The correlation of RANK, RANKL and TNFα expression with bone loss volume and polyethylene wear debris around hip implants

ARTICLE IN PRESS

0142-9612/$ - se

doi:10.1016/j.bi

�CorrespondE-mail addr

Biomaterials 27 (2006) 5212–5219

www.elsevier.com/locate/biomaterials

The correlation of RANK, RANKL and TNFa expression with boneloss volume and polyethylene wear debris around hip implants

Christopher A. Holdinga, David M. Findlayb,c, Roumen Stamenkovb, Susan D. Nealeb,Helen Lucasa, A.S.S.K. Dharmapatnia, Stuart A. Callaryb, Kush R. Shresthaa,b,c,

Gerald J. Atkinsb, Donald W. Howieb,c, David R. Haynesa,�

aDepartment of Pathology, The University of Adelaide, North Terrace, Adelaide, South Australia 5000, AustraliabDepartment of Orthopaedics and Trauma, The University of Adelaide, North Terrace, Adelaide, South Australia 5000, Australia

cHanson Institute, Adelaide, South Australia 5000, Australia

Received 22 January 2006; accepted 29 May 2006

Available online 27 June 2006

Abstract

This study investigates receptor activator NF-kB (RANK), RANK ligand (RANKL) and tumour necrosis factor (TNFa), key factors

regulating bone turnover, present in the tissues near peri-prosthetic osteolysis. Tissue was obtained from zones of peri-prosthetic

osteolysis from 11 patients undergoing revision of total hip prostheses, analysed preoperatively by high-resolution spiral multislice CT

using a metal artefact suppression protocol. Synovial tissue from 10 patients with osteoarthritis undergoing primary hip replacement was

used as control tissue. Immunohistochemical analysis of formalin fixed tissue sections demonstrated that RANK, RANKL and TNFawere strongly expressed by large multinucleated cells containing polyethylene wear debris in revision tissues. Control tissue stained

weakly for RANK, RANKL and TNFa. A strong statistical correlation (po0:02) was found between the five parameters, volume of

bone loss, polyethylene wear debris, RANK, RANKL and TNFa expression. Importantly, in vitro studies revealed that RANKL and

TNFa synergise to increase the volume of bone resorbed, by more than seven fold, when compared to the effect of either cytokine

treatment alone. This suggests that the interaction of TNFa and RANKL promotes osteoclast activity associated with polyethylene wear

and therapies targeting TNF activity may be useful to treat peri-implant osteolysis.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Osteolysis; Osteoclast; Wear debris; Polyethylene; Image analysis

1. Introduction

Failure of orthopaedic joint implants can be due tofactors such as infection, stress shielding, fracture, mechan-ical failure and poor surgical technique. However, the mostcommon reason for failure and revision surgery is asepticloosening [1], which is often associated with peri-prostheticosteolysis (PO). PO reduces bone around prostheses,leading to loss of fixation and this can make revision ofjoint replacements difficult, requiring complicated andexpensive surgery, and is frequently associated withconsiderable patient morbidity, and even mortality [2].PO is associated with wear particles liberated from the

e front matter r 2006 Elsevier Ltd. All rights reserved.

omaterials.2006.05.054

ing author. Tel.: +618 8303 3180; fax: +61 8 8303 4408.

ess: [email protected] (D.R. Haynes).

bearing surfaces of the prosthesis [3], stimulating aggressivegranulomatous lesions at the bone–cement or bone–prosthesis interface [4,5]. These lesions contain largenumbers of macrophages that have phagocytosed wearparticles, in turn producing mediators that enhanceosteoclast formation [6]. Moreover, osteoclasts arise fromcells of the monocyte/macrophage lineage [7] and cellsisolated from revision tissue readily form osteoclast-likecells capable of resorbing bone in culture [8].Polyethylene (PE) wear particles are thought to be the

most damaging particles stimulating peri-implant bone loss[9,10]. The activity of PE particles has been demonstratedin several animal models that mimic the human situation[11–14]. In vitro, macrophages exposed to PE particlesproduced increased amounts of many mediators that areassociated with the induction of osteoclast formation,

ARTICLE IN PRESSC.A. Holding et al. / Biomaterials 27 (2006) 5212–5219 5213

including PGE2, IL-1, IL-6, IL-8, MIP-1a, tumour necrosisfactor (TNFa) and RANKL [6,15–18].

Advances in our understanding of bone metabolismover the past decade have identified the activatedRANKL/RANK-receptor complex as a fundamentalfactor that promotes physiological osteoclast formation[19]. RANKL, expressed in bone on the surface ofosteoblast cells, interacts with RANK on the surface ofosteoclast precursor cells, which leads to their differentia-tion into mature osteoclasts, capable of resorption of thebone matrix. There is strong evidence that the activatedRANKL/RANK-receptor complex is also a significantfactor in stimulating bone loss in inflammatory diseases[20] and in peri-implant bone loss [21].

