Pure polyethylene particles (Ceridust VP 3620) were obtained from Clariant (Gersthofen, Germany). In our hands, the particle size was determined using scanning electron microscopy as previously detailed 6. The mean particle size was 5.14 μm ± 3.05 μm (median 4.39 μm; range 0.52 to 15.5 μm). More than 55% of the particles were less than 5 μm in size. All particles were washed in ethanol to remove endotoxins, and then dried in a dessicator 6. This treatment resulted in negative testing for endotoxins using a quantitative Limulus Amebocyte Lysate (LAL) Assay (Charles River, Margate, UK) at a detection level of <0.05 EU/mL. Particles were stored at 4°C before implantation.

Eleven week-old C57BL/6J female mice (n = 96) were purchased from Janvier Laboratory (Le Genest-Saint-Isle, France). All mice were handled in agreement with French and international guidelines for care and use of laboratory animals. The protocol was given ethical approval by the local Animal Care Committee. Forty-eight mice were subjected to bilateral ovariectomy (OVX group) via the dorsal approach under general anesthesia (isoflurane). Twenty-four mice were sham-ovariectomized (sham-OVX), that is, ovaries were exteriorized, but not removed. Twenty-four non-OVX mice constituted the normal control group. Animals were housed in quarantine under local vivarium conditions (24°C and 12 h/12 h light/dark cycle) for one week.

Surgical implantation of PE particles has been previously described 4. Briefly, mice were operated on under general anesthesia via inhaled isoflurane. All animals were 12 weeks old at surgery. A 0.5 × 0.5 cm² area of periostum was exposed by making a 10 mm midline sagittal incision over the calvaria. In PE-implanted groups, the particle powder (20 μg) was distributed uniformly over the intact periosteum using a sterile sharp surgical spoon. Table 1 shows the treatment of the different experimental groups of mice. Postoperatively, 24 OVX animals, including 12 sham-implanted mice and 12 mice implanted with particles, were given subcutaneous oestrogen replacement therapy five days per week (15 μg 17 beta-estradiol/kg body weight/day, Sigma Inc., St. Louis, MO, USA), and 24 OVX animals, including 12 sham-implanted mice and 12 mice implanted with particles, received subcutaneous vehicle only. Water and food were given ad libitum. Body weights were measured weekly, and the E injections were adjusted accordingly. Fourteen days postoperatively, the animals were sacrificed by an overdose of intraperitoneal sodium pentobarbital. The uteri were harvested and weighted to confirm oestrogen loss.

Seven calvariae per group were dissected after sacrifice. Specimens were freed of all soft tissues, and fixed in 4% neutralized paraformaldehyde. The skulls were analyzed with a high-resolution micro-computed tomograph (micro-CT) (Skyscan 1172; Skyscan, Aartselaar, Belgium) to perform qualitative and quantitative analyses of calvarial bone. Imaging analysis mainly focused on the osseous properties in the area of the skull sagittal suture. The radiographic projections (n = 210) were acquired at 80 kV and 100 μA with an exposure time of 100 ms. Eight frames were averaged for each rotation increment of 0.9° to increase the signal to noise ratio. 3D images were reconstructed with a voxel average size of 10 μm, using the manufacturer reconstruction software (NRecon, Skyscan). Qualitative and quantitative data were analyzed with a global fixed threshold 13. To minimize the bias from the 3D orientation of the calvaria, a spheric volume of interest (VOI of 2 mm diameter) with the midline suture of the skull in its center was defined, as previously described 6. For quantitative analysis of PE particle-induced osteolysis, the resident software (CTAn, Skyscan) was used to obtain the following measurements within the VOI: bone volume (BV, mm³), and trabecular thickness (Tb.Th., mm). To address bone tissue mineralization following oestrogen depletion/substitution, a relative density measurement was computed. For this purpose, the mean value of grey-level intensity, reflecting the density of bone tissue, was obtained in entire calvariae.

