S100A8/A9, a potent serum and molecular imaging biomarker for synovial inflammation and joint destruction in seronegative experimental arthritis

Male and female IL-1Ra^–/– mice from the BALB/c background were kindly provided by Dr. M. Nicklin (Sheffield, UK) and were generated as described previously [31]. Male and female BALB/c control mice (Janvier, France) were 4 weeks old upon arrival. Mice were housed under standard housing conditions: filter top cages, temperature 20–24 °C, 12 h light-dark cycle, and ad libitum access to animal chow and water.

Arthritis development in each hind paw in IL-1Ra^–/– mice was macroscopically scored weekly or every 2 weeks using an arbitrary scoring system on a scale of 0 to 2 per paw as described previously [32]. The following scoring criteria were used: 0, no redness and swelling; 0.25, slight redness; 0.5, slight redness and swelling; 0.75–1, mild redness and swelling; 1.25–1.5, moderate redness and swelling; 1.75–2, severe redness and swelling. Only hind paws were scored as arthritis rarely develops in the fore paws.

Blood was drawn from the retro-orbital plexus and collected in MiniCollect Serum Separator tubes (Greiner Bio-One). Serum concentrations of the cytokines IL-1β, IL-6, tumor necrosis factor (TNF), IL-17, IL-4, and interferon (IFN)-γ were determined using the Luminex multi-analyte technology on the Bio-Plex 100 system (Bio-Rad) in combination with the multiplex cytokine kit (Milliplex, Millipore, Amsterdam, the Netherlands). Serum was three-times diluted and sensitivity was >0.36 pg/ml. Serum S100A8/A9 concentrations were determined by an in-house sandwich ELISA specifically for mouse S100A8/A9 as described previously [23].

For assessment of joint inflammation and damage, total ankle joints were dissected and fixed in 4 % formalin for 4 days. Knees were decalcified in 5 % formic acid, dehydrated in a series of ethanol and embedded in paraffin. Sections of 7 μm were cut and stained with hematoxylin and eosin (H&E) for analysis of cell influx, bone erosion, and chondrocyte death, and Safranin O (SafO) staining for proteoglycan (PG) depletion and cartilage erosion. Each parameter was arbitrarily scored on a scale of 0 to 3 with steps of 0.25 by two independent observers in a blinded manner. Histological parameters were determined in the joints of the tibia and talus (tibio-talar joint) and of the talus and navicular bone (talo-navicular joint); three sections per ankle joint were scored and the mean score was determined.

Tissue sections (7 μm) from formalin-fixed, paraffin-embedded ankle joints of IL-1Ra^–/– mice and BALB/c control mice were digested with proteinase-free chondroitinase ABC (0.25 units/ml in 0.1 M Tris-HCl, pH 8.0; Sigma-Aldrich) for antigen retrieval. Tissue sections were incubated overnight with rabbit anti-S100A8 and anti-S100A9 [23] or with rabbit anti-VDIPEN for staining of MMP-mediated cartilage destruction. Sections were then incubated with biotinylated horseradish peroxidase-conjugated goat anti-rabbit IgG (Dako) as a second antibody followed by incubation with avidin-streptavidin-peroxidase (Elite-kit, Vector). Peroxidase activity was assessed by staining with 3,3′-diaminobenzidine (DAB; Powervision DAB, Immunologic, Duiven, the Netherlands) in the presence of H₂O₂ and all sections were counterstained with hematoxylin for S100A8 and S100A9 staining and with orange G (2 %) for VDIPEN staining.

To monitor local synovial S100A8 expression and activated MMPs in vivo, optical imaging was performed in 16-week-old IL-1Ra^–/– mice. Mice received an intravenous (i.v.) injection via a tail vein of the specific Cy7-labeled S100A8 polyclonal antibody (anti-S100A8-Cy7) or Cy7-labeled antibody of irrelevant specificity (Rabbit-IgG-Cy7) (2 nmol of Cy7 ~ 100 μg antibody) [23]. For MMP imaging, mice received an i.v. injection of 2 nmol Cy5.5-labeled AF489 (AF489-Cy5.5), a low-molecular weight MMP inhibitor that targets the active site of MMPs in vivo [33].

