Particle-Induced Osteolysis Is Not Accompanied by Systemic Remodeling But Is Reflected by Systemic Bone Biomarkers R.D. Ross,1 A.S. Virdi,1,2 S. Liu,1 K. Sena,1 D.R. Sumner1,2 1 Anatomy and Cell Biology, Rush University Medical Center, Chicago, Illinois 2Orthopedic Surgery, Rush University Medical Center, Chicago, Illinois
Received 20 December 2013; accepted 6 February 2014 Published online 6 March 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22607
ABSTRACT: Particle-induced osteolysis is caused by an imbalance in bone resorption and formation, often leading to loss of implant fixation. Bone remodeling biomarkers may be useful for identification of osteolysis and studying pathogenesis, but interpretation of biomarker data could be confounded if local osteolysis engenders systemic bone remodeling. Our goal was to determine if remote bone remodeling contributes to biomarker levels. Serum concentrations of eight biomarkers and bone remodeling rates at local (femur), contiguous (tibia), and remote (humerus and lumbar vertebra) sites were evaluated in a rat model of particle-induced osteolysis. Serum CTX-1, cathepsin K, PINP, and OPG were elevated and osteocalcin was suppressed in the osteolytic group, but RANKL, TRAP 5b, and sclerostin were not affected at the termination of the study at 12 weeks. The one marker tested longitudinally (CTX-1) was elevated by 3 weeks. We found increased bone resorption and decreased bone formation locally, subtle differences in contiguous sites, but no differences remotely at 12 weeks. Thus, the skeletal response to local particle challenge was not systemic, implying that the observed differences in serum biomarker levels reflect differences in local remodeling. ß 2014 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 32:967–973, 2014. Keywords: osteolysis; biomarkers; bone remodeling; implant fixation; aseptic loosening
Aseptic loosening is the most common cause for failure of hip and knee implants and is often closely associated with the development of particle-induced periimplant osteolysis.1 The host response to particles leads to the suppression of osteoblast-mediated bone formation and elevated osteoclast activity.2 Biomarkers of bone metabolism are receiving increased attention as tools for the improved diagnosis of osteoporosis,3 osteoarthritis,4 and bone metastases.5 Biomarkers may be useful in early detection of osteolysis,6 although none have yet been validated for this purpose. Clinically, biomarkers such as C-telopeptide of type I collagen (CTX-1), N-terminal propeptide of type I procollagen (PINP), tartrate-resistant acid phosphatase 5b (TRAP 5b), osteoprotegerin (OPG), and receptor activator of nuclear kappa-B ligand (RANKL) are among those used to study loosening, often with contradictory results.6 Other molecules such as cathepsin K, which has been proposed as a marker for osteoporosis7 and is implicated in the pathogenesis of peri-implant osteolysis,8,9 and sclerostin, a key regulator of bone remodeling,10 have not been examined in the context of aseptic loosening and osteolysis. Animal models provide an opportunity to screen potential diagnostic biomarkers without the confounding effects of co-morbidities and pharmacological treatments and also facilitate histomorphometric measurements of bone remodeling at multiple skeletal The present address of S. Liu is Department of Orthopaedics, University of Maryland, Baltimore, MD, USA The present address of K. Sena is Department of Periodontology, Kagoshima University, Kagoshima, Japan Conflict of interest: None. Grant sponsor: Grainger Foundation. Correspondence to: D.R. Sumner (T: 312-942-5501; F: 312-9425744; E-mail: [email protected]) # 2014 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
sites.11 Considerable biomarker research has been performed in animal models of osteoporosis, leading to improved understanding of pathogenesis and how the skeleton responds to pharmaceuticals.12 Only three osteolysis-related animal models have thus far reported circulating biomarker concentration changes,13–15 and none have determined the relative contributions of local and remote bone remodeling, even though it is well known that local skeletal insult can lead to elevated bone remodeling at remote sites.16,17 Our lab recently developed a rat model of particleinduced peri-implant osteolysis to study the pathophysiology of implant loosening.14 Using this model, we found that peri-implant bone resorption increases and bone formation decreases in the presence of lipopolysaccharide (LPS) doped polyethylene particles, causing a loss of implant fixation strength.14,18 Further, the addition of particles caused a dramatic decrease in the concentration of serum osteocalcin and a large increase in serum CTX-1,14 suggesting that remodeling differences at sites distant from the implant may have contributed to the differences in serum concentrations. Thus, in this study we determined the contributions of remote bone remodeling to serum biomarker levels.
MATERIALS AND METHODS Experimental Design Rat tissues were obtained from a previously reported IACUC-approved study in which 6-month-old (400 g) adult male Sprague-Dawley rats had cylindrical titanium implants placed longitudinally in the medullary cavity of both femurs.18 For the current study, two groups of 12 animals from the original study were investigated. Group 1 received weekly intra-articular injections of vehicle (6% rat serum) within both knee joints (Particles), while Group 2 received weekly intra-articular injections of 50 ml of LPS-doped polyethylene particles (þParticles). The rats were euthanized JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
967
968
ROSS ET AL.
