Arthritis Induces Lymphocytic Bone Marrow Inflammation and Endosteal Bone Formation


  • The authors have no conflict of interest


Arthritis can destroy the cortical bone barrier and expose bone marrow to synovial tissue. This study examines bone marrow changes in arthritis and its effects on cortical bone remodeling. Bone marrow next to arthritic lesions exhibits B-lymphocyte-rich infiltrates, which express BMPs and stimulate endosteal bone formation. Thus, bone marrow actively participates in the arthritic process.

Introduction: Imaging studies have shown that bone marrow changes occur in patients with rheumatoid arthritis (RA). To examine whether bone marrow is affected during arthritis, human TNF transgenic (hTNFtg) mice, which constitute an established animal model of human RA, were examined for bone marrow changes.

Materials and Methods: The hind paws (tarsal area) of 22 untreated hTNFtg mice, 5 hTNFtg mice treated with anti-TNF (infliximab), and 5 wildtype (WT) mice were examined histologically, immunohistochemically, and by means of mRNA in situ hybridization.

Results and Conclusions: All untreated hTNFtg mice with moderate (n = 10) and severe (n = 7) disease developed inflammatory bone marrow lesions during the course of disease, whereas no such lesions appeared in hTNFtg mice with mild disease (n = 5) and WT mice. Bone marrow infiltrates were almost exclusively composed of lymphocytes, and the overwhelming proportion (>80%) was B-cells. Presence and extent of bone marrow infiltrates were closely linked to severity of arthritis. In addition, blockade of TNF effectively reduced bone marrow inflammation. Interestingly, osteoblast numbers were increased at the endosteal surface in the vicinity of these lesions. Moreover, osteoid deposition; expression of bone matrix proteins, such as osteocalcin and osteopontin; and mineralization were enhanced, suggesting that inflammatory bone marrow infiltrates induce bone formation. Indeed, B-lymphocytes of these lesions expressed bone morphogenetic protein (BMP)-6 and −7, which are important stimulators of new bone formation. Thus, we conclude that bone marrow actively participates in destructive arthritis by generating B-lymphocyte-rich bone marrow lesions and inducing endosteal bone formation.


RHEUMATOID ARTHRITIS (RA) has unique properties for resorbing bone. Juxta-articular bone erosion is a hallmark of the disease and has become one of the diagnostic criteria of RA. It is also a surrogate marker for high disease activity and bad functional outcome.(1) The typical location of local bone erosion is the junction zone, which is the insertion of the synovial membrane at the interface between articular cartilage, bone, and the periosteum. This region is only separated by a thin layer of cortical bone from its adjacent bone marrow.

Local bone erosion is driven by inflammatory synovial tissue, which is hyperplastic and invades the joint space and juxtaarticular bone underlying the articular cartilage.(2,3) Generation of osteoclasts from mononuclear precursor cells located within synovial inflammatory tissue is a prerequisite for invasion of bone. Thus, local bone erosions in RA and animal models of RA harbor abundant amounts of mature osteoclasts. Furthermore, their essential role to induce such lesions was shown in osteoclast-free models of arthritis.(4,5) Longitudinal observations of radiological data of human RA and histopathological data of animal models of RA suggest that resorption of juxta-articular bone starts from surrounding periosteum and/or synovial membrane and proceeds to destroy cortical plates, which underlie articular cartilage and separate the synovial membrane from bone marrow.

Because bone erosions grow in size if disease control is inadequate and because cortical plates represent only relatively thin barriers, it can be assumed that it is just a matter of time and/or resorptive capacity until synovial inflammatory tissue reaches the bone marrow cavity. Data from MRI studies of RA joints suggest that bone marrow is indeed affected by adjacent synovial pathology, and this phenomenon is described as “bone marrow edema.”(6) However, the exact nature and the mechanism of formation of such bone marrow changes remain unclear, and no histological study has yet addressed this feature in detail.

In this study, we performed a systematic characterization of bone marrow changes associated with inflammatory arthritis. We used the human TNF transgenic (hTNFtg) mouse model of arthritis, which represents an established animal model of human RA. hTNFtg mice spontaneously develop inflammatory polyarthritis, which shares many features of human RA. Thus, arthritis is symmetrical, involves smaller joints, and is of a chronic progressive nature, which ultimately leads to synovial hyperplasia, cartilage damage, and most important, local bone erosion. We show the occurrence of inflammatory bone marrow infiltrates located adjacent to inflamed joints. These lesions consist predominantly of B-cells and only occur when the cortical bone barrier is disrupted by local bone erosion. Their major effect is a potent stimulation of endosteal bone formation.


