Inflammation is preceded by tumor necrosis factor-dependent infiltration of mesenchymal cells in experimental arthritis




To determine the involvement of mesenchymal progenitor cells in the induction of collagen-induced arthritis (CIA).


DBA/1 mice were immunized with type II collagen in adjuvant or adjuvant alone, and the presence of mesenchymal cells in the joints of prearthritic mice was studied by immunohistochemistry.


An analysis of the joints on day 10 postimmunization (at least 10 days before the onset of arthritis) revealed synovial hyperplasia without leukocytic infiltration. Large, round cells expressing bone morphogenetic protein receptors (BMPRs), which serve as markers for primitive mesenchymal cells, were present in increased numbers in the bone marrow adjacent to the joint, in the synovium itself, and within enlarged bone canals that connect the bone marrow to the synovium. Similar changes were observed in mice given adjuvant without collagen. Adjuvant-induced infiltration of BMPR+ cells and enlargement of bone canals were abrogated by anti-tumor necrosis factor (anti-TNF) treatment and were absent in TNFR p55/p75−/− mice. Increased numbers of bone marrow cells and enlarged bone canals were observed in nonimmunized TNF transgenic mice (which spontaneously develop arthritis).


These findings suggest that in CIA, there is an antigen-independent (innate) prearthritic phase that prepares the joint for the subsequent immune-mediated arthritis. The induction phase involves marrow-derived mesenchymal cells and requires the presence of TNF.

Rheumatoid arthritis (RA) is characterized by chronic inflammation of the synovial tissues and destruction of bone and cartilage, resulting in loss of joint function. Although much of the pathology of RA can be attributed to cells of the immune system, it is also possible that nonhematopoietic cells play a role in the initiation and/or perpetuation of the disease. For example, we have previously documented the presence of mesenchymal progenitor cells (MPCs) bearing receptors for bone morphogenetic proteins (BMPs) in the synovial membrane and synovial fluid of RA patients (1), although the pathophysiologic relevance of these cells is unclear. MPCs operate under the influence of BMPs, a group of 20 dimeric proteins belonging to the transforming growth factor β superfamily (2, 3). BMPs direct the differentiation of MPCs into osteoblasts, chondrocytes, and adipocytes and are endogenous regulators of skeletal development (3–5).

BMPs exert their diverse biologic effects through membrane receptors of 2 types, BMP receptor I (BMPRI) and BMPRII, on MPCs. BMPRI is further subclassified into BMPRIA and BMPRIB. The formation of heterodimeric complexes of BMPRI and BMPRII induces signal transduction following ligand binding (6, 7). Antibodies to BMPRs allow the identification of MPCs in situ. For example, during organogenesis of mouse embryos, cells expressing BMPRIA and BMPRIB can be demonstrated in areas of mesenchymal precartilage condensation and in premuscle or on cells in areas of developing cartilage and bone. In previous studies we have used anti-BMPR antibodies to characterize a population of MPCs in the blood of healthy individuals (8) and in the synovial tissue of patients with RA (1).

The presence of MPCs in the joints of patients with RA raises an important question. Do MPCs appear in the joint as a result of the inflammatory process or do they arrive before the onset of inflammatory changes? In this study, we used the murine collagen-induced arthritis (CIA) model to address this question and, in addition, to identify which factor(s) may be involved in recruiting MPCs to the joint. Our findings reveal that BMPR+ MPCs accumulate in the bone marrow and synovial tissue of mice prior to the onset of CIA and that their presence in prearthritic joints is regulated by tumor necrosis factor (TNF) and is associated with the opening of vascularized bone canals that link the bone marrow to the synovium.


Mice. Male DBA/1 mice (8–12 weeks of age) were used for induction of CIA. Female TNF transgenic mice (6–8 weeks of age) with an original background of C57BL6 × CBA (obtained from Dr. George Kollias, Hellenic Pasteur Institute, Athens, Greece) were backcrossed with DBA/1 mice (19 generations). Male TNFR p55/p75−/− (TNFR−/−) mice (6–8 weeks of age) were obtained from Dr. Jacques Peschon (Immunex, Seattle, WA) and were from a C57BL6 × 129 background.

Immunization. DBA/1 mice were immunized intradermally at the base of the tail with 200 μg of bovine type II collagen (CII) emulsified in Freund's complete adjuvant (CFA; Difco, West Molesley, UK). Alternatively, mice were immunized with CFA emulsified with an equal volume of phosphate buffered saline (PBS). Mice were inspected daily for signs of arthritis (erythema and/or joint swelling). They were killed 10–13 weeks postimmunization, and hind paws were removed above the ankle joint and processed for routine histologic and immunohistochemistry study.

