3 Partial loss of cartilage thickness, cartilage defects extending

down to >50% of cartilage depth as well as down to calcified layer4 Full-thickness cartilage loss with exposure of the subchondral bone

Table 2K–L osteoarthritis scoring system.

K–L grade Description

0 Normal bone, no joint space narrowing1 Doubtful narrowing of joint space and possible

osteophytic lipping2 Definite narrowing of joint space, osteophytes3 Definite narrowing of joint space, osteophytes, some

sclerosis and possible deformity of bone contour4 Marked narrowing of joint space, large osteophytes, severe

sclerosis and definite deformity of bone contour

1154 L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

taken from all specimens, using a stereo microscope (Discovery.V8,Zeiss). For each specimen, cartilage degeneration was graded from0 to 4 by three independent observers, according to the InternationalCartilage Repair Society (ICRS) score, as described by Marticke et al.[31] (Table 1).

In addition to the ICRS score, the Kellgren–Lawrence (K–L) scorewas used to classify the samples according to their OA grade. TheK–L score is determined from radiographs and takes into accountjoint space narrowing as a measure of cartilage thickness (Table 2).K–L scores were determined pre-operatively for the medial and later-al side of each tibia plateau separately, but not per bone–cartilagesample. Therefore, the K–L score is a more global measure of OA com-pared to the ICRS score, which was determined at each sample loca-tion. The results of the ICRS score and K–L score were compared toevaluate whether changes in bone matrix mineralization and bonevolume fraction are mainly driven by local cartilage degeneration orby more global factors.

Statistical analysis

To determine whether the bone matrix mineralization and thebone volume fraction were different for samples with different ICRSor K–L grades, we performed one-way ANOVA, with a Bonferronipost-hoc test at each depth level. To quantify the correlation between

a)Resorption

cavity

Mechanicalloading

Osteocytei

Osteoclast

Osteoblasts

SED ratesensation

Recruitmenstimulus

Stimuli from other osteocy

Boneresorption

Boneformation

FEA

Tissue mineraldensity

SEDdistribution

Bonearchitecture

Fig. 2.Mathematical model. (a) Schematic representation of the bone remodeling theory, adarticular cartilage and subchondral bone tissue.

bonematrix mineralization and bone volume fraction, we determinedPearson correlation coefficients and corresponding p-values. For com-parisons of means between two groups we performed Student t-tests.This was done to investigate differences between samples from themedial and lateral side of the tibia plateaus, and to investigate the dif-ference in bone volume fraction between the grade 0 and grade 4simulations. In the simulations we only used the mineralization datafrom the grade 0 and grade 4 ICRS groups, as those were the onlygroups which had statistically significant different mineralizationaccording to the experimental data.

To investigate trends for bonematrixmineralization andbonevolumefractionwith depth level, we fitted linear relationships between both thebone mineralization and the distance from the cartilage, and the bonevolume fraction and the distance from the cartilage. We determinedthese relationships for each sample individually, using a least-squaresmethod. Subsequently, we grouped the values obtained for the slopesof these relationships according to their ICRS grade and tested whetherthemean of the slopes for each of the ICRS gradeswas significantly differ-ent from zero, using Student t-tests. Finally, we tested whether therewere statistically significant differences in the slopes between the differ-ent ICRS grades, using one-way ANOVA with a Bonferroni post-hoc test.

Mathematical model

We used an established bone adaptationmodel based on the theoryof Huiskes et al. [32] to simulate the effect of changes in bone matrixmineralization on bone structure. This model was previously shown ca-pable of explaining several distinct bone adaptation features observedin animal and clinical studies, including the emergence and maturationof trabecular structures, load-adaptation, fracture remodeling, post-menopausal bone loss, and the development of cysts [32-38]). In themodel, osteocytes respond to the local strain energy density (SED)rate by promoting osteoblastic bone formation, while osteoclasts resorbbone near randomly occurring microcracks (Fig. 2a). The model equa-tions and parameter settings that we used are described in a previouspaper [39].

To represent the morphology and material properties of the tissues,and to calculate the local SED values, finite element analysis was used(Fig. 2b). Simulations were performed in a 2D domain that representspart of the articular cartilage and bone below the articular cartilage.We used a rectangular mesh of 200 x 100 elements, with an elementsize of 50 x 50 μm. The model consisted of 94 rows of bone tissue and6 rows of articular cartilage. Both cartilage and bone tissue were mod-eled as linear elastic materials as described previously [39]. To obtaina 2D trabecular structure, we started with a tissue with a uniformbone density in which osteocytes were randomly distributed. Themesh was loaded statically with 2 MPa compression in the vertical di-rection (perpendicular to the cartilage), and 1.2 MPa in the horizontaldirection, which for a linear elastic material represents the maximumSED rate of a dynamic load of 1 MPa and 0.6 MPa at 1 Hz respectively

cartilage bone marrowb)

t

tes

apted from Huiskes et al. [32]. (b) A two-dimensional finite element mesh representing

Table 3Sample data.

