Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory disease brought about by the activities of cells and inflammatory cytokines in the inflamed synovial membrane (for review [
1]). Cells found in diseased synovium are predominantly blood derived, with large numbers of activated T cells and macrophages, fewer plasma cells and dendritic cells, together with expanded numbers of resident fibroblasts and endothelial cells [
2]. Of the inflammatory cytokines, tumour necrosis factor (TNF)-α [
3] was demonstrated by us to be of pivotal importance in regulating the cytokine cascade in this tissue [
2]. These and other studies led to the hypothesis that TNF-α was a therapeutic candidate in RA [
4], and the demonstration that TNF-α blockade in murine arthritis ameliorated disease [
5] led to the testing of anti-TNF biologicals in humans [
6‐
10]. Anti-TNF biologicals are now licensed for use in RA and well over one million people worldwide have been treated with them. Patients with other chronic inflammatory diseases have also been treated with these drugs (for review [
11]). Despite the success of these anti-TNF biologicals, which protect joints from destruction [
12], they are not a cure, and blockade of TNF-α has the potential to compromise the immune response (for review [
13]).
In contrast to the clearly defined role of macrophage-derived cytokines in the pathogenesis of RA, the relevance and contribution of the T cells is not clear and has been challenged [
14]. In particular, the expectation that the increased T cells in the synovium are a result of clonal expansion to a given antigen has not been established. An human leucocyte antigen (HLA)-restricted T-cell response to antigen is suggested, because more than 80% of Caucasian RA sufferers have a shared epitope conserved across the HLA-DR1 and HLA-DR4 haplotypes (0101, 0401, 0404 and 1402) [
15]. However, no overall consensus has been reached on the potential autoantigens involved. T-cell responses to collagen type II, heat shock proteins and microbial antigens have been reported in a small proportion of RA patients (for review [
16]) and, more recently, autoantibodies to deiminated citrullinated peptides have been described, suggesting that they may be important autoantigens in this disease. This aside, the concordance for disease in identical twins is still less than 15%, suggesting that other factors are of major importance.
More recently, we explored the mechanism(s) underlying chronic production of TNF-α in RA synovium and found that the spontaneous TNF-α production from macrophages is dependent upon direct interaction with T cells found in this tissue [
17]. Several publications, principally from the Dayer group, describe the ability of activated T cells to induce inflammatory mediator production from monocytes in a contact-dependent manner [
18‐
21]. T cells found in the synovial joint tissue of RA patients are predominantly memory CD4
+ cells [
22], with unusual characteristics in that they are small, noncycling and nonapoptosing, but have features of activation [
23]. We hypothesized that these T cells are stimulated in the joint not as a consequence of antigen exposure but through bystander cytokine-driven activation. To test our hypothesis, we developed a surrogate model for RA T cells by stimulating resting normal human peripheral blood lymphocytes with a cocktail of cytokines (IL-2, IL-6 and TNF-α) for 8 days, as was previously reported [
24]. When fixed (to exclude mediators released by the T cells) and placed in co-culture with monocytes, proinflammatory cytokines and chemokines were produced in a cell contact-dependent manner [
25,
26]. Comparison of cytokine-activated T (Tck) cells with conventional T-cell receptor stimulated T (Ttcr) cells activated by immobilized anti-CD3 antibodies revealed the latter to be more activated and stimulate monocytes to produce both IL-10 and TNF-α [
27]. In contrast, Tck cells induced only TNF-α and not IL-10.
We postulated therefore that Tck cells are important in the perpetuation of rheumatoid disease [
25] because they induce an unbalanced, proinflammatory cytokine response from monocytes, and could thus form part of a vicious cycle. Furthermore, we demonstrated that Tck cells mimicked RA T cells in their effector function, because identical signalling pathways activated within the monocyte lead to TNF-α production [
17]. The monocyte phosphoinositide 3-kinase pathway and the transcription factor NFκB were utilised in a similar manner for induction of TNFα but was different from those induced by the Ttcr cells [
17]. More recently [
26], we demonstrated that contact-dependent induction of seven different chemokines in monocytes was also identical between Tck and RA T cells in the utilization of nuclear factor-κB in the monocyte signalling pathway but different to that by co-culture with Ttcr cells.
In this paper we investigate the phenotype of the day 8 Tck effector cells and, by using cell sorting methodologies, we determine the phenotype of Tck cells and their precursors in resting blood. Our results demonstrate that the most potent Tck effectors are from a starting population of CD4+CD45RO+ memory T cells. Furthermore, in culture these cells divide, become activated and express increased levels of activation, adhesion and integrin molecules. Comparison with T cells extracted from the RA joint indicate that day 8 Tck cells resemble RA T cells not only in their contact-dependent effector function but also in their active phenotype, with selective upregulation of adhesion/integrin molecules (including very late antigen [VLA]-4) that facilitate the extravasation of T cells into inflamed tissues, including RA synovium.