Animals and treatments

Heterozygous Tg197 hTNFtg mice (strain C57/Bl6), which develop a chronic inflammatory and destructive polyarthritis within 4-6 weeks after birth, have been described previously.(7) We investigated a total of 22 untreated hTNFtg mice: 10 mice that were 10 weeks of age, 5 mice that were 5 weeks of age, and 7 mice that were 15 weeks of age. In addition, five wildtype (WT) mice served as controls. Furthermore, five hTNFtg mice were treated with anti-TNF (infliximab), a chimeric monoclonal antibody (Centocor, Leiden, The Netherlands), at a dose of 10 mg/kg that was administered three times weekly through intraperitoneal injection. Therapy was initiated at the onset of symptoms (week 6) and lasted for 4 weeks. For assessment of bone matrix mineralization, the animals received calcein at a dose of 15 mg/kg, which was dissolved in 2% sodium bicarbonate and applied subcutaneously 3 and 1 days before death. Mice were killed by cervical dislocation, the blood was withdrawn by heart puncture, and the hind paws were dislocated for histological and immunohistochemical evaluation. All animal procedures were approved by the local ethical committee.

Preparation and histological evaluation of decalcified specimen

Right hind paws were fixed in 4.0% formalin overnight and decalcified in 14% EDTA (Sigma, St Louis, MO, USA) at 4°C (pH adjusted to 7.2 by addition of ammonium hydroxide; Sigma) until the bones were pliable. Serial paraffin sections (2 μm) of the right hind paw were used for the histochemical and immunohistochemical analyses. Beside hematoxylin-eosin (H&E) staining, TRACP staining (leukocyte acid phosphatase kit; Sigma)(8,9) for identification of osteoclasts and naphtol AS-D chloroacetate esterase (CAE) staining (Naphtol AS-D Chloroacetate Esterase and a-Naphtyl Acetate Esterase Kit; Sigma) for quantification of cells of the granulocytic lineage were performed.

Quantification of synovial inflammation and bone erosions in each tarsus was performed using H&E sections and a microscope (Zeiss Axioskop 2; Zeiss, Marburg, Germany) equipped with a digital camera and image analysis system (KS300; Zeiss). Calculations were performed as described previously.(4,8)

Preparation and histomorphometry of undecalcified specimen

Left hind paws were fixed in 70% ethanol, dehydrated in 100% methanol, and embedded undecalcified in methylmethacrylate (K-Plast; MDS GmbH, Buseck, Germany). Sections of 3 μm thickness were made on a Jung microtome (Jung, Heidelberg, Germany) and stained by Goldner's trichrome and von Kossa staining. One section of each paw was left unstained for fluorescence microscopy.

Histomorphometric measurement of the number of osteoblasts per bone perimeter (N.Ob/B.Pm), the fraction of bone surface covered by osteoblasts (ObS/BS), and the size of bone marrow infiltrates was done using the OsteoMeasure system (OsteoMetrics, Atlanta, GA, USA). Nomenclature of the histomorphometric parameters is according to international standards.(10) Measurement was performed in the vicinity of the infiltrates and on bone surfaces adjacent to infiltrate-free bone marrow in one metatarsal bone per paw as well as in the calcaneus.


For immunohistochemical detection of B-cells, deparaffined, ethanol-dehydrated tissue sections were blocked with PBS buffer containing 10% goat serum followed by incubation with a rat anti-mouse monoclonal CD45R/B220 antibody (diluted 1:300; BD Biosciences Pharmingen, San Jose, CA, USA) for 1 h at room temperature. For immunohistochemical staining of macrophages, T-lymphocytes, bone morphogenetic protein (BMP)-6, BMP-7, growth differentiation factor (GDF)-5, and GDF-6 deparaffined, ethanol-dehydrated tissue sections were boiled for 2 minutes in 10 mM sodium citrate buffer (pH 6.0) using a 700-W microwave oven, cooled to room temperature, and rinsed in detergent solution (0.5% Tween in PBS) for 10 minutes. Tissue sections were blocked for 20 minutes in PBS containing 20% rabbit serum followed by incubation for 1 h at room temperature with the following antibodies: rat monoclonal anti-macrophage (F4/80) Ab (diluted 1:80; Serotec, Raleigh, NC, USA), rat monoclonal anti-CD3 Ab (diluted 1:100; Novocastra, Newcastle, UK), rat monoclonal anti-fibroblast (diluted 1:40; Biogenesis, Dorset, UK), goat polyclonal anti-BMP-6 Ab (diluted 1:20; Santa Cruz Biotechnology, Santa Cruz, CA, USA), goat polyclonal anti-BMP-7 Ab (diluted 1:40; Santa Cruz Biotechnology), and goat polyclonal anti-GDF-5 Ab and anti-GDF-6 Ab (diluted 1:10; Santa Cruz Biotechnology). Immunohistochemical staining of the parathyroid hormone (PTH) receptor was done using deparaffined, ethanol-dehydrated tissue sections pretreated with trypsin (12 mg trypsin in 10 ml PBS) for 5 minutes at room temperature, followed by blocking for 20 minutes with PBS containing 20% rabbit serum and incubation for 1 h with goat polyclonal anti-PTH receptor Ab (diluted 1:20; Santa Cruz Biotechnology).