ICRS grade No. of samples Donor age (SD)

0 7 63 (8.4)1 20 68 (8.6)2 11 70 (7.6)3 6 70 (13.6)4 9 72 (15.3)

1155L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

[33]. The bonewas allowed to remodel for 5000 increments, afterwhicha stable trabecular structure was obtained, that did not change signifi-cantly with respect to bone volume fraction, trabecular number, trabec-ular thickness, and trabecular separation if remodeling was continuedfor 500 additional increments.

To evaluate whether the differences in bone volume fraction ob-served for different OA grades can be explained by the differences inmatrix mineralization, we used the experimentally determined miner-alization data as input for the model. In 1988, Currey found that the re-lation between bone mineralization and elastic modulus can beapproximated with a cubic power law [40]. This power law is includedin the adaptation model:

E x; tð Þ ¼ Ebρ x; tð Þ3

Here E(x,t) is the elasticmodulus at location x at time point t, and Eb isthe elastic modulus of healthy bone tissue (5 GPa) [41-43]. ρ(x,t) is themineralization at location x at time point t, which is normalized to the av-erage mineralization of healthy bone tissue. This means that a 10% de-crease in mineralization (i.e. ρ(x,t)=0.9), leads to a 27% lower elasticmodulus compared tohealthy bone tissue ((1–0.93)×100%). For our sim-ulations, we assumed that the samples for which cartilage degenerationwas scored as ICRS grade 0 represented healthy bone tissue. Therefore,to simulate bone adaptation to altered mineralization, the mean bonematrix mineralization measured for other OA grades was normalized tothe bone matrix mineralization measured for the grade 0 samples.

For different degrees of mineralization, we performedmultiple sim-ulation series. For each series, we startedwith a bone structure obtainedafter 5000 increments of remodeling under healthymineralization con-ditions. Subsequently, we changed the bone matrix mineralization in adepth-dependent manner according to the experimental data and weallowed the bone to remodel for 2500 increments. For each degree ofmineralization, we performed 10 simulation series, which only differedin their random distribution of osteocytes, thereby leading to differentbut similar bone structures. From the resulting bone structures, we de-termined the bone volume fraction at different depth levels, which wecompared to the experimental data. Similar to the experimentally de-termined parameters, bone volume fraction for the simulated structures

ICRS 0 ICRS

ICRS 3 ICRS

Fig. 3. ICRS grades. Samples classified according t

was determined for 2 mm thick slices, of which the center was at 1, 1.5,2, 2.5 and 3 mm distance from the cartilage.

Results

Cartilage degeneration according to ICRS score

For 29 of the 53 bone–cartilage specimens all three observers gradedcartilage degeneration the same, for 23 specimens two observers grad-ed cartilage degeneration the same while the third observer scored onegrade higher or lower, and for 1 specimen all three observers gradedcartilage degeneration differently. For this last specimen, the averagevalue of the three observers was used, while for the other specimensthe grade on which at least two observers agreed was used. For eachof the five ICRS grades, an example is shown in Fig. 3. In Table 3, thenumber of samples and the mean and standard deviation of the pa-tients' age at the time of surgery are shown for each ICRS grade. Therewere no statistically significant differences in the mean age of the pa-tients between any of the groups.

Because we also did not find any statistically significant differencesin either bone matrix mineralization or bone volume fraction betweenmedial and lateral samples with the same degree of cartilage degenera-tion (Student t-test, p>0.1), we did not make a distinction betweenmedial and lateral samples in our analyses. Furthermore, we decidedto treat medial and lateral specimens from the same plateau as inde-pendent, based on the assumption that locally, bonematrix mineraliza-tion and bone volume fraction aremore strongly related to the degree ofcartilage degeneration than to patient-specific global factors. This as-sumption was supported by the observation that generally the degree

1 ICRS 2

4

o the five ICRS cartilage degeneration grades.

1370190 [mg HA cm-3]

ICR

S 4

ICR

S 0

a) b)

Fig. 4. Typical example. (a) CT images of two samples of the same patient; one gradedICRS 0 (top), and one graded ICRS 4 (bottom). (b) Mineral distribution in cross-sectionsof these samples, revealing less and smaller areas of high mineralization in the scleroticbone of the ICRS grade 4 sample.

0 1 2 3 4−10

−5

0

5

10

15

20

slop

e T

MD

[mg

HA

cm

−3 /

mm

]

ICRS grade

Fig. 6. Boxplot of the slopes of the relationships between bone matrix mineralizationand the distance from the cartilage per sample for each of the five ICRS grades.

1156 L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

of cartilage degeneration differed markedly between the two sides ofone tibia plateau, together with bone matrix mineralization and bonevolume fraction.

