Matrix MetalloproteinaseDriven Endochondral ... - Wiley Online Library

JBMR

ORIGINAL ARTICLE

Matrix Metalloproteinase–Driven Endochondral Fracture Union Proceeds Independently of Osteoclast Activity Michelle M McDonald,1,2 Alyson Morse,1 Kathy Mikulec,1 Lauren Peacock,1 Paul A Baldock,2 Paul J Kostenuik,3 and David G Little1 1

The Children’s Hospital Westmead, Westmead, Australia Bone Biology Program, The Garvan Institute of Medical Research, Sydney, Australia 3 Amgen Inc., Thousand Oaks, CA, USA 2

ABSTRACT As new insights into the complexities of endochondral fracture repair emerge, the temporal role of osteoclast activity remains ambiguous. With numerous antiresorptive agents available to treat bone disease, understanding their impact on bone repair is vital. Further, in light of recent work suggesting osteoclast activity may not be necessary during early endochondral fracture union, we hypothesize instead a pivotal role of matrix metalloproteinase (MMP) secreting cells in driving this process. Although the role of MMPs in fracture healing has been examined, no directly comparative experiments exist. We examined a number of antiresorptive treatments to either block osteoclast activity, including the potent bisphosphonates zoledronic acid (ZA) and clodronate (CLOD), which work via differing mechanisms, or antagonize osteoclastogenesis with recombinant OPG (HuOPG‐Fc), comparing these directly to an inhibitor of MMP activity (MMI270). Endochondral ossification to union occurred normally in all antiresorptive groups. In contrast, MMP inhibition greatly impaired endochondral union, significantly delaying cartilage callus removal. MMP inhibition also produced smaller, denser hard calluses. Hard callus remodeling was, as expected, delayed with ZA, CLOD, and OPG treatment at 4 and 6 weeks, resulting in larger, more mineralized calluses at 6 weeks. As a result of reduced hard callus turnover, bone formation was reduced with antiresorptive agents at these time points. These results confirm that the achievement of endochondral fracture union occurs independently of osteoclast activity. Alternatively, MMP secretion by invading cells is obligatory to endochondral union. This study provides new insight into cellular contributions to bone repair and may abate concerns regarding antiresorptive therapies impeding initial fracture union. © 2013 American Society for Bone and Mineral Research. KEY WORDS: OSTEOCLAST; FRACTURE REPAIR; MATRIX METALLOPROTEINASES; ENDOCHONDRAL OSSIFICATION; REMODELING

Introduction

E

ndochondral bone repair comprises numerous stages of cellular recruitment, differentiation, tissue formation, removal, and remodeling. This complex sequence of events can be separated into four key stages to restore the injury to its original structure: the initial inflammatory response; soft callus (cartilage) formation; hard (bone) callus formation; and remodeling.(1,2) The biology of these stages has been well characterized; yet the temporal role that osteoclastic resorption plays requires clarification. With the widespread successful use of bisphosphonates (BPs) to address bone loss in numerous skeletal diseases, a number of investigations have examined the potential of these agents to augment healing or possibly impede repair.(3–10) BP treatment consistently leads to increases in fracture callus size and

resistance to refracture.(3–8,11) However, impaired resorption during the later stages of hard callus remodeling is consistently evident in these studies. In particular, hard callus remodeling is impeded during late fracture repair in the presence of both a single bolus dose of zoledronic acid (ZA) or continuous weekly dosing with ZA, with both regimes showing significant delays in hard callus remodeling, even as late as 26 weeks after fracture initiation.(3) In addition, long‐term treatment with continuous incadronate significantly altered hard callus remodeling in a similar rat model,(4) and although delayed and less‐frequent doses of ibandronate limited remodeling delays, they were still present in these modified dosing regimes.(9) Similarly, significant delays in hard callus remodeling were noted with both alendronate and denosumab; denosumab is a monoclonal antibody to receptor activator of NF‐kB ligand (RANKL) and hence an inhibitor of osteoclast formation.(10) Despite their

Received in original form October 2, 2012; revised form January 10, 2013; accepted January 23, 2013. Accepted manuscript online February 13, 2013. Address correspondence to: Michelle M McDonald, PhD, Bone Biology Program, The Garvan Institute of Medical Research, 384 Victoria Road, Darlinghurst, Sydney, NSW 2010, Australia. E‐mail: [email protected] Journal of Bone and Mineral Research, Vol. 28, No. 7, July 2013, pp 1550–1560 DOI: 10.1002/jbmr.1889 © 2013 American Society for Bone and Mineral Research

1550

ability to delay hard callus remodeling, BPs and denosumab did not, however, diminish callus structural strength, and in some cases led to enhanced biomechanical integrity at the expense of a larger callus.(3,5,6,8,11) Clearly, osteoclast resorption is pivotal to the hard callus remodeling process that serves to restore original bone architecture. Of interest to us from these many studies of antiresorptives in fracture repair is the fact that not one of these reports demonstrates evidence of impaired bony union. Thus, the preponderance of experimental evidence contrasts with the widely accepted assumption that osteoclasts are vital to the removal of unmineralized cartilaginous soft callus concomitant with the ensuing hard callus formation, in a process known as endochondral ossification.(12) Endochondral union was not impaired in the osteopetrotic ia/ia rat in which osteoclasts are nonfunctional,(13) nor was it impaired in mice treated with RANK; Fc (murine RANK fused to the fragment crystallizable, Fc region region of human IgG1).(14) Hence, although osteoclast resorption is key to hard calls remodeling, it is evident the osteoclasts role in soft callus removal may not be imperative. It therefore remains to be clarified which cells or processes may be compensating for osteoclasts in this process when antiresorptives are administered. Skeletal regeneration involves the intricate cooperation of numerous proteins, many of which have been assessed for their effect during fracture healing. Recently, several studies have alluded to a key role of matrix metalloproteinases (MMPs) in the process of endochondral fracture union.(15) MMPs comprise a family of more than 20 proteinases that are capable of modifying the extracellular matrix (ECM) in numerous developmental processes.(16) MMPs are highly associated with cellular migration in a number of remodeling environments, degrading ECM and in doing so releasing active growth factors.(17) Importantly to endochondral ossification at the growth plate, MMPs are secreted by vascular endothelial cells during angiogenesis and are vital to the degradation of unmineralized cartilage matrix.(18) More specifically, collagenase‐3 (MMP‐13) effectively degrades type II collagen, a major constituent of cartilage, in particular the cartilage soft callus of a healing limb. MMP‐13–deficient mice have altered endochondral growth and expanded growth plates.(19) Further, the temporal change in MMP‐13 expression during fracture repair coincides with the replacement of cartilage by bone.(20) Gelatinase‐9 (MMP‐9), on the other hand, removes denatured collagen II fragments along with collagens IV, IV, and XI. Initial studies uncovered the role of MMP‐9 in fetal development, with MMP‐9/ mice showing severe inhibition in the apoptosis of hypertrophic chondrocytes and hence expanded growth plates.(21) More recently MMP‐9 expression was shown throughout the entire process of adult skeletal regeneration, with MMP‐9/ adult mice displaying abnormal mechanisms of bone repair.(15) The most striking finding was that the cartilage callus in these mice appeared to be resistant to degradation, with larger unmineralized cartilage calluses that persisted into the latter part of healing, well after the unmineralized cartilage in the wild‐type mice had been replaced with bone. Yet healing in these mice eventually progressed by circumventing endochondral ossification and instead employing intramembranous ossification processes. Together, these findJournal of Bone and Mineral Research