Antibodies. Anti-TNF monoclonal antibody (mAb) TN3-19.12 (9) was a gift from Robert Schreiber (Washington University Medical School, St. Louis, MO), in conjunction with Celltech (Slough, UK). Anti-BMPR antibodies consisted of polyclonal rabbit antisera (obtained from Dr. C.-H. Heldin, Ludwig Institute for Cancer Research, Uppsala, Sweden) and were prepared against synthetic peptides corresponding to the intracellular juxtamembrane parts of the type IA, type IB, and type II receptors of BMP (10, 11). The antisera were affinity purified and tested for specificity by immunoprecipitation of crosslinked complexes of cultured cells transfected with the appropriate BMPR complementary DNA.

Immunohistochemistry studies. Immunolocalization of BMPR was carried out as previously described (1). Briefly, hind paws were embedded in OCT embedding matrix (Cell Path, Hemel Hemstead, UK) and then snap frozen. Sagittal sections (8 μm) were cut using a cryostat and a tungsten carbide knife, and transparent adhesive tape was used to attach the sections to the slide. Sections were fixed in ice-cold 4% phosphate buffered paraformaldehyde (pH 7.4) for 20 minutes and then washed in PBS. All further incubations and washes were carried out using PBS. After endogenous peroxidase activity was blocked with 0.1M sodium azide containing 1% H2O2, the specimens were incubated with 10% normal goat serum, 2% normal rat serum, and 1% bovine serum albumin (BSA) for 30 minutes at room temperature to eliminate nonspecific binding.

Specimens were then incubated with primary antibodies (1.2 μg/ml of anti-BMPRIA, 1.0 μg/ml of anti-BMPRIB, or 1.5 μg/ml of anti-BMPRII), followed by biotinylated secondary antibody (Vector, Burlingame, CA). The antibody-biotin conjugates were detected with an avidin-biotin-peroxidase complex (Vectastain Elite ABC; Vector). Color reaction was developed using 3-amino-9-ethyl-carbazole (Merck, Luttenworth, UK), and finally the specimens were lightly counterstained with Mayer's hematoxylin. Controls consisted of normal rabbit IgG (Vector), 1% BSA in PBS, or antibodies preabsorbed with the respective peptide used for immunization.

A staining procedure similar to that described above was used to identify cell surface molecules. To detect CD34+ cells, sections were stained with rat anti-mouse CD34 (PharMingen, San Diego, CA). Macrophages were detected with rat anti-mouse F4/80 (Serotec, Oxford, UK). CD4+ or CD8+ cells were detected with the appropriate rat anti-mouse antibody (PharMingen).

Microscopic analysis. Sections were examined using an Olympus BH2 microscope and characterized with a computer image analyzer (AnalySIS; Soft Imaging System GmbH, Munster, Germany), under blinded conditions. The entire tissue section of a whole paw was examined to localize and identify areas with positively stained cells. Synovial tissue was obtained from the metatarsophalangeal (MTP) joint and epiphyseal bone marrow regions of the same joint, and quantitative analysis was performed by color cell separation. For each high-power field, data were expressed as region of interest, i.e., the percent of the total area that exhibited positive staining. To ensure that approximately the same areas of tissue were studied for all of the BMP receptors, the staining was carried out in adjacent serial sections. Adjacent sections were also examined for the expression of CD4, CD8, F4/80, or CD34 antigens. In addition to immunohistochemical staining, sections from all hind limbs were stained with hematoxylin and eosin or Safranin O in order to assess histopathologic changes associated with the development of arthritis. The diameter of bone canals in the “bare zone” (an area between the articular margin and the junction of the capsule and periosteum) was measured using the image analyzer.

Statistical analysis. Histologic and immunohistologic data obtained by image analysis were subjected to analysis of variance followed by Tukey's test, to determine the significance of differences between groups.


Histopathologic changes prior to onset of CIA. Histologic analysis of joints removed 10 days after the mice were immunized with CII in CFA (n = 15 mice) revealed that the synovial tissues in 67% of proximal interphalangeal (PIP) and 61% of MTP joints had evidence of fibrin deposition and synovial hyperplasia. The changes were localized to the bare zone (Figure 1). No abnormalities were observed in distal interphalangeal (DIP) joints at this time. A 212% increase in the number of bone marrow cells was seen in the epiphyseal region of joints with synovial hyperplasia, compared with the same area in normal joints without synovial hyperplasia.