Bone matrix mineralization versus cartilage degeneration

For one patient, cartilage degeneration was graded ICRS 0 for the lat-eral sample and ICRS 4 for the medial sample. For these samples, CT-images and mineral distributions are shown in Fig. 4, as a typical exam-ple. Bone volume fraction is higher, while mineralization is lower in thegrade 4 specimen (Fig. 4). Especially themaximummineralization seemsto be decreased compared to the grade 0 specimen. The green areas inFig. 4b represent voxels that were partially filled with bone and partiallywith bone marrow. Therefore, these voxels were not taken into accountfor determining the mean bone matrix mineralization.

Between samples with ICRS 0, ICRS 1, ICRS 2, and ICRS 3, no statis-tically significant differences in bone matrix mineralization werefound (Fig. 5a). In samples with ICRS 4, mean bonematrix mineraliza-tion was significantly lower compared to samples with ICRS 0 to ICRS

1 1.5 2 2.5 3850

900

950

1000

TM

D [m

g H

A c

m-3

]

Depth [mm]

ICRS 0 ICRS 1

a)

********

********

****

****

*

Fig. 5. Bone matrix mineralization. (a) Bone matrix mineralization (TMD) at different distamineralization normalized to grade 0. ⁎pb0.05, ⁎⁎pb0.01, one-way ANOVA, Bonferroni pos

3 at lower depth levels. Compared to ICRS 0, this decrease varied be-tween 6% at 1 mm distance from the cartilage, to 4% at 3 mm distancefrom the cartilage (Fig. 5b).

The boxplot of the slopes of the relationships between bone min-eralization and depth indicates that for all ICRS grades, bone matrixmineralization increased with increasing distance from the cartilage(Fig. 6). For all grades, the slope was significantly higher than zero(right-tailed Student t-test, pb0.05). Although the difference inbone matrix mineralization between the samples with grade 4 andwith grade 0 cartilage degeneration seemed to decrease with increas-ing distance from the cartilage (Fig. 5b), no statistically significantdifference between the slopes of any of the groups could be demon-strated (one-way ANOVA, Bonferroni post-hoc test, pb0.05).

Bone volume fraction versus cartilage degeneration and bone matrixmineralization

Similar to the bonematrix mineralization, for the bone volume frac-tion there was no difference between ICRS grades 0 to 3 (Fig. 7a–b). ForICRS grade 4, bone volume fraction was significantly increased com-pared to the other ICRS grades at 1 mm, 1.5 mmand 2 mmdepth. Com-pared to grade 0, this increase varied between 69% at 1 mm distancefrom the cartilage and 24% at 3 mmdistance from the cartilage (Fig. 7b).

After obtaining the slope of the relationship between the bone vol-ume fraction and depth level for each sample, we grouped these slopesaccording the ICRS score. The boxplot of these slopes indicates that forall ICRS grades, bone volume fraction decreased with increasing dis-tance from the cartilage (Fig. 8). For ICRS 1, 2, 3, and 4 the slope was

1 1.5 2 2.5 30.92

0.94

0.96

0.98

1

1.02

Min

.BV

nor

mal

ized

to g

rade

0 [-

]

Depth [mm]

ICRS 2 ICRS 3 ICRS 4

b)

nces from the cartilage, for different grades of cartilage degeneration. (b) Bone matrixt-hoc test.

1 1.5 2 2.5 30.1

0.2

0.3

0.4

0.5

0.6

BV

/TV

[-]

Depth [mm]1 1.5 2 2.5 3

0.6

0.8

1

1.2

1.4

1.6

1.8

BV

/TV

nor

mal

ized

to g

rade

0 [-

]

Depth [mm]

***** *

** **

ICRS 0 ICRS 1 ICRS 2 ICRS 3 ICRS 4

a) b)

Fig. 7. Bone volume fraction. (a) Bone volume fraction (BV/TV) at different distances from the cartilage, for different grades of cartilage degeneration. (b) Bone volume fractionnormalized to grade 0. ⁎pb0.05, ⁎⁎pb0.01, one-way ANOVA, Bonferroni post-hoc test.

0 1 2 3 4−0.12

−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

slop

e B

V/T

V [/

mm

]

ICRS grade

Fig. 8. Boxplot of the slopes of the relationships between bone volume fraction and thedistance from the cartilage for each of the five ICRS grades.

1157L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

significantly lower than zero (left-tailed Student t-test, pb0.05). Onlythe slopes of the ICRS 1 and ICRS 4 groups differed significantly fromeach other (one-way ANOVA, Bonferroni post-hoc test, pb0.05). Be-cause the ICRS 0 group contained a far outlier, and because the slopesfor the ICRS 4 group were not normally distributed, we also performedWilcoxon Signed-Rank tests to investigate whether the slopes were dif-ferent from zero, and a Kruskall–Wallis test with a Bonferroni post-hoctest to compare the slopes from different groups. According to theWilcoxon Signed-Rank test the slopes for ICRS 1, 2, 3, and 4 were

850 900 950 10000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

TMD [mg HA cm-3]

BV

/TV

[-]

ICRS 0 ICRS 1

a) b)

Fig. 9. (a) Correlation between bone matrix mineralization (TMD) and bone volume fraction−0.62, pb0.01. (b) Correlation between bone matrix mineralization (Min.BV) and bone volcorrelation coefficient −0.92, pb0.01.

significantly lower than zero (pb0.05), and according to the Kruskall–Wallis test, the slopes of the ICRS 0 group and ICRS 4 group differedsignificantly (pb0.05).