The tissues involved in PO have also been shown toexpress high levels of TNFa [22] and TNFa itself has beenclaimed to induce osteoclast formation in the absence ofRANKL [23]. Since PO tissues typically contain bothTNFa and RANKL in abundance, their combinedactivities need to be considered. In this study we have forthe first time used state of the art CT to target collection oftissues from defined osteolytic lesions around hip pros-theses. We have shown the co-localisation of TNFa,RANK and PE particles in multinucleated giant cells inthese tissues, together with the synergistic activities ofTNFa and RANKL in vitro to promote osteoclastformation and resorption.

2. Methods

2.1. Patients

Eleven patients were selected on the basis of radiographic evidence of

pelvic osteolysis adjacent to the acetabular or femoral component of a

total hip arthroplasty (THA) implant. All patients had either a cementless

Harris-Galante (Zimmer Ltd, Warsaw USA) or PCA (Stryker Orthopae-

dics, Mahwah, USA) THA prosthesis. Osteoarthritic (OA) tissue was

obtained from OA patients undergoing primary hip replacement.

2.2. CT scanning

High-resolution spiral multislice CT scans (Somatom Volume Zoom,

Siemens, Munich, Germany), with a metal-artefact suppression protocol,

was used to measure the volume of acetabular and femoral periprosthetic

osteolytic lesions. These methods were previously validated in vitro [24].

Osteolysis was defined as demarcated lytic lesions that were not linear and

were more than 3mm in diameter. Details of the scanning procedure have

been described previously [24]. CT Scans of the hip were taken from the

top of the sacro–iliac joint to 2 cm below the end of the femoral prosthesis.

Scan parameters were as follows: 140 kV, 200mA and 0.75-s spiral-3mm

slice with 1mm feed. An extended scale technique [25], (window level up to

30710 HU) was used to suppress the resulting metal artefact. Coronal and

sagittal reconstruction images were generated from the axial scanning

data. The CT images were exported in DICOM format and a True Life

Anatomyr model (True Life Anatomy, Adelaide, South Australia) was

created from this data. Briefly, with a reconstruction interval of 0.8mm

and a slice thickness of 1.25mm, patient’s data was recorded on CD in

DICOM format and downloaded onto a Picker 6000 workstation for

analysis. On the display monitor, cross sections of the defects from the CT

scans in the axial plane were identified on corresponding slices and the

perimeter of each defect was determined by tracing the inner border of the

defect using the computer mouse. Using existing CAM/CAM software

(3dsMAX), static 3-D models were created from the imported CT scan

data. The software generates a 3-D surface rendered image that allows full

3-D anatomical visualisation of the size and location of osteolytic lesions

relative to the existing bone and the prosthesis.

2.3. Tissue collection

Soft tissue samples from sites of osteolysis were taken from patients

undergoing surgery for removal or replacement of failed hip prostheses.

Control samples of synovial tissue were obtained from patients with

osteoarthritis (OA) undergoing primary knee surgery and hip capsule

from OA patients undergoing primary hip surgery. This study was

approved by the Royal Adelaide Hospital Human Ethics Committee in

accordance with National Health and Medical Research Council of

Australia Guidelines.

2.4. Tissue preparation

Tissue samples were fixed immediately in 10% buffered formalin for

paraffin embedding. Paraffin blocks were stored at room temperature.

Five-mm sections were cut using a microtome and mounted on

3-aminopropyltriethoxy-silane (APTS) (Sigma, St. Louis, MO) coated

glass slides. For RANKL staining antigen retrieval was used. Sections

were deparaffinised then placed in 10mM Sodium Citrate pH 6, heated to

95 1C for 30min and slowly cooled before staining as normal.

2.5. Immunohistochemistry

Serial sections were stained with mouse monoclonal antibodies (mAb):

anti-human RANK (clone 80707, a gift from R&D Systems Inc.,

Minneapolis, MN, USA), anti-human sRANKL (mAb 500-M46, Pepro-

Tech Inc., Rocky Hill, NJ, USA) and anti-human TNFa (mAb 610, R&D

Systems). To eliminate variability in immunohistochemical staining, all

sections to be stained with a particular antibody were processed at the

same time. The absence of staining was identical to that obtained when an

isotype-matched negative control antibody was used. Specificity was

further shown by blocking the binding of the RANK, RANKL and TNFaantibodies by pre-incubation of tissue sections for 24 h with an excess of

human recombinant RANK-Fc, RANKL or TNFa (R&D Systems Inc.),

respectively. Immunohistochemical detection of RANK, RANKL or

TNFa protein was carried out using a double enhancement method as

described previously [21]. A validated, semi-quantitative, (SQA) scoring

system was carried out by two independent observers to evaluate the

percentage of cell staining, using a score of 0 (�) to 4 (++++), detailed

previously [21]. Briefly, a score of ‘‘�’’ indicated that there were 0–10%

positive cells, ‘‘+’’ indicated 11–25% cells positive, ‘‘++’’ indicated

26–50%, ‘‘+++’’3 indicated 51–75% and (++++) indicated more

than 75% cells positive. A similar scoring system was used for PE particles

assessed under polarised light. The highest numbers seen in any of the

tissues was given a score of ‘‘+++’’ (example in Fig. 1(f)) ranging to the

absence of particles given a score of ‘‘�’’. At least a few particles could be

seen in all revision tissues analysed.

