Background
Rheumatoid arthritis (RA) is a type of symmetric polyarticular arthritis, and T cell-mediated autoimmune responses to the articular synovium play a central role in the pathogenesis of RA [
1,
2]. A marked accumulation of CD4
+ T cells that lack CD28 expression is observed in patients with RA [
3]. CD4
+CD28
− T cells are autoreactive and could be regarded as effector memory T cells [
4]. T-cell activation requires antigen recognition along with a second signal delivered by interactions between the costimulatory receptor on T cells and its ligands on antigen-presenting cells (APCs) [
5]. Costimulatory molecules, including B7/CD28 and tumor necrosis factor (TNF)/tumor necrosis factor receptor (TNFR) superfamily members, are critical in antigen-specific T-cell responses. The absence of a second signal results in anergy and programmed cell death [
6,
7]. However, loss of CD28 does not induce apoptosis in CD4
+ T cells of patients with RA, suggesting that CD28-independent costimulatory pathways are involved in T-cell activation.
OX40 (CD134) and its ligand OX40L (CD252) are members of the TNF/TNFR superfamily. OX40 (TNFR superfamily member 4, TNFRSF4) is predominantly expressed by activated T cells, whereas the primary source of OX40L (TNF superfamily member 4, TNFSF4) is likely to be APCs. OX40 ligation augments the effector differentiation and the survival of memory T cells. The roles of OX40 and OX40L have been examined in most well-established autoimmune models [
8,
9]. Although TNFSF4 polymorphisms were shown not to be linked with RA susceptibility [
10], blockade of OX40/OX40L in collagen-induced arthritis (CIA) mice dramatically ameliorated disease severity [
11,
12]. This indicates that OX40 signal plays a critical role in the development of autoimmune arthritis. As an enhancer of effector memory T cells, OX40 may not be an initiator, but may be an accelerator, in the pathogenesis of RA. Association of OX40 expression with arthritis development should be further investigated. Meanwhile, whether OX40 signal is involved in the activation of CD4
+CD28
− T cells in RA should be clarified.
In view of this, potential correlations were analyzed between the CD4+CD28−OX40+ T-cell subset and clinicopathological characteristics of affected individuals among patients with RA and CIA mice. After cell sorting, the arthritis-associated pathogenic role of CD4+CD28−OX40+ T cells was investigated. Moreover, using OX40 blockade in vitro and in vivo, we revealed regulatory roles of OX40/OX40L signaling in autoimmune arthritis.
Methods
Human subjects
A total of 71 patients with RA diagnosed according to the 2010 rheumatoid arthritis classification criteria were recruited [
13]. Additionally, 44 sex- and age-matched patients with osteoarthritis (OA) and 47 healthy volunteers healthy controls (HCs) were included as control subjects. All subjects were recruited from the Third Affiliated Hospital of Soochow University, Jiangsu, China. Disease activity was evaluated according to the 28-joint Disease Activity Score (DAS28). The level of disease activity could be interpreted as rheumatoid arthritis with low disease activity (Lo-RA), rheumatoid arthritis with moderate disease activity (Mo-RA), rheumatoid arthritis with high disease activity (Hi-RA), or rheumatoid arthritis with remission (Re-RA), according to the European League Against Rheumatism criteria [
14]. Disease stages included early and late RA on the basis of criteria for early RA [
15]. Nine patients with RA received methotrexate (MTX; Shanghai Sine Pharmaceutical Laboratories Co., Shanghai, China) therapy (10 mg/week for 20 weeks by oral administration). None of the patients had received immunosuppressive drugs within 1 year before the study period. Peripheral blood (PB) and synovial fluid (SF) samples were obtained from subjects after informed consent was provided according to a protocol approved by the local ethics committee of Soochow University. Sampling was completed in accordance with the guidelines of the local institutional review boards. Data about the patients and HCs are presented in supplementary figures and tables (Additional file
1: Table S1).