Bone volume fraction and bone matrix mineralization were nega-tively correlated (Fig. 9). When for each ICRS grade the mean valuesper depth level were taken (i.e. the values displayed in Figs. 5a and7a), thereby averaging out the differences independent of OA gradeand depth level, the correlation increased (Fig. 9b).

Data classified according to K–L score

For two tibia plateaus, no K–L scores were determined preopera-tively, so the four samples taken from these plateaus were excludedfrom the analyses, leading to a total number of 49 samples (Table 4).

For both bone matrix mineralization and bone volume fraction, nostatistically significant differences between any of the K–L gradeswere found, at any depth level (Fig. 10a–b), which may at least partlybe explained by the uneven distribution of samples over the 5 K–Lgrades. What is interesting though, is that bone volume fractionseemed to increase gradually with increasing K–L grade (Fig. 10b),while for the ICRS score, a sudden increase was observed for grade4 only (Fig. 7). For bone matrix mineralization this trend was not soclear, and the differences between groups were even smaller thanthey were for the ICRS score. This is reflected in the Pearson correla-tion coefficient between bone matrix mineralization (Min.BV) andbone volume fraction (BV/TV) for the mean values per depth level

880 900 920 940 960 980 10000.1

0.15

0.2

0.25

0.3

0.35

0.4

TMD [mg HA cm-3]

BV

/TV

[-]

ICRS 2 ICRS 3 ICRS 4

(BV/TV), taking into account all individual data points. Pearson correlation coefficientume fraction (BV/TV) for the mean values per depth level for each ICRS grade. Pearson

Table 4Sample data after classification according to the K–L scoring system.

K–L grade No. of samples Donor age (SD)

0 3 65 (6.1)1 8 70 (8.2)2 3 73 (6.8)3 21 68 (11.0)4 14 71 (7.1)

1158 L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

for each K–L grade (Fig. 10c), which was lower than the coefficient wedetermined for the ICRS score (Fig. 9).

Bone adaptation simulations

Since bone matrix mineralization did not significantly differ be-tween grade 0 to grade 3 ICRS cartilage degeneration samples, onlyICRS 0 and ICRS 4 mineralization data were used as input for the sim-ulations. For the simulation of ICRS 4, a linear decrease in bone matrixmineralization of 6% at 1 mm depth to 4% at 3 mm depth was imple-mented based on the experimentally determined difference in meanbone matrix mineralization between ICRS grade 0 and ICRS grade 4cartilage degeneration (Fig. 5b). For the ICRS 0 simulation, the bonematrix mineralization was left unchanged.

For one of the ten simulation series, the bone structures are shownin Fig. 11. Fig. 11a shows the structure used as starting point for thesimulation of adaptation to the bone matrix mineralization of sam-

TM

D n

orm

aliz

ed to

gra

de 0

[-]

BV

/TV

nor

mal

ized

to g

rade

0 [-

]

1 1.5 2 2.5 30.94

0.96

0.98

1

1.02

1.04

Depth [mm]1 1.5

0.5

1

1.5

2

2.5

3

De

a) b)

K-L 0 K-L 1 K-L 2 K-L 3

Fig. 10. Experimental data organized according to K&L score. (a) Bone matrix mineralizationK–L 0. (b) Bone volume fraction (BV/TV) at different distances from the cartilage, for differe(TMD) and bone volume fraction (BV/TV) for the mean values per depth level for each K–L

cartilage bone marrow

a) b)

Fig. 11. Bone structures for the 2D simulations. (a) The starting point for the simulation of ggrade 4 OA. Blue indicates bone tissue that was only present at the end of the grade 0 simusimulation. It should be noted that green and blue areas do not indicate an increase in bonformed in the grade 4 simulation but not in the grade 0 simulation, or bone that was resor

ples with grade 0 and grade 4 cartilage degeneration, and inFig. 11b, the final structures for both situations are combined to high-light the differences between the bone structures. Although fromFig. 11b it can be seen that the final structures are different for theICRS grade 0 and grade 4 simulations, no subchondral sclerosisseems to have developed in the ICRS 4 simulation.

The decrease in bone matrix mineralization for the grade 4 simu-lations resulted in an increase in bone volume fraction compared tothe grade 0 simulations (Fig. 12). This increase varied between 9%at 1 mm depth and 4% at 3 mm depth (Fig. 12b). Although statisticallysignificant, this increase is much less than the experimentally ob-served increase of 69% at 1 mm depth to 24% at 3 mm depth (Fig. 7b).