Twenty-four or 3 h post i.v. injection of anti-S100A8-Cy7 and AF489-Cy5.5, respectively, mice were anesthetized (2.5 % isoflurane/oxygen) and placed in the light-tight chamber and imaged with the IVIS Lumina (Caliper Life Sciences, Hopkinton, MA, USA) for 1 min. For imaging of the Cy7-labeled antibodies, excitation and emission wavelengths were set at 710 and 810–885 nm, respectively, and for AF489-Cy5.5 at 640 and 695–770 nm. Regions of interest were set on the inflamed ankle joint and muscle tissue (background). Signal-to-noise ratio (SNR) was calculated as SNR = mean fluorescent intensity of ankle/standard deviation of background signal.

Nominal data are presented as mean values ± standard deviation, and ordinal data with box-and-whisker plot. Differences between groups were assessed by Student’s unpaired t test, or Mann-Whitney test when appropriate. Spearman rank correlation coefficients (rs) were calculated between serum levels of S100A8/A9 or pro-inflammatory cytokines and macroscopic and microscopic parameters for joint inflammation and destruction, and between fluorescent signal of in vivo imaging and macroscopic score for joint swelling. To determine the accuracy of serum S100A8/A9 as a biomarker for macroscopic joint swelling (score >0.5), receiver operating characteristic (ROC) curves were generated and the area under the curve (AUC) was calculated, where 0.5 represents random chance and 1.0 a perfect biomarker. Similarly, ROC and AUC were calculated to determine the accuracy of serum S100A8/A9 as biomarker for microscopic cell influx, bone erosion, cartilage erosion, PG depletion, and chondrocyte death (score >0.5). Statistical significance was set at P < 0.05 (two-tailed).

IL-1Ra^–/– mice (n = 26) developed a spontaneous and heterogeneous joint swelling in the hind paws starting at week 6, reaching a combined median joint swelling in both ankle joints of 0.6 on a scale of 0 to 4 with a disease incidence of 70 % at week 15 (Fig. 1a). Serum levels of S100A8/A9 were higher in mice with increased joint swelling and correlated significantly with the macroscopic score for joint swelling at week 15 (rs = 0.730, P < 0.0001; Fig. 1b). Moreover, at week 15, S100A8/A9 levels were able to discriminate low joint swelling (score 0–0.5) from mild and severe swelling (score 0.75–4) with high accuracy (AUC = 0.88, 95 % confidence interval (CI) 0.75–1.00; Additional file 1A). In contrast, serum levels of various pro-inflammatory cytokines did not correlate with joint swelling: IL-1β (rs = –0.352, P = 0.056), IL-6 (rs = –0.150, P = 0.454), and TNF (rs = –0.011, P = 0.956) (Fig. 1c). Serum levels of the T cell-related cytokines IL-17 (rs = 0.304, P = 0.102), IL-4 (rs = –0.004, P = 0.983), and IFN-γ (rs = –0.019, P = 0.921) also did not correlate with joint swelling (Additional file 2A).

To investigate whether serum levels of S100A8/A9 are prognostic for disease outcome, serum S100A8/A9 levels were measured in IL-1Ra^–/– mice (n = 37) every 2 weeks starting at week 4 when arthritis has not developed yet, and compared to levels in age-matched BALB/c control mice (n = 8). S100A8/A9 levels in IL-1Ra^–/– mice gradually increased during development and were significantly increased compared to control mice at 8 weeks of age (P = 0.029; Fig. 2a). From that time point on, serum S100A8/A9 levels correlated with the macroscopic score for joint swelling (rs = 0.489, P = 0.002; Fig. 2b) throughout disease development.

