Objectives To establish if changes in Th1/Th17 cell populations previously reported in experimental arthritis occur in patients with rheumatoid arthritis (RA) treated with anti-tumour necrosis factor α (TNFα) agents, and whether the therapeutic response to anti-TNFα is compromised in patients and mice because of elevated Th17/IL-17 levels. Finally, to assess the efficacy of combined blockade of anti-TNFα and anti-IL-17 in experimental arthritis.
Methods A longitudinal study of two independent cohorts (cohort 1, n=24; cohort 2, n=19) of patients with RA treated with anti-TNFα biological agents was carried out to assess their Th17/IL-17 levels before and after the start of anti-TNFα therapy. IL-12/23p40 production was assessed in plasma Peripheral blood lymphocytes (PBLs) and monocytes. Mice with collagen-induced arthritis (CIA) were treated with anti-TNFα alone, anti-IL17 alone or a combination of the two. Efficacy of treatment and response was assessed from changes in Disease Activity Score 28–erythrocyte sedimentation rate scores in patients, and in clinical scores and histological analysis in CIA.
Results Significant increases in circulating Th17 cells were observed in patients after anti-TNFα therapy and this was accompanied by increased production of IL-12/23p40. There was an inverse relationship between baseline Th17 levels and the subsequent response of patients with RA to anti-TNFα therapy. In addition, PBLs from non-responder patients showed evidence of increased IL-17 production. Similarly, in anti-TNFα-treated mice, there was a strong correlation between IL-17 production and clinical score. Finally combined blockade of TNFα and IL-17 in CIA was more effective than monotherapy, particularly with respect to the duration of the therapeutic effect.
Conclusions These findings, which need to be confirmed in a larger cohort, suggest that a Th17-targeted therapeutic approach may be useful for anti-TNFα non-responder patients or as an adjunct to anti-TNFα therapy, provided that safety concerns can be addressed.
Statistics from Altmetric.com
Tumour necrosis factor α (TNFα) antagonists are very successful in the treatment of patients with rheumatoid arthritis (RA), with approximately two-thirds of patients exhibiting a clinical response to treatment. However, we made the unexpected discovery that in collagen-induced arthritis (CIA), treatment with anti-TNFα leads to an increase in pathogenic Th1 and Th17 cells.1 Mechanistically, the increase in Th1 and Th17 cells in TNFR1−/− mice may be attributable in part to increased production of p40, the common subunit of interleukin (IL)-12 and IL-23, cytokines that are important for the differentiation/survival of Th1 and Th17 cells, respectively.1 ,2
Animal models of arthritis and other autoimmune diseases originally thought of as Th1-mediated diseases were instrumental in identifying Th17 cells as a pathogenic population after the discovery of IL-23, which shares the p40 subunit with IL-12. IL-23 was later identified as a cytokine necessary for the proliferation and maintenance of Th17 cells, and mice lacking either the IL-23p19 or IL-12/23p40 subunits are resistant to disease, whereas those lacking IL-12p35 are not.3 This suggested that Th1 cells may be protective, whereas Th17 cells are the principal pathogenic T cell population.4 In human RA, the presence of Th17 cells in the joints and their enrichment in peripheral blood has only recently been confirmed.5 ,6
Given our previous results showing increased Th1/Th17 responses following TNFα blockade in CIA, we set out to establish the importance of this finding in human RA by determining whether blockade of TNFα leads to changes in numbers of Th1/Th17 cells and p40 production. Using CIA as a tool, we also determined whether the combined blockade of TNFα and IL-17 is more effective than monotherapy, and whether long-term remission can be achieved when the increase in Th17 cells after anti-TNFα is prevented by blockade of IL-17.
Materials and methods
Two independent RA cohorts were recruited to this study (table 1). Patients with active RA were evaluated before and after administration of anti-TNFα therapy. Etanercept (50 mg) was given weekly, adalimumab (40 mg) fortnightly, and infliximab (3 mg/kg) at 0, 2, 4 and 8 week intervals. Peripheral blood was sampled before and 4 and 8–12 weeks after anti-TNFα. Patients were excluded if they had been previously treated with anti-TNFα, had received oral, intra-articular or intramuscular injection of steroids in the 4 weeks preceding the start of anti-TNFα therapy, or had intercurrent, active infections. Nine age- and sex-matched healthy volunteers were also recruited. Ethics approval was given by independent ethics committees before recruitment of patients (RREC 07/8070681 and REC 06/Q0605/8).