12 weeks after implant placement. Weight measurements were made at the onset of the experiment and weekly thereafter. Blood was collected via the tail vein at the time of surgery (week 0) and at weeks 3, 6, and 11 and then again at euthanasia through cardiac puncture. Rearing and ambulatory behaviors were measured at 8 and 11 weeks post surgery over a 30 min period using a previously reported technique.19 Animals received subcutaneous injections of calcein (10 mg/kg, Sigma-Aldrich, Saint Louis, MO) 10 and 2 days prior to euthanasia. Serum Biomarkers Serum was separated by centrifugation at 2,000 rpm (425g), placed in aliquots of 300 ml, and stored at 20˚C. Concentrations of osteocalcin (Immunodiagnostic Systems, Scottsdale, AZ, AC-12F1), N-terminal propeptide of type I procollagen (PINP, Immunodiagnostic Systems, AC-33F1), Cterminal telopeptides of type I collagen (CTX-1, Immunodiagnostic Systems, AC-06F1), tartrate-resistant acid phosphatase 5b (TRAP 5b, Immunodiagnostic Systems, SB-TR102), cathepsin K (Ucsn Life Science Inc., Wuhan, China, SEA267Ra), receptor activator of nuclear kappa-B ligand (RANKL, R&D Systems, Minneapolis, MN, MTR00), osteoprotegerin (OPG, Ucsn Life Science Inc., SEA108Ra), and sclerostin (R&D Systems, MSST00) were studied in the 12 weeks serum using commercially available ELISA kits, while CTX-1 measurements were also made longitudinally. Limited preliminary data made proper dilution difficult and limited serum volume did not allow for repeated dilution. Therefore, in several instances the concentrations of serum biomarkers reached the detection limit of the ELISA kit. In situations in which the upper detection limit was reached, the concentration of the biomarker is reported as this limit: 1,975 ng/ml for CTX-1; 26,250 pg/ml for cathepsin K; and 42 ng/ml for OPG. For osteocalcin, a lower detection limit (50 ng/ml) was reached; therefore, the concentration was set to one-half of the detection limit. microCT Scanning Femurs, tibiae, humeri, and the second lumbar (L2) vertebrae were dissected free of soft tissue and fixed in 10% neutral buffered formalin. Trabecular architecture and cortical geometry parameters were obtained using microCT (mCT 40, Scanco Medical, Bru¨ttisellen, Switzerland) at 70 kVp, 114 mA, and 300 ms integration time with a 16 mm isotropic voxel size. Trabecular data from the distal femurs were previously reported.18 Sample sites included cortical bone of the tibial and humeral diaphysis and trabecular bone in the distal femur, proximal tibia, and vertebra (Fig. 1). Trabecular bone volume per total volume (BV/TV), total cortical area (Tt.Ar), and medullarly area (Md.Ar) are reported within the main text, while specific trabecular architectural differences associated with BV/TV differences are presented in Supplemental Table 1. Histomorphometry Bone specimens were dehydrated in ethanol and embedded in PMMA (Sigma-Aldrich). Dynamic histomorphometry was performed under fluorescent microscopy (Nikon Eclipse 80i) using commercial software (OsteoMeasure, OsteoMetrics, Decatur, GA) to measure the bone formation rate per bone surface (BFR/BS). Tissues were then stained with basic fuchsin and toluidine blue to measure eroded surface per bone surface (ES/BS). All measurements were made and presented in accordance with standard conventions,20 using JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
Figure 1. Radiograph showing implant placement and sites where bone remodeling was measured. Local remodeling was measured in the distal femur (A), adjacent to the titanium implant ( ). Contiguous remodeling was measured in the proximal tibia (B) and the tibial diaphysis (C). Remote remodeling was assessed in the humeral diaphysis (D) and the second lumbar vertebra (E).
the same sample sites defined for microCT imaging (Fig. 1). The specific causes of the differences in BFR/BS are presented in Supplemental Table 2. Statistical Analysis SPSS (version 19) was used for all statistical analyses. Groups were compared using a non-parametric Mann–Whitney U-test to determine the effect of osteolysis (þParticles) to implant alone (Particles). The effect of time on the longitudinal CTX-1 data was assessed by Friedman’s test with each time point compared to 0 week with the Wilcoxson signed rank test. The between-group differences at each time point were assessed with a Student’s t-test following log transformation of the data. Receiver operating characteristic (ROC) analysis was performed to determine the accuracy of each of the biomarkers for differentiating the two groups. The accuracy is reported as the area under the ROC curve (AUC).