Discussion

We aimed to investigate the relationships between cartilage de-generation, bone matrix mineralization, and bone volume fraction ata local level. With regard to the relationship between bone matrixmineralization and cartilage degeneration, we found that mineraliza-tion was lowest for the samples with the highest ICRS score. Althoughdifferences in bone matrix mineralization between the grade 4 carti-lage degeneration samples and other groups were statistically signif-icant, especially close to the cartilage, the maximum difference inmineralization with the grade 0 group was only 6%. This moderate de-crease in mineralization is in concurrence with two studies from Liand Aspden [6,24], who found a decrease of 6% in bone material den-

2 2.5 3

pth [mm]

920 940 960 9800.1

0.15

0.2

0.25

0.3

0.35B

V/T

V [-

]c)

K-L 4 K-L 0 K-L 1 K-L 2 K-L 3 K-L 4

TMD [mg HA cm-3]

(TMD) at different distances from the cartilage, for different K&L scores, normalized tont K–L scores, normalized to K–L 0. (c) Correlation between bone matrix mineralizationgrade. Pearson correlation coefficient −0.71, pb0.01.

grade 0 grade 4

rade 0 and grade 4 OA. (b) Differences and overlap between simulations of grade 0 andlation, and green indicates bone tissue that was only present at the end of the grade 4e volume compared to the starting point. Green areas represent either bone that wasbed in the grade 0 simulation, but not in the grade 4 simulation.

a) b)

ICRS 0 ICRS 4

1 1.5 2 2.5 30.5

0.52

0.54

0.56

0.58

0.6

0.62

BV

/TV

[-]

Depth [mm]1 1.5 2 2.5 3

0.9

0.95

1

1.05

1.1

1.15

BV

/TV

nor

mal

ized

to m

ild O

A [-

]

Depth [mm]

**

**** *

*

Fig. 12. Simulated bone volume fractions. (a) Bone volume fraction (BV/TV) at different distances from the cartilage. (b) Bone volume fraction normalized to grade 0. ⁎pb0.05,⁎⁎pb0.01, Student t-test.

1159L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

sity for the subchondral plate of OA bone compared to controls and adecrease of 8% for cancellous bone, using gravimetric measurements.The decrease in mineralization that we observed with decreasing dis-tance to the cartilage is another indication that cartilage degenerationand bone demineralization might be related, together with the obser-vation that the difference in mineralization between the ICRS grade 4and grade 0 samples was highest close to the cartilage. However, nostatistically significant difference in the slope could be demonstratedbetween groups for the relationship between mineralization anddepth level. While the depth-dependency of bone mineralization inOA has not been studied before in the detail as presented here, twoother studies point into the same directions as our findings. In bothstudies, a depth-dependent difference in matrix mineralization wasreported between OA and control bones, with a lower mineralizationfor OA bone, and the largest difference close to the cartilage [25,30].

The relationship between bone volume fraction and cartilage de-generation was opposite from the relationship between mineraliza-tion and cartilage degeneration. While bone matrix mineralizationwas decreased for ICRS grade 4 samples, bone volume fraction was in-creased for this group compared to the other ICRS grades. In addition,bone volume fraction decreased with increasing depth level, wherebone mineralization increased with increasing depth level. Althoughbone volume fraction has been reported to decrease with increasingdistance from the cartilage in bone covered by healthy cartilage aswell [8,44], the slope of the relationship between bone volume frac-tion and depth level seemed to increase with increasing ICRS grade(Fig. 8). The 69% increase in bone volume fraction that we observedbetween the ICRS grade 0 and ICRS grade 4 group is comparable todifferences between OA and normal bone reported in two studieson human bone tissue [7,8]. In correspondence with our data, inthese studies the volume fraction of OA bone was described to

Initial grid structure Normal mine

Fig. 13. Bone structures for the 3D simulations. No significant sclerosis occurred after decrebone structure between the simulations with normal and decreased mineralization.

decrease with increasing distance from the cartilage [7,8]. Whetherthe difference in bone volume fraction between OA and controlbones was higher close to the cartilage, was not quantified.

As expected from their opposite relations with cartilage degenera-tion and depth level, bone volume fraction and bonematrixmineraliza-tion were correlated. Although the Pearson correlation coefficient of−0.62 indicates that only a moderate part of the variation in bone vol-ume fraction can be explained by variations in bone matrix mineraliza-tion, the correlation was statistically significant (pb0.01). It is notsurprising that the correlation coefficient is low, since it is known thatother factors such as genetics, age, body weight, and activity level influ-ence bone volume fraction as well. When effects not related to depthlevel or ICRS grade were averaged out by taking the mean values perdepth level for each ICRS grade, the correlation coefficient increasedto −0.92, indicating that 85% (−0.922×100%) of the remaining varia-tion in bone volume fraction could be explained by variations in bonematrix mineralization. This high correlation coefficient offers support-ing evidence for the hypothesis that bone demineralization could bethe cause of bone sclerosis in OA. However, it does not exclude the pos-sibility that both the decrease inmineralization and the increase in bonevolume fraction were independently caused by an external factor in-volved in OA. Sclerosis could even indirectly be the cause of the lowmineralization instead of the other way around, as the high remodelingrate associated with sclerosis might result in relatively younger andtherefore less-mineralized bone tissue [18].