All abovementioned procedures were followed by rinsing and blocking of endogenous peroxidase with 0.3-3% hydrogen peroxide in Tris-buffered saline (10 mM Tris-HCl, 140 mM NaCl, pH 7.4) for 10 minutes and subsequent 30-minute incubation with a biotinylated species-specific anti-IgG secondary antibody (for detection of CD3, F4/80, and fibroblasts; Vector, Burlingame, CA, USA; for detection of CD45R; BD Biosciences Pharmingen). Sections were incubated with the appropriate ABC-complex (VECTASTAIN@ABC reagent, vector for detection of CD3, F4/80, and fibroblasts; StreptABComplex/HRP; Dako, Glostrup, Denmark for detection of CD45R) for another 30 minutes using 3,3-diaminobenzidine (Sigma) as chromogen, resulting in brown staining of antigen expressing cells.

The composition of bone marrow infiltrates was quantitatively assessed by counting total cell numbers and the numbers of positively stained cells in each immunohistochemical staining. Cell counting was done at a magnification of 200× in the bone marrow infiltrates.

In situ hybridization

cDNA probes for the osteocalcin and osteopontin in situ hybridization were kindly provided by C Hartmann (Research Institute for Molecular Pathology, Vienna, Austria). Osteocalcin cDNA probes consisted of a 700- (osteocalcin) and 950-bp (osteopontin) insert cleaved with NotI (osteocalcin) and EcoRI (osteopontin) and transcribed with T3 (osteocalcin) and SP6 (osteopontin) RNA polymerase to generate antisense probes using digoxygenin labeling (DIG RNA Labeling Kit; Roche, Mannheim, Germany).

Slides were pretreated as follows. Sections were baked on a hot plate at 60°C for 1 h, deparaffinized and dehydrated, and post-fixed in 4% paraformaldehyde for 10 minutes at room temperature followed by rinsing in PBS buffer, digestion with Proteinase K (1 μg/ml) for 10 minutes at room temperature, another post-fixation step in 4% paraformaldehyde for 5 minutes at room temperature, acetylation in acetic anhydride (2.5 μl/ml) for 15 minutes at room temperature, and air drying at 37°C for 30 minutes. The digoxygenin labeled probe was diluted 1:100 in hybridization solution (10 mM Tris pH 7.5, 600 mM NaCl, 1 mM EDTA, 0.25% SDS, 10% dextrane sulfate, 1× Denhardt's, 200 μg/ml yeast tRNA, 50% formamide), heated to 85°C for 3 minutes, and incubated with the section overnight in an humidified chamber at 65°C, covered by a coverslip. After removal of the coverslips, the slides were rinsed with 1× SSC/50% formamide for 30 minutes at 65°C followed by digestion of single-stranded RNA by RNase A (20 μg/ml in TNE buffer) for 10 minutes at 37°C and washing steps in TNE buffer (10 mM Tris pH 7.5, 500 mM NaCl, 1 mM EDTA) for 10 minutes at 37°C and in 2× SSC and 0.2× SSC for 20 minutes each at 65°C.

Detection of the probe was performed by incubating the slides with MABT (100 mM maleic acid, 150 mM NaCl, 0,1% Tween 20, pH 7.5) two times for 5 minutes at room temperature, blocking for 1 h with 20% heat-inactivated sheep serum/MABT, and incubation with alkaline phosphatase-labeled anti-digoxygenin-antibody (1:2000; Roche) in 2% heat-inactivated sheep serum/MABT at 4°C overnight in an humidified chamber. For the color reaction, slides were incubated in nitroblue tetrazolium (0.25 μg/ml) and 5-bromo-4-chloro-3-indolyl phosphate (0.125 μg/ml) dissolved in NTMT buffer (100 mM NaCl, 100 mM Tris pH 9.5, 50 mM MgCl2, 0.1% Tween-20) for 3 h at room temperature in the dark, followed by washing steps in nitro-blue tetrazolium buffer, PBS buffer, and Aqua distilled water and counterstaining for 5 minutes with nuclear fast red.