RESULTS Animals in both groups gained weight over the course of the experiment, but the Particles group gained more weight than did the þParticles group (57 19 g vs. 23 19, p < 0.001). No differences were found in either the rearing or ambulatory activity of the two groups at 8 weeks post surgery. However, by 11 weeks the þParticles animals showed reduced activity when compared to the Particles group (117 51 vs. 319 90 and 261 89 vs. 709 131, rearing and ambulatory events per 30 min, respectively, p ¼ 0.001 and p < 0.001).
BIOMARKERS OF OSTEOLYSIS
969
Figure 2. Circulating serum biomarker concentrations: (a) Osteocalcin, (b) CTX-1, (c) cathepsin K, (d) OPG, (e) PINP, (f) TRAP 5b, (g) RANKL, and (h) sclerostin. Each datum point represents one animal; the horizontal bar represents the group mean. Significance is indicated by a black bar, and the resulting p-value is given. When necessary, the detection limit is given by the dotted line.
Serum Biomarkers Five of the eight biomarker concentrations examined showed robust between-group differences (Fig. 2). Osteocalcin was suppressed, while CTX-1, cathepsin K, OPG, and PINP were elevated in the þParticles group (p < 0.001 for each marker), with little to no between-group overlap. However, TRAP 5b, RANKL, and sclerostin were not different between groups. Osteocalcin was accurate in differentiating the animals in the Particles group (AUC ¼ 1.0), while CTX1, PINP, cathepsin K, and OPG were accurate in differentiating animals in the þParticles group (AUC ¼ 1.0, 1.0, 1.0, and 0.9, respectively). TRAP 5b,
Figure 3. CTX-1 concentrations measured at 0, 3, 6, and 11 weeks post surgery. A significant time effect existed in the þParticles group (p < 0.001), but not for the Particles group (p ¼ 0.145). Between-group differences are indicated by asterisks; p < 0.05, p < 0.01.
RANKL, and sclerostin had relatively poor accuracy (AUC ¼ 0.571, 0.571, and 0.429, respectively). Longitudinal measurements show elevated CTX-1 levels in the þParticles group at 3, 6, and 11 weeks post surgery (Fig. 3). For the Particle group, no effect was found with time (p ¼ 0.145), but in the þParticle group there was a significant time effect (p < 0.001), with each time point significantly elevated compared to 0 week (p ¼ 0.028). Local Remodeling Results from our previous study demonstrated that the addition of polyethylene particles caused significant peri-implant bone loss, leading to decreased implant pull-out strength.18 Specifically, this loss of peri-implant trabecular BV/TV was associated with decreased BFR/BS and increased ES/BS18 (re-plotted in Supplemental Fig. S1). Contiguous Remodeling The tibia was used as a contiguous skeletal site as it constitutes part of the knee joint, the site of particle injections. Compared to the Particles group, trabecular BV/TV in the proximal tibia was significantly lower (p ¼ 0.028, Fig. 4a), and BFR/BS was significantly higher (p ¼ 0.028, Fig. 4b) in the þParticles group. ES/ BS was not different between the two groups. In the cortical bone of the tibial diaphysis, neither the total area nor the periosteal ES/BS differed in the two groups, but there was a significant reduction in periosteal BFR/BS in the þParticles group (p < 0.001, Fig. 5a–c). The endocortical compartment was not affected by the addition of particles (Fig. 5d–f). JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
970
ROSS ET AL.
Figure 4. Bone architecture and remodeling parameters of the trabecular bone in the proximal tibia: (a) microCT-derived trabecular bone volume fraction (BV/TV), (b) trabecular bone formation rate (BFR/BS), and (c) eroded surface (ES/BS). Each datum point represents one animal, and the horizontal bar represents the mean of each group. Significance is indicated by the black bar, and the resulting p-value is given.
Remote Remodeling No differences were found between the two groups in either the periosteal or endocortical compartments of the humeral diaphysis (Supplemental Fig. S2) or the trabecular bone of the lumbar vertebrae (Supplemental Fig. S3).
