|Home | About | Journals | Submit | Contact Us | Français|
The effects of anti-resorptive agents (i.e. alendronate (Aln), osteoprotegerin (OPG)) on bone infection are unknown. Thus, their effects on implant-associated osteomyelitis (OM) were investigated in mice using PBS (placebo), gentamycin and etanercept (TNFR:Fc) controls. None of the drugs affected humoral immunity, angiogenesis, or chronic infection. However, the significant (p<0.05 vs. PBS) inhibition of cortical osteolysis and decreased draining lymph node size in Aln and OPG treated mice was associated with a significant (p<0.05) increase in the incidence of high-grade infections during the establishment of OM. In contrast, the high-grade infections in TNFR:Fc treated mice were associated with immunosuppression, as evidenced by the absence of granulomas and presence of Gram+ biofilm in the bone marrow. Collectively, these findings indicate that while anti-resorptive agents do not exacerbate chronic OM, they can increase the bacterial load during early infection by decreasing lymphatic drainage and preventing the removal of necrotic bone that harbors the bacteria.
Although the incidence of orthopaedic bone infection, known as osteomyelitis (OM), for joint prostheses and fracture-fixation devices have remained at 0.3–11% and 5–15% respectively over the last decade 1,2, the emergence of methicillin-resistant Staphylococcus aureus (MRSA) has dramatically impacted clinical management 3,4. Since OM induces osteolysis around the implant, which can lead to fractures and complicate revision surgery, orthopedists have anecdotally used anti-resorptive bisphosphonates in these patients to preserve bone stock without knowledge of their effects on chronic infection 5,6.
Another clinical condition that is associated with OM and bisphosphonate use is osteonecrosis of the jaw (ONJ) epidemic 7,8, which primarily affects cancer patients on immunosuppressive chemotherapy that also receive intravenous pamidronate or zoledronate for hypercalcaemia, and undergo unrelated oral surgery 9,10. However, ONJ has also been reported in healthy women taking oral alendronate for postmenopausal osteoporosis 9,10, raising serious concerns for the tens of millions of people taking these anti-resorptive agents. Although the etiology of ONJ remains unknown, careful reviews of all the available clinical and basic science data have derived two leading theories to explain this painful condition 11–13. The first is inhibition of angiogenesis, which is largely derived from the precedent of steroid-induced avascular necrosis of the hip, and new studies that have demonstrated bisphosphonate-inhibition of capillary tube formation, vessel sprouting and circulating endothelial cells 14–16. The other is the osteoclast-inhibiting effect of anti-resorptive agents that lead to a cessation of bone remodeling and mineral:matrix turnover 10. This renders the fatigued bone susceptible to involution of blood vessels, necrosis and infection. The impact of infection in ONJ is further highlighted by the finding that essentially all cases have histopathological evidence of OM 17,18, however, a cause and effect relationship has yet to be established.
To better understand the effects of bisphosphonates on OM, and shed light on potential mechanisms of ONJ, we investigated the interaction of anti-resorptive agents in an established murine trans-tibial infected pin model that emulates the biology of infected orthopaedic and periodontal implants 19. This model takes advantage of a gentamycin sensitive, luxA-E transformed stain of S. aureus (Xen29) 20, whose metabolic activity and bacterial load can be quantified in vivo using longitudinal bioluminescence imaging (BLI) and nuc quantitative real-time PCR (QRT-PCR) respectively 19. This model also has micro-CT outcome measures of osteolysis and vascularity, serological analyses of the protective IgG2b humoral response that appears by day 11, and traditional histology of osteoclasts and Gram-stained biofilm 19. Here we investigated two distinct classes of anti-resorptive drugs: the most widely prescribed bisphosphonate alendronate (Aln) 8, which induces osteoclast apoptosis via inhibition of the mevalonate pathway of isoprenoid biosynthesis 21; and the biologic antagonists osteoprotegerin (OPG) that inhibits osteoclast formation and induces osteoclast apoptosis by blocking receptor activator of nuclear factor kappa B (RANK) signaling pathway 22. While these studies demonstrated that neither Aln nor OPG could induce or exacerbate chronic OM, likely because they do not affect the protective humoral response nor inhibit angiogenesis in our model, both agents caused an increase in high-grade infections as defined by significant increases in the number of mice with BLI and nuc gene levels above maximum placebo levels. This finding coincided with a significant decrease in osteolysis and draining lymph node size, suggesting that anti-resorptive agents decrease efflux of marrow lymph during the establishment of OM. Thus, these effects could lead to a dramatic increase in intraosseous pressure, infarction and bone pain, which are the hallmarks of OM and ONJ.