Statistical analysis

Data are shown as means ± SE. Comparison of bone marrow cellularity and composition between hTNFtg and WT mice was done by one-way ANOVA, including Bonferroni's multiple comparison test. For comparison of N.Ob/B.Pm and ObS/BS between bone surfaces near and distant from infiltrates, for comparison of total cell number and B-cell number between untreated and anti-TNF-treated mice, and for comparison of the amount of inflammation and erosions between mice with and without bone marrow infiltrates, Mann-Whitney's U-test was used. For comparison of bone marrow infiltrate size between untreated, anti-TNF-treated, and WT mice, the Kruskal-Wallis test, including Dunn's multiple comparison test, was used. Correlation between infiltrate size and size of adjacent erosions was tested by means of the Spearman rank correlation test.


Arthritis leads to localized bone marrow inflammation

To investigate whether bone marrow is influenced by arthritis, we examined paws of hTNFtg mice for histological evidence of bone marrow changes. hTNFtg mice (n = 10) that are 10 weeks of age with established clinical signs of arthritis, such as joint swelling and loss of grip strength, were subjected to this analysis. All of these mice showed histological evidence for synovial hyperplasia (Figs. 1A and 1B) and bone erosion (Figs. 1E and 1F), which are hallmarks of destructive arthritis. WT control mice did not show histopathological signs of arthritis (Figs. 1C, 1D, 1G, and 1H). Interestingly, all hTNFtg mice showed patchy mononuclear cell infiltrates localized in the bone marrow (black arrows in Figs. 1A, 1B, 1E, 1F, 1I, and 1J), whereas WT mice did not show such lesions (Figs. 1C, 1D, 1G, 1H, 1K, and 1L). These lesions were only found at sites where the destructive inflammatory process penetrated juxta-articular bone entirely and thus exposed bone marrow to the inflamed synovial tissue. The infiltrates were located exactly at the interface between synovial inflammation tissue and bone marrow. In contrast, hTNFtg bone marrow at sites distant from arthritis, such as the tibial diaphysis, were free of mononuclear cell infiltrates (data not shown), indicating that these changes are closely linked to arthritis. Surprisingly, among all cell-specific labeling procedures, labeling with anti-CD45R most clearly visualized these bone marrow infiltrates, suggesting that these lesions consist of a considerable proportion of B-lymphocytes (Figs. 1I and 1J). In WT mice, no such aggregation of B-lymphocytes was found, and only scattered B-lymphocytes were present in the bone marrow (Figs. 1K and 1L).

Figure FIG. 1..

Arthritis leads to localized bone marrow inflammation. Hind paws of (A, B, E, F, I, and J) 10-week-old hTNFtg and (C, D, G, H, K, and L) WT mice were analyzed for bone marrow changes. Histological sections show the tarsal regions stained by (A-D) H&E, (E-H) TRACP, and (I-L) anti-CD45R antibody. (A and B) hTNFtg mice show synovial inflammation (white arrow) and localized bone marrow infiltrates (black arrow). (C and D) WT mice show no signs of inflammation and normal bone marrow composed of adipocytes and scattered hematopoietic cells but no infiltrates. (E and F) TRACP staining reveals bone erosions filled with numerous osteoclasts (white arrow) next to bone marrow infiltrates (black arrow). (G and H) WT mice are free of osteoclast-mediated bone erosions. (I and J) Immunohistochemistry for CD45R intensively labels bone marrow infiltrates (brown staining, black arrow) in hTNFtg mice. (K and L) Clusters of CD45R+ cells are absent in WT mice. A, C, E, G, I, and K: original magnification, ×25; B, D, F, H, J, and L: original magnification, ×200.