DISCUSSION Biomarkers of bone metabolism have received considerable attention for diagnosis of patients with systemic bone diseases such as osteoporosis3 and more localized
conditions such as osteoarthritis4 or bone metastases.5 In the context of joint replacement, bone biomarkers have been studied clinically primarily as a method to differentiate patients with stable from loosened implants with less specific focus on osteolysis,6 a major cause of loosening. Here, we showed that serum concentrations of CTX-1, cathepsin K, OPG, osteocalcin, and PINP are highly indicative of particle-induced peri-implant osteolysis. Further, altered bone remodeling following particle administration was limited to the local peri-implant site and the contiguous proximal
Figure 5. Cortical geometry and bone remodeling parameters of the tibial diaphysis: (a) microCT-derived total area, (b) periosteal bone formation rate (Ps.BFR/BS), (c) periosteal eroded surface (Ps.ES/BS), (d) microCT-derived medullary area, (e) endocortical bone formation rate (Ec.BFR/BS), and (f) endocortical eroded surface (Ec.ES/BS). Each datum point represents one animal, and the horizontal bar represents the mean of each group. Significance is indicated by the black bar, and the resulting p-value is given. JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
BIOMARKERS OF OSTEOLYSIS
tibial trabecular bone, with no remodeling differences at either the humeral or vertebral sites. For the one marker studied longitudinally (CTX-1), group differences were apparent as early as 3 weeks. Thus, our results show that serum concentrations of bone biomarkers reflect local peri-implant bone remodeling, providing further support for continued investigation of biomarkers as a means to diagnose and monitor particle-induced osteolysis. The elevated peri-implant bone resorption was consistent with the dramatic increases in serum concentrations of CTX-1 and cathepsin K. Previous studies of CTX-1 reported variable ability to discriminate between patients with stable and patients with loosened implants,21–23 while cathepsin K has not been studied as a biomarker for osteolysis. Interestingly, we found no difference in the concentration of TRAP 5b, which contradicts some previous clinical reports.24,25 TRAP 5b and cathepsin K are thought to be markers of osteoclast number. Cells other than osteoclasts were possibly responsible for the elevated cathepsin K expression observed in our study such as fibroblasts,26 mononuclear macrophages, and foreign body giant cells.8 RANKL concentration was not different between the þParticles and Particles groups, while OPG concentration was significantly elevated in the þParticles group. Clinically, OPG concentration is thought to increase, while RANKL concentration decreases during loosening.27 Yet, both in our study and a clinical study,28 the RANKL-to-OPG ratio decreased, which seems to be at odds with the consensus that increases in this ratio are associated with increased osteoclast differentiation.29 The presence of a decreased ratio suggests that serum levels of RANKL may be relatively insensitive to the tissue level of this signaling molecule, that a compensatory mechanism is induced by the particle challenge, or that alternative bone degradation pathways are active during particle-induced bone loss.26,30 The increase in PINP concentration reported here is different from the single clinical report of no difference,23 but is consistent with the elevated remodeling at the local and contiguous skeletal sites. Osteocalcin concentration was suppressed in the þParticles group, consistent with our previous report.14 The discordance between PINP and osteocalcin could imply that the bone formed during the peri-implant osteolysis process is largely undermineralized, a finding that was; however, not apparent in the histology. The high levels of cathepsin K or additional proteases that have been identified in the interfacial tissues of loose implants, including matrix metalloproteinases,31 could have degraded circulating osteocalcin into fragments32 that were not detected using the current assay. Serum sclerostin levels have not previously been studied in the context of peri-implant osteolysis, and we found no difference as a function of particle challenge in the current model.
971
Dynamic bone remodeling differences were restricted to sites in close proximity to the implant and location of particle administration. Specifically, we observed elevated bone resorption and suppressed bone formation in the peri-implant trabecular bone. A response also occurred in the proximal tibia, a contiguous skeletal site, presumably because the particles were injected into the synovial cavity of the knee joint. The suppressed bone formation rate at the periosteal surface of the tibial diaphysis may be a response to inflammatory signals induced at the knee joint, or perhaps caused by changes in hind limb loading. Indeed, we measured reduced activity outcomes following 11 weeks of particle injections, likely due to the continued inflammation in the peri-implant region. Dynamic remodeling effects were absent in the remote sites, indicating that the differences in circulating biomarkers reflect bone remodeling differences at the local and contiguous sites. A limitation of our study was that the volume of serum available for analysis prevented continued dilution of the highest biomarker responders: CTX-1, cathepsin K, and OPG. These markers reached upper detection limits, which made correlative statistical analysis impossible. However, we identified these biomarkers as being highly indicative of ongoing osteolysis and, therefore, future work will be directed at correlating biomarker concentrations to local disease pathogenesis. Similarly, the limited serum volume prevented analysis of additional candidate biomarkers that have been studied in the context of osteolysis,6 such as cross-linked N-telopeptide (NTX). The interpretation of the biomarker data would be enhanced by the addition of control groups in which non-doped and LPS-doped particles were injected into the knee joint of animals with and without implant placement. As described previously,14 LPS-doping was used to accelerate the development of osteolysis in vivo and to model the clinical situation in which endotoxin contamination is commonly found on orthopedic implant materials,33 as well as materials retrieved from patients diagnosed with aseptic loosening.34 Previous reports verified the accelerated osteolytic response to LPS-doped polyethylene35 and commercially pure titanium particles36,37 when compared to LPS-free controls. We did not compare LPS-doped to LPS-free particles, and we did not determine the effects of intra-articular injection of LPS in the absence of implants. These limitations in experimental design leave open the possibility that LPS contributed to the marker response by altering the ability to detect the markers independent of any effect on bone remodeling. These controls merit study, but the lack of a systemic remodeling effect in the animals treated with LPS-doped particles, the relatively rapid systemic clearance of unbound LPS,38 and the very low body burden14 lead us to interpret the marker levels as reflecting the direct effects of the particles on local bone-remodeling. JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
972
ROSS ET AL.
