PATIENTS AND METHODS

Patients

All patients fulfilled the revised criteria of the American College of Rheumatology for the diagnosis of RA.²⁴ Only patients with early disease (<1 year disease duration, as measured from the first clinical signs of arthritis) at the time of the baseline biopsy were enrolled in the study. Patients who currently were receiving or had previously received disease modifying antirheumatic drugs (DMARDs) or prednisone were excluded. The study protocol was approved by the ethical committees of the participating centres in the Netherlands and Australia.

Clinical, laboratory, and radiological evaluation

Clinical and laboratory assessment of the patients was performed on the day of the baseline synovial biopsy and after follow up. Clinical data included age, sex, duration of symptoms before the first biopsy procedure, use of DMARDs in the follow up period, tender and swollen joint counts, visual analogue scale for pain and morning stiffness, Health Assessment Questionnaire (HAQ), and disease activity score using 28 joint counts (DAS).²⁵ Laboratory measurements included the erythrocyte sedimentation rate (ESR), serum levels of C reactive protein (CRP), and RF. At baseline soluble granzyme B levels were determined in paired synovial fluid and blood serum samples, if available.^26,27 Radiographs of hands and feet were obtained at enrolment in the study and after follow up. All x ray findings were scored in random order by a “blinded” observer using the Larsen method²⁸ to assess radiological damage. Radiological deterioration was estimated by subtracting the scores at follow up from the entry Larsen scores. To compensate for the variable follow up we calculated the average change in Larsen score per year by dividing the delta change in Larsen score by the number of years of follow up.

Synovial tissue biopsies

Multiple synovial tissue samples (⩾6)²⁹ were obtained by blind needle biopsy (n = 18), or by small needle arthroscopy (n = 18) as described previously.³⁰ The biopsy specimens were immediately processed en bloc and embedded in Tissue Tek OCT (Miles Diagnostics, Elkhart, IN), snap frozen, and stored in liquid nitrogen. Cryosections 5 mm thick were cut from the frozen specimens and placed on glass slides, air dried overnight, wrapped in laboratory film, and stored at −80°C until stained.

Immunohistological staining

Serial sections were stained using mouse-antihuman antibodies against CD3 (Leu-4, Becton-Dickinson, San Jose, CA) to detect T cells, CD22 (CLB-B-Ly/1, Central Laboratory of the Netherlands Red Cross Blood Transfusion Service, Amsterdam, The Netherlands) to detect B cells, CD38 (Leu-17, Becton-Dickinson) to detect plasma cells, CD55 (Mab67, Serotec, Oxford, UK) to detect FLS, CD68 (EBM11, Dako, Glostrup, Denmark) to detect macrophages, and granzyme B (Gr7, Monosan, The Netherlands) to detect granzyme B+ cytotoxic cells. For all antibodies the staining was conducted, according to a three step immunoperoxidase method, as described previously.³¹ Figure 1 provides an example of a sections showing granzyme B+ cells, T cells, and FLS.

Digital image analysis of synovial tissue

Only synovial tissue samples in which the intimal lining layer was identifiable were included in the study. All sections were coded and analysed in a random order by an independent observer who was unaware of the clinical data (MCK), as described previously.³² One separate representative region including the intimal lining layer and synovial sublining was chosen for the evaluation of each section. From this region 20 consecutive high power fields (HPFs) were captured and digitised covering an area of 2.1 mm². The HPF images were acquired on a fully automated Leica DMRXA microscope (Leica, Wetzlar, Germany) with a Prior stage table, captured using a three chip CCD (Charged Coupled Device) videocamera (Sony, Tokyo, Japan), and digitised with a Matrox 32 bit colour video digitiser card, using a highly standardised macro program written in the program language QUIPS for the Leica Qwin image analysis software (Leica, Cambridge, UK). The resultant colour images were in a 740×570 pixel RGB format with a 24 bit resolution, enabling the use of 16 581 375 colours. For each acquisition session the microscope, camera, and computer were calibrated according to a standardised procedure, and for each individual marker all images were acquired in one single session. The images obtained were stored using tagged image file compression on a writable CD ROM.