B-lymphocytes predominate bone marrow infiltrates

We next addressed the cellular composition of the bone marrow infiltrates using cell-specific markers for B- and T-lymphocytes, macrophages, granulocytes, osteoclasts, and fibroblasts,. In fact, quantitative analysis of cellular composition showed that B-lymphocytes account for the overwhelming proportion of cells in bone marrow infiltrates (83.57 ± 3.06%; Figs. 2A and 2B). T-lymphocytes were far less common (5.44 ± 0.98%, Fig. 2C). Also, macrophages (2.86 ± 0.87%) and granulocytes (4.06 ± 0.79%) were only scarcely present in these lesions (Figs. 2D and 2E), and if seen, such cells were localized at its margins adjacent to synovial inflammatory tissue. Osteoclast precursors and mature osteoclasts were not found in bone marrow infiltrates and were confined to the surface of bone (Fig. 2F). Also, no fibroblast-like synoviocytes were found in these lesions (Fig. 2G). Thus, the almost exclusive lymphocytic nature of bone marrow infiltrates makes it easily distinguishable from synovial inflammatory tissue, which is predominantly composed of macrophages, fibroblast-like synoviocytes, granulocytes, and osteoclasts, but only a few lymphocytes (Figs. 2B–2G).

Figure FIG. 2..

B-lymphocytes predominate in bone marrow infiltrates. (A) Percentage (mean ± SE) of the individual cell types forming bone marrow infiltrates (MΦ, macrophages; PMN, polymorphonuclear neutrophils = granulocytes). Note that the great majority (83 ± 3%) are B-cells. (B-H) Histological images of a representative bone marrow infiltrate: (B) anti-CD45R labeling showing B-cells (brown), (C) anti-CD3 labeling showing T-cells (brown), (D) anti-F4/80 labeling showing macrophages (brown), (E) naphtol AS-D chloroacetate esterase staining of granulocytes (red), (F) TRACP staining of osteoclasts (purple), (G) anti-fibroblast labeling showing fibroblast-like synoviocytes (brown), and (H) H&E control stain. Original magnification, ×200.

Bone marrow infiltrates are associated with high inflammatory and erosive disease activity

Next, we were interested in the histopathological changes associated with formation of bone marrow infiltrates. As indicated before, normal synovial membrane as found in WT mice, was not associated with bone marrow infiltrates. Also, hTNFtg mice with mild disease and without bone erosion (5 weeks of age) did not show such lesions (Fig. 3A). However, when moderate (10 weeks of age) or severe (15 weeks of age) degrees of arthritis were present, all mice showed evidence for bone marrow infiltrates (Fig. 3A). A close relationship to disease severity was also evident from quantitative analysis of synovial inflammation and bone erosion in hTNFtg mice with and without bone marrow inflammation (Fig. 3B); the former showed significantly (p < 0.05) more synovial inflammation (1.52 ± 0.15 versus 0.67 ± 0.32 mm2). Furthermore, bone erosions were significantly (p < 0.001) more common in hTNFtg mice with bone marrow infiltrates than hTNFtg mice without such lesions (0.47 ± 0.06 versus 0.04 ± 0.02 mm2), suggesting that increased disease severity and formation of bone marrow infiltrates are closely linked.

Figure FIG. 3..

Presence and size of bone marrow infiltrates are associated with disease activity. (A) Frequency of bone marrow infiltrates in WT mice (10 weeks of age) and hTNFtg mice that were 5, 10, and 15 weeks of age. (B) Extent (mean ± SE) of inflammatory tissue and bone erosion in hTNFtg mice with bone marrow infiltrates (BMI) compared with mice without such lesions. Significant differences: *p < 0.05; **p < 0.001. (C) Correlation between the size of individual bone marrow infiltrates and the size of adjacent bone erosions (r = 0.496, p < 0.05).

Because infiltrates occurred immediately adjacent to penetrating erosions and were located at the base of the erosion where the pannus is in contact with the bone marrow, we also correlated the size of the individual bone marrow infiltrates with the size of the adjacent erosions. A significant positive correlation between infiltrate size and erosion size was evident (r = 0.496, p < 0.05, Fig. 3C). Taken together, these results show that the presence of bone marrow infiltrates is associated with high disease activity and that the extent of adjacent bone erosion is linked to bone marrow infiltrates.