Our data suggest that localized remodeling due to particle-induced peri-implant osteolysis is detectable using systemic biomarkers, even though the involved region constituted
Received 20 December 2013; accepted 6 February 2014 Published online 6 March 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22607
ABSTRACT: Particle-induced osteolysis is caused by an imbalance in bone resorption and formation, often leading to loss of implant fixation. Bone remodeling biomarkers may be useful for identification of osteolysis and studying pathogenesis, but interpretation of biomarker data could be confounded if local osteolysis engenders systemic bone remodeling. Our goal was to determine if remote bone remodeling contributes to biomarker levels. Serum concentrations of eight biomarkers and bone remodeling rates at local (femur), contiguous (tibia), and remote (humerus and lumbar vertebra) sites were evaluated in a rat model of particle-induced osteolysis. Serum CTX-1, cathepsin K, PINP, and OPG were elevated and osteocalcin was suppressed in the osteolytic group, but RANKL, TRAP 5b, and sclerostin were not affected at the termination of the study at 12 weeks. The one marker tested longitudinally (CTX-1) was elevated by 3 weeks. We found increased bone resorption and decreased bone formation locally, subtle differences in contiguous sites, but no differences remotely at 12 weeks. Thus, the skeletal response to local particle challenge was not systemic, implying that the observed differences in serum biomarker levels reflect differences in local remodeling. ß 2014 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 32:967–973, 2014. Keywords: osteolysis; biomarkers; bone remodeling; implant fixation; aseptic loosening
Aseptic loosening is the most common cause for failure of hip and knee implants and is often closely associated with the development of particle-induced periimplant osteolysis.1 The host response to particles leads to the suppression of osteoblast-mediated bone formation and elevated osteoclast activity.2 Biomarkers of bone metabolism are receiving increased attention as tools for the improved diagnosis of osteoporosis,3 osteoarthritis,4 and bone metastases.5 Biomarkers may be useful in early detection of osteolysis,6 although none have yet been validated for this purpose. Clinically, biomarkers such as C-telopeptide of type I collagen (CTX-1), N-terminal propeptide of type I procollagen (PINP), tartrate-resistant acid phosphatase 5b (TRAP 5b), osteoprotegerin (OPG), and receptor activator of nuclear kappa-B ligand (RANKL) are among those used to study loosening, often with contradictory results.6 Other molecules such as cathepsin K, which has been proposed as a marker for osteoporosis7 and is implicated in the pathogenesis of peri-implant osteolysis,8,9 and sclerostin, a key regulator of bone remodeling,10 have not been examined in the context of aseptic loosening and osteolysis. Animal models provide an opportunity to screen potential diagnostic biomarkers without the confounding effects of co-morbidities and pharmacological treatments and also facilitate histomorphometric measurements of bone remodeling at multiple skeletal The present address of S. Liu is Department of Orthopaedics, University of Maryland, Baltimore, MD, USA The present address of K. Sena is Department of Periodontology, Kagoshima University, Kagoshima, Japan Conflict of interest: None. Grant sponsor: Grainger Foundation. Correspondence to: D.R. Sumner (T: 312-942-5501; F: 312-9425744; E-mail: [email protected]) # 2014 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
sites.11 Considerable biomarker research has been performed in animal models of osteoporosis, leading to improved understanding of pathogenesis and how the skeleton responds to pharmaceuticals.12 Only three osteolysis-related animal models have thus far reported circulating biomarker concentration changes,13–15 and none have determined the relative contributions of local and remote bone remodeling, even though it is well known that local skeletal insult can lead to elevated bone remodeling at remote sites.16,17 Our lab recently developed a rat model of particleinduced peri-implant osteolysis to study the pathophysiology of implant loosening.14 Using this model, we found that peri-implant bone resorption increases and bone formation decreases in the presence of lipopolysaccharide (LPS) doped polyethylene particles, causing a loss of implant fixation strength.14,18 Further, the addition of particles caused a dramatic decrease in the concentration of serum osteocalcin and a large increase in serum CTX-1,14 suggesting that remodeling differences at sites distant from the implant may have contributed to the differences in serum concentrations. Thus, in this study we determined the contributions of remote bone remodeling to serum biomarker levels.
MATERIALS AND METHODS Experimental Design Rat tissues were obtained from a previously reported IACUC-approved study in which 6-month-old (400 g) adult male Sprague-Dawley rats had cylindrical titanium implants placed longitudinally in the medullary cavity of both femurs.18 For the current study, two groups of 12 animals from the original study were investigated. Group 1 received weekly intra-articular injections of vehicle (6% rat serum) within both knee joints (Particles), while Group 2 received weekly intra-articular injections of 50 ml of LPS-doped polyethylene particles (þParticles). The rats were euthanized JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
967
968
ROSS ET AL.