Calvariae were processed undecalcified after embedding in polymethyl-methacrylate according to a method previously described 6. Sections (10 μm) were obtained at the depth where the presence of particles was detected within the calvaria tissue. These sections were stained with Stevenel Blue and picrofuchsine and a serial section stained for tartrate-specific acid phosphatase-positive (TRAP) osteoclasts (Acid Phosphatase kit, Sigma-Aldrich, Steinheim, Germany). Using a magnification of 20×, the histomorphometric analysis of each calvaria cap was performed on the most central section and on four adjacent sections. The region of interest was defined as previously recommended 4. The sagittal suture area (SSA) was determined by tracing the area of soft tissue between the parietal bones. It included resorption pits on the superior surface of the calvaria visible in the same field as the midline suture. To determine bone thickness, sections were divided using a digital caliper in four 100 μm steps to the left and right sides of the midline suture respectively. The measurements were expressed as a percentage of the mean ratio of bone thickness to the total tissue thickness (BT/TT) of the nine regional measurements in the five adjacent sections. To quantify bone loss, results were also expressed as a percentage of the mean difference of BT/TT (Δ(BT/TT)) between sham-operated and particle-implanted animals, in normal controls, sham-OVX, OVX, and OVX + E mice groups. Osteoclasts were identified as large multinucleated cells located on the bone perimeter within a resorption lacuna. Their localization was confirmed in a serial section stained for tartrate-specific acid phosphatase. The osteoclast number was measured in the region of interest of the five consecutive calvaria sections.

Calvariae were removed en bloc under sterile conditions from five animals per group randomly assigned for culture. Each calvaria was placed into one well of 12 wells/plate and cultured with 1 ml serum-and phenol-free Dulbecco's modified Eagle's medium (DMEM) with glutamine (Invitrogen, Paisley, UK), and 1% penicillin and streptomycin at 37°C and 5% CO2 for 24 h. Twenty four hours later, the culture medium was collected and stored at -80°C for the assay of IL-1β, IL-6, TNF-α and RANKL secretion. Then, calvariae were calcinated and ashes weighted to normalize the production of cytokines.

To measure IL-6 concentration in serum, blood was obtained by cardiac puncture before death, and collected in heparinized tubes. Blood samples were then centrifuged at 3,000 × g for 10 minutes, aliquoted and frozen at -80°C until assayed. Serum measurements of IL-6 were performed using quantitative, noncompetitive, sandwich enzyme-linked immunosorbent assay kit for detecting mouse IL-6 (Quantikine, R&D Systems Europe, Ltd, Abingdon, UK). Similarly, ELISA kits for detecting mouse IL-6, IL-1β, TNF-α and RANKL (R&D Systems) were employed to measure cytokines released into organ culture supernatant. The detection limits of the assay were 1.6 pg/mL for IL-6, 3 pg/mL for IL-1β, 5.1 pg/mL for TNF-α, and 5 pg/mL for RANKL. Cytokines levels lower than the detection limit were considered to be 0 pg/ml. Absorbance was read in an ELISA reader (Micro-Quant, Bio-Tek Instruments, Colmar, France) at 490 and 540 nm as per manufacturer's instructions.

Calvariae were removed from culture medium, and immediately snap-frozen in liquid nitrogen, and pulverized to powder in a stainless steel mortar. RNA was extracted with 1 mL Trizol^® reagent (Invitrogen, Carlsbad, CA, USA), using the manufacturer's protocol. Quantity and purity of RNA was determined by absorbance on a NanoDrop™ 1000 Spectrophotometer (Labtech France, Palaiseau, France) at 260 and 280 nm. All samples had ratios >1.7 and were accepted for analysis.

Samples of total RNA were reverse-transcribed using the first-strand synthesis kit of SuperScript™ II Reverse Transcriptase (Invitrogen). Quantitative gene expression analyses were carried out using Real-time PCR by means of the TaqMan™ Gene Expression Assays Protocol (mouse RANK: Mm00437135_m1, mouse RANKL: Mm01313943_m1, mouse OPG: Mm01205928_m1, and human 18S rRNA: Hs99999901_s1; Applied Biosystems, Foster City, CA, USA) and the iCycler thermal cycler RT-PCR Detection System, coupled to MyiQ™ Single-Color software (Bio-Rad, Hercules, CA, USA). The absolute number of gene copies was normalized using 18s rRNA and the relative quantification of the genes was calculated using the "comparative C_T (threshold cycle) method" as per manufacturer's instructions (Applied Biosystems).

