Rheumatoid arthritis (RA) is traditionally considered a synovial-centred disease, with pannus invasion causing bone erosion. However, since the advent of magnetic resonance imaging (MRI) scanning, additional pathology within the subchondral bone has been recognised in the form of bone oedema. This feature is characterised by increased signal on fat-suppressed T2-weighted (T2w) images indicating the presence of mobile hydrogen ions (most likely as H₂O) where there is normally low-signal fat within trabecular bone.1 MRI bone oedema is clinically important as it is frequently present in patients with early disease,2 and is a key predictor of plain radiographic erosive joint damage.3^–5

Recently, two separate studies comparing MRI appearances with bone histology have confirmed that bone oedema represents a cellular infiltrate indicating the presence of osteitis within the subchondral bone.6 7 The cellular components of osteitis related to MRI bone oedema have not yet been characterised. However, pathological studies of subchondral bone have provided some clues to the potential factors. Lymphocytic aggregates containing CD45RO+ T cells and B cells within bone marrow have been identified in subchondral bone in RA,8 and a recent immunohistochemical study reported an association between bone marrow inflammation and osteoclasts in RA subchondral bone.9 These studies have raised the possibility that subchondral bone marrow inflammation plays a role in perpetuating inflammation and local joint damage in RA.

The aim of the current study was to determine the cellular characteristics of osteitis associated with MRI bone oedema, and further clarify the relationship between MRI bone oedema and bone erosion in RA.

PATIENTS AND METHODS

Patients

From February 2005 to September 2007, 11 patients fulfilling criteria for RA,10 scheduled for orthopaedic surgery to the hands/wrists or feet/ankles, were recruited according to local ethics committee requirements. We have previously described the histopathology reports (but not immunohistochemistry results) from four patients in this study.7 Clinical assessments and MRI scans were performed 1–2 days before surgery.

Magnetic resonance imaging scans

Preoperative gadolinium-enhanced MRI scans were obtained using a 1.5-T scanner (Siemens) with a dedicated extremity coil, as previously described.7 The presence or absence of MRI bone oedema in the relevant bone was recorded by a musculoskeletal radiologist (AD) who was blinded to all histological scores.

Immunohistochemistry of bone specimens

At the time of surgery, bone samples were collected and transported to the laboratory. Samples were labelled with careful recording of the site and orientation of the bone to allow later comparison between corresponding MRI site and histological slide. Bone samples were processed as previously described.11 Antigen retrieval was achieved with Dako Target Retrieval Solution pH 6.0 (Dako, Carpinteria, California, USA) or, for RANKL, EDTA pH 7.5 (10 mM).

For all assays except for RANKL detection, the EnVision+ System-HRP (DAB) Kit (Dako) was used. The following mouse antihuman primary antibodies were used for 60 min at room temperature: anti-CD68 (1:50, Dako, clone PG-M1), anti-plasma cell (1:50, Dako, clone US38c), anti-CD8 (1:50, Dako, clone C8/144B), anti-CD20 (1:200, Dako, clone L26), and anti-tartrate-resistant acid phosphatase (TRAP) (1:100, Zymed, San Francisco, clone ZY9C5), or appropriate isotype-matched antibody. In the case of RANKL staining, the ABC staining kit (Santa Cruz Biotechnology, Santa Cruz, California, USA) was used, with goat antihuman RANKL polyclonal antibody (dilution 1:35, Santa Cruz Biotechnology, clone C-20) or isotype-matched polyclonal antibody.

Quantification of staining

In order to quantify the number of macrophages, CD8+ T cells, plasma cells and B cells, areas of bone marrow were identified in each slide. For each cell type a single observer who was blinded to the MRI findings, counted the number of positively staining cells in five high-power fields throughout the slide at ×20 magnification, and for each slide the mean number of positively staining cells was recorded for analysis. The number of osteoclast-like cells was quantified by recording the mean number of TRAP positively staining multinucleated cells (containing at least three nuclei) adjacent to trabecular bone in 10 high power fields at ×20 magnification.

For RANKL staining, a semiquantitative approach was used for this staining. Two observers, both blinded to the MRI findings, independently scored for degree of staining (0–4; with 0,<5% staining and 4, >75% staining) within bone marrow in the entire slide at ×20 magnification, based on a validated method.12 Scores were compared and where differences occurred, a consensus opinion was obtained for each slide.

Statistical analysis

MRI bone oedema was classified as absent or present, and generalised linear mixed models with a logit link that allow for the correlations between measurements on the same patient were used. For the quantitative immunohistochemistry analysis, the data were log transformed to reduce the influence of outliers in the data. The relationship between the number of osteoclasts, inflammatory cell numbers and RANKL expression was analysed by Spearman correlations.

