The RANKL/RANK/OPG Signaling Pathway Mediates Medial Arterial Calcification in Diabetic Charcot Neuroarthropathy

Approval from the local research ethics committee was granted for the use of human tissue, and procedures were in accordance with institutional guidelines and the Declaration of Helsinki.

CN was confirmed clinically, supported by radiologic features as described previously (18). Serum was obtained via the antecubital fossa from 12 patients with CN at the time of diagnosis of acute stage disease, 10 patients with diabetes, and 5 nondiabetic volunteers, all matched for age, sex, and renal function. Ten milliliters of venous blood was withdrawn aseptically into sterile, heparinized tubes and centrifuged at 3,200 rpm for 15 min at 4°C. The resultant sera were measured in aliquots and stored at −80°C until future use. A recent analysis of patient serum, the Rancho Bernardo study, reported that storage at −70°C for up to 11 years did not affect RANKL or OPG concentrations (19).

Serum RANKL and OPG levels were measured using a sandwich ELISA technique according to the manufacturer’s protocol. According to the manufacturer, the intra-assay and interassay coefficients of variation for the soluble RANKL (ampli-sRANKL, BI-20452, lot no. P83C) ELISA assay are 8–9 and 3–6%, respectively, whereas the same values for the OPG assay (OPG, BI-20402 lot no. 873aC) are 4–40 and 7–8%, respectively (Biomedica Gruppe, Vienna, Austria). Absorbance was measured at 450 and 560 nm for the OPG and RANKL, respectively.

A Bio-plex angiogenesis panel (“Pro-Human Angiogenic 9-plex plate” catalog number 171-304060) obtained from Bio-Rad Laboratories Inc. (Hercules, CA) was used according to the manufacturer’s protocol for analysis of angiopoietin, vascular endothelial growth factor, hepatocyte growth factor, leptin, platelet endothelial cell adhesion molecule or CD-31, and follistatin and two inflammatory cytokines, granulocyte colony–stimulating factor (GCSF) and interleukin-8 (IL-8). The intra-assay and interassay coefficients of variation of the Bioplex angiogenesis assay were ≤15 and ≤25%, respectively, with an accuracy of 70–130%. Data were analyzed using Bio-Plex Manager software version 5.0.

Segments of tibial artery were obtained from patients undergoing surgery (amputation) and used for imunohistochemistry and explantation of VSMCs as described previously (20). A panel of internal mammary artery (IMA) (deemed surplus to requirements from patients undergoing cardiac artery bypass graft surgery) was used as noncalcified control artery. Arterial sections were stained using rabbit anti-human RANKL antibody (dilution 1:200 catalog number ab9957; AbCAM, Cambridge, U.K.) and a nonimmune serum (IgG) as negative control as described previously (20).

Human VSMCs were obtained by tissue explantation (21) using smooth muscle cell growth medium obtained from PromoCell (PromoCell GmbH, Heidelberg, Germany) and subsequently maintained in regular growth media (Dulbecco’s modified Eagle’s medium, high glucose, 4.5 g/L) containing 5% FBS, 10 mmol/L sodium pyruvate, 1.8 mmol/L CaCl₂, 100 U/mL penicillin, and 100 mg/mL streptomycin. All cells were used between passages three and eight.

VSMC mineralization was induced using a modification of the method described by Petrova et al. (22). Briefly, VSMCs were seeded in a 24-well plate at densities of 5 × 10⁴ cells/cm² and cultured in regular growth media until they reached 80% confluence. Cells were then switched to osteogenic media, i.e., regular growth medium containing 2.6 mmol/L CaCl₂, and 5 mmol/L β-glycerophosphate for up to 7 days. Cells were treated with 5% serum from 1) healthy volunteers, 2) diabetic patients, or 3) patients with CN. Recombinant OPG (20 pmol/L) (catalog number GF120; Millipore Corporation, Hayward, CA) was added in a parallel set of experiments when cells were changed to osteogenic media and subsequently added each time the medium was replaced, as were vehicle control subjects. All experiments were performed in triplicate using three populations of VSMC, each from different patients. Alkaline phosphatase (ALP) activity assays and Alizarin red staining were carried out as described previously (20,21).