All values are presented as means ± SEM. Quantitative parameters obtained from micro-CT (BV and Tb.Th) were analyzed by two-tailed Student's t-test within each group. Comparisons across unpaired groups were made according to the Mann-Whitney U test. Data concerning osteoclast number, sagittal suture area, and bone thickness were initially analyzed using a one-way analysis of variance (ANOVA). Post hoc comparison between groups used the Fisher's protected least significant difference. Two-way ANOVA was performed to determine whether there was a significant effect of either polyethylene particles or ovariectomy or oestrogen supplementation on osteoclast number, sagittal suture area and bone thickness. Cytokines concentrations and mRNAs expression were analyzed using one-way ANOVA followed by the Bonferroni/Dunnet's test. The level of statistical significance was set at P < 0.05.

All mice had increased body-weight by the end of the experiment. The final body-weight gain was +18% in the OVX group as compared to +6.8% (P < 0.0001), +3.5% (P < 0.0001), and +11% (P = 0.0001) in normal controls, sham-OVX mice, and OVX+E mice, respectively.

At sacrifice, atrophy of uterine tissue was noted, indicating the effectiveness of oestrogen privation following ovariectomy. There was a significant difference in uterine weight in OVX mice group (mean, 24.5 ± 3.6 mg), as compared to normal control animals (mean, 107.9 ± 12 mg, P < 0.0001), sham-OVX mice (mean, 86.4 ± 17 mg, P < 0.0001) and OVX+E mice (mean, 80.2 ± 10.9 mg, P < 0.0001). In contrast, uterine weight was notably less altered after oestrogen supplementation in OVX animals as compared to normal controls (80.2 ± 10.9 mg versus 107.9 ± 12 mg, P = 0.002).

Ovariectomy induced a consistent, although non-significant, decrease in mean grey-level intensity in calvariae without PE particles as compared to normal controls (Figure 1). Specifically, mean grey-level values were 117.6 ± 7.7 (range, 107 to 48), 115 ± 2 (range, 107 to 120) and 110 ± 0.3 (range, 109 to 111), in normal controls, sham-OVX mice and OVX mice, respectively. In contrast, grey-level intensity significantly increased in OVX animals after oestrogen supplementation (124 ± 7.8), as compared to OVX mice (P = 0.02), while it was similar to normal controls ({ P = 0.22) and sham-OVX mice (P = 0.07). The presence of PE particles induced profound changes in calvarial bone microarchitecture in all groups (Figure 2). In particle-implanted animals, BV significantly decreased within the VOI as compared to sham-implanted animals in normal control group (P = 0.004), in sham-OVX group (P = 0.003), in OVX group (P = 0.04) and in OVX+E group (P = 0.001) (Figure 3). However, decrease in BV following particle implantation was consistently less marked in the OVX group (-7.8%, as compared to -21.2% in normal control group, P = 0.04, to -19.9% in sham-OVX group, P = 0.003, and to -21.4% in OVX+E group, P = 0.005). Polyethylene particles did not produce any significant decrease in bone thickness (Tb.Th) in OVX mice (+3.8% as compared to sham-implanted OVX mice, P = 0.08), in contrast with normal controls (-13.6% as compared to sham-implanted normal controls, P = 0.01), sham-OVX mice (-11.1% as compared to sham-implanted sham-OVX mice, P = 0.008), and OVX+E group (-11.6% as compared to sham-implanted OVX+E mice, P = 0.005) (Figure 4).

Histological sections showed a consistent erosion of the calvarial bone, associated to fibrous and granulomatous scar tissue layer on the periosteal side of the calvaria in particle-implanted mice, in all groups. The most dramatic lesions were observed when PE particles were implanted in the OVX+E group (Figure 5). However, the tissue response to particles appeared limited in OVX animals without E supplementation. Using polarized light, many particles were found intracellularly within macrophages and foreign-body giant cells. TRAP staining confirmed the presence of osteoclasts located in resorptive lacunae.