ings suggest that the early stages of fracture repair rely heavily on the activity of a number of MMPs, but do not clarify the cell(s) vital for delivering these enzymes. In addition to cartilage degradation, MMPs contribute significantly to the activities of osteoclasts, participating in organic matrix solubilization at specific skeletal sites, as well as regulating the initiation and termination of bone resorptive processes.(22) Interestingly, mice lacking either RANK or RANKL, hence presenting with no osteoclasts, demonstrate similar alterations in endochondral bone growth as do MMP deficient mice. This suggests osteoclast MMP production may be important to growth plate cartilage removal; however, in light of results from antiresorptive investigations in bone repair,(3,5,6,8,11) we speculate that this is not the case during endochondral fracture union. Taken together, these investigations suggest that the stages of early fracture union may proceed in the absence of osteoclast activity, whereas MMP activity is vital to cartilage callus degradation. Published data, however, cannot rule out osteoclastic involvement in the process, particularly with the BPs that can block their bone resorptive capacity but not cartilage resorption via production of MMPs. Further, osteoclasts or chondroclasts are commonly located at the chondro‐osseous junction and can actively participate in cartilage degradation through secretion of a number of proteases, including MMPs.(23,24) Hence, in MMP knockout (KO) fracture studies, osteoclasts are still present and may also be impaired in their cartilage remodeling capacity; again, these studies do not specifically point to the cell type essential to this process. Finally, when mice were treated with RANK:Fc to abolish osteoclast formation, endochondral fracture union was not impaired.(14) Therefore, a level of ambiguity pertains to the osteoclasts’ exact role in achieving endochondral fracture repair, and in the current climate of extensive BP administration, and more recently the emerging use of denosumab in patients with fractures, clarity on this issue is of high clinical importance. In this study, we therefore sought to directly determine the necessity of either osteoclastic resorption or MMP activity in endochondral fracture union in one controlled investigation. Using a rat closed fracture model, we assessed the process of endochondral ossification under the influence of antiresorptive BP administration versus the specific blockade of osteoclast formation by osteoprotegerin (OPG‐Fc), and compared these treatments to the effects of a broad spectrum MMP inhibitor, MMI270.

Subjects and Methods Study design All animal experiments were performed after approval from the Westmead Hospital Animal Ethics Committee, protocol number 4081. Nine‐week‐old male Wistar rats were randomly assigned into the five treatment groups of 30, as outlined in Table 1. Two BPs were included in the study: ZA (kindly provided by Novartis Pharma), a third‐generation potent BP that inhibits osteoclast activity primarily via the mevalonate pathway without necessarily causing apoptosis; and clodronate (CLOD), an earlier compound shown to induce osteoclast apoptosis upon uptake via creation

MMP DRIVEN ENDOCHONDRAL UNION PROCEEDS IN THE ABSENCE OF OSTEOCLASTS

1551

Table 1. Rat Fracture Study Treatment groups Saline Clodronate Zoledronic acid huOPG‐Fc MMI270

– 45.0 mg/kg twice per week 0.01 mg/kg twice per week 10 mg/kg twice per week 120 mg/kg daily

2 weeks

4 weeks

6 weeks

Total

10 10 10 10 10

10 10 10 10 10

10 10 10 10 10

150

Values are number of rats in each group; a total of 150 rats were studied. huOPG‐Fc ¼ human osteoprotegerin; MMI270 ¼ inhibitor of matrix metalloproteinase activity.

of an ATP analogue.(25,26) ZA and CLOD were both administered twice per week at a dose of 0.01 mg/kg and 45 mg/kg, respectively, via subcutaneous (s.c.) injection. These doses were designed to approximate clinically relevant doses over a 6‐week period. Human recombinant osteoprotegerin (HuOPG‐Fc) (OPG; kindly provided by Amgen Inc.) was also administered twice per week via s.c. injection at a dose of 10 mg/kg, a dose confirmed to abrogate osteoclasts in rats.(27) Finally, inhibition of MMP activity was achieved using a broad spectrum MMP inhibitor (MMI270; kindly provided by Novartis Pharma) at a dose of 120 mg/kg via daily s.c. injections, a dose determined by a pilot dose‐finding study in our laboratory that produced extensive growth‐plate widening (data not shown). A saline injection group was used as a control. All treatment agents were commenced 2 days prior to fracture and continued until the harvest time points. Rats were euthanized at 2, 4, and 6 weeks postfracture from each of the five treatment groups, with 10 rats at each time point (Table 1). Throughout the experiment, the rats were housed in independent cages and were administered with a pellet diet and water ad libitum.

Fracture healing Closed femoral fractures were induced using a modified version of the Einhorn apparatus as used previously by our group.(3,28) Rats were anesthetized by intraperitoneal injection of ketamine (75 mg/kg) and xylazine (9 mg/kg). A small incision was made at the knee and a 0.5‐mm Kirschner wire was inserted into the medullary canal of the femur. A fracture was produced by three‐ point bending and the position and lack of comminution confirmed by X‐ray using a Faxitron MX‐20 digital X‐ray system (Faxitron X‐ray Corp., Wheeling, IL, USA). In order to measure dynamic bone formation rates, 10 mg/kg calcein was administered 10 and 3 days prior to euthanasia. Serum samples were obtained at baseline and at 2‐week, 4‐week, and 6‐week time points to examine serum tartrate‐resistant acid phosphatase 5b (TRAP5b) levels in the saline and OPG‐treated rats.