Cells were seeded at a density of 5 × 10⁴ cells/cm² into four-well chamber slides (Laboratory-Tek Chamber Slides, Fisher Scientific UK LTD, Leicestershire, U.K.). At confluence, they were subcultured for 1 h in osteogenic medium (OM) containing 5% Charcot serum, 20 pmol/L human recombinant soluble RANKL, and 20 pmol/L human recombinant soluble RANKL plus 20 pmol/L human recombinant OPG. Cells were fixed in 4% paraformaldehyde, blocked in 5% goat serum/PBS and followed by incubation with primary rabbit anti–NF-κB antibody (1:500 dilution in 5% goat serum/PBS) and an Alexa Fluor 488-conjugated goat anti-rabbit IgG secondary antibody (Molecular Probes, Invitrogen, Paisley, U.K.). Rabbit IgG was used as intra-assay negative control. Cells were counterstained with 4 µg/mL DAPI and mounted in Vectashield mounting medium (Vector Laboratories, Inc., Burlingame, CA).

Each experiment was performed in triplicate with three different populations of human VSMCs explanted from different patients. Data were expressed as mean ± SD for the ELISA, Bio-plex, and cell culture assays (optical density [OD] and ALP activity). Differences between groups were analyzed using an unpaired Student t test (two groups) or a one-way ANOVA (three or more groups) as appropriate. Where multiple samples were compared using ANOVA, a post hoc Bonferroni analysis was performed to enable comparison between groups. Data that were not normally distributed were log-transformed before performing statistical tests. A P value < 0.05 was considered statistically significant.

The general clinical characteristics of the different patient groups are presented in Table 1.

To determine whether local RANKL expression was upregulated in calcified versus noncalcified arteries, sections of calcified tibial artery and noncalcified IMA were stained immunohistochemically with anti-human RANKL antibody and Alizarin red. Figure 1 shows a representative section of calcified human tibial artery and a noncalcified IMA. The tibial artery displayed both medial and intimal calcification with positive RANKL staining detected within the vicinity of the calcified areas (B1, B2, and B3; C1 and C2). There was a distinct lack of positive RANKL staining in noncalcified areas (D1 and D2) and control IMA segments (A1, A2, and A3).

ELISA was performed to determine serum levels of RANKL and OPG in the two patient groups and control subjects. Serum OPG levels were higher in patients with CN (8.2 ± 2.7 pmol/L, ANOVA: P = 0.031) compared with diabetic patients (7.7 ± 3.3 pmol/L) and healthy control subjects (4.2 ± 0.7 pmol/L) (Fig. 2A). The Bonferroni post hoc test showed that OPG levels were higher in patients with CN versus control subjects (P = 0.01) and in diabetic patients versus control subjects (P = 0.029), but not significantly different between patients with CN and patients with diabetes (P = 0.638). In a similar trend, RANKL levels in patients with CN were 0.3 ± 0.042 pmol/L, approximately 8- and 15-fold higher than in patients with diabetes (0.04 ± 0.05 pmol/L) and healthy control subjects (0.02 ± 0.01 pmol/L), respectively (ANOVA test, P = 0.055). The Bonferroni test showed that patients with CN had higher serum levels of RANKL than diabetic patients (P = 0.033) or healthy control subjects (P = 0.063), but there were no significant differences between diabetic patients and control subjects (P = 0.895) (Fig. 2B). Most important, the RANKL/OPG ratio was significantly higher in patients with CN ([36.8 ± 43.1] × 10⁻³, ANOVA P = 0.033) compared with those with diabetes ([5.2 ± 4.9] × 10⁻³) or healthy control subjects ([4.6 ± 3.1] × 10⁻³) (Fig. 2C). The Bonferroni test showed that the RANKL/OPG ratio was higher in patients with CN than in diabetic patients (P = 0.019) and in patients with CN than in control subjects (P = 0.051), but differences between diabetic patients and control subjects were not significant (P = 0.974).

To determine whether an inflammatory milieu/stimulus may be modulated in patients with CN and thus driving elevated RANKL expression and subsequent secretion into systemic circulation in this group, we analyzed a panel of nine cytokines using a Bio-plex multiarray suspension assay (Fig. 3). We demonstrated that two proinflammatory cytokines (GCSF and IL-8) were significantly higher in patients with CN compared with diabetic patients and control subjects. This may well reflect an elevated inflammatory environment in patients with CN acting as the stimulus for elevated RANKL secretion. The serum levels of noninflammatory cytokines were not significantly different among the patients with CN, compared with patients with diabetes, and control patients.

Figure 4A shows results from the culture of human VSMCs at day 7 under the different conditions. Cells treated with Charcot serum displayed increased calcification as evidenced by Alizarin red staining. After dye elution and quantification, there were overall differences (ANOVA: P < 0.0001) in the extent of mineralization between the different culture conditions (Fig. 4C). Mineralization was higher in human VSMCs coincubated in Charcot serum (OD = 0.16 ± 0.01 vs. OM control subjects: 0.09 ± 0.003, P = 0.0003) compared with diabetic serum (OD = 0.15 ± 0.01 vs. OM, P = 0.0004) or healthy serum (OD = 0.14 ± 0.03 vs. OM, P = 0.037) (Fig. 4C).