CIA induction and corticosteroid treatment
Male DBA/1 mice at 8–10 weeks of age were obtained from the Shanghai Laboratory Animal Center and were maintained in a specific pathogen-free animal facility at Soochow University. All animal experiments conducted in this study were approved by the University Committee on the Use of Live Animals in Teaching and Research at our institution. Male DBA/1 mice were immunized intradermally at the base of the tail with 200 μg of bovine collagen type II (CII; Chondrex, Redmond, WA, USA) dissolved in 100 μl of 0.05 M acetic acid and mixed with an equal volume of complete Freund’s adjuvant containing heat-killed
Mycobacterium tuberculosis (H37Ra strain, 4 mg/ml; Chondrex). Three weeks later, animals were reimmunized with 200 μg of CII emulsified in incomplete Freund’s adjuvant (Chondrex). Mice were scored for clinical signs as follows (per paw): 0, paws with no swelling; 1, paws with swelling of finger joints or focal redness; 2, paws with mild swelling of the wrist or ankle joints; 3, paws with severe swelling of the entire paw; and 4, paws with deformity or ankylosis. CIA mice were grouped into acute collagen-induced arthritis (A-CIA) and chronic collagen-induced arthritis (C-CIA) stages according to the criteria described by Thornton et al. [
16]. On day 35 after the first immunization, dexamethasone (Dex; Tianjin Pharmaceutical Jiaozuo Co., Tianjin, China) was intraperitoneally injected for 7 days, including low dose (L-dose, 0.5 mg/kg/day), high dose (H-dose, 2 mg/kg/day), and a PBS control.
Sample preparation and flow cytometry
For PB samples, the fluorochrome-labeled monoclonal antibodies (mAbs) antihuman CD4, CD28, and OX40 were added to 80 μl of whole blood before erythrocyte lysis was performed. Synovial fluid mononuclear cells (SFMCs) were isolated after SF samples were treated with hyaluronidase (10 μg/ml; Sigma-Aldrich, St. Louis, MO, USA). The mAbs described above were added to 100-μl SFMC suspensions (2 × 10
6 cells/ml). For CIA mice, fluorochrome-labeled antimouse mAbs were added to 100-μl splenocyte suspensions (1 × 10
6 cells/ml). Cells were incubated and evaluated using a COULTER EPICS XL flow cytometer (Beckman Coulter, Brea, CA, USA). The gating strategy for the CD4
+CD28
−OX40
+ T-cell subset is described in supplementary figures and tables (Additional file
1: Fig. S1). For intracellular staining, peripheral blood mononuclear cells (PBMCs; 3 × 10
6/well) were stimulated with phorbol 12-myristate 13-acetate (PMA, 50 ng/ml; eBioscience, San Diego, CA, USA) and ionomycin (1 μg/ml; eBioscience). Antihuman CD4, CD28, and OX40 mAbs were added before fixation and permeabilization, followed by the addition of phycoerythrin (PE)-cyanine 7 (Cy7)-conjugated antihuman interferon (IFN)-γ (clone 4S.B3; BioLegend, San Diego, CA, USA), interleukin (IL)-4 (clone MP4-25D2; BioLegend), or IL-17A (clone BL168; BioLegend) mAbs. Intracellular cytokine production was assessed using an FC 500 analyzer (Beckman Coulter). Information about all antibodies is provided in supplementary figures and tables (Additional file
1: Table S2).
Adoptive transfer of CD4+CD28−OX40+ T cells
On day 28 after the second immunization, CIA mice were killed, and mononuclear splenocytes were stimulated in vitro with CII (30 μg/ml) for 72 h. CD4+ T cells were selected from mononuclear splenocytes using CD4+ microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) to over 97% purity. After CD28 and OX40 antimouse mAbs were added, T-cell subsets were sorted using a FACSAria™ II cell sorter (BD Biosciences, Franklin Lakes, NJ, USA). Sorted T cells (1 × 106 cells in 200 μl of PBS) were injected intravenously into CIA mice on day 0 after the second immunization. After being fixed and decalcified, mouse ankles were embedded in paraffin and sectioned at 5-μm thickness before being stained with hematoxylin and eosin (H&E).
OX40/OX40L blockage in vitro
On day 28 after the second immunization, mononuclear splenocytes (1 × 105/well) of CIA mice were seeded onto 96-well tissue culture plates (Corning, Corning, NY, USA) in 10% FBS/RPMI 1640 medium and stimulated with CII (30 μg/ml) or anti-CD3 mAb (clone 145-2C11, 1 μg/ml; BioLegend) in the presence of antimouse OX40L mAb (clone RM134L; BioLegend). Rat immunoglobulin G (IgG) (clone RTK4530; BioLegend) was added as a control. After 72 h of incubation, cell proliferation was determined using a Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Mashikimachi, Japan). Supernatant cytokine levels were assessed using a mouse Th1/Th2/Th17 cytokine kit (BD Biosciences) according to the manufacturer’s instructions.