Radiological analysis Plain radiographs of harvested femurs were taken and examined by a blinded observer to assess radiological union. A score of 0 (no bridging), 1 (partial bridging; ie, one cortex bridged), or 2 (complete union; ie, two cortices bridged) was given to each sample and outcomes between groups compared. Femur length was also determined from radiographs of intact left femurs.

1552

MCDONALD ET AL.

Quantitative computed tomography (QCT) scans were then performed on all fractured samples using a Stratec XCT Research SA scanner (Stratec Medizintechnik GmbH, Pforzheim, Germany) to determine mineralized tissue content (BMC), volumetric bone mineral density (vBMD), and bone volume of each callus. Intact left femurs were also scanned in the region corresponding to the fracture callus on the right limb to allow for systemic assessment of agents on intact cortical bone. Furthermore, the proximal regions of left tibias were scanned to assess systemic effects of agents on metaphyseal bone. One sample, selected as representative of the mean from QCT data, from each group at the 6‐week time point was then allocated to be scanned using high‐resolution micro‐CT (µCT) scanning using a desktop microtomographic imaging system (Skyscan 1172; Skyscan NV, Kontich, Belgium). Three‐dimensional (3D) analysis of each representative sample and 3D imaging were performed using VGStudio Max imaging software (Version 1.2; Volume Graphics, Heidelberg, Germany).

Histological analysis Once radiographic analyses were completed, all fracture samples were bisected in the sagittal plane to produce lateral and medial halves. Samples were fixed in 4% paraformaldehyde (PFA) for 4 to 5 hours at room temperature (RT) and then incubated overnight at 4°C before being transferred to 70% ethanol for analysis. Medial halves were decalcified using 0.5 M EDTA, 2.5% PFA for 3 days at 4°C, followed by 4 to 6 weeks in 0.34 M EDTA at 4°C, before being processed to paraffin. Lateral halves were processed undecalcified to resin for histological analysis. Sections 5 µm thick were obtained from both halves of each callus at the cut surface to obtain central sections, ideally with the empty intramedullary pin site in view. Paraffin sections were stained using Safranin O/Light Green to examine cartilage callus content and determine histological union rates, and TRAP stain was used to localize TRAP‐positive osteoclastic cells. All calluses were graded as a united or not united and the area of each callus that contained avascular cartilage or fibrous tissue was calculated and presented as a percentage of the entire callus size. Resin sections were stained with von Kossa for mineralized tissue morphology and architecture or cleared to assess fluorescently labeled bone.

Statistical analysis Statistical analysis was performed using one‐way analysis of variance (ANOVA) with least squared differences for n ¼ 10 or Journal of Bone and Mineral Research

more. For categorical analysis of union rates, Fisher’s exact test was used (SPSS Inc., Chicago, IL, USA).

Results Radiographs Union rates were not different across ZA, CLOD, or HuOPG‐Fc (OPG) treatment groups when compared to saline at each time point examined, with all fractures united by 6 weeks. However, MMI270 treatment produced significant reductions in radiological union rates at both 4 and 6 weeks postfracture as compared to saline treatment, with less than 10% united at 4 weeks and only 30% of fractures united at 6 weeks (p < 0.01, Fisher’s exact test, Fig. 1A, E).

QCT QCT scans of fracture samples measured changes in callus BMC and vBMD, and in mineralized callus volume (volume) at each of

the three time points. At 4 weeks, ZA and CLOD treatment induced a 30% to 54% increase in callus BMC and a 35% to 44% increase in callus volume as compared to saline (p < 0.01; Fig. 1B, C). OPG treatment led to increases of 92% in BMC and 68% in volume compared to saline at this time (p < 0.01; Fig. 1B, C). Furthermore, OPG increased callus vBMD by 12% compared to saline (p < 0.01; Fig. 1D). By 6 weeks, ZA treatment caused a 78% increase in callus BMC and a 64% increase in callus volume (p < 0.01; Fig. 1B, C). Interestingly, however, callus vBMD was increased 9% with ZA by this stage (p < 0.01, Fig. 1D). OPG treatment at 6 weeks showed only a 97% increase in callus BMC and a 66% increase in callus volume compared to saline, leading to a 19% increase in callus vBMD (p < 0.01; Fig. 1B–D). In contrast, treatment of fractures with the MMP inhibitor, MMI270, led to a 26% reduction in BMC and 42% less callus volume compared to saline at 2 weeks (p < 0.01; Fig. 1B, C). At 4 weeks, this 26% reduction in volume remained (p < 0.05; Fig. 1C). By 6 weeks, however, no differences in volume or BMC were noted, compared to saline. However, callus vBMD was

Fig. 1. (A) Bar chart of the percentage of samples united from radiographic analysis of samples harvested at 2, 4, and 6 weeks. By 6 weeks all samples in all groups except MMI270 had achieved union. At 4 and 6 weeks MMI270 led to decreased union rates compared to all other treatment groups (p < 0.01, Fisher’s exact test). Bar chart of (B) mean callus bone mineral content (BMC), (C) mean callus bone volume, (D) mean callus volumetric bone mineral density (vBMD) at 2, 4, and 6 weeks. p < 0.01 compared to saline, †p < 0.05 compared to saline. Error bars are 1 SD. (E) Representative radiographs of samples from each group at 6 weeks showing delayed union in MMI270 treated samples and enhanced mineralized callus density and volume at 6 weeks in OPG, CLOD, and ZA treatment groups compared to Saline.

Journal of Bone and Mineral Research


1553

higher in MMI270‐treated samples as compared to saline at every time point, with increases of 26%, 9%, and 12% observed at 2, 4, and 6 weeks, respectively (p < 0.01; Fig. 1D). µCT scans were performed on one representative sample, as determined from the mean value for each group from QCT data at 6 weeks. 3D reconstructions of these scans clearly demonstrate the results obtained from radiological assessment and QCT scans, with delayed union in MMI270‐treated samples, and enhanced bone volume and density with OPG‐treated, ZA‐ treated, and CLOD‐treated samples compared to saline (Fig. 2).