All sections were examined using a computer (Qwin Pro V2.5, Leica, Cambridge, UK) and computer assisted colour video image analysis system operating a specialised algorithm (SYNDIA v1.1) written in the program QUIPS.³²

Statistical analysis

The distribution of each continuous variable was checked for normality. Non-parametric methods were used for data failing tests for normality. The patients were divided into two groups based on the scores for radiological deterioration for each year of follow up (a cut off value for the change in Larsen’s score per year of two points was used). The groups were compared using the Mann-Whitney U test. To assess the baseline characteristics, such as sex, presence of RF, age, and histological markers to predict radiological progression, we used a generalised linear regression model with binary family and log link. Relative risk is for the immunohistological markers given per difference of 100 cells/mm². Each variable was included and considered significant when the two tailed p value was <0.05. Variables that were found to be significant were then entered jointly into a generalised linear regression model. Data were analysed using STAT version 7 for Windows.

RESULTS

Clinical features

Thirty six patients (15 (42%) men, 21 (58%) women), participated in the study; mean age at study entry was 64 years (range 28–86). The baseline biopsy was performed at a mean disease duration of 5 months (range 1–12). Follow up examination was performed after a mean of 58 months (range 38–72). Patients were divided into two groups based upon the radiological progression for each year of follow up: group I with minimal progression (n = 20; Larsen <2 points/year); group II with pronounced progression (n = 16; Larsen ⩾2 points/year). During the follow up period, 7 patients (group I n = 5, group II n = 2) did not receive any DMARDs, 29 patients (group I n = 15, group II n = 14) received at least one DMARD. Of these 29 patients, 16 patients (group I n = 6, group II n = 10) received two or more, and of these 16 patients, 5 patients (group I n = 2, group II n = 3) received three or more DMARDs. None of the patients were treated with biological agents. Table 1 presents the clinical, laboratory, and radiological findings. Twenty one (58%) patients were RF positive and 15 patients (42%) RF negative. In keeping with previous studies,³³ half of the patients had radiological signs of joint damage at study entry (Larsen at study entry 4.4 (0.9) (mean (SEM)); three patients had no signs of damage at the end of follow up. On average, all measures of disease activity had improved at the time of the follow up evaluation. The mean (SEM) DAS score was reduced from 5.8 (0.2) at enrolment to 3.6 (0.3) at follow up (p<0.0001). Despite clinical improvement radiological deterioration was seen in almost all patients (Larsen score at follow up 19.2 (3.8); p<0.0001).

Immunohistological findings in relation to progression of joint damage

Table 2 summarises the demographic, clinical, immunohistological, and radiological data. Generally, there was increased cellularity in the synovial tissue of patients with more severe progression of joint destruction. There were significantly more FLS in the synovium of patients with more destructive disease than in patients with less progression of radiological signs of joint damage (change in Larsen score <2 points/year: 158 (2–1473) FLS/mm² (median (range)); change in Larsen score ⩾2 points/year: 441 (35–2405) FLS/mm²; p = 0.03). Although the numbers of B cells, plasma cells, and granzyme B+ cells tended to be increased in group II, comparison by Mann-Whitney U test did not show statistical significance. Of interest, there was no clear cut difference in macrophage infiltration between the two patient groups.

Regression analysis was used to identify prognostic markers of joint damage, as described previously.^21,34 Regression analysis using a univariate model identified female sex (relative risk (RR) = 10.7; p = 0.015), FLS (RR = 1.7; p = 0.02), T cells (RR = 2.7; p = 0.013), and granzyme B+ cells (RR = 7.2; p = 0.003), as discriminators for unfavourable radiological outcome (group II, Larsen score ⩾2 points per year).

In addition, a multivariate model identified female sex (RR = 78.2; p = 0.017), FLS (RR = 1.5; p = 0.013), and T cells (RR = 1.2; p = 0.015) as independent predictors of unfavourable radiological outcome (table 3).

Soluble granzyme B levels in relation to progression of joint damage

To provide more insight into the possible role of granzyme B+ cytotoxic cells, we measured soluble granzyme B (produced by these cells) in body fluids. Paired synovial fluid and serum samples were available from seven patients participating in the biopsy study. In three patients with minimal progression (group I) the median serum level was 362 µg/ml (range 160–357) and the median granzyme B level in synovial fluid was 253 µg/ml (30–308) (fig 2). In these patients granzyme B+ cytotoxic cells were absent in synovial tissue (median 0 cells/mm², range 0–1). In four patients with more severe radiological progression (group II), granzyme B serum levels were similar to those in group I (median 362 µg/ml, range 121–939), but granzyme B levels in synovial fluid were markedly increased (median 489 µg/ml, range 55–1337). Moreover, these patients had increased numbers of granzyme B+ cytotoxic cells in the synovium (19 cells/mm² (1–47)). Thus, although statistical analysis could not be performed owing to the small sample size, these data are consistent with the results from the larger biopsy study and support the notion that granzyme B+ cytotoxic cells are associated with progression of joint destruction.