All animal studies were performed under University of Rochester Committee for Animal Resources approved protocols. Pathogenic challenges with S. aureus Xen29 were performed as previously described 19. Drug treatments started on the day of surgery, except for Aln, which started 3 days before surgery. Aln (Calbiochem, San Diego, CA) was given at 10μg/kg i.p. once every 3 days. OPG (R&D Systems, Minneapolis, MN) was given at 200μg i.p. once every 3 days. TNFR:Fc (Amgen, Thousand Oaks, CA) was given at 200μg i.p. once every 3 days. Placebo was given as 200μl PBS i.p. once every 3 days. Gentamycin (Calbiochem, San Diego, CA) was given daily (0.1 mg/kg i.p.).
Longitudinal osteolysis was assessed radiographically using a Faxitron Cabinet x-ray system (Faxitron, Wheeling, IL, USA) 19. Bioluminescent imaging (BLI) of mice infected with Xen29 was performed on day 0, 4, 7, 11, 14, 18 post-infection as previously described 19. Micro-CT and vascular micro-CT was performed as previously described 23,24. CE-MRI was performed as previously described 25,26.
After micro-CT, the tibial samples were processed for decalcified histology and stained with Orange G/alcian blue (H&E), Gram-stained, or stained for tartrate-resistant acid phosphatase (TRAP) activity as we have described previously 27. Immunohistochemistry to identify T-cells, B-cells, macrophages and neutrophils was performed as described previously 28, using primary antibodies specific for CD3 (diluted 1:100; Novocastra, Newcastle, UK), CD45R/B220 (diluted 1:300; BD Biosciences Pharmingen, San Jose, CA, USA), F4/80 (diluted 1:80; AbD Serotec, Raleigh, NC, USA), and Neu7/4 (clone# MCA771G, diluted 1:2000, AbD Serotec, Raleigh, NC) respectively. Osteoclasts were quantified as the mean +/− SD number of TRAP+ multinucleated cells in the 10x field centered on the pin tract from three contiguous 3 μm sections 500μm apart, as previously described 29.
Total DNA was extracted from infected tibiae and analyzed by RTQ-PCR for the S. aureus-specific nuc gene standardized to mouse β-actin as previously described 19.
All values are presented as means ± SD. Significant differences between groups in the histology, micro-CT, serology, MRI and PLN weight analyses were determined by ANOVA with Bonferroni correction. A Fisher’s exact test was performed to determine significant differences in the incidence of infection and high-grade infection. The correlation between BLI vs. nuc was estimated using Pearson’s correlation coefficient and tested for significance using a two-sided t-test. In all cases, p<0.05 was considered significant.
In order to assess the effects of anti-resorptive agents on the establishment of implant-associated OM, we investigated five experimental groups in the Xen29-infected trans-tibial pin model 19: Group 1 received effective gentamycin prophylaxis that prevents infection (Gent), Group 2 received placebo (PBS), Group 3 received OPG, Group 4 received Aln, and Group 5 received the tumor necrosis factor (TNF) antagonist etanercept (TNFR:Fc) as a control that is known to inhibit osteoclasts 30 and exacerbate infections by inhibiting granuloma formation 31. Upon initial inspection of the x-rays taken on day 18, effects of all three anti-resorptive treatments displayed remarkable large soft-tissue abscesses compared to the Gent and PBS controls (Fig. 1A–E). Histological analyses of these tibiae also confirmed the significant effects of these drugs on the reduction of osteoclasts numbers at the bone-implant interface (Fig. 1F–O). Although TNF blockade also reduced osteoclast numbers to a certain extent, it is noteworthy that the effects of TNFR:Fc on bone histology were significantly different from all of the other groups, indicating a different mechanism of action. Consistent with the osteoclast results, we found that OPG and Aln treated mice had large amounts of unresorbed necrotic cortical bone adjacent to the pin tract (Fig. 1H–I, M–N), which served as a nidus for infection as evidenced by the presence of Gram+ biofilm in parallel sections (Fig. 1R–S). In contrast, the extensive osteolysis in the PBS and TNFR:Fc treated mice appeared to have removed all of the necrotic bone around the pin hole (Fig. 1G, J), and there was no evidence of bacteria at these sites in parallel sections (Fig. 1L, O). However, infection in these animals was evident from Gram+ biofilm within necrotic bone fragments that were likely generated during implantation of the pin (Fig. 1Q, T). As expected from effective prophylaxis, Gent treated mice had no evidence of osteolysis or infection (Fig. 1A,F,K,P).