Figure 1.

Presence of bone canals prior to onset of collagen-induced arthritis. Shown is a sagittal section of proximal interphalangeal joint from a mouse with synovial hyperplasia and an enlarged bone canal, connecting the synovium (S) to the bone marrow (BM) at the bare zone. The specimen was obtained 10 days after immunization with collagen in adjuvant, before there was any clinical evidence of arthritis. B = bone; C = cartilage (hematoxylin and eosin stained; original magnification × 40).

Normal mouse joints have minute (∼25 μm in diameter) canals in the bare zone connecting the bone marrow to the synovium. In joints with synovial hyperplasia, there was a significant increase in the size of these canals compared with that in joints without hyperplasia (P < 0.01). In addition, the bone canals in joints with synovial hyperplasia appeared more vascularized than those in nonimmunized joints and were found to contain large, round cells with a morphology similar to that of cells in the bone marrow (Figure 2). No leukocytic infiltrates and no CD4+ or CD8+ lymphocytes were observed in either normal or hyperplastic synovium 10 days after immunization.

Figure 2.

Presence of bone marrow cells in bone canals. A, An enlarged bone canal in the bare zone of a proximal interphalangeal (PIP) joint prior to the onset of collagen-induced arthritis. In the canal are blood vessels (BV) containing large, round cells. B, A normal bone canal in an unaffected PIP joint. Arrows indicate the margins of the canals at the junction with the synovium (S). B = bone; BM = bone marrow. (Hematoxylin and eosin stained; original magnification × 200.)

Histopathologic changes in the joints of mice immunized with CFA alone. To distinguish the separate contributions of antigen (CII) and adjuvant (CFA) in the induction of synovial hyperplasia, mouse joints were analyzed histologically 12 days after the administration of CFA alone. Approximately 38% of PIP and 52% of MTP joints exhibited synovial hyperplasia and fibrin deposition. Hyperplastic changes were not present in the synovium of DIP joints. The epiphyseal areas of hyperplastic PIP and MTP joints showed an accumulation of bone marrow cells. Moreover, joints that showed synovial hyperplasia also had larger canals in the bare zone, similar to those seen in the joints of mice immunized with CFA plus collagen. In joints with synovial hyperplasia, the canals were 2.5 times larger than in normal joints (P < 0.01). Thus, canal enlargement (>35–40 μm) was equivalent in mice that had been immunized with collagen in adjuvant and those immunized with adjuvant alone, and in both cases the canals were seen to contain marrow elements.

Effect of immunization on accumulation of BMPR+ cells. In sections from nonimmunized mice, weak expression of BMPRIA and BMPRII on periosteal cells and osteoblasts of cortical bone was occasionally detected by immunohistochemistry, but no BMPR+ cells were observed in the synovial membrane. Some chondrocytes in the articular cartilage situated close to the epiphyses showed moderate staining for BMPRIA, BMPRIB, and BMPRII. However, strong staining for BMPRIA and BMPRII was observed exclusively on large, round cells in the bone marrow. Ten days after immunization with either CFA alone or CFA plus collagen, the mice that developed synovial hyperplasia had increased BMPRIA and BMPRII expression on cells of the periosteum and on osteoblasts of cortical bone. At the same time, large, round bone marrow cells staining strongly for BMPRIA and BMPRII were found to be accumulated in the epiphyseal region. Cells with the same morphology were also seen in the enlarged canals in the bare zone and in hyperplastic synovium (Figure 3). BMPRIA+ and BMPRII+ cells in the hyperplastic synovium were localized primarily in the subintima, although a proportion were found in the synovial lining.

Figure 3.

Bone morphogenetic protein receptor-positive (BMPR+) cells in collagen-induced arthritis. BMPRIA+ cells are evident in the synovial tissue (S), bone marrow (BM), and the enlarged canals in the bare zone and periosteal area 10 days after immunization with collagen in adjuvant, prior to the onset of arthritis. Arrows indicate the location of bone canals. B = bone; C = cartilage; JS = joint space (hematoxylin and eosin stained; original magnification × 100).