DISCUSSION

We confirmed in 36 patients with recent onset RA (symptoms <1 year) that female sex is predictive for an unfavourable disease outcome after a mean follow up of 54 months. In the synovium retrieved in the first year of the disease we found that the presence of granzyme B+ cells, T cells, and FLS at baseline is associated with the severity of radiological deterioration over the follow up period.

The most challenging observation in this study is perhaps the relationship between the initial number of T cells and unfavourable disease outcome in RA. Until recently, the possible role of T cells in the pathogenesis of joint destruction has been controversial.² The recently proposed interaction between RANKL+ T cells and RANK+ osteoclasts^4,35,36 might provide the mechanism by which T cells promote erosive disease. Consistent with this view, RANKL expression was seen predominantly in synovial tissue from patients with active RA, particularly within areas of lymphocyte infiltration.^6,37 In addition, the expression of OPG, which inhibits osteoclastogenesis by acting as a soluble decoy receptor for RANKL, is very low in the synovium of patients with RA.³⁷

We also observed a relationship between the number of FLS in the initial synovial biopsy samples and joint destruction. Of interest, FLS also express RANKL and might thus activate osteoclasts by the same mechanism as that described above.³⁶ Previous work has shown that cultured rheumatoid FLS may efficiently induce osteoclastogenesis in association with up regulated expression of RANKL and decreased production of OPG.³⁸ In addition to stimulating osteoclasts, FLS might also directly promote joint degradation.^12,15 They serve as source of factors that mediate joint destruction: cytokines, metalloproteinases, serine proteinases, and cathepsins, together with various small molecules that increase vascular permeability and enhance the inflammatory response. It has been suggested that the interaction between FLS and macrophages through ligand pairs like CD97/CD55³⁹ and the expression of adhesion molecules such as VCAM-1 facilitating attachment to extracellular matrix⁴⁰ might be involved in the destructive process.

Granzyme B+ cytotoxic cells were the third cell type found to be associated with the development of destructive disease. Granzyme B is a serine protease that is stored in the granules of effector memory T cells, such as activated cytotoxic T cells and natural killer cells. The results from this study were supported by the association between increased expression of granzyme B+ cells (in the group with more destructive disease) and raised synovial fluid levels of extracellular granzyme B. Previously, we have shown that soluble granzyme levels are specifically increased in the serum and synovial fluid of patients with established RA compared with disease controls.²⁷ In addition, the number of granzyme B+ cells is specifically increased in rheumatoid synovial tissue.³¹ The relationship between granzyme B and destruction might be explained by the extracellular activity of the enzyme when released from cytotoxic cells. It has been shown previously that granzyme B can degrade extracellular matrix.^41,42 Of interest, granzyme B+ cells were also shown to be present at the invasive front, the pannus-cartilage junction.⁴² Further support for the role of granzyme B producing cells in the destructive process comes from the recent observation that granzyme B levels are highest in patients with early RF positive RA when compared with other arthritides, and this is independently predictive of the development of erosions.⁴³

Previous work suggested a relationship between synovial tissue macrophages and an unfavourable course of RA.^10,44,45 We could not confirm these results. The present study is larger than previous studies and we selected patients with earlier disease at baseline. In addition, immunohistological markers for FLS and cytotoxic cells were included that were not previously studied, and investigated in a prospective study design. Moreover, we investigated the simultaneous associations of immunohistological variables with a specific outcome variable instead of analysing each variable separately, because these variable are highly correlated with each other. Therefore, we believe that the results shown here are valid. Of interest, the same approach did disclose a relationship between synovial macrophage infiltration and scores for local disease activity,⁹ supporting the view that the pathogenesis of synovial inflammation and joint destruction might be partly different.¹⁰

Taken together, this prospective study shows a positive relationship between cell infiltration by T cells and FLS, on the one hand, and joint destruction, on the other, which might be explained in part by stimulation of osteoclasts. In addition, we found increased granzyme B+ cells in patients with more destructive disease; this might be explained by the role of granzyme B in degradation of extracellular matrix. Interference with these pathways may help to protect the joints against destruction.