Peripheral blood mononuclear cell (PBMC) isolation
Patient blood was collected in tubes with EDTA for the collection of PBMCs and plasma. PBMCs were purified using Ficoll gradient centrifugation (Cedarlane, Ontario, Canada), followed by two washes in phosphate-buffered saline. Cells were counted and seeded at 1.5×106 or 0.75×106 per well and stimulated with Phorbol 12-myristate 13-acetate (PMA) and ionomycin for intracellular fluorescence-activated cell sorting (FACS) analysis or ELISA, respectively.
For the study of monocytes and T lymphocytes, cells were enriched by negative selection using RosetteSep isolation kits (StemCell Technologies, Grenoble, France). Monocytes were stimulated with 1 μg/ml lipopolysaccharide (LPS; Sigma-Aldrich Dorset, UK). T lymphocytes were stimulated with 10 μg/ml plate-bound CD3 monoclonal antibody (clone OKT3) and with soluble CD28 monoclonal antibody (clone 28.2; eBiosciences, Hatfield, UK) for 48 h. Cytokine production was quantified using Meso Scale Discovery (MSD) kits (Meso Scale Diagnostics LLC, Baltimore, Maryland, USA).
For extracellular staining, stimulated or non-stimulated cells were stained with anti-human or anti-mouse CD4 (BD Biosciences)for 30 min at 4°C before fixation with Cytofix (BD Biosciences). For intracellular cytokine staining, cells were stimulated for up to 5 h with 400 pg/ml PMA and 2 mM ionomycin in the presence of 250 pg/ml brefeldin A. After extracellular staining, cells were permeabilised with phosphate-buffered saline containing 1% bovine serum albumin and 0.05% saponin and stained with phycoerythrin-conjugated anti-human or anti-mouse IL-17 (BD Biosciences) for 30 min at room temperature. Cells were acquired and analysed on FACS Canto II using FACSDIVA software (BD Biosciences).
Supernatants from LPS-stimulated PBMCs and blood plasma were collected for analysis of human IL-12/23p40 levels by ELISA. Plates were coated with purified anti-human p40 (BD Biosciences, Oxford, UK). Bound cytokine was detected using biotinylated anti-human p40 (BD Biosciences). Cytokine standards were obtained from Peprotech (London, UK). Mouse IL-17 was detected using a DuoSet from R&D systems (Abingdon, UK) following the manufacturer's instructions.
Comparative mRNA levels of RAR-related orphan receptor gamma (RORγt) were assessed in patients and healthy controls. PBMCs were lysed in RNA Lysis Buffer (RLT) lysis buffer with dithiothreitol, and RNA was extracted using the RNAeasy kit (QIAGEN, Crawley, UK). Reverse transcription was performed using the ABI High capacity reverse transcription system. Amplification of cDNA was performed on an ABI AB7900HT real-time PCR machine. Human acidic ribosomal protein (HuPO) was used as a control. The relative concentration of each gene of interest was calculated using the ΔΔCt method11 and expressed as relative units using a healthy control as a calibrator after normalisation against HuPO.
CIA was induced in DBA/1 and C57BL/6 mice (Harlan, UK) as previously described.7 After the onset of arthritis, mice were treated with anti-TNFα (clone TN3-19.12; Dr R D Schreiber, Washington University St Louis, Michigan, USA) and/or anti-IL17A (eBio17CK15A5; eBiosciences). Disease progression was monitored every day and mice were culled at day 10 after onset. For long-term withdrawal experiments, treatment was stopped at day 10 after therapy, and disease was monitored for 30 days. Clinical assessment was based on the following scoring system: 0 = normal; 1 = slight swelling and/or erythema; 2 = pronounced swelling; 3 = ankylosis. All four limbs were scored, giving a maximum score of 12. All animal experiments were performed according to the guidelines of the Animals Scientific Procedures Act 1986 and the approval of the UK Home Office.