2.6. Identification of polyethylene particles

Oil Red O staining was used to detect small polyethylene particles that

could not be seen under polarised light. This was a variation on a method

previously published by Hansen et al. [26]. A 0.5% w/v solution was

prepared by dissolving 0.5 g of Oil Red O (Gurr BDH Chemicals, Poole,

England) in 100ml propylene glycol (1,2-propanediol; MP Biomedicals,

Aurora, OH, USA) by continuous stirring on a magnetic stirrer at 95 1C.

The solution was then filtered twice and stored at 60 1C. Tissue sections

were firstly de-waxed in xylene and rehydrated in a gradient series of

alcohol. Sections were then incubated in 100% propylene glycol for 5min

for a total of 2 changes. To rule out the possibility that the Oil Red O

ARTICLE IN PRESS

Fig. 1. CT and immunohistology images. (a) An example of a sagittal CT image of patient 3, demonstrating osteolytic lesions (arrows), from which tissue

was obtained for immunostaining. (b) The 3D-reconstructions of the prosthesis (grey) and osteolytic zones (red) generated from the digital CT image. (c)

RANK protein expression (red) in typical tissue obtained from the revision patients. The arrow denotes a large multinucleated cell expressing RANK. (d)

RANKL protein expression (red) in typical tissue obtained from revision patients. The arrow denotes a large multinucleated cell expressing RANKL. (e)

TNFa protein expression (red) in typical tissue obtained from the revision patients. The arrow denotes a large multinucleated cell expressing TNFa. (f) Thesame field of view as (c) but taken under polarised light to show bio-refringent polyethylene particles engulfed by the multinucleated cells (arrow). (g) Lack

of RANK protein expression in synovial tissue obtained from an OA patient. (h) Oil Red O staining of polyethylene particles (red) in tissue obtained from

a patient undergoing revision hip replacement surgery.

C.A. Holding et al. / Biomaterials 27 (2006) 5212–52195214

staining of particle storing cells is the result of lipid accumulation, tissue

sections were subjected to de-lipidation by incubating sections in

chloroform:methanol (2:1) for 1 h prior to staining. There was no change

in the level of Oil Red O stain (data not shown), indicating polyethylene

associated staining rather than lipid associated staining. Sections were

then stained with Oil Red O for 7min at room temperature. Staining was

then followed by differentiation in a solution of 85% propylene glycol in

distilled water for 2min, and then in distilled water for 5min. Finally,

tissue sections were counterstained with haematoxylin and mounted with

aqueous mounting medium.

2.7. In vitro osteoclast assay

Human peripheral blood mononuclear cells (PBMC) from 6 healthy

normal donors (The Australian Red Cross Blood Service, South Australia,

Australia) were isolated by diluting blood 1:3 in Hank’s Balanced Salt

Solution (HBSS), layering onto Ficoll-Paque separation media (Pharma-

cia Biotech, Uppsala, Sweden) and performing gradient centrifugation as

previously described [27]. Cells were seeded into 16mm 24-well trays at a

concentration of 2� 106 cells/ml in 0.5ml volumes onto sterilised 13mm

diameter (0.1mm thick) whale tooth dentine slices (a gift from The

Australian Customs Service, Canberra, Australia). Cells were allowed to

attach for 24 h at 37 1C, 5% CO2, after which nonadherent PBMC were

removed and remaining cells used to generate osteoclasts. Cultures were

maintained in 1ml of a-minimal essential medium (Invitrogen, Melbourne,

Victoria, Australia) containing 10% foetal calf serum (Invitrogen), 2mML-

glutamine (Invitrogen), 5 mg/ml penicillin and 50U/ml streptomycin

(Invitrogen). The medium was additionally supplemented with 10�8mol/L

1a, 25(OH)2D3, (Novachem, Melbourne, Victoria, Australia), and

10�8mol/L Dexamethasone (Fauldings, Adelaide, SA, Australia) and

25 ng/ml recombinant human M-CSF (Chemicon, Victoria, Australia), for

the 14 days of culture, with medium replenished every 3 days. In some

experiments, medium was supplemented with recombinant human

RANKL (Research Diagnostics Inc., Concord, MA, USA) and/or TNFa(R&D Systems) at the concentrations indicated.

2.8. Analysis of bone resorption in vitro

Functional evidence of osteoclast activity was determined by lacunar

resorption on whale tooth dentine slices. Following 14 days culture,

medium was removed and dentine washed with HBSS. Cells were removed

by addition of 0.05% (w/v) Trypsin in PBS/EDTA and incubated for at

least 1 h at 37 1C, 5% CO2. Slices were then washed with distilled water,

air-dried overnight, before being mounted onto stubs and carbon coated

for visualisation using a Philips XL-20 scanning electron microscope.