Histological union At 4 weeks, histological union rates were greater than 80% for each of the ZA, CLOD, OPG, and saline groups (data not shown). By 6 weeks, all fractures treated with ZA, CLOD, OPG, and saline had united, supporting the radiological findings. However, at both 4 and 6 weeks, MMI270 treatment drastically inhibited fracture union, with only approximately 20% union at both 4 and 6 weeks. Representative safranin O stained sections reflecting this data are shown in Figure 3A, with higher power images depcting differences with OPG and MMI270 in particular (Fig. 3B, C). Similarly, only MMI270 showed significant effects on the percentage of callus ossified tissue, with significant reductions at each of the three time points examined (Figs. 3A; 4A). On Safranin O/Light Green–stained sections, only MMI270 treatment affected the callus area of cartilage tissue compared to saline, with significant retention of avascular cartilage callus in the MMI270 group at both 4 and 6 weeks postfracture (p < 0.02; Figs. 3A–C; 4B).

Histological hard callus morphology Upon closer inspection, distinct differences in hard callus morphology were noted in the histological sections. Although calluses treated with OPG had reached union, a large amount of Safranin O–stained cartilage matrix was retained in the calluses, even after 6 weeks postfracture (Fig. 3B, C). To confirm that this matrix was mineralized cartilage as compared with that retained in the MMI270 calluses, undecalcified sections of the corresponding one‐half of each callus were stained with von Kossa for mineralized tissue formation (Fig. 3B). As expected, the OPG‐treated calluses were densely mineralized, even in areas where cartilage matrix remnants had been retained (Fig. 3C). Unlike in MMI270‐treated samples, no chondrocytes remained in the OPG‐treated samples. This confirms that the morphological

change of excessive mineralized cartilage callus retention with OPG was due to inhibition of remodeling of this primary callus by osteoclasts, and not due to negative effects on avascular unmineralized cartilage tissue removal. Such retention of primary mineralized callus has been previously demonstrated by our group with BP treatment in this rat model, and was noted again in the current study; however, ZA and CLOD did not show retained primary callus to the same extent as was seen with OPG (Fig. 3B). In contrast, the MMI270 samples retained a large area of unmineralized soft callus (cartilage tissue) (Fig. 3B, C; Fig. 4A, B). Because extensive delays in hard callus remodeling were noted with the OPG, ZA, and CLOD groups, we examined mineralized tissue properties on histology sections stained with von Kossa. Hard callus bone volume fraction (BV/TV) was significantly increased by 56% at 2 and 4 weeks and 65% at 6 weeks in OPG‐treated groups compared to saline (p < 0.01; Fig. 4C). Similarly, ZA increased hard callus BV/TV by 39% at 4 weeks and 42% at 6 weeks and CLOD by 21% at 4 and 6 weeks compared to saline (p < 0.01; Fig. 4C). These results correlated well with QCT data for these groups. Interestingly, although endochondral union was delayed with MMI270, hard callus BV/TV was increased 45% at 4 weeks and 67% at 6 weeks compared to saline (p < 0.01; Fig. 4C). These results again correlate well with increased vBMD from QCT scans and are suggestive of impaired hard callus remodeling with MMI270, even in the smaller delayed healing calluses resulting from this agent. When examining trabecular bone architecture in the hard callus, differences in callus trabecular thickness (Tb.Th) and number (Tb.N) were seen with the four treatment groups (Fig. 4D, E). Tb.Th was increased by 36% at 2 weeks, 40% at 4 weeks, and 42% at 6 weeks with OPG treatment compared to saline (p < 0.05). Tb.N with OPG treatment was increased 15% at 2 weeks (p < 0.01) then returned to control levels. ZA and CLOD treatment did not increase Tb.Th; however, Tb.N was increased by 37% at 4 weeks and 56% at 6 weeks with ZA, and 13% at 2 weeks (p < 0.05) and 52% at 6 weeks with CLOD (p < 0.01). MMI270 increased Tb.Th by 20% at 6 weeks compared to saline (p < 0.01). Clearly, hard callus remodeling was hindered in all treatment groups compared to saline in this study. Analysis of osteoclast parameters at 4 and 6 week time points confirmed this, whereas analysis at week 2 was not possible due to the early stage of repair and small numbers of active osteoclasts on mature bone surfaces. Osteoclast number/bone surface (Oc.N/BS) was

Fig. 2. Representative 3D reconstructions of samples µCT‐scanned from each group at 6 weeks. Note the failure of union in MMI270 group and enhanced callus volume and density with OPG, ZA, and CLOD treatment compared to Saline.

1554

MCDONALD ET AL.


Fig. 3. (A) Representative images of Safranin O/Light Green–stained sections from each group at 2, 4, and 6 weeks and von Kossa–stained sections from each group at 6 weeks only. Note the clear retention of red‐stained cartilage soft callus at 4 and 6 weeks in MMI270‐treated samples compared to all other groups. At 6 weeks von Kossa sections show bony bridging (mineralized tissue in black) in all groups except for MMI270 with unmineralized callus persisting at the fracture site. Original magnification 2.0. (B) Higher magnification images of Safranin O/Light Green–stained and von Kossa–stained representative sections from MMI270‐treated and OPG‐treated samples at 6 weeks. Red‐stained unmineralized cartilage callus is retained in MMI270 calluses, whereas the red‐stained cartilage remnants of primary bony callus are mineralized (black) on von Kossa–stained sections. Original magnifications 2.5 and 10. (C) Representative images of Safranin O/Light Green–stained sections at higher magnification, demonstrating the distinct differences between primary spongiosa with retained calcified cartilage matrix in the OPG‐treated sample and unmineralized, avascular cartilage soft callus in the MMI270‐treated sample. Original magnification 40.



1555

Fig. 4. Bar charts showing mean values for (A) percent ossified tissue, (B) avascular callus area, (C) hard callus BV/TV, (D) callus trabecular thickness. and (E) callus trabecular number. Error bars are 1 SD. p < 0.01 compared to saline, †p < 0.05 compared to saline.