To investigate the potential of S100A8/A9 as a prognostic biomarker, we investigated whether increased S100A8/A9 serum levels at early time points were associated with increased joint swelling at the 16-week end point. As development of arthritis in IL-1Ra^–/– mice is variable, mice were stratified at end point (week 16) into three groups, i.e., mice with low (score 0–0.5), mild (score 0.75–2), and severe macroscopic swelling (score 2.25–4). Next, serum S100A8/A9 levels in the three groups were investigated throughout disease development. Whereas serum S100A8/A9 levels between low and mild arthritic mice did not differ, severe arthritic mice showed increased serum S100A8/A9 levels compared to both low and mild arthritic mice at week 10 (P = 0.009 and P = 0.004, respectively), indicating that high serum S100A8/A9 levels at week 10 are prognostic for severe macroscopic swelling (score 2.25–4) at week 16 (Fig. 2c). The prognostic value of serum S100A8/A9 in IL-1Ra^–/– mice at week 10 is further strengthened by its correlation with the macroscopic score for swelling at week 16 (rs = 0.605, P < 0.0001; Fig. 2d). Furthermore, while at weeks 4, 6, and 8 serum S100A8/A9 levels were not able to predict severe joint swelling (score 2.25–4) at week 16 (AUC = 0.52, 0.61, and 0.61, 95 % CIs 0.32–0.71, 0.42–0.79 and 0.43–0.79, respectively; Additional file 1B), serum S100A8/A9 levels were an accurate prognostic biomarker at weeks 10, 12, and 14 (AUC = 0.82, 0.88, and 0.81, 95 % CIs 0.68–0.96, 0.77–0.99, and 0.67–0.95, respectively; Additional file 1B).

We next investigated whether serum levels of S100A8/A9 at 16 weeks, apart from macroscopic score for joint swelling, also correlated with microscopic parameters for joint inflammation. Synovial cell influx was determined in the tibio-talar and talo-navicular joints using an arbitrary scale of 0 to 3, and the combined score of both ankle joints was determined. Ankle joints of 16-week-old arthritic IL-1Ra^–/– mice showed increased amounts of infiltrated inflammatory cells (≈60 % neutrophils and 40 % monocytes/macrophage) in the synovium, which correlated closely with the macroscopic score of swelling (rs = 0.930, P < 0.0001; Additional file 2B) and serum levels of S100A8/A9 (rs = 0.763, P < 0.0001; Fig. 3a). Additionally, serum S100A8/A9 was able to discriminate increased cell influx (score >0.5) with a high accuracy (AUC = 0.93, 95 % CI 0.82–1.00; Additional file 1C).

A prominent expression of S100A8 and S100A9 within the activated synovium and infiltrating cells in the surrounding tissue of the ankle joints of 16-week-old IL-1Ra^–/– mice was visualized by immunohistochemistry using a specific antibody against murine S100A8 and S100A9 in serial sections (Fig. 3b). No expression of S100A8 or S100A9 was observed in the synovia of IL-1Ra^–/– mice with no swelling (score 0), whereas in mice with a macroscopic score for joint swelling of 1 and 2 a clear increased S100A8 and S100A9 expression was observed in the inflamed ankle joint (Fig. 3b). Expression of S100A8 and S100A9 within the inflamed ankle joint of IL-1Ra^–/– mice show a similar distribution and this is co-expressed in monocytes and neutrophils (Fig. 3b), indicating the inflamed joint as the source of increased serum S100A8/A9 levels in high arthritic IL-1Ra^–/– mice. No staining with isotype control IgG was found on ankle joint sections of arthritic IL-1Ra^–/– mice (inserts in Fig. 3b).

Since expression of both S100A8 and S100A9 is similar in ankle joints of arthritic IL-1Ra^–/– mice, we continued examining S100A8 only in the following experiments. To investigate whether local expression of S100A8 in high arthritic IL-1Ra^–/– mice could also be visualized in vivo we used a specific anti-S100A8 antibody coupled to Cy7. Targeting of anti-S100A8-Cy7 was assessed in 16-week-old IL-1Ra^–/– mice using optical imaging, and a fluorescent signal was observed in inflamed ankle joints (mean macroscopic score of 1.4 ± 0.3, n = 6; Fig. 3c). Specificity of the anti-S100A8-Cy7 signal was confirmed by comparison to the signal of an irrelevant rabbit-IgG-Cy7 in IL-1Ra^–/– mice with comparable joint swelling (mean macroscopic score of 1.3 ± 0.4, n = 6) which was significantly lower (P = 0.0002; Fig. 3c).