Materials and methods
Reagents
IL-6 was kindly gifted by Novartis (Vienna, Austria), IL-2 from the US National Institutes of Health, IL-10 from Schering Plough (Kenilworth, NJ, USA), TNF-α from Boehringer Ingelheim (Biberach, Germany) and IFN-γ from Bender (Wien, Austria). Lipopolysaccharide (Salmonella typhimurium) was purchased from Sigma (Gillingham, UK) and carboxyfluoroscein succinimidyl ester (CFSE) from Molecular Probes (Leiden, The Netherlands). Flow cytometry antibodies were purchased from BD Pharmingen (Erembodegem, Belgium). Blocking antibodies against CD18 (clone TS1/18) and CD11a (clone HI111) were from BioLegend (Cambridge, UK), anti-CD69 (clone TP1.55.3) was from Beckman Coulter (High Wycombe, Buckinghamshire, UK) and anti-CD49d (clone TA-2) was from Endogen (Rockford, Illinois, USA). Isotype control antibody (mouse IgG1) was purchased from BD Pharmingen. Cells cultured in RPMI/glutamine (PAA Laboratories, Pasching, Germany) were supplemented with 10% normal AB human serum (Sigma Aldrich, Dorset, UK) or 5% foetal calf serum (Biowest, Nuaillé, France), as appropriate. All reagents had under 0.1 units/ml of endotoxin, by LAL (limulus amebocyte lysate) assay (BioWhittaker, Walkersville, MD, USA).
Isolation of lymphocyte subpopulations by magnetic beads
Buffy coats were purchased from the North London Blood Transfusion Service (Colindale, UK) and peripheral blood mononuclear cells isolated by Lymphoprep™ (Cedarlane Laboratories Ltd, Ontario, Canada) centrifugation followed by elutriation, as described previously [
17]. Lymphocyte subsets were isolated using CD4 or CD8 isolation kits (Dynal, Invitrogen, Paisley, UK), in accordance with the manufacturer's instructions. CD45RO
+ cells were isolated negatively using memory CD4
+ T-cell isolation kit from Miltenyi Biotec (Guildford, Surrey, UK), in accordance with the manufacturer's instructions.
Isolation of T cell sub-populations by cell sorting (fluorescence-activated cell sorting)
A cocktail of anti-CD4 with either anti-CD45RA or anti-CD45RO antibodies was used to sort negatively for memory or naïve populations, respectively. A cocktail of anti-CD4, anti-CD45RA and anti-CCR7 (anti-CC chemokine receptor 7) antibodies was used to sort the CD4+ central (CD45RA-CCR7-) and effector memory (CD45RA-CCR7+) populations on day 0. For the day 8 sort, CCR7 antibody was used to sort positively the central and effector memory cells from the CD4+CD45RO+ Tck population that was previously isolated at day 0 using the memory CD4+ T-cell isolation kit from Miltenyi Biotec, as described above. The resulting cell purity was routinely assessed at more than 95%. Cells were sorted on a FACSVantage SE (BD Biosciences, Cambridge, UK).
Tck generation and monocyte co-culture
Tck cells were generated from resting blood with IL-2, IL-6 and TNF-α, as previously described [
17]. Supernatants were harvested and T-cell-derived cytokine (IFN-γ, granulocyte macrophage colony-stimulating factor, lymphotoxin-α and IL-10) levels were measured using sandwich ELISA (BD Pharmingen). Tck cells were co-cultured with human peripheral blood monocytes for 18 hours, as previously described [
17], and TNF-α levels in supernatants were assayed by sandwich ELISA (Pharmingen, San Diego, CA).
Isolation of RA synovial membrane mononuclear cells and phenotypic analysis
Mononuclear cells were obtained from RA synovial tissue, as previously described [
17]. Tissues were from joint replacement surgery specimens provided by the Orthopedic/Plastic Surgery Department, Charing Cross Hospital, London, UK (covered by ethics approval RREC 1752), after receiving signed and informed patient consent and anonymization. Peripheral blood lymphocytes or RA synovial mononuclear cells were incubated with fluorochrome-labelled antibodies and analyzed on a Becton-Dickinson LSR I flow cytometer (BD Biosciences). A total of 5,000 synovial lymphocytes were collected and results were analyzed using CellQuest Pro software (BD Biosciences).
CFSE assay of cell proliferation
Lymphocytes were resuspended in phosphate-buffered saline with 2.5 to 5 μmol/l CFSE at room temperature for 10 minutes and washed extensively before culture. At day 8, cells were counterstained with antibodies to cell surface markers. In all, 50,000 events were routinely collected by flow cytometry.
