Dynamic activation of bone morphogenetic protein signaling in collagen-induced arthritis supports their role in joint homeostasis and disease

Eight-week-old male DBA/1J mice were purchased from Janvier Laboratories (Le Genest-Saint-Isle, France). All experiments were approved by the Ethics Committee for Animal Research (Katholieke Universiteit Leuven, Leuven, Belgium). For induction of arthritis, chicken sternal cartilage CII (Sigma-Aldrich, Bornem, Belgium) was dissolved at 2 mg/mL in phosphate-buffered saline (PBS)/0.1 M acetic acid, stirred overnight (O/N) at 6°C, and emulsified with an equal volume of CFA (1 mg/mL) (Sigma-Aldrich). One hundred microliters of the emulsion (0.1 mg of CII) was injected intradermally at the base of the tail. At day 21 after immunization, mice received an intraperitoneal booster injection of 100 μL of CII (2 mg/mL). At day 25, the onset of arthritis was synchronized by an intraperitoneal injection of 100 μL of lipopolysaccharide (500 μg/mL in PBS) (Sigma-Aldrich) [25]. Mice were sacrificed at different time points (day 0, 20, 27, 33, 40, and 47 after immunization) for immunohistochemistry and protein and RNA expression assays. In additional experiments, mice were injected daily with 100 μL of soluble tumor necrosis factor-alpha (TNFα) receptor etanercept/PBS (250 μg/mL) (Wyeth Pharmaceuticals, Louvain-la-Neuve, Belgium) intraperitoneally (or PBS alone as negative control) from day 29 onwards. The severity score was determined daily according to the scoring system of Backlund and colleagues [26]. Mice were sacrificed at day 35.

At each time point and at the end of the TNFα blocking experiment, synovium and cartilage samples were dissected, separated, and used for RNA extraction. RNA was isolated using a Nucleospin RNAII kit (Macherey-Nagel, Düren, Germany) and reverse-transcribed using a Revert-Aid H Minus First strand cDNA synthesis kit (Fermentas, St. Leon-Rot, Germany) according to the manufacturers' instructions. For quantitative analysis, real-time polymerase chain reaction (PCR) was performed in duplicate using the Rotor-gene 6000 detection system (Corbett Research, Westburg, Leusden, The Netherlands). Gene expression of mouse BMP2, BMP4, BMP6, BMP7, growth and differentiation factor-5(GDF5), Noggin (NOG), and TNFα were studied using assay-on-demand primer/probe sets (Applied Biosystems, Lennik, Belgium). Expression was normalized to mouse housekeeping gene GAPDH (glyceraldehydes-3-phosphate dehydrogenase) using the comparative threshold method [27]. In kinetic experiments, data were further normalized to baseline levels.

At each time point, three sets of three pooled knees were used for protein extraction. Whole knees were weighed and homogenized (CAT homogenizer X120; CAT Ingenieurbüro M. Zipperer GmbH, Staufen, Germany) in 1 mL of cell extraction buffer (Biosource Europe, Nivelles, Belgium) supplemented with 5% Proteinase Inhibitor Cocktail (Sigma-Aldrich) and 1 mM PMSF (phenylmethylsulfonyl fluoride) (Sigma-Aldrich). Protein extracts were normalized to wet weight in an appropriate volume of cell extraction buffer. Samples were analyzed under reduced conditions (0.1 M DTT [1,4-dithiothreitol]). Samples were boiled for 5 minutes at 95°C, chilled on ice, and loaded onto a 4% to 12% Bis-Tris gel (Invitrogen Corporation, Carlsbad, CA, USA). Electrophoresis was carried out into a commercially available running buffer (NuPage MES SDS Running buffer; Invitrogen Corporation) at 130 V for 10 minutes in the beginning, followed by 25 minutes at 200 V. Proteins were transferred on a prewet PVDF (polyvinylidene difluoride) membrane (Millipore S.A./N.V., Brussels, Belgium) for 70 minutes at 30 V in a transfer buffer containing 0.4 M glycine, 0.5 M Tris base, 0.01 M SDS, and 200 mL/L methanol. Nonspecific binding sites were blocked in Tris-buffered saline/0.1% Tween (TBST) (wash buffer) with 5% milkpowder (BlottoA) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) for 1 hour at room temperature (RT). Blots were then probed O/N at 4°C with polyclonal antibody against phosphorylated SMAD1/5 (P-SMAD1/5) or SMAD5 (Cell Signaling Technology, Inc., Danvers, MA, USA) (1:1,000 in TBST/5% bovine serum albumin [BSA]) and thereafter incubated with an appropriate horseradish peroxidase (HRP)-conjugated secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) at a dilution of 1:5,000 in TBST/5% milkpowder for 1 hour at RT. For detection, a chemiluminescent substrate (Western Lightning; PerkinElmer Life and Analytical Sciences, Inc., Waltham, MA, USA) was applied on the membrane. Blots were visualized using an LAS-3000 mini CCD (charge-coupled device) camera using an exposure time of 15 minutes. Densitometry measurements were done using digital image densitometry analysis (ImageJ; National Institutes of Health, Bethesda, MD, USA). As positive controls, mouse mesenchymal progenitor C2C12 cells were stimulated with recombinant BMP2 (300 ng/mL for 30 minutes). As negative controls, blots were incubated with secondary antibody alone (data not shown).