reduced by 99% to 100% in OPG samples at 4 and 6 weeks compared to control (p < 0.01), such that osteoclasts were essentially absent at all time points (Fig. 5A). CLOD treatment resulted in decreases in Oc.N/BS of 62% at 4 weeks and 81% at 6 weeks compared to saline (p < 0.01). In contrast, at 4 weeks ZA treatment increased Oc.N/BS by 60% compared to saline (p < 0.01). This increase was reversed by week 6, but at no time point was Oc.N/BS reduced by ZA. Finally, MMI270 treatment did not alter Oc.N/BS compared to saline. Histological assessment of bone resorption correlated well with serum measures of TRAP5b levels in OPG‐treated animals. OPG treatment for 6 weeks more than halved baseline serum levels of TRAP5b from 5.31 U/L to 2.34 U/L, compared to a much smaller decrease in this time frame in saline‐treated samples from 3.94 U/L to 3.49 U/L. Assessment of bone formation revealed transient alterations in the percentage of fluorescently labeled bone area, ie, the

percent of callus surface actively forming bone, within the hard callus of saline‐treated samples. As can be seen in Fig. 5, bone formation was robust at 2 weeks, when modeling is responsible for new bone formation rather than remodeling. Interestingly, treatments had minimal effects on overall labeling at this time point. Bone formation at later time points was, however, markedly reduced compared to 2 weeks in the saline group, presumably reflecting remodeling‐based bone formation. At these later stages when bone formation was remodeling dependent, antiresorptive treatments significantly reduced the ratio of labeled bone in the hard callus. ZA treatment reduced the percent of labeled bone area by 42% at 4 weeks and 56% at 6 weeks compared to saline (p < 0.01; Fig. 5B). CLOD treatment reduced the percent of labeled bone area by 50% at 4 weeks compared to saline (p < 0.01). OPG treatment reduced the percent of labeled bone tissue area by 41% at 4 weeks and 64%

Fig. 5. Bar charts of mean values for (A) callus osteoclast number/bone surface at 4 and 6 weeks only and (B) percentage of labeled bone area of bone area at 2, 4, and 6 weeks. Error bars are 1 SD. p < 0.01 compared to saline, †p < 0.05 compared to saline.

1556

MCDONALD ET AL.


at 6 weeks compared to saline (p < 0.01). Finally, MMI270 treatment did not reduce bone formation, instead increasing it by 21% at 4 weeks compared to saline (p < 0.01; Fig. 5B).

postmenopausal osteoporosis patients also had no negative impact on union or fracture healing despite potent inhibition of osteoclasts.(31)

Discussion

MMP activity from non‐osteoclastic cells is essential to achieving endochondral union

Cartilage callus remodeling proceeds in the absence of osteoclasts but not in the absence of active MMPs The removal of avascular cartilage tissue and its replacement with vascularized bone are pivotal to achieving bony union at a fracture site. Until recently, the osteoclast or TRAP‐positive chondroclast, were viewed as the key cells driving this process, with mice lacking OPG and hence showing an abundance of TRAP‐positive cells exhibiting advanced cartilage callus removal.(29) Current research has challenged this theory, showing no negative effects of potent antiresorptive agents on soft cartilage callus remodeling to achieve union.(3–10) In addition, a lack of resorbing osteoclasts in ia/ia rats, and impaired osteoclast formation in mice treated with RANK:Fc both failed to alter normal cartilage callus remodeling.(13,14) In the current study we tested agents with three distinct mechanisms of action. First, two BPs, ZA and CLOD, which ablated osteoclast activity with uncertain effects on their MMP production; second, recombinant OPG, which abolished osteoclast formation and hence all residual MMPs that osteoclasts produce; and last, MMI270, which blocked MMPs from all potential cell sources. Although ZA treatment did not reduce osteoclast number, the resorptive capacity of these cells was markedly reduced as evidenced by the increased hard callus bone volume (Figs. 1C; 2). Further, clodronate treatment, which did reduce osteoclast number, also produced a larger mineralized callus compared to saline (Figs. 1C; 2). Finally, OPG treatment, which abrogated TRAP‐positive cells, reducing serum TRAP5b levels by over fivefold at 6 weeks compared to saline treatment, was associated with a near twofold increase in hard callus bone volume compared to saline (Figs. 1C; 2). Importantly, in all cases of impaired osteoclast activity and formation, the removal and replacement of the soft cartilage callus tissue with primary hard callus was not impaired. With ZA, CLOD, and OPG treatment, union was achieved at a similar rate to saline treatment (Figs. 1E; 2), and the extent of cartilage callus was also progressively reduced at a similar rate (Figs. 3; 4A, B). In contrast, MMP inhibition across all cells impaired endochondral union, with cartilage callus remaining in many samples when all control mice were united. Hence, this pivotal stage of fracture repair is impaired when MMPs are blocked, but proceeds normally both in the presence of inactive osteoclasts that could still produce MMPs, and with a complete absence of these cells and their MMPs with OPG treatment. These findings are strongly supported by numerous preclinical investigations, although we are the first to perform such a direct comparison in one study.(3,4,6,8,24) Notably, these findings also support clinical data indicating the absence of fracture healing complications with BPs or the RANKL inhibitor denosumab during clinical fracture repair at the doses tested. As such, one recent trial examining single‐dose ZA in patients with tibial osteotomies showed no negative effects on union rate.(30) Denosumab treatment of Journal of Bone and Mineral Research

In the absence of osteoclastic resorption, we hypothesized that compensatory ECM remodeling during endochondral ossification occurred by the MMP‐dependent vascular endothelial cell. Positioned at the vanguard of vascular invasion into the unmineralized cartilage, vascular endothelial cells can degrade collagen matrix, allowing subsequent bone formation.(21,32) The most abundant component of the ECM, fibrillar collagen, is cleaved by collagenolytic MMPs, such as MMP‐9 and MMP‐13, driving angiogenesis and cartilage degradation during the early stages of fracture repair.(19,20) Previous work has linked several MMPs to endochondral repair with increasing levels of these proteases shown up to postfracture day 14.(33) Further, mice lacking even one member of this protease family demonstrate clear delays in endochondral fracture repair.(15,19,34) Treatment of rat femoral fractures with the potent broad spectrum MMP inhibitor MMI270, in the current study, did indeed impair endochondral fracture repair, significantly hindering union (Figs. 1E; 2). Excessive retention of unmineralized cartilage callus tissue was noted in this group and hence this led to a reduced percent of mineralized callus at the 6‐week end point (Figs. 1A; 3A–C; 4A, B). In contrast, OPG treatment had no negative effects on union rates or soft callus remodeling, despite near‐total ablation of osteoclasts (and whatever MMPs they would normally produce). This strongly suggests that non‐ osteoclastic cells are an important source of MMPs during early fracture healing, and that MMPs themselves, rather than osteoclasts per se, are critical for these functions. Confirmation of this outcome could be achieved using genetic mouse technology with conditional deletion of MMPs in specific cell populations. Colnot and colleagues(15) have examined in detail specific MMP activity during endochondral fracture repair, using genetic mouse models null for these proteases, illustrating that both MMP‐9 and MMP‐13 are pivotal to the degradation and removal of unmineralized cartilage matrix to achieve union.(15,19,34) More recently, however, the absence of MMP‐2 was associated with normal endochondral union, but significant delays in hard callus remodeling.(35) This indicates that some but not all MMPs may be essential for driving cartilage callus remodeling, others more specifically involved in bone resorption by osteoclasts.(36) After numerous investigations pinpointed a role for MMPs during endochondral fracture repair, recent attention has turned to how MMP levels can be used clinically to assess the progression of fracture repair.(37)