Likewise, one-way ANOVA test showed that the early osteogenic differentiation marker, ALP, was significantly different (P < 0.0001) in human VSMCs subcultured in any of the three patient sera compared with OM that contained FBS (same concentration) in place of patient serum. The effect was most marked in Charcot than in diabetes serum. At day 7, the ALP activity of human VSMCs cultured in Charcot serum (49.9 ± 7.3) was approximately twofold higher than that induced by diabetes serum (29.8 ± 3.7 µmol/L phosphate/mg protein/min, P = 0.013) and threefold higher than either healthy serum (17.6 ± 10.1, P = 0.011) or OM control subjects (15.3 ± 5.4 μmol/L phosphate/mg protein/min, P = 0.003) (Fig. 4B).

To establish whether factors in Charcot serum or RANKL triggered the mineralization process per se or needed the presence of the elevated calcium and phosphate in the osteogenic conditions, cells were grown in ordinary growth medium to which the Charcot serum or RANKL was added. The VSMCs were shown to have a greater capacity to mineralize when cultured in the presence of Charcot serum (OD: 0.07 ± 0.003, P = 0.0009) or recombinant RANKL (OD 0.072 ± 0.003, P = 0.0002) compared with ordinary growth medium only (OD = 0.059 ± 0.002) (data not shown). The high levels of mineralization detected in the VSMCs cultured in the presence of Charcot serum or recombinant RANKL were similar to those induced by OM (OD = 0.075 ± 0.002).

The anti-RANKL effect of OPG on VSMC osteoblastic differentiation and mineralization induced by Charcot and diabetes sera is shown in Fig. 5A–C. Compared with VSMCs subcultured in Charcot serum alone, coincubation of VSMCs with Charcot serum plus OPG resulted in an attenuation of osteoblastic differentiation (ALP activity: 49.9 ± 7.3 vs. 12.2 ± 8.6, P = 0.004) and mineralization (OD: 0.16 ± 0.01 vs. 0.13 ± 0.003, P = 0.018) of these cells (Fig. 5B and C).

However, VSMCs coincubated with diabetes serum and OPG showed only a trend toward less mineralization (OD = 0.14 ± 0.004) and reduced osteoblastic differentiation (ALP activity, 20.9 ± 10.9) compared with those cultured in diabetes serum alone (Fig. 5B and C); the difference failed to achieve statistical significance.

To establish the downstream effects of Charcot serum or RANKL on transcription factors that may be involved in the mineralization of human VSMCs, we investigated the activation of the NF-κB pathway. VSMCs were subcultured for 1 h in Charcot serum (5% v/v) or recombinant RANKL (20 pmol/L) and stained immunocytochemically for the p65 component of NF-κB to detect its intracellular localization. Representative photomicrographs obtained after immunofluorescent staining for NF-κB are shown in Fig. 6A and B. After 1 h of culture of VSMCs in osteogenic media to which Charcot serum was added, NF-κB translocated from its cytoplasmic location to a predominantly nuclear and perinuclear location (Fig. 6A, images D and G). When VSMCs were incubated with recombinant RANKL in lieu of Charcot serum, nuclear translocation of p65 component of NF-κB was equally observed (Fig. 6B, images D and G).

To validate that the RANKL/OPG signaling involved translocation of NF-κB during mineralization of VSMCs, immunostaining was performed on subcultures of VSMCs (Charcot or RANKL) coincubated with OPG. The presence of OPG abrogated the NF-κB nuclear translocation induced by Charcot serum (Fig. 6A, images E and H) or recombinant RANKL (Fig. 6B, images E and H), thus confirming that NF-κB activation induced by Charcot serum directly involved RANKL/RANK/OPG signaling.

Our results have demonstrated that RANKL/RANK/OPG signaling is implicated in human MAC in people with diabetes and CN. The RANKL immunolocalization data suggest a direct association between localized tissue expression of RANKL and MAC. Patients with CN also displayed high levels of inflammatory cytokines (IL-8, GCSF), OPG, RANKL, and particularly elevated RANKL/OPG ratio compared with diabetic patients or healthy volunteers. The high systemic level of RANKL enables it to bind to its receptor, RANK, and drive vascular smooth muscles cells into an osteoblastic differentiation pathway resulting in accelerated deposition of a mineralized matrix leading to MAC. This differentiation process is mediated by RANKL-induced translocation of NF-κB from the cytoplasm into the nucleus (Fig. 6A and B), leading to downstream nuclear mechanisms culminating in upregulated osteoblastic genes and increased ALP activity (Fig. 4B). The specific decoy receptor for RANKL, OPG, abrogates the osteoblastic differentiation of human VSMCs.