To further explore whether S100A8 imaging can be used as a biomarker tool to distinguish disease severity in IL-1Ra^–/– mice, anti-S100A8-Cy7 targeting was imaged in ankle joints of IL-1Ra^–/– mice with various degrees of joint swelling and showed a significant correlation with macroscopic score for joint swelling (rs = 0.731, P = 0.001; Fig. 3d). These data indicate that S100A8 in the inflamed joint of IL-1Ra^–/– mice can be visualized non-invasively by molecular imaging using anti-S100A8-Cy7, and may be a relevant biomarker tool to assess disease severity.

Next, we studied the relationship between S100A8/A9 and joint destruction in 16-week-old IL-1Ra^–/– mice. First, we determined bone erosion on eight bone surface areas of the tibia, talus, and navicular bone. The combined mean bone erosion score of both ankle joints correlated with serum levels of S100A8/A9 in IL-1Ra^–/– mice (rs = 0.712, P < 0.0001; Fig. 4a), and at week 16 serum S100A8/A9 was able to discriminate increased bone erosion (score >0.5) with high accuracy (AUC = 0.87, 95 % CI 0.72–1.00; Additional file 1C).

Subsequently, various parameters for cartilage destruction were studied on the four cartilage surfaces of the tibio-talar and talo-navicular joints, and the combined mean score of both ankle joints was determined. All parameters correlated with serum levels of S100A8/A9, i.e., PG depletion (rs = 0.656, P = 0.0003; Fig. 4b), chondrocyte death (rs = 0.767, P < 0.0001; Fig. 4c), and cartilage erosion (rs = 0.726, P < 0.0001; Fig. 4d), and serum S100A8/A9 levels were able to discriminate for increased PG depletion, chondrocyte death, and cartilage erosion (score >0.5) with high accuracy (AUC = 0.84, 0.92, and 0.87, 95 % CIs 0.69–1.00, 0.81–1.00, and 0.72–1.00, respectively; Additional file 1C).

As MMPs are major enzymes involved in mediating cartilage destruction, we additionally investigated their link with the increased S100A8 levels that were observed in ankle joints of arthritic IL-1Ra^–/– mice. First, MMP-generated VDIPEN aggrecan neoepitopes were determined within the cartilage of ankle joint sections of arthritic IL-1Ra^–/– mice using immunohistochemistry. Increased MMP-mediated VDIPEN staining in the cartilage was associated with increased expression of S100A8 in the synovium, as shown by S100A8 and VDIPEN immunostaining in directly adjacent sections (Fig. 5a). VDIPEN neoepitopes were mainly found in the extracellular matrix surrounding the chondrocytes in the articular cartilage (Fig. 5a). In IL-1Ra^–/– mice with no joint swelling, little VDIPEN staining was visible around the chondrocytes, while in mice with increased joint swelling a gradual increase in VDIPEN staining was also observed in the extracellular matrix. In IL-1Ra^–/– mice with a maximum score for joint swelling of 2, all chondrocytes and the cartilage matrix were VDIPEN-positive (Fig. 5a), indicative of severe MMP-mediated cartilage breakdown. No staining with isotype control IgG was found on ankle joint sections of arthritic IL-1Ra^–/– mice (insert Fig. 5a).

Finally, to investigate the relationship between S100A8/S100A9 and MMP activity within the inflamed joints of IL-1Ra^–/– mice, we measured activated MMPs in IL-1Ra^–/– mice using molecular imaging. To this end, we used the low-molecular weight MMP inhibitor AF-489 which has been shown to bind to the active sites of several gelatinases and collagenases involved in cartilage breakdown, i.e., MMP-2, -9 and -13, with IC₅₀ values in the nanomolar range [33]. Targeting of the MMP inhibitor AF489-Cy5.5 was clearly visible in inflamed ankle joints of IL-1Ra^–/– mice, which correlated with increased macroscopic swelling of the ankle joint (rs = 0.751, P = 0.0008; Fig. 5b ). Increased levels of active MMPs can be visualized in vivo by optical imaging and is associated with increased cartilage damage and VDIPEN neoepitopes in chondrocytes of IL-1Ra^–/– mice which are linked to increased levels of S100A8 expression in the synovium.