A Biorad MRC 1000UV (Hemel Hempstead, UK.) confocal micro-

scope was used to further analyse 8 pits from each disc to measure for

depth and diameter. Briefly, the confocal microscope laser scans through

the material at 0.4 mm increments (z series) to define the pit. The individual

images from the series were then smoothed using Confocal Assistant 40

version 4.02 (freeware). Then using the MPL-1000 program (Biorad,

Hemel Hempstead, UK.) the z series were used to draw a profile line that

represented the material surface across the centre of the imprint that was

used to calculate the depth and diameter of the resorption pits. The

diameter measurements were taken twice for each pit along what were

judged to be the longest and shortest sides and then averaged.

ARTICLE IN PRESSC.A. Holding et al. / Biomaterials 27 (2006) 5212–5219 5215

2.9. Statistical analysis

Differences in variables observed (pit depth, pit diameter and pit

number) between different treatments were analysed using nonparametric

Kruskall Wallis test in SPSS version 11.5 (Chicago, USA). Scores were

converted to RANK and the mean RANK for each group was compared.

In the case of significant differences was observed with Kruskall Wallis

test, non-parametric analysis Mann–Whitney U test then was used to

analyse differences between two independent groups (Table 2, Fig. 2).

Bivariate correlation among parameters observed (RANK SQA, TNF

SQA, PE debris and defect size) was investigated using Kendall’s tau b

test. Statistical significance was accepted when po0:05.

3. Results

3.1. Clinical and demographic features

Clinical and demographic data the patients undergoingrevision are presented in Table 1. There were a total of 5males and 6 females with and a mean age of 63 years. The

Fig. 2. Scanning electron microscopy was used to determine the number of pits

plus (a) 10 ng/ml RANKL, (b) 50 ng/ml RANKL, or (c) 10 ng/ml each of TNFaprofile of individual pits to calculate pit depth and volume. Typical pit profiles o

treated group, and (f), from the 10 ng/ml each of TNFa and RANKL-treated

Table 1

Patient and tissue data including time since primary hip replacement (TIA tim

Patient Sex Age (years) THA time

(years)

Tissue site D

(

1 M 66 13.6 Acetabular 3

2 M 58 12 Acetabular 1

3 F 50 12.7 Acetabular 8

4 M 73 12.9 Acetabular 7

5 F 45 14 Femoral 3

6 F 45 13 Acetabular 3

7 F 79 16.8 Femoral 2

8 M 69 16.2 Acetabular 1

9 F 82 15 Acetabular 9

10 M 63 15.6 Acetabular 4

11 F 64 12.8 Acetabular 3

Also shown are the corresponding scores for PE debris, of RANK, RANKL

time since implantation of the prosthesis was between 12and 16.8 years with a mean time of 14 years.

3.2. Peri-prosthetic osteolysis and tissue selection

CT was used to determine the volume of bone lossadjacent to hip THA implants. These CTs were also used,for the first time to our knowledge, to preoperatively targetsites of osteolysis for tissue samples collection. An exampleof such a CT scan and the derived 3-D reconstruction fromone patient is shown in Fig. 1. Tissue from these sites hadhigh levels of RANK, RANKL, TNFa, and polyethyleneparticles associated with large multinucleated cells, asdescribed below.

3.3. Immunohistochemistry

RANK protein was expressed at high levels in revisiontissues adjacent to osteolytic lesions (Fig. 1(c)) compared to

on dentine slices produced by PBMCs after 14 days treatment with M-CSF

and RANKL. Then Confocal microscope images (d) were used to define a

btained from the scans are shown in panels (e), from the 10 ng/ml RANKL

group.

e) and volume of peri-prosthetic osteolytic lesions (defect size)

efect size

cm3)

PE debris

score

RANK

score

TNFa score RANKL

score

.66 + + + +

.56 + + + +

.32 +++ +++ +++ ++

.37 ++ ++ + +

.51 + + � +

1.11 ++ ++ ++ +++

9.1 +++ +++ +++ ++

3.06 +++ +++ +++ +++

.29 ++ ++ ++ ++

7.14 +++ +++ ++ +++

3.12 +++ +++ +++ +++

and TNFa expression which were calculated as described in Section 2.

ARTICLE IN PRESSC.A. Holding et al. / Biomaterials 27 (2006) 5212–52195216

no staining for the presence of RANK protein in thecontrol OA tissues (Fig. 1(g)). Staining for RANK proteinin the revision tissues was strongly associated with largemultinucleated osteoclast-like cells. Under polarised light,polyethylene particles were visible within many of themultinucleated osteoclast-like cells expressing RANK(Fig. 1(f)).