OX40/OX40L blockade in vivo
On days 1–4 after the second immunization, randomized mice were injected intraperitoneally with antimouse OX40L mAb (100 μg/mouse/day) or IgG as controls. In another experiment, mice with similar arthritis scores were treated with anti-OX40L mAb or IgG on days 14–17 after the second immunization. On days 28 and 39 after anti-OX40L mAb early treatment, mononuclear splenocytes (1 × 105/well) were stimulated with CII (30 μg/ml) for 72 h. Cell proliferation, supernatant cytokine levels, and intracellular staining were evaluated as described above. The PE-Cy7-conjugated antimouse mAbs used were as follows: IFN-γ (clone XMG1.2), IL-4 (clone 11B11), TNF-α (clone MP6-XT22), and IL-17A (clone TC11-18H10.1) (all from BioLegend).
Statistical analysis
All statistical analyses were performed using IBM SPSS Statistics for Windows version 20.0 software (IBM, Armonk, NY, USA). All quantitative data are presented as the mean ± SD. Student’s t test or nonparametric Mann-Whitney U test was used for independent samples. A paired-samples t test or a nonparametric Wilcoxon signed-rank test was performed for paired samples. For multiple comparisons, one-way analysis of variance or the Kruskal-Wallis test was performed. A Spearman’s r value was calculated for correlation analyses. A P value less than 0.05 was considered to denote a significant difference.
Discussion
Recently, the self-activation mechanisms of CD4
+CD28
− T cells have been explored, including allelic associations [
17], cytokine production [
18,
19], cell receptor stimulation [
20,
21], and DNA methylation [
22,
23]. However, in the absence of CD28, which costimulatory molecules mediate the second signal is not yet well known. Researchers in a recent study reported that OX40 expression was upregulated on CD4
+CD28
− T cells in patients with acute coronary syndrome and that the secretion of IFN-γ and TNF-α was suppressed by OX40 blockade [
24]. However, whether abnormal OX40 signaling is involved in the self-responses of CD4
+CD28
− T cells in patients with RA remains unclear. In this study, a CD4
+CD28
−OX40
+ T-cell subset was established to be clinicopathologically significant, and its autoreactivation and pathogenicity were demonstrated in patients with RA and CIA mice. Additionally, the regulatory roles of OX40 signaling in CD4
+CD28
− T cells were verified in vitro and in vivo.
In accord with previous studies [
25], abnormal expression of OX40 in patients with RA and CIA mice suggested that an OX40 costimulatory signal was involved in the pathological development of autoimmune arthritis. Further upregulated expression of OX40 in SF of patients with RA implied that OX40 signal may promote abnormal activation of autoreactive T cells after migration of CD4
+ T into the inflammatory microenvironment. In patients with RA, enhanced OX40 expression was observed not only on CD4
+CD28
− but also on CD4
+CD28
+ T cells. OX40 expression levels did not depend on CD28 expression in RA, suggesting that OX40 costimulated T cells in a CD28-independent way. In addition, CD4
+CD28
− T cells with enhanced CD45RO expression suggested the characteristics of memory T cells, and this population is enriched in patients with RA, probably because of OX40 overexpression. Elevated levels of IFN-γ and IL-4 suggest that CD4
+CD28
−OX40
+ T cells may be involved in Th1/Th2 responses and help to produce autoantibody, such as RF [
26]. However, CD4
+CD28
−OX40
− may just function in a Th1 pathway, and for CD4
+CD28
+OX40
+ T cells, a Th2 pathway may be involved. These results implied that CD4
+CD28
−OX40
+ T cells were distinct from the other T-cell subsets and may exhibit specific characteristics in autoimmune arthritis. Upon OX40 costimulation, effector differentiation and long survival may emerge in memory CD4
+CD28
− T cells, which may increase disease activity and lead to a protracted course of disease, as described in another study [
25]. As a consequence of OX40 overexpression, CD4
+CD28
− T cells may differentiate diversely in response to multiple autoantigens, suggesting extraarticular manifestations, as mentioned in previous studies [
3,
27]. Immunosuppressive agents regulate immune responses by changing the expression and function of costimulatory molecules [
28,
29], and this regulatory role could be effectively reversed by OX40-mediated costimulatory signals [
30]. Reduced OX40 signals by MTX or Dex treatment may suppress the autoactivation of CD4
+CD28
− T cells and lead to arthritis remission. However, arthritis relapses after medicine withdrawal suggest that new specific approaches for therapy must be explored.