Knees and ankles were dissected, formaldehyde/PBS-fixed O/N, decalcified with Decal (3 days at RT) (Serva, Heidelberg, Germany) or EDTA (ethylenediaminetetraacetic acid) (0.5 M in PBS, pH 7.5) (10 changes at 4°C), and embedded in paraffin. For immunohistochemistry, paraffin sections (5 μm thick) were deparaffinized with Histo-Clear (National Diagnostics, Atlanta, GA, USA) and methanol. For antigen retrieval, sections were incubated 2 hours at RT in 0.1 M sodium citrate/0.1 M citric acid. Endogenous peroxidase was quenched by incubating the slides for 10 minutes in 3% H₂O₂ in water (BMP7 and P-SMAD1/5) or 3% H₂O₂ in methanol (BMP2). Sections were washed three times in PBS/0.1% Triton (BMP7 and P-SMAD1/5) or in TBST (wash buffer) (BMP2) and blocked 30 minutes at RT in 20% donkey serum (BMP7, BMP2) or 20% goat serum (P-SMAD1/5) in wash buffer and were incubated O/N at 4°C with primary antibody at a final concentration of 10 μg/mL chicken anti-BMP7 (Pfizer Inc, New York, NY, USA), 2 μg/mL goat anti-BMP2 (Santa Cruz Biotechnology, Inc.), 1:50 dilution of rabbit anti-P-SMAD1/5 (Cell Signaling Technology, Inc.) or with isotype control (chicken, goat, and rabbit IgG) (Santa Cruz Biotechnology, Inc.) or serum (Dako, Glostrup, Denmark) at an appropriate concentration in wash buffer. Sections were then washed three times with wash buffer and incubated for 30 minutes at RT with secondary antibody. For BMP7 immunostaining, the secondary antibody was an HRP-conjugated anti-chicken (Jackson ImmunoResearch Europe Ltd, Newmarket, Suffolk, UK) diluted 1:100. For BMP2 immunostaining, a biotinylated donkey anti-goat at 1:400 dilution was used, followed by streptavidin anti-HRP (LSAB kit) (Dako) (30 minutes at RT). For P-SMAD1/5 immunostaining, an ABC kit (Vecta stain rabbit IgG [Vector laboratories Ltd., Peterborough, UK]) was used for signal amplification. Liquid DAB (3,3'-diaminobenzidine) substrate chromogen system (Dako) was used as a peroxidase substrate. Sections were counterstained with hematoxylin. Adjacent sections were stained with hematoxylin and eosin (H&E). GDF5 immunohistochemistry was not performed on these samples as we found that different commercially available antibodies showed a lack of specificity.

Where appropriate (n > 3), results were analyzed with SPSS 15.0 (SPSS Inc., Chicago, IL, USA) with the unpaired non-parametric Mann-Whitney U test. Statistically significant differences were defined as P values of less than 0.05.

Activation of BMP signaling during the course of CIA was visualized at different time points by immunohistochemistry for P-SMAD1/5 (Figure 1b,d,f,h). Histo-morphological scoring of adjacent H&E-stained sections was used to assess the severity of disease (Figure 1a,c,e,g,i). In the initial phase of CIA (exsudate score 1 and infiltration and pannus formation score 0), only a few P-SMAD1/5-positive cells were visible, mostly in the lining layer of the synovium (Figure 1b). With increasing severity, characterized by infiltration of inflammatory cells and synovial hyperplasia (exsudate score 1, infiltration score 3, and pannus formation score 1), P-SMAD1/5-positive cells became apparent in the subintimal zone of the synovium. Also, a few blood vessels and articular chondrocytes showed nuclear P-SMAD1/5 staining (Figure 1d–f). Eventually, in the destructive phase (pathology exsudate score 2, infiltration score 3, and pannus formation score 2), P-SMAD1/5-positive cells were identified in the invading pannus tissue (Figure 1h). In contrast, healthy knees showed almost no P-SMAD1/5-positive cells (data not shown) and IgG controls were negative (Figure 1j).