Blockade of CD18, CD69, CD49d and CD11a on Tck cells and RA synovial mononuclear cells
Day 8 Tck cells were fixed in 1% paraformaldehyde at 4°C before incubation with 1.25 μg per million cells of blocking antibody against CD18, CD69, CD49d, or CD11a. Isotype control antibody was also included. After 30 minutes, Tck cells were washed to remove excess antibodies and co-cultured with fresh monocytes at a ratio of 8:1 in the presence of purified human IgG (5 μg/ml). Supernatants were collected after 18 hours and TNF-α was assayed by ELISA. RA synovial mononuclear cells (MNCs) were incubated in triplicate for 24 hours at 0.5 × 106/ml in the presence or absence of 5 μg/ml of blocking antibody against CD18, CD69, CD49d, or isotype control. Supernatants were collected for TNF-α analysis by ELISA.
Statistical analysis
Results were analysed using Prism 4 software (GraphPad Software Inc., San Diego, CA, USA) and statistical differences between groups were analyzed using Wilcoxon rank test, Student's t-test, or one-way analysis of variance with Bonferroni post hoc corrections as appropriate. Correlation association between groups of data was analysed using nonparametric (Spearman) correlation analysis.
Discussion
Although TNF-α is now acknowledged to be important in the pathology of RA [
11], the mechanisms that lead to its dysregulated production remain unknown but are of importance to the development of improved therapies. Several pieces of evidence from our group and others have emerged to indicate that macrophage-derived TNF-α in inflamed synovial tissue is T cell and contact dependent [
17,
21,
30‐
32]. In order to study this mechanism in more detail, we developed a human surrogate model system based on cytokine activation of peripheral blood lymphocytes, and found that after 8 days these cytokine-activated T cells (but not those activated by crosslinked anti-CD3) mimicked the effector function of rheumatoid T cells [
17,
25]. Importantly, the differences between Tck cells and those derived from antigen activation provided an opportunity for selective inhibition of pathogenic T cells. In this report we characterize the generation of day 8 Tck cells in terms of the induced lymphocyte proliferation, cell surface phenotype and cytokine production. Furthermore, by a series of cell sorting experiments, we not only identify the phenotype of the most potent effector cell at day 8 but also, based upon this knowledge, uncover the precursor phenotype of Tck cells within normal resting peripheral blood lymphocyte population. This may have implications for our understanding of the generation of pathogenic T cells in the highly proinflammatory environment of the rheumatoid synovium.
Analysis of the lymphocytes expanded from human peripheral blood with the cytokine cocktail in our day 8 Tck cultures revealed a significant expansion in CD3
-CD56
+ cells (NK population), with a lower overall change in the number of CD3
+CD56
- T cells over the 8 days, despite cell division within the CD8
+ population. The lack of change in total cell number at day 8 is apparently due to apoptosis in approximately 30% of the CD3
+ cells (data not shown), most likely because of depletion of growth factors in these cultures, particularly when expanded together with the rapidly proliferating NK cells. Indeed, if the human peripheral blood lymphocytes were expanded with IL-15 alone (previously to generate effective Tcks cells [
25]), the expansion of NK cells at day 8 was even more marked. Furthermore, if the NK cells were separated at day 8 from the CD3
+ T cells, they also exhibited contact effector function on monocytes but in a phosphoinositide 3-kinase dependent manner (unpublished data), similar to that induced by anti-CD3 antibody stimulated peripheral blood lymphocytes [
17,
26,
33]. Only the CD3
+ cells within the day 8 Tck cells (and not the CD3
-CD56
+) exhibited effector function identical to that of RA synovial T cells, and thus we focused our attention on these cells.
Closer analysis and enrichment of the CD3
+ subset phenotype (at either day 0 or day 8) revealed that the most efficient effector cell resided within the CD3
+CD4
+CD45RO
+ memory subset, and more specifically within the effector memory population (CCR7
-). The day 8 Tck cells resembled RA synovial T cells, particularly in terms of the increased expression of several adhesion molecules including VLA-4. The latter is of particular interest because, based on its overall expression at day 8, it was the only ligand to correlate positively with the ability to induce monocyte production of TNF-α (Table
2). Synovial T cells express high levels of both the alpha (CD49d) and beta (CD29) chains of VLA-4 and exhibit
ex vivo an enhanced ability to bind to fibronectin fragments (a VLA-4 ligand) [
34]. Furthermore, an alternate VLA-4 ligand, namely VCAM-1, is expressed selectively on RA endothelium and macrophages in the lining layer of the synovium, and is inhibited following anti-TNF-α therapy in the clinic [
35]. It has been hypothesized that VLA-4 facilitates extravasation of lymphocytes across endothelium (via VCAM-1) and their retention in the joint tissue (via fibronectin binding and/or VCAM-1). To investigate this further, we intend to explore the ability of VLA-4
high Tck cells to extravasate endothelium in
ex vivo cultures. Furthermore (and potentially of greater interest), we will determine the ability of CD4
+CD45RO
+VLA-4
high Tck cells to traffic to synovial grafts in the RA/severe combined immunodeficient mouse model, together with their potential to upregulate cytokines locally in this tissue.