In order to assess the effects of Aln, OPG and TNFR:Fc on Xen29 metabolism and bacterial load during the establishment of implant-associated OM, we performed longitudinal BLI before, during and after the establishment of protective humoral immunity (Days 7, 11, 18, respectively) in our model 19 (Fig. 2A–C). While these experiments failed to demonstrate any significant differences between the averages of the groups (p>0.05), indicating that the drugs cannot exacerbate infection by themselves, we did observe a few mice in the OPG, Aln and TNFR:Fc groups that had remarkably high BLI values, suggestive of “high-grade” infection. To test this we performed Fisher’s exact tests comparing the number of uninfected, infected and high-grade infected mice in each group (Fig. 2A–C). These results demonstrated that while Aln and OPG did not increase the incidence of infection at any time they caused a significant increase in the number of mice with high-grade infections early on (Day 7 and 11). Interestingly, the effects of TNFR:Fc were similar with the exception that TNF inhibition caused a 100% infection rate on Day 7.
In order to confirm our observation that anti-resorptive drugs induce high-grade infections during the establishment of OM, we quantified the bacterial load on day 7 and day 18 (Figs. 2D–E). Interestingly, while no differences were observed early on, the OPG, Aln and TNFR:Fc groups all had significantly more high-grade chronic infections on day 18. In order to demonstrate that the high-grade infections determined by BLI and RTQ-PCR were in the same animals, we performed a linear regression analysis of all the day 18 data (Fig. 2F). This significant correlation (R2=0.64, p<0.05) validates our findings that anti-resorptive agents can increase the incidence of high-grade OM.
In order to better understand the mechanism by which anti-resorptive drugs increase the incidence of high-grade OM, we investigated their effects on bone volume, vasculature and host immune responses during the infection. First, we performed micro-CT analyses on the infected tibiae to assess bone and vasculature around the pin on day 18. While the micro-CT images of the bone corroborated the x-ray and histology findings (Fig. 1), the most striking result was the zone of inhibited reactive bone formation around the infected pins of Aln and OPG treated mice (Fig. 3A). This zone appeared as an almost perfect ring ~2mm in diameter, and is the first formal demonstration that OM-induced osteolysis observed in radiographs is largely due to inhibition of bone formation rather than pure osteoclastic bone resorption. Independent quantification of the osteolytic area in the cortical and reactive bone demonstrated that Aln and OPG completely inhibit cortical resorption, but have no effects on reactive bone formation (Fig. 3B). In contrast, despite its ability to decrease osteoclast numbers and promote high-grade OM, TNFR:Fc did not significantly inhibit cortical osteolysis compared to PBS controls.
Given that avascular necrosis is often associated with OM, we also quantified vascularity in the infected tibiae via 3D vascular micro-CT of lead chromate perfused vessels. These studies failed to detect any significant differences between any of the five groups in terms of vasculature volume, vessel thickness, vessel density, vessel spacing, and degree of anisotropy (data not shown). Thus, none of the drug effects in our model can be explained by changes in angiogenesis.
To test if drug induced immunosuppression of humoral immunity was responsible for the high-grade infections, we examined total Ig levels in sera from the animals over the course of infection. No remarkable serologic differences between groups were observed during the course of OM (data not shown). These data confirmed the induction of the previously characterized IgG2b response in all of the mice, and demonstrate that none of the drugs inhibited humoral immunity against Xen29. This finding is consistent with the 100% survival rate from a bacterial challenge that would cause lethal sepsis in immunodeficient mice. We also reanalyzed the histology to assess drug effects on granuloma formation (Fig. 4A–H). Although Aln and OPG could not prevent infection of necrotic bone adjacent to the implant, we found no evidence of biofilm formation anywhere else (Fig. 4A, B). In fact, we could readily identify active granulomas in immediate proximity to the pin that were completely Gram negative (Fig. 4C, D), indicating that anti-resorptive drugs do not interfere with innate immunity. In contrast, mice treated with TNFR:Fc had Gram positive biofilm in both necrotic bone and soft tissue adjacent to the pin (Fig. 4E–H). Furthermore, there was no evidence of an organized cellular immune response around the infected soft tissue (Fig. 4H). These results are consistent with the critical role for TNF in innate immunity and granuloma formation.