Role of TNF in accumulation of BMPR+ cells, enlargement of bone canals, and synovial hyperplasia. Previously, we and others have shown that TNF blockade during the induction phase of CIA delays the onset of arthritis and reduces its subsequent clinical severity (12–14). To evaluate the possible role played by TNF in the subclinical changes observed in this study, mice were immunized on day 0 with CFA alone, then injected intraperitoneally with 300 μg of anti-TNF mAb (TN3-19.12) or PBS (6 animals in each group) on days 7, 9, and 11. On day 13 the animals were killed and their hind paws analyzed. Anti-TNF treatment caused reductions in the proportion of joints with hyperplasia and in the size of bone canals (Figure 4). This treatment also led to reductions in the numbers of BMPRIA+ and BMPRII+ cells in the bone marrow and synovium (Figure 5).

Figure 4.

Inhibition of bone canal enlargement by tumor necrosis factor (TNF) blockade. Freund's complete adjuvant (CFA)-immunized mice were treated on days 7, 9, and 11 with anti-TNF monoclonal antibody or phosphate buffered saline (n = 6 per group). On day 13, the mice were killed and the diameter of the canals was determined by image analysis. Open bars = bone canals in proximal interphalangeal joints; solid bars = bone canals in metatarsophalangeal joints. Values are the mean and SEM.

Figure 5.

Inhibition of accumulation of bone morphogenetic protein receptor-positive (BMPR+) cells by tumor necrosis factor (TNF) blockade. Freund's complete adjuvant (CFA)-immunized mice were treated on days 7, 9, and 11 with anti-TNF monoclonal antibody or phosphate buffered saline (n = 6 per group). On day 13, the mice were killed and BMPR expression in the bone marrow (A) and in the synovium (B) was analyzed by immunochemistry, using antibodies specific for BMPRIA (open bars), BMPRIB (not shown), and BMPRII (solid bars). Sections were studied by image analysis and the data are expressed as the percent of the total area showing positive staining (mean and SEM).

To further establish the importance of TNF in the subclinical changes, TNFR−/− mice (n = 12) were immunized with CFA alone. On day 12, neither the degree of synovitis nor the size of the canals differed in CFA-immunized TNFR−/− mice compared with nonimmunized animals, and there was no evidence of bone marrow cell accumulation in the epiphyseal areas. Furthermore, the expression of BMPRIA and BMPRII on bone marrow cells at the epiphysis in TNFR−/− mice immunized with CFA was similar to that in control, nonimmunized mice. Strikingly, no BMPRIA+ or BMPRII cells were identified in the synovium of CFA-immunized TNFR−/− mice.

Finally, to confirm the role of TNF in promoting marrow cellularity and enlargement of canals, joints from nonimmunized human TNF (HuTNF) transgenic mice, which spontaneously develop arthritis at ∼3 weeks of age, were examined. Interphalangeal joints from 6–8-week-old HuTNF transgenic mice exhibited a dramatic increase in the numbers of bone marrow cells at the epiphysis compared with nontransgenic littermates and even compared with DBA/1 mice immunized with collagen in CFA or with CFA alone. For example, there was a 5–8-fold increase in the numbers of bone marrow cells at the epiphysis in HuTHF transgenic mice compared with mice given CFA alone or CFA plus collagen (data not shown). Similarly, the size of the bone canals was almost double the size of the canals in the hyperplastic joints of mice given collagen in CFA or CFA alone (data not shown).

Origin of BMPR+ cells in the bone marrow and synovium. Sequential sections of synovium and bone marrow from nonimmunized mice or from mice immunized with collagen in CFA were examined immunohistologically (on day 10 after immunization) for coexpression of BMPRIA/BMPRII and CD34, a marker for hematopoietic progenitor cells. In the bone marrow, CD34+ cells and BMPRIA+ cells or BMPRII+ cells were in proximity to each other, but no overlapping staining was observed. In the synovium, CD34 expression was seen only on vascular endothelial cells in the marginal bare zone, while BMPRIA and BMPRII expression was limited to cells with a large, round or fibroblast-like morphology. Therefore, it appears that BMPR+ cells arise from mesenchymal, rather than hematopoietic, progenitor cells.