To obtain a better insight in the role that bone demineralizationmayplay in the increase in bone volume fraction, we performed bone adap-tation simulations. In these simulations, a depth-dependent decrease inmineralization led to a depth-dependent increase in bone volume frac-tion, in agreement with our experimental observations. However, de-spite the cubic relation between matrix mineralization and elastic

ralization Decreased mineralization

asing the matrix mineralization by 5% as can be observed from the small differences in

1160 L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

modulus of bone tissue, themaximum increase in bone volume fractionfor the grade 4 simulations compared to the grade 0 simulationswas 9%,which is much less than the 69% increase that we found in the experi-ments. To verify that the low increase in bone volume fraction in oursimulations was not a result of using 2D analyses or a high initial bonevolume fraction, we also investigated the effect of decreased minerali-zation in 3D, for a small cube (1.35 mm×1.35 mm×1.35 mm) of bonetissue with a physiologic bone volume fraction (Fig. 13). In this simula-tion, a 5% decrease in bone matrix mineralization led to an average in-crease of 7% in bone volume fraction, similar to the 2D simulations. Ina previously published 3D simulation study, bone adaptation in re-sponse to a tissue stiffness decrease of 20% led to an increase in bonevolume fraction of approximately 10% [28], which is quite similar toour simulation results.

As the model is theoretical, it cannot prove or disprove whether theincrease in bone volume fraction was caused by mechanoregulatedbone adaptation in response to decreased mineralization. However,the simulations do indicate that it is highly unlikely that the 6%–4% de-crease in bone demineralization can be responsible for the 69%–24% in-crease in bone volume fraction that we observed in the experiments.Therefore, we assume that other factors than bone demineralization,such as high joint loading or increased bone turnover contribute signif-icantly to the development of subchondral sclerosis. This idea is sup-ported by the differences in the changes in bone volume fraction thatwe observed when we classified the specimens according to ICRS andK–L scores. Locally, severe cartilage degeneration may cause additionalsclerosis, while on a more global scale, bone volume fraction increasesalready at a relatively early stage of the disease process. An exampleof local sclerosis underneath an area of full-thickness cartilage erosionthatwe observed in one of the specimens is shown in Fig. 14. A possibleexplanation for this localized depth-dependent increase in bone volumecould be the enhanced delivery of biochemical signals from the OA sy-novial fluid to the denuded subchondral bone.

A point of discussion in our study is that we treated samples fromthe medial and lateral side of the same joint as independent samples.Our reason for treating them as independent samples is that generally,the degree of cartilage degeneration and bone structural changes in OAdiffered markedly between the two sides of one tibia plateau. We con-sidered pair-wise comparison for our study, but since we were notable to obtain samples from both sides for 11 plateaus—due to limitedheight of one side of these plateaus as a result of the surgical cuttingplane or the presence of cysts—and because the difference in severityof OAwas not the same for all plateaus, we chose to treat them as inde-pendent samples instead.

To summarize our results, we found that local changes in sub-chondral bone mineralization and bone volume fraction only oc-curred underneath severely degenerated cartilage (defined by ICRSgrade 4), while at the joint scale bone volume fraction seemed to in-crease independent of decreased bone mineralization, and in earlierstages of the disease process (defined by low K–L grade). We did

Fig. 14. Osteochondral plug. Localized sclerosis is present underneath the area of full-thickness cartilage erosion.

not investigate the nature of the relationships between both mineral-ization and bone volume fraction with cartilage degeneration, but it ispossible the biochemical signaling from the degenerating cartilage orsynovial fluid might be involved. This may for example explain thedepth-dependency in the bone changes. Furthermore, we found thatbone volume fraction and bone matrix mineralization were inverselycorrelated. As a possible mechanism for the relationship betweenthese variables we considered mechanoregulated bone adaptation.Both the experimental data and the simulations indicated that adepth-dependent decrease in mineralization could cause a depth-dependent increase in bone volume fraction. However, quantitativelywe showed that it can probably only explain part of the subchondralsclerosis observed in OA. Therefore, we conclude that the sclerosis inOA bone is most likely the result of the combined effect of global fac-tors at the joint scale, and local factors at the tissue scale. Quantita-tively, the global factors seem to contribute more to the changes inbone volume fraction observed in OA patients.

Acknowledgments

This project is funded by the Royal Netherlands Academy of Artsand Sciences.

References

[1] Goldring SR. Role of bone in osteoarthritis pathogenesis. Med Clin North Am2009;93(1):25–35 xv.