Size of bone marrow infiltrates is influenced by therapy with anti-TNF

Because presence and extent of bone marrow infiltrates were associated with disease severity and adjacent bone erosion, we hypothesized that therapeutic inhibition of arthritis would reduce bone marrow infiltrates. To test this, we treated hTNFtg mice (10 weeks of age) with a blocking antibody against hTNF. Indeed, inhibition of synovial inflammation and bone erosion achieved by anti-TNF treatment was accompanied by a reduction in occurrence and size of bone marrow infiltrates. Whereas all untreated hTNFtg mice displayed bone marrow infiltrates, only 3/5 (60%) of hTNFtg mice treated with anti-TNF antibody showed such lesions. Comparing their size, bone marrow infiltrates were significantly (p < 0.05) larger in untreated hTNFtg mice (0.0084 ± 0.0012 mm2) compared with anti-TNF-treated hTNFtg mice (0.0037 ± 0.0013 mm2, Fig. 4A). This effect was based exclusively on a reduced amount of B-cells in such lesions, whereas the number of the other cell types scarcely present in such lesions, namely T-cells, macrophages, and granulocytes, did not significantly change on anti-TNF treatment (Fig. 4B). Thus, bone marrow infiltrates are linked to disease activity of inflammatory arthritis and respond to therapy that inhibits arthritis.

Figure FIG. 4..

TNF blockade inhibits formation of bone marrow infiltrates. (A) Size of bone marrow infiltrates in untreated hTNFtg mice, hTNFtg mice treated with anti-TNF antibody, and WT mice. Note a significant reduction (*p < 0.05) of size of bone marrow infiltrates on TNF blockade. (B) Numbers of various cell types within bone marrow infiltrates in untreated hTNFtg mice and hTNFtg mice treated with anti-TNF antibody. Note a significant reduction (*p < 0.05) of total cell numbers and B-cell numbers on TNF blockade. Data are given as mean ± SE.

Bone marrow infiltrates stimulate endosteal osteoblast recruitment

Searching for a potential function of bone marrow infiltrates and considering their close relation to inflammatory bone erosion, we were interested in whether such lesions potentially affect adjacent cortical bone. On close inspection of the endosteal bone surface, we noticed accumulation of cuboid-shaped cells with typical histomorphological criteria of osteoblasts on bone surfaces next to bone marrow infiltrates (Fig. 5A). In contrast, endosteal surface in areas distant from such lesions showed only scarce amounts of osteoblasts (Fig. 5B). Beside characteristic histomorphological features, specific molecular signs, such as PTH receptor and RANKL, also identified these cells as being of osteoblastic nature (Figs. 5C and 5E). Conversely, expression of these molecules was virtually absent at the endosteal surface distant from bone marrow infiltrates (Figs. 5D and 5F). We further quantified the fraction of bone surface covered by osteoblasts (ObS/BS) as well as the numbers of osteoblasts per bone perimeter (N.Ob/B.Pm) and compared results from the endosteal surface at sites adjacent to bone marrow infiltrates with those distant from these lesions (Figs. 5G and 5H). Both ObS/BS and N.Ob/B.Pm were significantly (p < 0.001 for ObS/BS and p < 0.0001 for N.Ob/B.Pm) higher in the vicinity of bone marrow infiltrates compared with infiltrate-free areas (33.3 ± 4.4% versus 2.7 ± 1.9% and 22.8 ± 3.5/mm versus 1.7 ± 0.8/mm, respectively). Thus, these data indicate that bone marrow infiltrates are associated with an accumulation of osteoblasts in the neighboring endosteal lining.

Figure FIG. 5..

Bone marrow infiltrates lead to endosteal osteoblast recruitment. Endosteal surface in the vicinity of bone marrow infiltrates show an increased number of osteoblasts (A, C, and E) compared with normal endosteal surface (B, D, and F). (A and B) Goldner trichrome stained osteoblasts (arrows) appear as cuboid-shaped cells (red) close to bone marrow infiltrates (*) and are virtually absent on normal endosteal surface. (C-F) Osteoblasts show expression of the PTH receptor (C, black arrows, brown-stained cells) and RANKL (E, black arrows, brown-stained cells) that are not expressed in the cells lining the bone surfaces adjacent to infiltrate-free bone marrow (D and F). Original magnification, ×1000. (G) Fraction of osteoblast-covered surface (ObS/BS) at endosteal sites adjacent to bone marrow infiltrates (BMI) and adjacent to BMI-free bone marrow. (H) Number of osteoblasts per millimeter bone perimeter (N.Ob/B.Pm) at endosteal sites adjacent to bone marrow infiltrates (BMI) and adjacent to BMI-free bone marrow. Data are given as mean ± SE; *p < 0.001, **p < 0.0001.