Hard callus remodeling was impaired with anti‐osteoclastic agents Although endochondral union was reached exclusive of osteoclast activity with ZA, CLOD, and OPG treatment, delays in hard callus remodeling were extensive, in particular with OPG treatment. Mineralized callus tissue properties were increased by


1557

almost twofold with OPG treatment at the 6‐week time point (Figs. 1C; 2; 4). The woven bone callus formed up to the time of union often contains cores of mineralized cartilage remnants. These remnants are seen stained red and surrounded by green bone on Safranin O–stained paraffin sections, and as they are mineralized they stain black on von Kossa sections (Fig. 3B, C). This primitive callus resembles that of the primary spongiosa of the metaphysis, which is subsequently remodeled into more mature trabeculae, and in the case of fracture, further remodeled into lamellar bone. As reported by our group and others, a lack of osteoclastic remodeling leads to the persistence of immature callus, delaying its remodeling into lamellar bone.(3,4,7–9,38) Further, stress‐fracture remodeling under treatment with the BP risedronate was significantly hindered, delaying final repair.(39) Figure 3A demonstrates the effects of delayed remodeling in our ZA‐treated and CLOD‐treated animals, but interestingly OPG treatment appeared to abrogate this hard callus remodeling further (Fig. 3B, C). Given the mechanism of action of OPG at the high dose used, largely abolishing osteoclast formation, such severely impaired callus remodeling would be expected. Doses of human OPG in the current study were excessive, relative to previous clinical dosing, as a means of maintaining adequate drug exposure in rats that can mount immune responses against the foreign human protein. It is important to highlight here that this retained immature callus is mineralized and is not suggesting delayed bony union. These outcomes are similar to those recently demonstrated with denosumab treatment, which also abolished osteoclast populations.(10) Although radiological union rates were not affected with denosumab, the authors provided histological images pointing to retention of unmineralized cartilage callus with this treatment. Data to quantify this retention was, however, lacking and only representative images were shown. It could be speculated that this result is due to a lack of chondroclastic cells, TRAP‐positive cells resorb unmineralized cartilage tissue; however, in the current study OPG treatment abolished this population of cells also and failed to show any delay in this process. Alternatively, it is possible that RANKL inhibition has direct effects on cartilage tissue differentiation, with RANKL‐deficient mice showing extensive expansion of the hypertrophic chondrocyte zone of the growth plate.(40) Hence, inhibiting RANKL in these fractures may have directly increased cartilage callus formation, thereby delaying cartilage callus removal. Nevertheless, we have quantified in detail the amount of persisting unmineralized cartilage callus in our study and do not see alterations in this with OPG, ZA, or CLOD treatment, only delayed remodeling of the primary hard callus. Although MMP inhibition impaired endochondral union, hard callus BV/TV was in fact increased with MMI270 treatment compared to saline, although not to the extent of BP and OPG treatment, suggestive of slightly delayed hard callus remodeling. MMP‐9 and MMP‐2 are secreted by osteoclasts to resorb collagen‐based matrices of both cartilage and bone.(23,36) Hence, along with inhibiting osteoclastic nonspecific remodeling of the cartilage soft callus, MMP blockade also impaired osteoclast‐ specific remodeling of the hard callus. The osteoclast is evidently active in remodeling of both soft and hard calluses, but nonspecific MMP‐driven resorption of the cartilage callus by

1558

MCDONALD ET AL.

osteoclasts is clearly not essential, given that the process of cartilage callus resorption was not impaired in the current study after complete osteoclast ablation with OPG treatment.

Bone formation was altered with reduced hard callus remodeling Although the direct effects of antiresorptive agents on bone formation in vivo remain speculative, it is clear in this study that bone formation during hard callus remodeling was reduced in the presence of these agents. ZA, CLOD, and OPG treatment all led to reduced percentage of labeled bone area in the hard callus (Fig. 5B). Similar reductions in bone formation parameters were recently reported in ibandronate‐treated fractures at 6 weeks.(9) Largely noted at the 4‐week and 6‐week time points, these reductions clearly coincide with hard callus remodeling. Hence, we propose that this altered bone formation is an indirect response to reduced tissue remodeling. Interestingly, MMI270‐ treated calluses showed no reduction of bone formation, signifying early bone formation is independent of cartilage callus removal. Further, this also suggests that the smaller reductions in hard callus remodeling seen with MMI270 treatment do not manifest in altered bone formation. The reduced callus volume in this group is likely due to delayed cartilage callus ossification.

Limitations This study is limited in its assessment of CT and histological outcomes only. A lack of union would indeed negatively impact mechanical properties of a healing callus, an important outcome when translating these studies to the clinic. Further, based on previously published work,(3,5,6,27,39,41) although the larger calluses produced by 6 weeks with ZA, CLOD, and OPG would be stronger in resistance to load, the material properties of the largely unremodeled woven bone callus is likely inferior to remodeled bone in the saline‐treated callus. One possible implication of these biomechanical data is that callus integrity in animals or patients treated with antiresorptives might be maintained at the expense of a larger callus. Additional studies assessing mechanical outcomes are required to address these limitations. In regard to ZA administration, the weekly dose administered in this study does not reflect well the yearly ZA doses administered to osteoporotic patients. Although the intention was to show no adverse effects of weekly ZA on union, it should be highlighted that the delays in hard callus remodeling shown here would be less evident in patients treated yearly with ZA. With MMP inhibition in fractures, although delayed union was demonstrated, a number of fractures still achieved union by 6 weeks. This would suggest that either the dose of MMI270 was not high enough to block all MMP activity, or intramembranous bone formation proceeded to a level adequate to eventually achieve union.