[2] Mansell JP, Collins C, Bailey AJ. Bone, not cartilage, should be the major focus inosteoarthritis. Nat Clin Pract Rheumatol 2007;3(6):306–7.

[3] Karsdal MA, Leeming DJ, Dam EB, Henriksen K, Alexandersen P, Pastoureau P,Altman RD, Christiansen C. Should subchondral bone turnover be targeted whentreating osteoarthritis? Osteoarthritis Cartilage 2008;16(6):638–46.

[4] Hunter DJ, Spector TD. The role of bone metabolism in osteoarthritis. Curr Rheu-matol Rep 2003;5(1):15–9.

[5] Lajeunesse D, Reboul P. Subchondral bone in osteoarthritis: a biologic link with artic-ular cartilage leading to abnormal remodeling. Curr Opin Rheumatol 2003;15(5):628–33.

[6] Li B, Aspden RM. Composition and mechanical properties of cancellous bone fromthe femoral head of patients with osteoporosis or osteoarthritis. J Bone Miner Res1997;12(4):641–51.

[7] Bobinac D, Spanjol J, Zoricic S, Maric I. Changes in articular cartilage and subchon-dral bone histomorphometry in osteoarthritic knee joints in humans. Bone2003;32(3):284–90.

[8] Kamibayashi L, Wyss UP, Cooke TD, Zee B. Trabecular microstructure in the medialcondyle of the proximal tibia of patients with knee osteoarthritis. Bone1995;17(1):27–35.

[9] Ding M, Danielsen CC, Hvid I. Bone density does not reflect mechanical propertiesin early-stage arthrosis. Acta Orthop Scand 2001;72(2):181–5.

[10] Buckland-Wright C, Patel N. Pattern of advancement in the zone of calcified cartilagedetected in hand osteoarthritis. Osteoarthritis Cartilage 2000;8(Suppl A):S41–4.

[11] Pastoureau P, Leduc S, Chomel A, de Ceuninck F. Quantitative assessment of artic-ular cartilage and subchondral bone histology in the meniscectomized guinea pigmodel of osteoarthritis. Osteoarthritis Cartilage 2003;11(6):412–23.

[12] Lahm A, Kreuz PC, Oberst M, Haberstroh J, Uhl M, Maier D. Subchondral and tra-becular bone remodeling in canine experimental osteoarthritis. Arch OrthopTrauma Surg 2006;126(9):582–7.

[13] Ewald FC, Poss R, Pugh J, Schiller AL, Sledge CB. Hip cartilage supported by meth-acrylate in canine arthroplasty. Clin Orthop Relat Res 1982;171:273–9.

[14] Brandt KD, Myers SL, Burr D, Albrecht M. Osteoarthritic changes in canine articu-lar cartilage, subchondral bone, and synovium fifty-four months after transectionof the anterior cruciate ligament. Arthritis Rheum 1991;34(12):1560–70.

[15] Radin EL, Paul IL, Rose RM. Role of mechanical factors in pathogenesis of primaryosteoarthritis. Lancet 1972;1(7749):519–22.

[16] Mansell JP, Tarlton JF, Bailey AJ. Biochemical evidence for altered subchondral bonecollagen metabolism in osteoarthritis of the hip. Br J Rheumatol 1997;36(1):16–9.

[17] Lavigne P, Benderdour M, Lajeunesse D, Reboul P, Shi Q, Pelletier JP, Martel-PelletierJ, Fernandes JC. Subchondral and trabecular bone metabolism regulation in canineexperimental knee osteoarthritis. Osteoarthritis Cartilage 2005;13(4):310–7.

[18] Day JS, van der Linden JC, Bank RA, Ding M, Hvid I, Sumner DR, Weinans H. Adap-tation of subchondral bone in osteoarthritis. Biorheology 2004;41(3–4):359–68.

[19] Brandt KD. Transection of the anterior cruciate ligament in the dog: a model of os-teoarthritis. Semin Arthritis Rheum 1991;21(3, Suppl. 2):22–32.

[20] Neilson M, White A, Malik U, Morrison E, McGill PE, McDonald SW. Changes inbone architecture in the femoral head and neck in osteoarthritis. Clin Anat2004;17(5):378–91.

[21] Quasnichka HL, Anderson-MacKenzie JM, Bailey AJ. Subchondral bone and liga-ment changes precede cartilage degradation in guinea pig osteoarthritis. Bior-heology 2006;43(3–4):389–97.

1161L.G.E. Cox et al. / Bone 50 (2012) 1152–1161

[22] Hayami T, Pickarski M, Zhuo Y, Wesolowski GA, Rodan GA, Duong le T. Character-ization of articular cartilage and subchondral bone changes in the rat anterior cru-ciate ligament transection and meniscectomized models of osteoarthritis. Bone2006;38(2):234–43.

[23] Muehleman C, Green J, Williams JM, Kuettner KE, Thonar EJ, Sumner DR. The ef-fect of bone remodeling inhibition by zoledronic acid in an animal model of carti-lage matrix damage. Osteoarthritis Cartilage 2002;10(3):226–33.