Bone marrow infiltrates are associated with bone formation in the endosteal lining

To address whether the observed recruitment of osteoblasts is followed by stimulation of bone formation, we searched for signs of newly formed bone in the endosteum next to bone marrow infiltrates. von Kossa staining revealed thickened seams of uncalcified bone matrix (osteoid) covering the bone surface underneath osteoblasts at sites where the endosteal bone surface is exposed to bone marrow infiltrates (Fig. 6A). No such deposition was found in the endosteum next to normal bone marrow (Fig. 6B). Synthesis of key matrix molecules, such as osteocalcin and osteopontin, was increased in areas covered by osteoid and exposed to bone marrow infiltrates, and this expression was absent in the normal endosteal bone surface (Figs. 6C–6F). Furthermore, as evident from calcein-labeling experiments and visualization of fluorescence double labeling, the endosteal layer next to bone marrow infiltrates was characterized by mineral apposition onto newly formed bone matrix (Fig. 6G). In contrast, no labeling of the bone matrix adjacent to infiltrate-free marrow was detected (Fig. 6H). Thus, bone marrow infiltrates are associated with the entire process of new bone formation in the adjacent endosteal lining layer.

Figure FIG. 6..

Increased endosteal bone formation next to bone marrow infiltrates. (A, C, E, and G) Endosteal surface adjacent to bone marrow infiltrates (*). (B, D, F, and H) Endosteal surface adjacent to normal bone marrow. (A and B) von Kossa staining showing prominent osteoid seam (green arrow) underneath the osteoblasts (black arrow, A), which is absent next to normal bone marrow (B). (C-F) In situ hybridization showing expression of (C) osteocalcin mRNA and (E) osteopontin mRNA in osteoblasts (black arrows) adjacent to the bone marrow infiltrates. (G and H) Fluorescence microscopy shows an osteoid seam (dark band, green arrow) covered by osteoblasts (white arrows). Below, a brightly labeled fluorescence double band (arrowheads) indicates mineralization of the newly formed bone matrix. This is absent on the endosteal surface next to normal bone marrow. Original magnification, ×1000.

Bone marrow infiltrates express BMPs

After finding evidence for excess bone formation in the vicinity of bone marrow infiltrates, we were interested in how cells in these lesions might stimulate bone formation. We therefore examined bone marrow infiltrates for the expression of BMP-6, BMP-7, GDF-5, and GDF-6, which are centrally involved in building up skeletal tissue. Interestingly, BMP-7, also termed osteogenic protein (OP)-1, was expressed by 13.14 ± 1.41% of cells of bone marrow infiltrates (Fig. 7A). Also, BMP-6 expression was occasionally found in such lesions, although the proportion of cells expressing BMP-6 (2.03 ± 0.36%) was considerably lower (Fig. 7B). Expression of GDF-5 and GDF-6 was absent in bone marrow infiltrates (Figs. 7C and 7D). Double immunofluorescence revealed B-cells as the predominant cell type expressing BMP-7 (Fig. 7E). These data indicate that bone anabolic factors, such as BMP-7, are expressed by bone marrow infiltrates in the vicinity of arthritis.

Figure FIG. 7..

B-cells of bone marrow infiltrates express BMPs. Bone marrow infiltrates were stained for (A) BMP-7, (B) BMP-6, (C) GDF-5, and (D) GDF-6 expression. BMP-7 expression (white arrows, brown-stained cells) was more abundant than BMP-6 expression, which was scarcely found in bone marrow infiltrates. No expression of GDFs was detectable. (E) Immunohistochemical double labeling with anti-BMP-7 (black arrows, brown-stained cells) and anti-CD3 (white arrows, blue-stained cells) shows that expression of BMP-7 is confined to CD3 cells. (F) Negative control for the double labeling (serial section), nuclei counterstained with hematoxylin. Original magnification, ×1000.


This investigation shows that inflammatory arthritis has the ability to penetrate the entire cortical barrier that exposes bone marrow to inflammatory synovial tissue. This unphysiological interaction is followed by generation of a mononuclear cell infiltrate at the interface of synovial tissue and normal bone marrow, which predominantly consists of B-cells, and to a smaller extent, T-cells. This bone marrow infiltrate separates synovial tissue from normal bone marrow. Its almost exclusively lymphocytic nature clearly distinguishes it from neighboring synovial tissue, which consists of a mixed cell population with interspersed B- and T-cells, and from normal bone marrow, which contains adipocytes. Interestingly, such lesions were absent in WT mice and also in hTNFtg mice with mild synovial inflammation and shallow erosions, which did not break through cortical bone. In contrast, such lesions were found next to deep erosions, which penetrated the entire cortical barrier and exposed bone marrow to synovial inflammatory tissue. Indeed, synovial tissue has the potential to form lymphocyte aggregates and follicle-like structures.(11,12) Although bone marrow lesions constitute lymphocytic aggregates rather than follicle-like structures, the mechanisms leading to B-cell accumulation in the bone marrow may share similarities to the generation of follicle-like structures in the synovial membrane. The close contact of B-cells with T-cells, as well as with synovial-fibroblast like cells, is typical for both lesions. Earlier investigations of follicle-like structures have elegantly shown that both T-cells and synovial fibroblast-like cells provide stimuli for B-cell proliferation and/or survival.(13–15)