12 weeks after implant placement. Weight measurements were made at the onset of the experiment and weekly thereafter. Blood was collected via the tail vein at the time of surgery (week 0) and at weeks 3, 6, and 11 and then again at euthanasia through cardiac puncture. Rearing and ambulatory behaviors were measured at 8 and 11 weeks post surgery over a 30 min period using a previously reported technique.19 Animals received subcutaneous injections of calcein (10 mg/kg, Sigma-Aldrich, Saint Louis, MO) 10 and 2 days prior to euthanasia. Serum Biomarkers Serum was separated by centrifugation at 2,000 rpm (425g), placed in aliquots of 300 ml, and stored at 20˚C. Concentrations of osteocalcin (Immunodiagnostic Systems, Scottsdale, AZ, AC-12F1), N-terminal propeptide of type I procollagen (PINP, Immunodiagnostic Systems, AC-33F1), Cterminal telopeptides of type I collagen (CTX-1, Immunodiagnostic Systems, AC-06F1), tartrate-resistant acid phosphatase 5b (TRAP 5b, Immunodiagnostic Systems, SB-TR102), cathepsin K (Ucsn Life Science Inc., Wuhan, China, SEA267Ra), receptor activator of nuclear kappa-B ligand (RANKL, R&D Systems, Minneapolis, MN, MTR00), osteoprotegerin (OPG, Ucsn Life Science Inc., SEA108Ra), and sclerostin (R&D Systems, MSST00) were studied in the 12 weeks serum using commercially available ELISA kits, while CTX-1 measurements were also made longitudinally. Limited preliminary data made proper dilution difficult and limited serum volume did not allow for repeated dilution. Therefore, in several instances the concentrations of serum biomarkers reached the detection limit of the ELISA kit. In situations in which the upper detection limit was reached, the concentration of the biomarker is reported as this limit: 1,975 ng/ml for CTX-1; 26,250 pg/ml for cathepsin K; and 42 ng/ml for OPG. For osteocalcin, a lower detection limit (50 ng/ml) was reached; therefore, the concentration was set to one-half of the detection limit. microCT Scanning Femurs, tibiae, humeri, and the second lumbar (L2) vertebrae were dissected free of soft tissue and fixed in 10% neutral buffered formalin. Trabecular architecture and cortical geometry parameters were obtained using microCT (mCT 40, Scanco Medical, Bru¨ttisellen, Switzerland) at 70 kVp, 114 mA, and 300 ms integration time with a 16 mm isotropic voxel size. Trabecular data from the distal femurs were previously reported.18 Sample sites included cortical bone of the tibial and humeral diaphysis and trabecular bone in the distal femur, proximal tibia, and vertebra (Fig. 1). Trabecular bone volume per total volume (BV/TV), total cortical area (Tt.Ar), and medullarly area (Md.Ar) are reported within the main text, while specific trabecular architectural differences associated with BV/TV differences are presented in Supplemental Table 1. Histomorphometry Bone specimens were dehydrated in ethanol and embedded in PMMA (Sigma-Aldrich). Dynamic histomorphometry was performed under fluorescent microscopy (Nikon Eclipse 80i) using commercial software (OsteoMeasure, OsteoMetrics, Decatur, GA) to measure the bone formation rate per bone surface (BFR/BS). Tissues were then stained with basic fuchsin and toluidine blue to measure eroded surface per bone surface (ES/BS). All measurements were made and presented in accordance with standard conventions,20 using JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
Figure 1. Radiograph showing implant placement and sites where bone remodeling was measured. Local remodeling was measured in the distal femur (A), adjacent to the titanium implant ( ). Contiguous remodeling was measured in the proximal tibia (B) and the tibial diaphysis (C). Remote remodeling was assessed in the humeral diaphysis (D) and the second lumbar vertebra (E).
the same sample sites defined for microCT imaging (Fig. 1). The specific causes of the differences in BFR/BS are presented in Supplemental Table 2. Statistical Analysis SPSS (version 19) was used for all statistical analyses. Groups were compared using a non-parametric Mann–Whitney U-test to determine the effect of osteolysis (þParticles) to implant alone (Particles). The effect of time on the longitudinal CTX-1 data was assessed by Friedman’s test with each time point compared to 0 week with the Wilcoxson signed rank test. The between-group differences at each time point were assessed with a Student’s t-test following log transformation of the data. Receiver operating characteristic (ROC) analysis was performed to determine the accuracy of each of the biomarkers for differentiating the two groups. The accuracy is reported as the area under the ROC curve (AUC).
RESULTS Animals in both groups gained weight over the course of the experiment, but the Particles group gained more weight than did the þParticles group (57 19 g vs. 23 19, p < 0.001). No differences were found in either the rearing or ambulatory activity of the two groups at 8 weeks post surgery. However, by 11 weeks the þParticles animals showed reduced activity when compared to the Particles group (117 51 vs. 319 90 and 261 89 vs. 709 131, rearing and ambulatory events per 30 min, respectively, p ¼ 0.001 and p < 0.001).