Conclusions We have directly compared anti‐osteoclastic agents with an inhibitor of MMP‐driven matrix degradation during endochondral fracture repair. Treatment with either ZA or CLOD Journal of Bone and Mineral Research

significantly reduced osteoclast activity; nevertheless, neither agent impeded the achievement of bony union, as shown previously. Moreover, OPG treatment, which abolished TRAP‐ positive cell production, also failed to impair endochondral union in rats. On the other hand, MMP inhibition potently impaired endochondral union. These outcomes support our hypothesis that nonspecific osteoclast‐driven cartilage callus removal is not essential to the achievement of endochondral fracture union; nevertheless, osteoclasts are pivotal to hard callus remodeling. In agreement with previous authors examining growth‐plate biology,(18,21) this novel investigation suggests that vascular endothelial cells and TRAP‐positive cells can collaborate to drive endochondral fracture union, but MMP‐driven degradation by non‐osteoclastic cells is crucial. Hence, this work challenges the current dogma for the temporal role of osteoclasts during bone repair, in that initial soft callus remodeling was proven capable of proceeding independently of osteoclast activity. Importantly, this adds significantly to the growing literature of support for the safe use of BPs and denosumab during early fracture repair.

Disclosures

restoration of strength in a rat model of fracture repair. J Orthop Res. 2005 Sep;23(5):1029–34. 6. Amanat N, McDonald M, Godfrey C, Bilston L, Little DG. Optimal timing of a single dose of zoledronic acid to increase strength in rat fracture repair. J Bone Miner Res. 2007 Jun;22(6):867–76. 7. Li J, Mori S, Kaji Y, Mashiba T, Kawanishi J, Norimatsu H. Effect of bisphosphonate (incadronate) on fracture healing of long bones in rats. J Bone Miner Res. 1999 Jun;14(6):969–79. 8. Cao Y, Mori S, Mashiba T, Westmore MS, Ma L, Sato M, Akiyama T, Shi L, Komatsubara S, Miyamoto K, Norimatsu H. Raloxifene, estrogen, and alendronate affect the processes of fracture repair differently in ovariectomized rats. J Bone Miner Res. 2002 Dec;17(12):2237–46. 9. Manabe T, Mori S, Mashiba T, Kaji Y, Iwata K, Komatsubara S, Yamamoto T. Effect of dosing interval duration of intermittent ibandronate treatment on the healing process of femoral osteotomy in a rat fracture model. Calcif Tissue Int. 2012;90:193–201. 10. Gerstenfeld LC, Sacks DJ, Pelis M, Mason ZD, Graves DT, Barrero M, Ominsky MS, Kostenuik PJ, Morgan EF, Einhorn TA. Comparison of effects of the bisphosphonate alendronate versus the RANKL inhibitor denosumab on murine fracture healing. J Bone Miner Res. 2009;24(2):196–208. 11. Bukata SV. Systemic administration of pharmacological agents and bone repair: what can we expect. Injury. 2011 Jun;42(6):605–8. 12. Einhorn TA. The science of fracture healing. J Orthop Trauma. 2005 Nov‐Dec;19(10 Suppl):S4–6.

PJK is an Amgen employee. DGL has received research support from Amgen, Novartis, Celgene, Lilly, and N8 Medical. All of the other authors state that they have no conflicts of interest.

Acknowledgments Novartis Pharma provided zoledronic acid and MMI270 for the study; Amgen Inc. provided the recombinant human osteoprotegerin (HuOPG‐Fc). We also acknowledge Mr. Sean Hogan for his support developing the closed fracture model used in this study and Expedite Publishing for assistance with manuscript preparation. Authors’ roles: Study Design: MMM and DGL. Study conduct: MMM, AM, KM, LP, PAB, PJK. Data collection: MMM, AM, KM, LP. Data analysis: MMM and AM. Data Interpretation: MMM and DGL. Drafting manuscript: MMM and DGL. Revising manuscript content: MMM, DGL, PJK. Approving final version of manuscript: MMM, DGL. MMM takes responsibility for the integrity of the data analysis.

References 1. Schindeler A, McDonald MM, Bokko P, Little DG. Bone remodeling during fracture repair: the cellular picture. Semin Cell Dev Biol. 2008 Oct;19(5):459–66. 2. Marsell R, Einhorn TA. The biology of fracture healing. Injury. 2011;42 (6):551–5. 3. McDonald MM, Dulai S, Godfrey C, Amanat N, Sztynda T, Little DG. Bolus or weekly zoledronic acid administration does not delay endochondral fracture repair but weekly dosing enhances delays in hard callus remodeling. Bone. 2008 Oct;43(4):653–62.

13. McDonald MM, Morse A, Peacock L, Mikulec K, Schindeler A, Little DG. Characterization of the bone phenotype and fracture repair in osteopetrotic incisors absent rats. J Orthop Res. 2011 May;29(5): 726–33. 14. Flick LM, Weaver JM, Ulrich‐Vinther M, Abuzzahab F, Zhang X, Dougall WC, Anderson D, O Keefe RJ, Schwarz EM. Effects of receptor activator of NFjB (RANK) signaling blockade on fracture healing. J Orthop Res. 2003; (21):676–84. 15. Colnot C, Thompson Z, Miclau T, Werb Z, Helms JA. Altered fracture repair in the absence of MMP9. Development. 2003;130(17):4123–33. 16. Stamenkovic I. Extracellular matrix remodelling: the role of matrix metalloproteinases. J Pathol. 2003 Jul;200(4):448–64. 17. Rundhaug JE. Matrix metalloproteinases and angiogenesis. J Cell Mol Med. 2005 Apr‐Jun;9(2):267–85. 18. Ortega N, Wang K, Ferrara N, Werb Z, Vu TH. Complementary interplay between matrix metalloproteinase‐9, vascular endothelial growth factor and osteoclast function drives endochondral bone formation. Dis Model Mech. 2010;3:224–35. 19. Behonick DJ, Xing Z, Lieu S, Buckley JM, Lotz JC, Marcucio RS, Werb Z, Miclau T, Colnot C. Role of matrix metalloproteinase 13 in both endochondral and intramembranous ossification during skeletal regeneration. PLoS One. 2007 Nov 7;2(11):e1150. 20. Yamagiwa H, Tokunaga K, Haylani T, Hatano H, Uchidna M, Endo N, Takahashi HE. Expression of metalloproteinase‐13 (collagenase‐3) is induced during fracture healing in mice. Bone. 1999;25(2):197–203. 21. Vu TH, Shipley JM, Bergers G, Berger JE, Helms JA, Hanahan D, Shapiro SD, Senior RM, Werb Z. MMP‐9/gelatinase B is a key regulator of growth plate angiogenesis and apoptosis of hypertrophic chondrocytes. Cell. 1998 May 1;93(3):411–22. 22. Everts V, Korper W, Hoeben KA, Jansen ID, Bromme D, Cleutjens KB, Heeneman S, Peters C, Reinheckel T, Saftig P, Beertsen W. Osteoclastic bone degradation and the role of different cysteine proteinases and matrix metalloproteinases: differences between calvaria and long bone. J Bone Miner Res. 2006 Sep;21(9):1399–408.