It should be emphasized that this was an in vitro study, and we did not seek to correlate the clinical stage of CN to the extent of MAC or the serum level of RANKL/OPG. This is an obvious limitation that will be addressed through future correlative clinical studies.

The bone changes seen in CN are customarily thought to be triggered by events that enhance an inflammatory milieu, inducing monocytes into an osteoclastic differentiation program (10,22,23). We show that IL-8 and GCSF are two proinflammatory cytokines that are elevated in the serum of patients with CN. Inflammatory cytokines induce activated T cells to secrete RANKL (24,25), and attendant to this, OPG levels increase to counteract the effects of RANKL (26). Figure 2 illustrates that both RANKL and OPG levels are higher in patients with CN. The current study is the first to provide data on systemic levels of RANKL and OPG in patients with CN. There is evidence in the literature reporting elevated levels of RANKL and OPG in other disease conditions associated with high bone turnover and vascular calcification, notably rheumatoid arthritis (27,28), Paget’s disease (29,30), and osteoporosis (31–33). In particular, most studies reveal that the RANKL/OPG ratio is the more significant predictor of these disease conditions. Our data showing that the ratio of RANKL/OPG is significantly higher in patients with CN than in diabetic patients or nondiabetic control subjects (Fig. 2C) support these findings. Whether there is a cutoff ratio that may be diagnostic of CN or the ratio can be used to monitor disease activity remains to be elucidated. However, the high levels of the RANKL/OPG ratio can be anticipated to orchestrate deleterious effects on the vasculature as discussed below.

Local RANKL expression within the medial layer of the vasculature coincided with areas of MAC. Intense RANKL staining was observed in the vicinity of calcified areas confirmed by alizarin staining and a lack of staining in the negative tissue control (shown here as IMA) and the nonimmune IgG intra-assay control (results not shown). These findings agree with and support results from previous studies demonstrating increased RANKL protein and mRNA expression patterns in MAC (Monckeberg’s sclerosis) (34,35). Schoppet et al. (35) also found that OPG was predominantly adjacent to areas of calcification and areas of apoptosis throughout the whole circumference of their arterial specimens. However, the data for OPG immunostaining are less consistent in the literature. A study by Kaden et al. (13) on calcific aortic stenosis confirmed the positive correlation between RANKL immunostaining and calcification, but showed an inverse relationship between OPG immunostaining and calcified aortic stenosis. However, in the study by Kaden et al., decalcification was performed by placing tissue sections for 48 h in 270 mmol/L EDTA, a protocol that has not been replicated elsewhere and that may affect epitope exposure.

Our data demonstrate that explanted resident VSMCs undergo an osteogenic differentiation when cultured in media containing Charcot serum (Fig. 4B), a milieu that we show contains a high RANKL/OPG ratio (Fig. 2C). Our data suggest that systemic OPG is increased in patients with CN probably in response to and to mitigate the effects of high levels of RANKL. It is worth noting that although both OPG and RANKL levels were higher in diabetic patients compared with healthy control subjects, the RANKL/OPG ratio was not different between these two groups. It is therefore likely that elevated OPG levels in diabetic patients do compensate for elevated RANKL levels such that the ratio remains comparable to that of nondiabetic control subjects.

To further support the hypothesis that RANKL mediates MAC in patients with CN, immunocytochemical studies reveal that both RANKL- and Charcot-induced differentiation of human VSMCs are associated with the translocation of NF-κB from the cytoplasm into the nucleus (Fig. 6).

Of note, RANKL and Charcot serum enhanced mineralization of human VSMCs even in nonosteogenic conditions, i.e., normal growth medium without any added β-glycerophosphate or calcium. This finding refutes commonly held suggestions that MAC is merely a direct precipitation of inorganic (calcium and phosphate) ions when their levels are high in systemic circulation. Our data provide direct evidence of mineralization of human VSMCs in vitro by using nonosteogenic culture conditions. In fact, vascular calcification is an intricate and highly organized process involving cell differentiation, matrix deposition, and secondary ossification resulting in the formation of hydroxyapatite. In this study, we show that the cells are not only depositing a mineralized matrix but also actually switching their phenotype into osteoblastic cells as suggested by the elevated ALP activity, which is a marker of early osteogenic differentiation.