The results of this study show that the initiation of synovial hyperplasia in CIA is accompanied by 2 early events: 1) the enlargement of vascularized bone canals that link the bone marrow and the synovium and 2) the accumulation of BMPR+ cells in the synovial membrane and bone marrow, in advance of inflammatory cell accumulation and clinical onset of arthritis. These same changes were observed in mice given CFA alone; therefore, we may assume that bone canal enlargement and accumulation of BMPR+ cells are not dependent on CII-specific immune responses. Moreover, the findings that these changes are inhibited by anti-TNF treatment, are absent in TNFR−/− mice, and occur spontaneously in TNF transgenic mice confirm the important role played by TNF in these preinflammatory events. This may explain the previous observation, by 3 independent groups (12, 13, 15), that TNF blockade during the induction phase of CIA (prior to disease onset) reduces the subsequent severity of disease.

The simultaneous presence of BMPR+ mesenchymal cells in the bone marrow and in the hyperplastic synovium suggests a connection between these 2 juxtaposed tissues. Conventionally, the cellular elements that accumulate in inflamed joint tissues are thought to come from the circulation. Alternative routes are seldom considered, although blood vessels and mesenchymal elements that arise in the perichondrium and extend into the epiphysis during the late stages of embryogenesis of the appendicular skeleton in the mouse have been well described (16, 17).

The notion of a bone marrow-synovial connection is not new. Nakagawa et al proposed a role for bone marrow cells in the pathogenesis of RA and tested the idea in a CIA model (18). Stromal cells from the bone marrow of rats were isolated by plastic adherence in tissue culture, labeled, and then injected intraperitoneally into syngeneic rats 2 days before immunization with collagen. At week 1, labeled cells were observed only in the marrow, but at 2 weeks, labeled cells were found “proliferating in the bone marrow and migrating into the joint cavity through the canals at the bare zone.” In this study, we have similarly demonstrated the presence of BMPR+ mesenchymal cells in enlarged canals in the bare zone and in the hyperplastic synovium.

Bone canal enlargement is not simply the result of vasodilatation, as can be seen from our measurements (Figure 4), but how can enlargement develop so quickly? Using dual x-ray absorptiometry, Takagi et al demonstrated a significant local decrease in bone mineral density in the knee joints of rats 7 days after CFA injection, but 1 week before arthritis was clinically apparent (19). More important, this was not a generalized response to adjuvant, because only the distal femur and proximal tibia showed changes in bone mineral density at this early stage. Those authors hypothesized that the differences could be explained by the fact that periarticular bone is primarily trabecular bone, and the magnitude and rate of trabecular bone loss are greater and faster than those of cortical bone loss (19).

The actual mechanism remains unexplained, but the well-documented role of TNF in osteoporosis (20) and in bone resorption in periodontal disease (21) indicates that TNF has an important function in bone remodeling. Furthermore, the observed effects of TNF blockade and the changes in TNF transgenic mice constitute strong evidence that this cytokine is a major participant in prearthritic epiphyseal bone remodeling and expansion of bone canals. One further observation that may explain the mechanism of opening up of bone canals was the presence of activated osteoclasts in close proximity to the enlarged canals (results not shown) suggesting that it is an osteoclast-mediated process.

Histologic evidence of synovial hyperplasia antedating clinical arthritis was described in an early report on light and electron microscopy findings in the knee joints of Wistar and Sprague-Dawley rats immunized with collagen (22). The investigators proposed 2 stages in the development of CIA: an initial stage starting around day 12 consists of fibrin deposition and synovial hyperplasia in the joints, whether the animals develop arthritis or not; later, a second stage occurs in a minority of animals, in which there is a severe inflammatory synovial reaction with bone and cartilage destruction. Mesenchymal cells bearing BMPRs are the only new cells in the hyperplastic prearthritic synovium. Their significance remains to be defined, but since similar cells in bone marrow and in rheumatoid synovial tissue constitutively produce stromal cell-derived factor (1), a potent chemokine for B cells, T lymphocytes, and monocytes (23), the role of primitive mesenchymal cells in the induction phase of CIA may be to promote the accumulation of immunocompetent cells into the joint and hence initiate the inflammatory phase.

In summary, we have shown that BMPR+ mesenchymal cells are among the earliest cells to arrive at the prearthritic joint in CIA and are strongly associated with early hyperplastic changes in the synovium. Although this study focused on early events in CIA, we have previously shown that BMPR+ mesenchymal cells are also present in patients with established RA (1), suggesting that they are involved not only in early synovial hyperplasia, but also in the chronic phase of arthritis. Finally, these findings further elaborate the complex role played by TNF in inflammatory arthritis.


The authors are indebted to Robert Shreiber and Celltech for providing TN3-19.12, to George Kollias for providing HuTNF transgenic mice, and to Jacques Peschon for providing TNFR−/− mice.