The pattern of RANKL protein expression was verysimilar to that seen for RANK and was also stronglyassociated with large multinucleated osteoclast-like cells(Fig. 1(d)). In contrast, no staining for the presence ofRANKL was seen in the control OA tissues as previouslydescribed [21] (not shown).

The pattern of TNFa protein expression was very similarto that seen for RANK. TNFa was also strongly associatedwith large multinucleated osteoclast-like cells (Fig. 1(e)).Again, large numbers of polyethylene wear particles werepresent within the cytoplasm of these cells. In contrast, nostaining for the presence of TNFa was seen in the controlOA tissue (not shown). The intensity of staining forRANK, RANKL, TNFa, and the presence of particles issummarised for each patient in Table 1.

3.4. Oil red O staining for polyethylene

Compared to polarised light analysis there was asignificantly greater amount of PE debris observed intissue sections stained with Oil Red O indicating thepresence of many small PE particles not visible under lightmicroscopy (Fig. 1(h)). However, both Oil Red O andpolarised light gave similar results when analysing relativenumbers of PE particles. Like polarised light, analysisusing Oil Red O staining was associated with multi-nucleated osteoclast-like cells and macrophages. No stain-ing with Oil Red O was observed in any of the tissues fromOA patients (data not shown).

3.5. Correlation of defect size, PE debris, RANK and TNFaexpression

There were strong correlations between defect size, PEdebris, RANK, RANKL and TNFa expression (Table 2).The ‘‘r’’ values ranged from 0.65 to 0.72. All thecorrelations were statistically significant with many beinghighly significant (po0:01).

Table 2

Statistical correlation of the defect size, PE debris, RANK expression, TNFa

Defect size PE debris

Defect size r ¼ 0:67PE debris po0:01RANK po0:01 po0:01TNFa p ¼ 0:01 po0:01RANKL po0:01 p ¼ 0:016

The top right half of the table shows correlation coefficients and the bottom left

as described in Section 2.

3.6. Interaction of TNFa with RANKL in resorption

In vitro experiments were performed to evaluate theability of TNFa to stimulate osteoclast formation andactivity. Human PBMCs were grown for 14 days in thepresence of 25 ng/ml M-CSF and 10 ng/ml TNFa, or10 ng/ml RANKL or 50 ng/ml RANKL, or both TNFaand RANKL at 10 ng/ml. Resorption was analysed byscanning electron microscopy of dentine surfaces forevidence of pit formation.The analysis of numbers and morphology of the

resorption pits are summarised in Fig. 3. Treatment with50 ng/ml RANKL induced the largest number of pits, with10 ng/ml RANKL producing slightly less. When 10 ng/mlRANKL was added with 10 ng/ml TNFa significantlyfewer pits were observed than with 10 ng/ml RANKLalone. However, the morphology of the pits varied suchthat pits produced by the RANKL and TNFa co-treatmentwere significantly larger in both diameter (po0:05) anddepth (po0:05) than with the other treatments. Examplesof typical resorption pits formed under various conditionsseen under scanning election microscopy are shown inFigs. 2(a–c). Fig. 2(d) shows an image from confocalmicroscopy. The confocal microscopy profiles of typicalpits demonstrating the marked increase in diameter anddepth induced by the addition of TNFa are shown inFigs. 2(e and f). Assuming the shape of the pit to be asegment of a sphere (‘‘crater/dome like’’ shape) then thevolume of the pit can be calculated by the formulað3r2 þ d2

Þph=6. Since the pits produced by co-treatmentwith RANKL and TNFa were more than 2.5 times larger indiameter and 2.2 times deeper than the 10ng/ml RANKLtreatment group, the average volume resorbed in each pit inthe RANKL and TNFa co-treatment was more than 24times the volume for 10ng/ml RANKL treatment alone.Taking into account the numbers of pits, the overall volumeof dentine removed in the co-treatment group was more thatseven-fold greater with RANKL and TNFa co-treatmentthan for 10ng/ml RANKL alone and more than two-foldgreater than for 50ng/ml RANKL treatment (Fig. 3).

4. Discussion

It has previously been difficult to be certain that tissueobtained by surgeons at revision surgery for loose implants

expression and RANKL expression described in Table 1

RANK TNFa RANKL

r ¼ 0:67 r ¼ 0:65 r ¼ 0:81r ¼ 1:00 r ¼ 0:93 r ¼ 0:68

r ¼ 0:93 r ¼ 0:68po0:01 r ¼ 0:72p ¼ 0:016 p ¼ 0:01

shows the corresponding p values for the individual correlations calculated

ARTICLE IN PRESS

Fig. 3. Average pit number, diameter, depth, and volume of dentine

resorbed by PBMCs in the presence of added M-CSF together with

RANKL, at 10 or 50ng/ml or RANKL at 10ng/ml+TNF-a at 10ng/ml.

** ¼ po0:005 and * ¼ po0:05 when compared to TNFa (10ng/ml)

+RANKL (10 ng/ml) treatment.