Histomorphometric data are shown in Table 2. One-way ANOVA revealed a significant difference for the sagittal suture area (P < 0.0001). The presence of PE particles induced a significant increase in SSA in normal control group (P < 0.001), in sham-OVX group (P < 0.001), in OVX group (P < 0.001), and in OVX+E group (P < 0.0001). No significant difference was found in SSA between sham-implanted normal controls group and sham-implanted OVX group (P = 0.9), between PE-implanted normal control mice and PE-implanted OVX mice (P = 0.99), and between sham-implanted normal control mice and sham-implanted OVX+E mice (P = 1). Two-way ANOVA revealed a significant effect of PE particles on SSA (P < 0.0001). There was no independent effect of OVX on SSA (P = 1) in the absence of particles. One-way ANOVA revealed a significant difference for osteoclast number (P < 0.0001). The presence of PE particles induced a consistent increase in osteoclast number in normal control group (P < 0.0001), in sham-OVX group (P < 0.001), and in OVX+E group (P < 0.0001), but not in OVX group (P = 0.99). There was also a significant effect of ovariectomy on osteoclast number (P = 0.047) in the absence of particles.

One-way ANOVA showed a significant difference for bone thickness (P < 0.0001). Implantation of PE particles resulted in a marked decrease in BT/TT in normal control group (P = 0.001), in sham-OVX group (P < 0.001), and in OVX+E group (P < 0.0001). However, the presence of PE particles did not influence BT/TT in OVX group (P = 0.088). Also, two-way ANOVA showed an independent effect of ovariectomy on BT/TT (P < 0.001) in the absence of particles.

Serum levels of IL-6 significantly increased after PE particles implantation in all mice groups, as compared with sham-implanted internal controls (P = 0.008 in normal control group, P = 0.01 in sham-OVX mice, and P = 0.02 in OVX+E mice), except in OVX mice group (P = 0.57) (Figure 6). Ovariectomy did not significantly increase serum IL-6 concentrations in the absence of particle implantation (sham-implanted OVX group versus sham-implanted normal controls, P = 0.78, and sham-implanted OVX+E group versus sham-implanted normal control mice, P = 0.84).

The presence of PE particles induced a significant increase in IL-1 β concentration in all mice groups (Figure 7). However, particles implantation did not significantly influence IL-6 media rates as compared to internal controls (Figure 7). In OVX mice without particles, the levels of TNF-α were significantly lower than in normal controls (P = 0.02). In addition, in this group, PE particles did not increase the rates of TNF-α (P = 1), which remained considerably lower than in normal controls and in OVX+E mice with PE particles (P = 0.002 and P = 0.002, respectively) (Figure 7). Significant differences in RANKL basal release (absence of particles implantation) were found between OVX mice and each other mice groups (normal controls, P = 0.047, sham-OVX mice, P = 0.016, OVX+E mice, P = 0.02). Noteworthy, RANKL local production appeared to be stimulated by PE particles in normal controls, in sham-OVX mice, and in OVX+E mice, while in OVX group, no significant modification in RANKL rates could be detected following particles implantation (P = 0.95) (Figure 7).

As shown in Figure 8, the combined effect of ovariectomy and PE particles implantation stimulated RANKL mRNA expression in calvariae, in all mice groups. However, PE-implanted calvariae in normal controls, sham-OVX and OVX+E mice expressed RANKL mRNA at levels more than twice higher than OVX mice. OPG mRNA expression was not significantly different between groups (Figure 8). As a result, the RANKL/OPG mRNA ratio significantly increased in all mice groups after PE implantation, but appeared consistently downregulated in OVX mice (Figure 8). The RANKL/OPG mRNA ratio significantly differed between OVX mice and each other mice group in the absence of particles (P = 0.047 as compared to normal controls, P = 0.034 as compared to sham-OVX mice, and P = 0.049 as compared to OVX+E mice). In addition, no significant difference could be observed in the basal RANKL/OPG mRNA ratio between normal controls and OVX+E mice, indicating that E supplementation reestablished RANKL mRNA expression relative to OPG mRNA.