Histopathological evidence for a direct contact between synovial inflammatory tissue and bone marrow resulting from local bone erosion is scarce. Bone is usually not present in synovectomy specimens, and even histological investigations of resected metatarsal heads were almost exclusively focused on changes of the junction zone but not the bone marrow cavity. However, Mohr et al.(16) described in 1986 that synovial pannus can gain access and invade bone marrow, although systematic analyses have not been performed. Bone marrow changes, however, are a well-known feature of magnetic resonance tomographies (MRT) of joints of RA patients. Thus, “bone marrow edema” is a localized increase of water content and blood flow, which points to replacement of fat-rich normal bone marrow by an inflammatory infiltrate. Interestingly, circumstances that lead to bone marrow edema in MRT scans of humans are similar to those triggering bone marrow infiltration in our animal model: Thus, lesions (1) are linked to joint inflammation, but do not occur in non-inflammatory conditions(17); (2) are closely linked to erosive disease(18); (3) disappear on remission of disease(19); and (4) increase in frequency on longer disease duration.(17)

In this study, we also show enhanced endosteal bone formation, including osteoblast recruitment, bone matrix formation, and mineralization on bone surfaces in the immediate vicinity of the bone marrow infiltrates. Thus, arthritic bone erosions penetrating the entire cortical barrier also induce a kind of repair mechanism by inducing bone formation in the endosteal lining. Whether this is a direct effect through coupling mechanisms between synovial osteoclasts and endosteal osteoblasts or is triggered by the lymphocytic bone marrow infiltrate remains to be clarified. Recently, one in vitro study has shown that differentiation of bone marrow stromal cells into mature osteoblasts is induced by inflammatory T-cells.(20) Thus, stimulation of endosteal bone formation by T-lymphocytes from bone marrow infiltrates seems feasible. However, in the aforementioned study, the T-cell factors, which are responsible for the formation of osteoblasts, have not been characterized exactly. In addition, T-cells comprise only a very small proportion of the bone marrow infiltrates, which consist mainly of B-cells. In further search for the mechanisms that lead to enhanced bone formation, we examined these infiltrates immunohistochemically for expression of BMP-6, BMP-7, GDF-5, and GDF-6. These proteins have potent osteoinductive properties and also initiate cartilage deposition and enchondral bone formation in mesenchymal tissues, which mirrors their key role in embryonic skeletogenesis and fracture healing.(21,22) BMPs are also believed to play an important role in bone remodeling in adults by inducing the differentiation of mesenchymal stem cells of the bone marrow into osteoprogenitors and osteoblasts.(23,24) We found that B- but not T-lymphocytes in bone marrow infiltrates express BMP-7 and to a lesser extent also BMP-6. These findings extend earlier in vitro findings by Detmer et al.,(25) which showed expression of BMP-7 in normal circulating B- and T-lymphocytes and expression of BMP-6 in a B-lymphocytic cell line. These data also suggest that B-cell depletion in arthritis might substantially affect bone. Thus, we are currently performing the analysis of TNF-induced arthritis in a B-cell-depleted model to define the role of B-cells in arthritic bone changes more extensively.

In summary, this study describes the nature of bone marrow involvement in arthritis. It shows that bone marrow cannot be regarded as an “innocent bystander” in arthritis but that it actively participates in this inflammatory disease. The generation of a lymphocytic infiltrate, which faces invading synovial tissue and is associated with endosteal bone formation, may be an attempt by the bone marrow to close and repair bone damage caused by arthritis. Of course, this TNF-based arthritis model needs further confirmation from other animal models of arthritis to strengthen its relevance for human RA. In addition, future experiments are needed to examine whether anti-TNF reverses endosteal bone formation mediated by B-cells and whether blockade of BMPs is effective in arresting this repair mechanism.


This study was supported by the START prize of the Austrian Science Fund (to GS), the Center of Molecular Medicine of the Austrian Federal Ministry for Education, Science and Culture and the City of Vienna, and Austrian National Bank Grant 8715 (to GS).