C.A. Holding et al. / Biomaterials 27 (2006) 5212–5219 5217

is specifically from the osteolytic site identified radiologi-cally. For the first time, to our knowledge, we have beenable to precisely target tissue of interest preoperatively,

using newly developed CT with metal artefact suppression.In cases where sequential CT scans have been performed, ithas also been possible to identify sites of active osteolysisnear loose implants and therefore to sample from theseregions at surgery. Thus, the development of this metho-dology has allowed us, with the aid of the surgeon, toconfidently obtain aggressive osteolytic tissue for study.Our previous work [8,21] has demonstrated that

RANKL mRNA and protein is highly expressed inperiprosthetic tissue associated with osteolysis and thiswas confirmed in tissue obtained in this study. Theinteraction of RANKL with its cognate receptor, RANK,promotes osteoclast formation and activity and theabundant expression of RANK in multinucleated giantcells near sites of osteolysis is consistent with a role for thismolecule in the development of osteoclasts at sites of peri-implant osteolysis. Demonstration of the presence ofRANK protein in cells has been limited by the availabilityof suitable anti-RANK antibodies. Although mRNAexpression has been identified in periprosthetic tissuesfrom sites of osteolysis [28], some doubt remained as to thebiological significance because of the observation byAnderson and co-workers [29] that RANK mRNAexpression did not always correlate with RANK proteinexpression. We now show, with the use of specific anti-RANK antibodies, high levels of RANK expression inthese tissues. Cells that had engulfed large PE wearparticles were previously shown to express RANK mRNA[21] and this study confirms those observations at the levelof RANK protein. TNFa and RANKL were alsoassociated with large multi-nucleated cells containing largeand numerous small PE particles. Taken together, thesefindings suggest that large multinucleated cells containingPE can differentiate into osteoclast-like cells in the absenceof osteoblasts as they not only possess RANK receptorsbut also produce RANKL and abundant TNFa.It is widely believed that the RANKL/RANK interac-

tion, together with M-CSF are required for the formationof osteoclasts during normal bone turnover but this maynot be the case at sites of pathological osteoclastogenesis,such as that adjacent to orthopaedic implants. TNFa hasbeen demonstrated to stimulate osteoclast differentiationthrough a RANK independent mechanism [30] andSabokbar and colleagues [31] have shown that TNF inthe presence of M-CSF may stimulate osteoclast formationfrom cells derived from peri-implant tissues in vitro in theabsence of RANKL/RANK interaction. Other inflamma-tory mediators, such as IL-1b, have also been suggested tohave a direct role in enhancing the action of TNFa [31]. Anumber of other cytokines associated with osteoclastogen-esis are also found in peri-implant tissues [22], some ofwhich are likely to enhance osteoclast formation, directly,or indirectly by stimulating RANKL expression.Our demonstration of an interaction between TNFa and

RANKL to enhance human osteoclast activity is similar tothat described by Lam et al. [32] for murine osteoclasts. Inthis mouse system, while TNFa alone did not induce

ARTICLE IN PRESSC.A. Holding et al. / Biomaterials 27 (2006) 5212–52195218

osteoclastogenesis, TNFa did induce osteoclastogenesisboth in vitro and in vivo in the presence of permissive levelsof RANKL [32]. It is of interest that pH electrode studieshave shown the interface membrane around loosened totalhip replacement prostheses to be acidic and that acid-induced, cell-mediated bone resorption involves bothRANKL and TNFa [33]. The present study extends thiswork further by showing that the volume of mineralisedtissue resorbed by osteoclasts stimulated with RANKL andTNFa is much greater than for each factor alone and eventhan high levels of RANKL (eg. 50 ng/ml). The increasedresorption volume is consistent with the ability of TNFa toprolong the survival of osteoclasts, similar to that reportedto occur with IL-1b [34]. Thus, although it has also beenreported that TNFa inhibits the apoptotic effects ofbisphosphonates on osteoclasts [35], targeting TNFa mayreduce bone resorption in pathological bone resorptionaround implants.

Significantly, this study was able to show, for the firsttime, that there is strong correlation between the volume ofbone loss, the presence of large numbers of PE wearparticles and the presence of key osteolytic factors, RANK,RANKL and TNFa. This finding is consistent with anumber of other clinical reports and with growing evidencein vivo [12,13] and in vitro [6,15–18] for a major role for PEparticles as causative agents in PO. In vitro, macrophagesexposed to PE particles produced increased amounts of anumber of inflammatory mediators including TNFa [6].The frequent finding of PE particles within multinucleatedcells staining for TNFa, RANKL and RANK is consistentwith the correlation of the four parameters described inTable 2, and the engulfment of PE wear particles bymacrophages, as a causative agent in particle-inducedosteolysis.