[24] Li B, Aspden RM. Mechanical and material properties of the subchondral boneplate from the femoral head of patients with osteoarthritis or osteoporosis. AnnRheum Dis 1997;56(4):247–54.

[25] Mansell JP, Bailey AJ. Abnormal cancellous bone collagen metabolism in osteoar-thritis. J Clin Invest 1998;101(8):1596–603.

[26] Day JS, Ding M, van der Linden JC, Hvid I, Sumner DR, Weinans H. A decreasedsubchondral trabecular bone tissue elastic modulus is associated with pre-arthritic cartilage damage. J Orthop Res 2001;19(5):914–8.

[27] Behets C,Williams JM, ChappardD,Devogelaer JP,Manicourt DH. Effects of calcitoninon subchondral trabecular bone changes and on osteoarthritic cartilage lesions afteracute anterior cruciate ligament deficiency. J Bone Miner Res 2004;19(11):1821–6.

[28] van der Linden JC, Day JS, Verhaar JA, Weinans H. Altered tissue properties inducechanges in cancellous bone architecture in aging and diseases. J Biomech2004;37(3):367–74.

[29] Cox LG, van Rietbergen B, van Donkelaar CC, Ito K. Bone structural changes in os-teoarthritis as a result of mechanoregulated bone adaptation: a modeling ap-proach. Osteoarthritis Cartilage 2011;19(6):676–82.

[30] Grynpas MD, Alpert B, Katz I, Lieberman I, Pritzker KP. Subchondral bone in oste-oarthritis. Calcif Tissue Int 1991;49(1):20–6.

[31] Marticke JK, Hosselbarth A, Hoffmeier KL, Marintschev I, Otto S, Lange M,Plettenberg HK, Spahn G, Hofmann GO. How do visual, spectroscopic and biome-chanical changes of cartilage correlate in osteoarthritic knee joints? Clin Biomech2010;25(4):332–40 (Bristol., Avon.).

[32] Huiskes R, Ruimerman R, van Lenthe GH, Janssen JD. Effects of mechanical forces onmaintenance and adaptation of form in trabecular bone. Nature 2000;504(6787):704–6.

[33] Ruimerman R, Huiskes R, van Lenthe GH, Janssen JD. A computer-simulationmodel relating bone–cell metabolism to mechanical adaptation of trabecular ar-chitecture. Comput Methods Biomech Biomed Engin 2001;4(5):433–48.

[34] Ruimerman R, van Rietbergen B, Hilbers P, Huiskes R. A 3-dimensional computermodel to simulate trabecular bone metabolism. Biorheology 2003;40(1–3):315–20.

[35] Ruimerman R, Hilbers P, van Rietbergen B, Huiskes R. A theoretical framework forstrain-related trabecular bone maintenance and adaptation. J Biomech 2005;38(4):931–41.

[36] Tanck E, Hannink G, Ruimerman R, Buma P, Burger EH, Huiskes R. Cortical bonedevelopment under the growth plate is regulated by mechanical load transfer. JAnat 2006;208(1):73–9.

[37] Isaksson H, Grongroft I, Wilson W, van Donkelaar CC, van Rietbergen B, Tami A,Huiskes R, Ito K. Remodeling of fracture callus in mice is consistent with mechan-ical loading and bone remodeling theory. J Orthop Res 2009;27(5):664–72.

[38] Cox LG, Lagemaat MW, van Donkelaar CC, van Rietbergen B, Reilingh ML,Blankevoort L, van Dijk CN, Ito K. Bone 2011;49(4):762–8.

[39] Cox LG, van Rietbergen B, van Donkelaar CC, Ito K. Analysis of bone architecturesensitivity for changes in mechanical loading, cellular activity, mechanotransduc-tion, and tissue properties. Biomech Model Mechanobiol 2011;10(5):701–12.

[40] Currey JD. The effect of porosity and mineral content on the young's modulus ofelasticity of compact bone. J Biomech 1988;21(2):131–9.

[41] Rho JY, Ashman RB, Turner CH. Young's modulus of trabecular and cortical bonema-terial: ultrasonic and microtensile measurements. J Biomech 1993;26(2):111–9.

[42] Choi K, Kuhn JL, Ciarelli MJ, Goldstein SA. The elastic moduli of human subchon-dral, trabecular, and cortical bone tissue and the size-dependency of corticalbone modulus. J Biomech 1990;23(11):1103–13.

[43] van Rietbergen B, Weinans H, Huiskes R, Odgaard A. A new method to determinetrabecular bone elastic properties and loading using micromechanical finite-element models. J Biomech 1995;28(1):69–81.

[44] Patel V, Issever AS, Burghardt A, Laib A, Ries M, Majumdar S. Microct evaluation ofnormal and osteoarthritic bone structure in human knee specimens. J Orthop Res2003;21(1):6–13.