4. Li C, Mori S, Li J, Kaji Y, Akiyama T, Kawanishi J, Norimatsu H. Long‐term effect of incadronate disodium (YM‐175) on fracture healing of femoral shaft in growing rats. J Bone Miner Res. 2001; 16:429–36.

23. Sawae Y, Sahara T, Sasaki T. Osteoclast differentiation at growth plate cartilage‐trabecular bone junction in newborn rat femur. J Electron Microsc (Tokyo). 2003;52(6):493–502.

5. Amanat N, Brown R, Bilston LE, Little DG. A single systemic dose of pamidronate improves bone mineral content and accelerates

24. Pelletier JP, Boileau C, Brunet J, Boily M, Lajeunesse D, Reboul P, Laufer S, Martel‐Pelletier J. The inhibition of subchondral bone



1559

resorption in the early phase of experimental dog osteoarthritis by licofelone is associated with a reduction in the synthesis of MMP‐13 and cathepsin K. Bone. 2004 March;34(3):527–38. 25. Russell RG, Rogers MJ. Bisphosphonates: from the laboratory to the clinic and back again. Bone. 1999 Jul;25(1):97–106. 26. Kellinsalmi M, Mönkkönen H, Mönkkönen J, Leskelä HV, Parikka V, Hämäläinen M, Lehenkari P. In vitro comparison of clodronate, pamidronate and zoledronic acid effects on rat osteoclasts and human stem cell‐derived osteoblasts. Basic Clin Pharmacol Toxicol. 2005 Dec;97(6):382–91. 27. Ulrich‐Vinther M, Andreassen TT. Osteoprotegerin treatment impairs remodeling and apparent material properties of callus tissue without influencing structural fracture strength. Calcif Tissue Int. 2005;76:280–6. 28. Bonnarens F, Einhorn TA. Production of a standard closed fracture in laboratory animal bone. J Orthop Res. 1984;2(1):97–101. 29. Ota N, Takaishi H, Kosaki N, Takito J, Yoda M, Tohmonda T, Kimura T, Okada Y, Yasuda H, Kawaguchi H, Matsumoto M, Chiba K, Ikegami H, Toyama Y. Accelerated cartilage resorption by chondroclasts during bone fracture healing in osteoprotegerin‐deficient mice. Endocrinology. 2009 Nov;150(11):4823–34. 30. Harding AK, W‐Dahl A, Geijer M, Toksvig‐Larsen S, Tägil M. A single bisphosphonate infusion does not accelerate fracture healing in high tibial osteotomies: a randomized study of 46 patients. Acta Orthopaedica. 2011;82(4):465–70. 31. Silvano Adami, Cesar Libanati, Steven Boonen, Steven RCummings, Pei‐Ran Ho, Andrea Wang, Ethel Siris, Joseph Lane. Denosumab Treatment in Postmenopausal Women With Osteoporosis Does Not Interfere With Fracture Healing: Results From the FREEDOM Trial. JBJS. Epub Oct. 24 2012. 32. Zijlstra A, Aimes RT, Zhu D, Regazzoni K, Kupriyanova T, Seandel M, Deryugina EI, Quigley JP. Collagenolysis‐dependent angiogenesis mediated by matrix metalloproteinase‐13 (collagenase‐3). J Biol Chem. 2004;279(26):27633–45.

1560

MCDONALD ET AL.

33. Uusitalo H, Hiltunen A, Soderstrom M, Aro HT, Vuorio E. Expression of cathepsins B, H, K, L, and S and matrix metalloproteinases 9 and 13 during chondrocyte hypertrophy and endochondral ossification in mouse fracture callus. Calcif Tissue Int. 2000;67:382–90. 34. Kosaki N, Takaishi H, Kamekura S, Kimura T, Okada Y, Minqi L, Amizuka N, Chung U, Nakamura K, Kawaguchi H, Toyama Y, D’Armiento J. Impaired bone fracture healing in matrix metalloproteinase‐13 deficient mice. Biochem Biophys Res Commun. 2007;354:846–51. 35. Lieu S, Hansen E, Dedini R, Behonick D, Werb Z, Miclau T, Marcucio R, Colnot C. Impaired remodeling phase of fracture repair in the absence of matrix metalloproteinase‐2. Dis Model Mech. 2011;4:203–11. 36. Sorsa T, Ramamurthy NS, Vernillo AT, Zhang X, Konttinen YT, Rifkin BR, Golub LM. Functional sites of chemically modified tetracyclines: inhibition of the oxidative activation of human neutrophil and chicken osteoclast pro‐matrix metalloproteinases. J Rheumatol. 1998 May;25(5):975–82. 37. Wigner NA, Kulkarni N, Yakavonis M, Young M, Tinsley B, Meeks B, Einhorn TA, Gerstenfeld LC. Urine matrix metalloproteinases (MMPs) as biomarkers for the progression of fracture healing. Injury. 2012 Mar;43(3):274–8. 38. Smith EJ, McEvoy A, Little DG, Baldock PA, Eisman JA, Gardiner EM. Transient retention of endochondral cartilaginous matrix with bisphosphonate treatment in a long‐term rabbit model of distraction osteogenesis. J Bone Miner Res. 2004 Oct;19(10):1698–705. 39. Kidd LJ, Cowling NR, Wu AC, Kelly WL, Forwood MR. Bisphosphonate treatment delays stress fracture remodeling in the rat ulna. J Orthop Res. 2011 Dec;29(12):1827–33. 40. Kim N, Odgren PR, Kim DK, Marks SC Jr, Choi Y. Diverse roles of the tumor necrosis factor family member TRANCE in skeletal physiology revealed by TRANCE deficiency and partial rescue by a lymphocyte‐ expressed TRANCE transgene. Proc Natl Acad Sci U S A. 2000 Sep 26;97(20):10905–10. 41. Mori S. Fracture healing with anti‐resorptive agents. J Musculoskelet Neuronal Interact. 2003;3(4):314–6.