5. Conclusion

This study identifies TNFa as a key mediator thatmarkedly enhances RANKL/RANK stimulated osteoclastformation in tissue adjacent to sites of aggressive peri-implant osteolysis. The results reinforce the findings ofSabokbar et al. [31] in identifying TNFa as a potentialtarget for therapy of osteolytic peri-implant lesions. Theuse of drugs that inhibit the action of TNFa is widelyaccepted for the treatment of rheumatoid arthritis inpatients that fail to respond to more traditional anti-inflammatory therapies. Recently a small clinical study hasbeen carried out in patients with peri-implant osteolysiswith the TNFa inhibitor, etanercept, however, due to thelow statistical power of the study the effect of anti-TNFatreatment on lysis could not be properly evaluated [36].Our findings suggest that a larger study may show abeneficial effect of blocking TNFa. While anti-TNFatherapy is not without side effects, it is possible that theseagents could provide effective treatment for peri-implantosteolysis in some cases. Targeting RANKL or TNFa

activity around an implant may reduce the need forrevision surgery.

Acknowledgements

The authors wish to thank the surgeons and nursing staffof the Department of Orthopaedic Surgery and Trauma,Royal Adelaide Hospital, for the provision of surgical bonesamples. This work was supported by grants from theNational Health and Medical Research Council ofAustralia, and the Department of Orthopaedics andTrauma, Royal Adelaide Hospital.

References

[1] Graves SE, Davidson D, Ingerson L, Ryan P, Griffith EC,

McDermott BF, et al. The Australian Orthopaedic Association

National Joint Replacement Registry. Med J Aust 2004;180:S31–4.

[2] Harris WH. The problem is osteolysis. Clin Orthop Relat Res 1995:

46–53.

[3] Howie DW, Cornish BL, Vernon-Roberts B. Resurfacing hip

arthroplasty. Classification of loosening and the role of prosthesis

wear particles. Clin Orthop Relat Res 1990:144–59.

[4] Santavirta S, Hoikka V, Eskola A, Konttinen YT, Paavilainen T,

Tallroth K. Aggressive granulomatous lesions in cementless total hip

arthroplasty. J Bone Joint Surg Br 1990;72:980–4.

[5] al Saffar N, Revell PA. Interleukin-1 production by activated

macrophages surrounding loosened orthopaedic implants: a potential

role in osteolysis. Br J Rheumatol 1994;33:309–16.

[6] Matthews JB, Green TR, Stone MH, Wroblewski BM, Fisher J,

Ingham E. Comparison of the response of three human monocytic

cell lines to challenge with polyethylene particles of known size and

dose. J Mater Sci Mater Med 2001;12:249–58.

[7] Quinn JM, Neale S, Fujikawa Y, McGee JO, Athanasou NA. Human

osteoclast formation from blood monocytes, peritoneal macrophages,

and bone marrow cells. Calcif Tissue Int 1998;62:527–31.

[8] Haynes DR, Crotti TN, Potter AE, Loric M, Atkins GJ, Howie DW,

et al. The osteoclastogenic molecules RANKL and RANK are

associated with periprosthetic osteolysis. J Bone Joint Surg Br 2001;

83:B902–11.

[9] Looney RJ, Boyd A, Totterman S, Seo GS, Tamez-Pena J, Campbell

D, et al. Volumetric computerized tomography as a measurement of

periprosthetic acetabular osteolysis and its correlation with wear.

Arthritis Res 2002;4:59–63.

[10] Orishimo KF, Claus AM, Sychterz CJ, Engh CA. Relationship

between polyethylene wear and osteolysis in hips with a second-

generation porous-coated cementless cup after seven years of follow-

up. J Bone Joint Surg Am 2003;85-A:1095–9.

[11] Howie DW, Vernon-Roberts B, Oakeshott R, Manthey B. A rat

model of resorption of bone at the cement–bone interface in the

presence of polyethylene wear particles. J Bone Joint Surg 1988;70-

A:257–63.

[12] Kim KJ, Kobayashi Y, Itoh T. Osteolysis model with continuous

infusion of polyethylene particles. Clin Orthop Relat Res 1998:46–52.

[13] Brooks RA, Sharpe JR, Wimhurst JA, Myer BJ, Dawes EN, Rushton

N. The effects of the concentration of high-density polyethylene

particles on the bone–implant interface. J Bone Joint Surg Br 2000;

82:595–600.

[14] Ren W, Yang SY, Wooley PH. A novel murine model of orthopaedic

wear-debris associated osteolysis. Scand J Rheumatol 2004;33:

349–57.

[15] Horowitz SM, Gonzales JB. Effects of polyethylene on macrophages.

J Orthop Res 1997;15:50–6.

[16] Rader CP, Sterner T, Jakob F, Schutze N, Eulert J. Cytokine

response of human macrophage-like cells after contact with

ARTICLE IN PRESSC.A. Holding et al. / Biomaterials 27 (2006) 5212–5219 5219

polyethylene and pure titanium particles. J Arthroplasty 1999;14:

840–8.

[17] Baumann B, Rader CP, Seufert J, Noth U, Rolf O, Eulert J, et al.