Activation of BMP signaling in CIA was further semi-quantified by Western blot on pooled protein extracts from whole knees at different time points (day 0, 20, 27, 33, 40, and 47 of the experiment). Positive controls were obtained by stimulating C2C12 cells with recombinant BMP2 (300 ng/mL) for 30 minutes. Mouse knee extracts (three per time point, each pooled from three different knees) were normalized to wet weight. A specific 60-kDa band was found in all samples (Figure 2a). Negative controls showed no band of this size (data not shown). C2C12 cells stimulated with BMP2 had a higher expression of P-SMAD1/5 than unstimulated cells (Figure 2b). We further studied the relative amount of phosphorylated and therefore activated SMAD1/5 as compared with total SMAD5 levels. Densities of P-SMAD1/5 and SMAD5 bands of mouse samples were quantified using digital image analysis densitometry software. (ImageJ, National Institutes of Health, Bethesda, MD, USA) Normalized density of P-SMAD1/5 bands per time point was consistently upregulated (three samples versus three samples) from day 33 onwards as compared with baseline (Figure 2c). These semi-quantitative Western blot data thus corroborated our immunohistochemical results, supporting more active BMP signaling in the joint as CIA progresses.

Quantitative PCR for different BMP ligands and BMP antagonist NOG in both synovium and cartilage extracts (three sets per time point, extracts from two mice pooled per set) showed a dynamic expression pattern with upregulation or downregulation of these molecules during the course of CIA (day 0, 20, 27, 33, and 40) as compared with healthy controls (Figure 3). As expected, expression of TNFα in the synovium gradually increased during the course of the disease process. BMP7 was upregulated with peak expression around day 20. The expression levels of BMP2, BMP4, BMP6, and NOG were stable around day 20 and decreased gradually, in contrast to GDF5 expression, which was downregulated abruptly around day 20. As gene expression levels are normalized to a housekeeping gene, they are also likely to be influenced by the increased number of cells in the synovium. In the articular cartilage, an increase in TNFα was visualized, though to a milder extent, but we found a striking increase in GDF5 expression, whereas the other molecules showed a more or less stable expression pattern. At day 40, BMP2 was very mildly upregulated in the articular cartilage. Again, due to normalization, expression levels are likely to be influenced by the decreased number of cells in the cartilage, due to chondrocyte cell death.

The gene expression data prompted us to further study BMP2 and BMP7 protein levels in the arthritic joint. These prototype BMPs belong to different subfamilies preferentially binding to distinct type I and type II receptor complexes. Immunohistochemical analysis of affected mouse knees revealed BMP2 expression predominantly in the hyperplastic lining layer of synovium and in the articular cartilage (Figure 4b,h). BMP2 was less expressed but still present in some cells in the subintimal zone (Figure 4d,f). The BMP7 staining pattern in mouse knee joint was influenced by disease severity. Mild inflammation (Figure 5a) was associated with discrete BMP7 staining in the lining layer of synovium (Figure 5b) and in superficial chondrocytes in the articular cartilage (Figure 5b). With increasing severity (Figure 5c,e), BMP7 positivity became more pronounced in the lining layer and subintima (Figure 5d,f). Eventually, in the destructive phase (Figure 5g), BMP7 positivity appeared more and more in the subintima, away from the lining layer (Figure 5h).

As BMP expression was recently reported in the human TNF transgenic mouse model, we further investigated the effect of blocking TNFα on expression of BMPs in synovium and cartilage, using daily intraperitoneal injections of etanercept or PBS (as control) during clinically apparent CIA (n = 9 mice per group). Six-day treatment reduced severity of arthritis, as shown by the cumulative severity scores (Figure 6a). At endpoint, synovium and cartilage were collected and mRNA expression levels of BMP2, BMP7, and GDF5 were compared between the two groups. In the synovium, expression levels of BMP2, BMP7, and GDF5 were not significantly changed. In cartilage, however, etanercept treatment raised the expression levels of BMP7 significantly (P = 0.003). BMP2 and GDF5 expression was not affected (Figure 6b). Immunohistochemistry confirmed the increased levels of BMP7 in the cartilage (Figure 6c).