There is some controversy regarding the origin of cells that undergo osteogenic differentiation in MAC. Although some authors have suggested that circulating osteogenic precursor cells migrate from the blood through the vessel wall to differentiate locally into osteoblastic cells, others have maintained that these osteoblastic cells originate from residential human VSMCs displaying marked phenotypic plasticity under specific stimuli. Our study supports the latter hypothesis because it demonstrates using an in vitro model that explanted human VSMCs of medial origin directly differentiate into an osteoblastic phenotype. Our data support those of Schoppet et al. (35), who found that the medial tissue surrounding calcified lesions in MAC contained only smooth muscle cells, whereas macrophages and lymphocytes were absent.

By using VSMCs from rat aortas, Panizo et al. (15) demonstrated that RANKL induced mineralization in a dose-dependent manner, an effect abrogated by OPG. Data reported in this study support and extend those of Panizo et al., implicating RANKL/OPG signaling in MAC by using smooth muscle cells from human origin.

This study showed that OPG, a specific decoy receptor for RANKL, attenuated mineralization and osteoblastic differentiation of VSMCs induced by Charcot serum or RANKL. Our data also suggest a more pronounced effect of OPG on Charcot-mediated differentiation and mineralization of VSMCs compared with that induced by diabetes serum. Because OPG directly and specifically blocks the action of RANKL, these data provide additional evidence of RANKL/RANK signal modulation in CN-induced MAC. The nuclear translocation of NF-κB was also abrogated by coincubation with OPG, thus lending support to the key mechanistic role played by NF-kB in RANKL-induced mineralization. In fact, binding of RANKL to its receptor RANK may activate the canonical or the alternative pathway of NF-κB signaling (36,37).

Tseng et al. (38) recently used a murine model of calcification to show that atherosclerotic calcification involves bone remodeling with simultaneous activation of both osteoclastic and osteoblastic differentiation. Although OPG attenuated Forskolin-induced osteoclastic differentiation, thus confirming similar reports by Mabilleau et al. (10), Tseng et al. reported that calcification was not abrogated by OPG and concluded that pathways different from RANKL/OPG signaling were implicated. This is not surprising because murine models of calcification, and likewise, atherosclerotic calcification, may involve different pathways than in humans. However, we provide evidence for the direct involvement of RANKL/OPG in human MAC as typified in CN.

In summary, patients with CN have high levels of inflammatory cytokines potentially modulating the RANKL/OPG signaling pathway. They further display elevated systemic (serum) RANKL levels and high RANKL/OPG ratio, driving VSMCs into an osteoblastic differentiation pathway, resulting in deposition of a mineralized matrix in vitro. Differentiation of VSMCs occurs via the RANKL/RANK signaling cascade, setting intracellular mechanisms involving the nuclear translocation of NF-κB. This study provides additional data demonstrating that MAC in CN is independent of the high calcium-phosphate microenvironment. Furthermore, OPG is shown to prevent nuclear translocation of NF-κB and abrogate differentiation and mineralization of VSMCs induced by RANKL or Charcot serum. On the basis of the findings from this study, a working model of MAC as occurs in patients with CN is proposed (Fig. 7).

When interpreted in the light of other studies confirming the crucial role of RANKL/OPG signaling in osteoclastic bone resorption (10), our data provide a potential unifying hypothesis for the underlying pathogenetic mechanisms resulting in diabetic CN. Abnormal RANKL/OPG signaling, therefore, may be proposed as the mechanism underpinning the paradoxical osteolysis and MAC seen in CN and similar disease conditions. This holds promise for the treatment of CN, especially because new drugs specifically targeting RANKL are already marketed for use in osteoporosis. Clinical trials are warranted to assess the efficacy of anti-RANKL therapy in the treatment or prevention of CN and vascular calcification.

This study was funded partly by Diabetes UK project grants RD07/0003493 (Foot Problems in Dialysis Patients with Diabetes) and 03/0002584 (A Randomized Double-Blind Placebo-Controlled Trial of the Oral Bisphosphonate, Alendronate plus intravenous Pamidronate, in Active Diabetic Charcot Neuroarthropathy). Support was received from the Manchester National Institute for Health Research Biomedical Research Centre and Manchester Academic Health Science Centre.

No potential conflicts of interest relevant to this article were reported.

A.N. designed the study, performed all the experiments, researched data, and wrote and edited the manuscript. A.W. researched data. E.B.J., F.S.-I., S.R., and J.V.S. researched data and reviewed the manuscript. A.J.M.B. and M.Y.A. designed and supervised the study, researched data, and reviewed and edited the manuscript.