While considering alternative mechanisms responsible for the high-grade OM, we reflected on an explanation as to why the host chooses to mount such a vigorous osteolytic response to infection in which a much greater bone void is generated than what would be expected based on the margins of the biofilm (e.g. >2mm3 vs. <0.2mm3). The simplest explanation for this is that efficient drainage of the infection requires a large cavity for efflux of pus out of the medullary canal, which is under enormous intraosseous pressure from the recruitment and proliferation of immune cells and fluid. Thus, we reasoned that inhibition of cortical resorption would compromise this process as evidenced by a significant decrease in draining lymph node volume. To test this hypothesis we performed in vivo MRI to quantify the popliteal lymph node (PLN) volume on day 9 (data not shown), which demonstrated that both OPG, Aln and TNFR:Fc significant decrease PLN volume during the peak of infection. We also analyzed the tibiae and PLN of these mice ex vivo, which demonstrated two remarkable observations. First, there was no macroscopic evidence of pus draining from the pin tract of the OPG and Aln treated mice (Fig. 4I). This was to be expected based on the micro-CT data, which demonstrated the cortical hole is approximately the same size as the implant (Fig. 3). Secondly, OPG and Aln significantly inhibited the expansion of the draining PLN, but had no effects on the contralateral PLN, as determined by weight (Fig. 4J). Immunohistochemical analysis of these PLN with antibodies specific for B-cells (B220), T-cells (CD3), neutrophils (Neu7/4) and macrophages (F4/40) failed to generate any remarkable findings (data not shown), as the PBS samples simply appeared larger. We also performed flow cytometry to detect B220, CD4, CD8, CD11b, CD11c, and CD80 positive cells in PLN from infected and contralateral legs of mice treated with PBS or Aln (data not shown). The only significant finding was that Aln decreased the percentage of CD80+ cells (activated antigen presenting cells) by 20% versus PBS controls (p < 0.02). These results suggest that reduced lymphangiogenesis in the local draining lymph nodes after anti-resorptive therapy may be responsible for high-grade infections, but future studies are required to formally demonstrate the relationship between lymphatic egress from the infected medullary canal and PLN expansion during the establishment of OM.
Although infection rates following bone-implant surgery remain fairly low, the total number of OM cases is high3, and the recent surge in MRSA further underscores the significance of this clinical problem4. At best, this catastrophic outcome requires a challenging revision surgery that is further complicated by deteriorated bone stock due to osteolysis. Thus, an important unanswered question regarding the treatment of patients with implant-associated OM is whether or not they would benefit from anti-resorptive therapy. Despite the successful case reports in the literature5,6, our study indicates that this practice may be contraindicated since Aln and OPG: i) increase the amount of necrotic cortical bone around the implant that serves as a nidus for infection (Fig. 1), ii) significantly increases the incidence of high-grade infections in mice during the establishment of OM (Fig. 2), and iii) reduce the cortical hole (Fig. 3) through which the opsonized bacteria and lymph must drain out of the wound (Fig. 4I, J).