In adoptive transfer assays, although similar arthritis scores were observed on day 34 in the duration curves, CD4
+CD28
−OX40
+ T cells demonstrated a more powerful pathogenic role than the CD4
+CD28
−OX40
−, CD4
+CD28
+OX40
−, and CD4
+CD28
+OX40
+ T cells in both early (days 1–33) and late (days 35–64) arthritis. This finding indicates that, in early arthritis, activation of CD4
+CD28
−OX40
+ T cells occurs in a short-lived effector- and autoantigen-dependent manner but probably acts in a long-lived memory- and autoantigen-independent fashion in late arthritis, in agreement with the OX40 properties described by Croft [
8,
9]. Probably owing to natural killer cell or chemokine receptors expressed on CD4
+CD28
− T cells [
20,
21], mice that received transfers with CD4
+CD28
−OX40
− T cells exhibited deteriorated arthritis progression, especially in early arthritis. On one hand, this meant that, in early arthritis, additional mechanisms were involved in activation of CD4
+CD28
− T cells, except for OX40 costimulation, On the other hand, long-lived memory characteristics of OX40 [
8,
9] may take effects in late arthritis. Autoantigen-specific CD4
+CD28
−OX40
+ T cells resulted in persistent autoimmune injuries and protracted joint damage. However, the absence of OX40 markedly reduced the survival of memory T cells, and late arthritis development was alleviated in the mice that received transfers with CD4
+CD28
+OX40
− T cells. Recent studies showed that OX40 costimulatory signals promoted differentiation of effector T lymphocytes in a CD28-independent way [
31] and that “superstimulation” of combined CD28 and OX40 did not outperform CD28 by itself [
32]. Additionally, the presence of OX40 may neutralize costimulation of CD28 [
33], which may impair arthritis pathogenic roles of CD4
+CD28
+OX40
+ T cells. A distinct pathogenic role has been further confirmed on the basis of different characteristics of CD28
− and CD28
+ T cells [
4]. The robust pathogenicity of CD4
+CD28
− T cells may reflect OX40 characteristics of costimulating memory T cells in a CD28-independent manner, as described in a report by Demirci et al. [
34].
In vitro different time points resulted in findings inconsistent with those of Yoshioka et al. [
11], probably because of the diverse roles of OX40 in short-lived effector and long-lived memory cells [
8,
9]. In our study, the blockade time point was day 28 after the second immunization, whereas in the study by Yoshioka et al., it was day 7. On day 7, OX40
+ T cells may have occurred in the absence of short-lived effector responses, whereas OX40 memory characteristics may be generated on days 14–28. In vitro OX40 blockade on day 7 may have little effect. On the basis of in vitro studies, early treatment on days 1–4 in vivo showed results similar to those of the studies by Yoshioka et al. [
11] and Gwyer Findlay et al. [
12], probably owing to suppression of short-lived effector responses of OX40
+ T cells. However, inconsistent with the findings of Yoshioka et al. [
11], delayed treatment on days 14–17 also distinctly ameliorated arthritis symptoms, mainly owing to the different time points. On day 14 (blockade time point in our study), generation of memory T cells may be reduced by OX40 blockage, whereas short-lived effector responses resulting from OX40 may have faded on day 7 (blockade time point in study by Yoshioka et al.). Changes in cytokine levels further corroborated the findings in analysis of human samples. In the immunopathogenesis of RA, Th1 and Th2 responses are involved in different phases [
2]. As observed in transfer assays, Th1/Th2-polarized CD4
+CD28
−OX40
+ T cells played central roles in all phases of autoimmune arthritis. However, CD4
+CD28
+OX40
− and CD4
+CD28
+OX40
+ T cells contributed to outcomes only in the early phases of the pathological process. For CD4
+CD28
−OX40
− T cells, although early Th1 response aggravated arthritic development in an OX40-independent way, progression faded in late-phase arthritis.