RESULTS

Clinical characteristics

We analysed 28 bones from 11 patients obtained from a total of 13 separate surgical procedures. The clinical characteristics of each patient, surgical procedures and bone specimens are summarised in supplementary table 1. The median (range) disease duration was 23 (4–43) years and DAS28 was 3.34 (1.49–6.47). There were nine (82%) patients seropositive for rheumatoid factor, eight (73%) patients on methotrexate, and one patient on anti-tumour necrosis factor therapy. The majority of the bones analysed were phalangeal and metatarsal bones of the feet (supplementary table 1). Fourteen of the 28 (50%) bones analysed had MRI bone oedema.

Analysis of osteoclasts within samples with magnetic resonance imaging bone oedema

Samples with high-grade MRI bone oedema had numerous TRAP-positive multinucleated cells, typical of osteoclasts, adjacent to trabecular bone (fig 1). Analysis of all samples showed those bones with MRI bone oedema had greater numbers of osteoclasts, than those without MRI bone oedema (table 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Analysis of osteoclasts in samples with corresponding magnetic resonance imaging (MRI) bone oedema. Axial fat-suppressed T2w MRI scans showing: (A) the fourth metatarsal head (circle), with no MRI bone oedema; (B) the first metatarsal head (circle), with high-grade MRI bone oedema. Bone samples stained for tartrate-resistant acid phosphatase showing: (C) the fourth metatarsal head with no osteoclasts and minimal bone marrow inflammatory infiltrate (corresponds to MRI scan in A), and (D) the first metatarsal head with numerous osteoclasts and intense bone marrow inflammatory infiltrate (corresponds to MRI scan in B). Scale bar represents 10 μm.

Immunohistochemical characterisation of magnetic resonance imaging bone oedema

The bone samples with MRI bone oedema also had more abundant inflammatory cells within bone marrow than those samples without MRI bone oedema (table 1, fig 2). Numerous CD68-positive macrophages and plasma cells were present in bone marrow in those samples with MRI bone oedema. CD8+ T cells were also increased, but at much lower numbers. B cells were present only in small numbers, but in some samples with MRI bone oedema, B cell aggregates were present in bone marrow (fig 2). In addition, there were significant correlations between osteoclast number and the number of other inflammatory cells within bone marrow, including macrophages (r = 0.54, p = 0.003), plasma cells (r = 0.61, p = 0.005) and CD8+ T cells (r = 0.38, p = 0.046).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Immunohistochemical characterisation of magnetic resonance imaging (MRI) bone oedema. Representative examples of CD68 (A, B), plasma cell (C, D), CD8 (E, F), CD20 (G, H) and RANKL (I, J) staining. (A,C,E,G,I) Examples of bone samples with no corresponding MRI bone oedema. (B,D,F,H,J) Examples of bone samples with MRI bone oedema. Scale bar represents 10 μm.

Analysis of RANKL expression within samples with magnetic resonance imaging bone oedema

Samples with high-grade MRI bone oedema had extensive expression of RANKL (fig 2). There was a trend to increased RANKL expression within bone marrow in all samples with MRI bone oedema (table 1). There was a strong correlation between the bone marrow RANKL score and the number of osteoclasts within the bone samples (r = 0.59, p = 0.004).

DISCUSSION

This study is the first to characterise histologically the cellular composition of MRI bone oedema in RA. We have shown bone marrow infiltration by a number of cell types, including macrophages, plasma cells, CD8+ T cells and B cells. Furthermore, MRI bone oedema is strongly associated with the presence of osteoclasts, sitting in resorption pits on the surfaces of bony trabeculae, suggesting a potential explanation for the well-documented relationship between MRI bone oedema and the development of bone erosion. This study ties imaging data, showing a convincing link between MRI bone oedema and progression of radiographic erosions,3^–5 with cellular data, at the level of the individual rheumatoid joint and examined directly by immunohistochemistry.

The relationship between MRI bone oedema and osteoclast development and function in patients with RA has not been previously reported. However, consistent with our results, MRI bone oedema in tumour necrosis factor-transgenic mice is associated with high numbers of osteoclast precursor cells within bone marrow.13 Together, these studies suggest an important pathway for the development of bone erosion, starting from an inflammatory process within bone marrow, leading to activation of the RANKL/OPG pathway, with subsequent osteoclast formation and activation, and ultimately bone resorption.

We acknowledge that this study has some limitations. The sample number was relatively small (although it is larger than most studies of bone pathology in RA).6 9 We have undertaken a rigorous statistical analysis of the samples, with the use of generalised linear mixed models to correct for repeated measures in the same patient. These factors may have reduced the power of the study to identify small differences between groups. Most patients had long-standing disease; therefore, we are unable to comment on the histological appearances of MRI bone oedema in early disease. In addition, other cells that have been identified within the rheumatoid synovium, including CD4+ T cells and dendritic cells, were not analysed in this study.

Despite these limitations, this work has provided insights into the mechanisms of bone damage in RA, implicating chronic inflammation and RANKL-induced osteoclastogenesis within bone marrow. These findings support the hypothesis that bone erosion in RA occurs through activation of local bone resorption mechanisms within subchondral bone as well as through synovial invasion into bone.