Equally important to our positive findings are results that refute claims regarding the anti-angiogenic and immunosuppressive effects of anti-resorptive therapies. In search of the etiology of ONJ, others have reported that bisphosphonates inhibit capillary tube formation, vessel sprouting and circulating endothelial cells 14–16, and cited this as a mechanistic factor in the condition. In contrast, our vascular micro-CT analyses failed to demonstrate any drug effects on perfusable vessels >10μm (data not shown). Thus, our findings are more consistent with reports from others who have argued against this theory based on the fact that drugs such as thalidomide, with far greater anti-angiogenic capacity, do not result in ONJ 14. Similarly, the findings that RANK signaling is required for lymph node development 32,33 and immunoregulatory co-stimulation 34,35 have led to a suspicion that RANK-ligand inhibition is immunosuppressive. However, our studies corroborate the prior findings of Stolina et al, who demonstrated that OPG treatment does not affect cell-mediated immune responses and granuloma formation 36, and the recent finding that humans genetically deficient in RANKL have severe osteopetrosis without immunodeficiency 37. Therefore, we conclude that RANKL inhibition has insignificant immunoregulatory effects in the setting of in vivo pathogenic challenge, and that its ability to increase the incidence of high-grade OM can be entirely attributed to its effects on osteoclastic bone resorption, as these effects could be replicated with Aln (Figs. 1–3). To further abate concerns regarding the clinical use of a biologic RANKL inhibitor (i.e. denosumab 38), we demonstrated that OPG and TNFR:Fc mediate similar effects on OM via different mechanisms, as evidenced by their differential effects on osteolysis, granuloma formation, and lymphatics. Given the remarkable clinical success of anti-TNF therapy over the last decade, and its much greater effects on the immune system compared to OPG, it is becoming more apparent that immunosuppression is unlikely to be a major limitation of biologic anti-resorptive therapy.
The most surprising results of our study was the observation that the massive bone void around the transtibial implant was largely due to inhibition of reactive bone formation rather than osteoclastic bone resorption, as evidenced by the micro-CT images of the Aln and OPG treated samples (Fig. 3A). Most interestingly is that this inhibition of bone formation appears as a perfectly symmetrical ring that extend >1mm from the biofilm adjacent to the pin. Thus, it is clear that this is mediated by an indirect mechanism that likely involves a secreted factor (i.e. noggin, DKK1, SOST) that has a gradient expression pattern originating from the infected bone. Elucidation of this factor(s) is now the subject of ongoing investigation, as it could also prove to be important for preventing inflammation mediated bone loss in aseptic conditions such as erosive arthritis and periprosthetic osteolysis.
While we do not want to overstate the implications of our findings as they relate to the emerging ONJ epidemic, we feel that there are several parallels that warrant discussion. The first and most frustrating is the absence of etiologic factors that are necessary and sufficient to faithfully reproduce OM and ONJ in similar individuals. Even though a correlation between ONJ and bisphosphonate use in cancer patients following oral surgery has been established, its overall incidence in this population is >10% 9,10. Similarly, despite our best efforts in a controlled experiment model, we have been unable to establish conditions that lead to equivalent OM in genetically identical littermates, in which many of the mice that receive a Xen29 coated pin remain uninfected (Fig. 2). This indicates that stochastic epigenetic factors, such as the amount of necrotic bone available for bacterial adhesion, and the temporally restricted period for this event to occur, may be critical to the establishment of both OM and ONJ.
The other parallel that deserves mention is the potential association between inhibited lymphatics during the establishment of bone infection and bone pain. One potential mechanism by which anti-resorptive agents could inhibit lymphangiogenesis is by reducing the amount of VEGF-C produced during the infection through its effects on osteoclast precursors and osteoclasts, which produce high levels in response to RANKL39. Clinically, OM and ONJ are primarily discovered as a result of symptomatic bone pain, whose nature is largely unknown. Our novel finding that OPG and Aln drug inhibition of resorption is associated with significant decreased lymphatics during the establishment of bone infection implies that there must be a marked increase in intraosseous pressure at this time, which would be expected to induce bone pain characteristic of that seen in OM and ONJ. Interestingly, recent clinical studies have demonstrated that surgery-induced lymphedema is a major cause of local pain, and improvement of lymphatic drainage is an effective therapy in these patients 40. Thus, the association of decreased lymph node size and drainage in our model of OM warrants further investigation.
This work was supported by research grants from the National Institutes of Health PHS awards AR48681, DE17096, AR52674, AR51469, AR46545, AR54041 and AR53459.
The authors would like to thank Dr. Christopher Beck of the Department of Biostatistics, for his assistance with the statistical analyses. We also thank Laura Yanoso for technical assistance with the micro-CT and Krista Scorsone for technical assistance with the histology. This work was supported by research grants for the US Dept. of Defense (ERMS No.06136016 and UID No. 99853) and the National Institutes of Health PHS awards AR48681, DE17096, AR52674, AR51469, AR46545, AR54041 and AR53459.