The fact that the resting CD4
+CD45RO
+ T cells from normal blood are unable to induce TNF-α from monocytes indicates that this effector function is acquired either due to
de novo gene transcription and/or conformational change of a surface molecule during the differentiation process. Several cell surface molecules, including integrins and adhesion molecules, have previously been shown to be of importance in contact-dependent effector function, including CD69 [
19], lymphocyte function-associated antigen-1 (CD11a and CD18) and others [
18,
36]. We confirmed that blockade of CD18 and CD69 on Tck cells reduced monocyte TNF-α production in the co-culture system, which is in agreement with previous studies [
19,
21,
36]; also, in one single experiment blockade of CD49d (VLA-4) also reduced monocyte TNF-α production. More importantly, the role of these cell surface ligands was verified in the present study using the
ex vivo RA synovial MNC culture system, reinforcing the relevance of Tck cells as surrogates for RA synovial T cells, and highlighting the potential role of these ligands in contributing to chronicity in RA. However, it is noteworthy that none of these ligation events were dominant in either system, and several different ligation events are potentially 'involved' in these contact-mediated events in inflammatory tissue, as was previously described [
19].
VLA-4 also probably plays a role in trafficking of T cells to the synovial microenvironment (as discussed above). Interestingly, the chemokine receptor CXCR4 previously identified on RA synovial T cells [
37] was also increased on day 8 Tck cells. CXCR4 plays a key role in trafficking T cells to inflamed synovium in response to ligands including stromal cell-derived factor-1 [
37,
38]. Therefore, antigen-independent activation by proinflammatory cytokines could not only induce cells capable of perpetuating disease by cellular contact with monocytes but also facilitate the accumulation of these pathogenic cells at the site of disease. We are currently exploring the role of VLA-4 (as discussed above), together with CXCR4 and stromal cell-derived factor-1, in migration of these Tck cells
in vivo in the RA/SCID mouse model.
Although the series of cell sorting experiments indicated that enriching for CD4
+CD45RO
+ T cells increased the Tck effector potency, variation between different experiments with different donors was very evident. One possible explanation is that regulatory T cells (or factors) are present in the cultures and suppress Tck effector function, either by modulating the generation of effector Tck cells and/or by modulating the ability of the monocytes to produce TNF-α after contact activation [
39]. Consistent with this hypothesis was our observation that the proportion of CD4
+CD25
+ cells increased in the Tck cultures (Table
1) and were found within the dividing population of CD4
+CD45RO
+ cells (Figure
4a). We have preliminary data showing that CD25
+Foxp3
+CD127
- regulatory T cells are present in these Tck cultures and that they expand over the 8-day culture period at a faster rate than the Tck effector cells. Therefore, although the phenotype of the divided CD4
+CD45RO
+ cells (Figure
4a) expressed more surface activation and adhesion molecules, including CD49d (Figure
4e), they may not necessarily be more potent effectors than the nondivided CD4
+CD45RO
+ cells because CD25
+Foxp3
+CD127
- regulatory T cells are also present within these populations. We are currently determining in a separate and extensive study the modulatory effect of these CD25
+Foxp3
+CD127
- regulatory T cells upon the expansion of Tck cells, the acquisition of effector function and the direct effect on the monocytes in the co-culture model. In the RA synovial tissue we examined in this report (Figure
5), the proportion of CD4
+CD25
+ cells was relatively low compared with that in other published reports. However, there is inconsistency in the published reports, with several describing an increase in the proportion of CD25
+Foxp3
+CD127
- regulatory T cells in the synovial fluid [
40,
41] and others describing a reduction in synovium compared with synovial fluid [
42].
Competing interests
All authors concur with the submission and assert that the material submitted for publication has not previously been reported and is not under consideration for publication elsewhere.
Authors' contributions
FMB initiated the study, reviewed analyzed data, created figures and wrote the manuscript. NMGS designed the study, acquired data, created figures and wrote the manuscript. SO designed the study, acquired the data, created figures and wrote the manuscript. CL acquired data, created figures and wrote the manuscript. PA acquired the data and created figures. PG acquired data. AA acquired data. ACP acquired data. PH acquired data. AF acquired data. JTB designed the study, analyzed the data, assembled and created figures, and wrote the manuscript. MF designed the study, acquired data and wrote the manuscript.