Effects of polyethylene and TiAlV wear particles on expression of

RANK, RANKL and OPG mRNA. Acta Orthop Scand 2004;75:

295–302.

[18] Garrigues GE, Cho DR, Rubash HE, Goldring SR, Herndon JH,

Shanbhag AS. Gene expression clustering using self-organizing maps:

analysis of the macrophage response to particulate biomaterials.

Biomaterials 2005;26:2933–45.

[19] Hofbauer LC, Khosla S, Dunstan CR, Lacey DL, Boyle WJ, Riggs

BL. The roles of osteoprotegerin and osteoprotegerin ligand in the

paracrine regulation of bone resorption. J Bone Miner Res 2000;15:

2–12.

[20] Haynes DR. Bone lysis and inflammation. Inflamm Res 2004;53:

596–600.

[21] Crotti TN, Smith MD, Findlay DM, Zreiqat H, Ahern MJ, Weedon

H, et al. Factors regulating osteoclast formation in human tissues

adjacent to peri-implant bone loss: expression of receptor activator

NFkappaB, RANK ligand and osteoprotegerin. Biomaterials

2004;25:565–73.

[22] Jiranek WA, Machado M, Jasty M, Jevsevar D, Wolfe HJ, Goldring

SR, et al. Production of cytokines around loosened cemented

acetabular components. J Bone Joint Surg 1993;75-A:863–79.

[23] Fuller K, Murphy PM, Kirstein B, Fox SW, Chambers TJ.

TNFalpha potently activates osteoclasts, through a direct action

independent of and strongly synergistic with RANKL. Endocrinol-

ogy 2002;143:1108–18.

[24] Stamenkov R, Howie D, Taylor J, Findlay D, McGee M, Kourlis G,

et al. Measurement of bone defects adjacent to acetabular

components of hip replacement. Clin Orthop Relat Res 2003:

117–24.

[25] Link TM, Berning W, Scherf S, Joosten U, Joist A, Engelke K, et al.

CT of metal implants: reduction of artifacts using an extended CT

scale technique. J Comput Assist Tomogr 2000;24:165–72.

[26] Hansen T, Otto M, Gaumann A, Petrow PK, Delank KS,

Kirkpatrick CJ, et al. Cathepsin K in aseptic hip prosthesis loosening:

expression in osteoclasts without polyethylene wear particles.

J Rheumatol 2001;28:1615–9.

[27] Haynes DR, Atkins GJ, Loric M, Crotti TN, Geary SM, Findlay

DM. Bidirectional signaling between stromal and hemopoietic cells

regulates interleukin-1 expression during human osteoclast forma-

tion. Bone 1999;25:269–78.

[28] Haynes DR, Crotti TN, Zreiqat H. Regulation of osteoclast activity

in peri-implant tissues. Biomaterials 2004;25:4877–85.

[29] Anderson DM, Marakovsky E, Billinglsey WL. A homologue of the

TNF receptor and its ligand enhance T-cell growth and dendritic

function. Nature 1997;390:175–9.

[30] Kobayashi K, Takahashi N, Jimi E, Udagawa N, Takami M, Kotake

S, et al. Tumor necrosis factor a stimulates osteoclast differentiation

by a mechanism independent of the ODF/RANKL-RANK interac-

tion. J Exp Med 2000;191:275–85.

[31] Sabokbar A, Kudo O, Athanasou NA. Two distinct cellular

mechanisms of osteoclast formation and bone resorption in

periprosthetic osteolysis. J Orthop Res 2003;21:73–80.

[32] Lam J, Takeshita S, Barker JE, Kanagawa O, Ross FP, Teitelbaum

SL. TNF-alpha induces osteoclastogenesis by direct stimulation of

macrophages exposed to permissive levels of RANK ligand. J Clin

Invest 2000;106:1481–8.

[33] Frick KK, LaPlante K, Bushinsky DA. RANK ligand and TNF-

alpha mediate acid-induced bone calcium efflux in vitro. Am J

Physiol Renal Physiol 2005;289:F1005–11.

[34] Lee ZH, Lee SE, Kim CW, Lee SH, Kim SW, Kwack K, et al. IL-

1alpha stimulation of osteoclast survival through the PI 3-kinase/Akt

and ERK pathways. J Biochem (Tokyo) 2002;131:161–6.

[35] Zhang Q, Badell IR, Schwarz EM, Boulukos KE, Yao Z, Boyce BF,

et al. Tumor necrosis factor prevents alendronate-induced osteoclast

apoptosis in vivo by stimulating Bcl-xL expression through Ets-2.

Arthritis Rheum 2005;52:2708–18.

[36] Schwarz EM, Campbell D, Totterman S, Boyd A, O’Keefe RJ,

Looney RJ. Use of volumetric computerized tomography as a

primary outcome measure to evaluate drug efficacy in the prevention

of peri-prosthetic osteolysis: a 1-year clinical pilot of etanercept vs.

placebo. J Orthop Res 2003;21:1049–55.