BIOMARKERS OF OSTEOLYSIS
969
Figure 2. Circulating serum biomarker concentrations: (a) Osteocalcin, (b) CTX-1, (c) cathepsin K, (d) OPG, (e) PINP, (f) TRAP 5b, (g) RANKL, and (h) sclerostin. Each datum point represents one animal; the horizontal bar represents the group mean. Significance is indicated by a black bar, and the resulting p-value is given. When necessary, the detection limit is given by the dotted line.
Serum Biomarkers Five of the eight biomarker concentrations examined showed robust between-group differences (Fig. 2). Osteocalcin was suppressed, while CTX-1, cathepsin K, OPG, and PINP were elevated in the þParticles group (p < 0.001 for each marker), with little to no between-group overlap. However, TRAP 5b, RANKL, and sclerostin were not different between groups. Osteocalcin was accurate in differentiating the animals in the Particles group (AUC ¼ 1.0), while CTX1, PINP, cathepsin K, and OPG were accurate in differentiating animals in the þParticles group (AUC ¼ 1.0, 1.0, 1.0, and 0.9, respectively). TRAP 5b,
Figure 3. CTX-1 concentrations measured at 0, 3, 6, and 11 weeks post surgery. A significant time effect existed in the þParticles group (p < 0.001), but not for the Particles group (p ¼ 0.145). Between-group differences are indicated by asterisks; p < 0.05, p < 0.01.
RANKL, and sclerostin had relatively poor accuracy (AUC ¼ 0.571, 0.571, and 0.429, respectively). Longitudinal measurements show elevated CTX-1 levels in the þParticles group at 3, 6, and 11 weeks post surgery (Fig. 3). For the Particle group, no effect was found with time (p ¼ 0.145), but in the þParticle group there was a significant time effect (p < 0.001), with each time point significantly elevated compared to 0 week (p ¼ 0.028). Local Remodeling Results from our previous study demonstrated that the addition of polyethylene particles caused significant peri-implant bone loss, leading to decreased implant pull-out strength.18 Specifically, this loss of peri-implant trabecular BV/TV was associated with decreased BFR/BS and increased ES/BS18 (re-plotted in Supplemental Fig. S1). Contiguous Remodeling The tibia was used as a contiguous skeletal site as it constitutes part of the knee joint, the site of particle injections. Compared to the Particles group, trabecular BV/TV in the proximal tibia was significantly lower (p ¼ 0.028, Fig. 4a), and BFR/BS was significantly higher (p ¼ 0.028, Fig. 4b) in the þParticles group. ES/ BS was not different between the two groups. In the cortical bone of the tibial diaphysis, neither the total area nor the periosteal ES/BS differed in the two groups, but there was a significant reduction in periosteal BFR/BS in the þParticles group (p < 0.001, Fig. 5a–c). The endocortical compartment was not affected by the addition of particles (Fig. 5d–f). JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
970
ROSS ET AL.
Figure 4. Bone architecture and remodeling parameters of the trabecular bone in the proximal tibia: (a) microCT-derived trabecular bone volume fraction (BV/TV), (b) trabecular bone formation rate (BFR/BS), and (c) eroded surface (ES/BS). Each datum point represents one animal, and the horizontal bar represents the mean of each group. Significance is indicated by the black bar, and the resulting p-value is given.
Remote Remodeling No differences were found between the two groups in either the periosteal or endocortical compartments of the humeral diaphysis (Supplemental Fig. S2) or the trabecular bone of the lumbar vertebrae (Supplemental Fig. S3).
DISCUSSION Biomarkers of bone metabolism have received considerable attention for diagnosis of patients with systemic bone diseases such as osteoporosis3 and more localized
conditions such as osteoarthritis4 or bone metastases.5 In the context of joint replacement, bone biomarkers have been studied clinically primarily as a method to differentiate patients with stable from loosened implants with less specific focus on osteolysis,6 a major cause of loosening. Here, we showed that serum concentrations of CTX-1, cathepsin K, OPG, osteocalcin, and PINP are highly indicative of particle-induced peri-implant osteolysis. Further, altered bone remodeling following particle administration was limited to the local peri-implant site and the contiguous proximal
Figure 5. Cortical geometry and bone remodeling parameters of the tibial diaphysis: (a) microCT-derived total area, (b) periosteal bone formation rate (Ps.BFR/BS), (c) periosteal eroded surface (Ps.ES/BS), (d) microCT-derived medullary area, (e) endocortical bone formation rate (Ec.BFR/BS), and (f) endocortical eroded surface (Ec.ES/BS). Each datum point represents one animal, and the horizontal bar represents the mean of each group. Significance is indicated by the black bar, and the resulting p-value is given. JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
BIOMARKERS OF OSTEOLYSIS
tibial trabecular bone, with no remodeling differences at either the humeral or vertebral sites. For the one marker studied longitudinally (CTX-1), group differences were apparent as early as 3 weeks. Thus, our results show that serum concentrations of bone biomarkers reflect local peri-implant bone remodeling, providing further support for continued investigation of biomarkers as a means to diagnose and monitor particle-induced osteolysis. The elevated peri-implant bone resorption was consistent with the dramatic increases in serum concentrations of CTX-1 and cathepsin K. Previous studies of CTX-1 reported variable ability to discriminate between patients with stable and patients with loosened implants,21–23 while cathepsin K has not been studied as a biomarker for osteolysis. Interestingly, we found no difference in the concentration of TRAP 5b, which contradicts some previous clinical reports.24,25 TRAP 5b and cathepsin K are thought to be markers of osteoclast number. Cells other than osteoclasts were possibly responsible for the elevated cathepsin K expression observed in our study such as fibroblasts,26 mononuclear macrophages, and foreign body giant cells.8 RANKL concentration was not different between the þParticles and Particles groups, while OPG concentration was significantly elevated in the þParticles group. Clinically, OPG concentration is thought to increase, while RANKL concentration decreases during loosening.27 Yet, both in our study and a clinical study,28 the RANKL-to-OPG ratio decreased, which seems to be at odds with the consensus that increases in this ratio are associated with increased osteoclast differentiation.29 The presence of a decreased ratio suggests that serum levels of RANKL may be relatively insensitive to the tissue level of this signaling molecule, that a compensatory mechanism is induced by the particle challenge, or that alternative bone degradation pathways are active during particle-induced bone loss.26,30 The increase in PINP concentration reported here is different from the single clinical report of no difference,23 but is consistent with the elevated remodeling at the local and contiguous skeletal sites. Osteocalcin concentration was suppressed in the þParticles group, consistent with our previous report.14 The discordance between PINP and osteocalcin could imply that the bone formed during the peri-implant osteolysis process is largely undermineralized, a finding that was; however, not apparent in the histology. The high levels of cathepsin K or additional proteases that have been identified in the interfacial tissues of loose implants, including matrix metalloproteinases,31 could have degraded circulating osteocalcin into fragments32 that were not detected using the current assay. Serum sclerostin levels have not previously been studied in the context of peri-implant osteolysis, and we found no difference as a function of particle challenge in the current model.
971
Dynamic bone remodeling differences were restricted to sites in close proximity to the implant and location of particle administration. Specifically, we observed elevated bone resorption and suppressed bone formation in the peri-implant trabecular bone. A response also occurred in the proximal tibia, a contiguous skeletal site, presumably because the particles were injected into the synovial cavity of the knee joint. The suppressed bone formation rate at the periosteal surface of the tibial diaphysis may be a response to inflammatory signals induced at the knee joint, or perhaps caused by changes in hind limb loading. Indeed, we measured reduced activity outcomes following 11 weeks of particle injections, likely due to the continued inflammation in the peri-implant region. Dynamic remodeling effects were absent in the remote sites, indicating that the differences in circulating biomarkers reflect bone remodeling differences at the local and contiguous sites. A limitation of our study was that the volume of serum available for analysis prevented continued dilution of the highest biomarker responders: CTX-1, cathepsin K, and OPG. These markers reached upper detection limits, which made correlative statistical analysis impossible. However, we identified these biomarkers as being highly indicative of ongoing osteolysis and, therefore, future work will be directed at correlating biomarker concentrations to local disease pathogenesis. Similarly, the limited serum volume prevented analysis of additional candidate biomarkers that have been studied in the context of osteolysis,6 such as cross-linked N-telopeptide (NTX). The interpretation of the biomarker data would be enhanced by the addition of control groups in which non-doped and LPS-doped particles were injected into the knee joint of animals with and without implant placement. As described previously,14 LPS-doping was used to accelerate the development of osteolysis in vivo and to model the clinical situation in which endotoxin contamination is commonly found on orthopedic implant materials,33 as well as materials retrieved from patients diagnosed with aseptic loosening.34 Previous reports verified the accelerated osteolytic response to LPS-doped polyethylene35 and commercially pure titanium particles36,37 when compared to LPS-free controls. We did not compare LPS-doped to LPS-free particles, and we did not determine the effects of intra-articular injection of LPS in the absence of implants. These limitations in experimental design leave open the possibility that LPS contributed to the marker response by altering the ability to detect the markers independent of any effect on bone remodeling. These controls merit study, but the lack of a systemic remodeling effect in the animals treated with LPS-doped particles, the relatively rapid systemic clearance of unbound LPS,38 and the very low body burden14 lead us to interpret the marker levels as reflecting the direct effects of the particles on local bone-remodeling. JOURNAL OF ORTHOPAEDIC RESEARCH JULY 2014
972
ROSS ET AL.
Our data suggest that localized remodeling due to particle-induced peri-implant osteolysis is detectable using systemic biomarkers, even though the involved region constituted