Isolation of murine cells for cytokine analyses
Murine PBMCs were collected from heparinised blood after red cell removal with RBC-lysis buffer (Sigma-Aldrich, Poole, UK). PBMCs were stimulated with 50 μg/ml chicken collagen type II ex vivo for 48 h, and supernatants were collected for cytokine analysis. Joint-infiltrating cells were released from the hind paws of arthritic mice in the presence of 100 pg/ml DNAse and 300 pg/ml Liberase TL (Roche, Welwyn Garden City, UK) for 90 min. Cells were stimulated for 5 h with PMA and ionomycin.
In all statistical analyses, p values are denoted as *p<0.05, **p<0.01, ***p<0.001. Statistical analyses for patients with RA were performed using repeated-measures analysis of variance, followed by Dunnet's multiple comparison test. For the CIA, statistical analyses were performed using one-way analysis of variance except for the withdrawal experiment, which used a repeated-measures analysis of variance followed by Bonferroni multiple comparisons test. Analysis of monocyte-derived dendritic cells (MoDCs) (online supplementary figure 1) used the Wilcoxon matched pair t test.
Patient characteristics and response to therapy
Two cohorts of anti-TNFα-naive patients with RA who were about to start therapy were recruited from Imperial College Healthcare NHS Trust (24 patients; cohort 1) and Barts and the London NHS Trust (19 patients; cohort 2). Blood was collected before treatment, and then at 4 and 8–12 weeks after the start of treatment. However, it was not possible to perform all of the assays on all samples. The baseline patient characteristics and responses to therapy are shown in table 1. Using EULAR response criteria, we classified 31 patients as good to moderate responders (collectively termed hereafter as ‘responders’) and 12 as non-responders.
TNFα blockade increases Th17 responses in RA
We previously reported that TNFα blockade, or genetic deletion of tumour necrosis factor receptor 1 (TNFR1), in mice with CIA results in an increase in Th17 and Th1 cells in the draining lymph nodes1 and peripheral blood (unpublished data). This led us to investigate whether the use of TNFα inhibitors would result in a comparable increase in Th1/Th17 cells in the blood of patients with RA. A group of age- and sex-matched healthy volunteers (n=9) was also bled to establish the normal range of Th1/Th17 cells in blood. PBMCs were stimulated with PMA/ionomycin to determine numbers of Th1/Th17 cells by FACS. In parallel, enriched T lymphocytes were stimulated with anti-CD3+anti-CD28 antibodies to quantify interferon (IFN)γ and IL-17 production. The first finding to emerge was that the percentage of Th17 cells (figure 1A), but not Th1 cells (data not shown), was ∼2.5 times higher in patients with RA than in healthy controls. This is in agreement with previous findings6 and provides clear evidence of heightened Th17 responses in patients with RA. Next, we found that the percentage and total number of Th17 cells after 4 weeks of therapy was significantly increased, and further increased after 8–12 weeks (figure 1A,B). Numbers of Th1 cells and levels of IFNγ did not change significantly (data not shown). We therefore focused further work exclusively on Th17 responses.
In agreement with the increased number of Th17 cells, a trend towards increased production of IL-17 by anti-CD3+anti-CD28-stimulated T cells was observed at 4 and then 8–12 weeks after therapy, although it did not reach statistical significance (figure 1C). Similarly, there was a progressive trend towards increased levels of mRNA encoding the Th17-specific transcription factor, RORγt, after anti-TNFα treatment (figure 1D).
Despite increased Th17 responses in the first cohort after anti-TNFα treatment, 22 out of 24 patients responded to treatment at 4 weeks, and this response was sustained at 8–12 weeks in 18 of the 24 patients (table 1).
TNFα blockade increases p40 production in vitro and in vivo
Our reported findings in mice showed that the increase in Th17 cells observed after anti-TNFα therapy was partly due to increased production of p40.1 ,2 This was based on observations that TNFα inhibits p40 production in vitro and that blockade of p40 inhibits the expansion of Th17 cells in vivo. In order to establish the relevance of these findings in man, MoDCs from 19 healthy donors were treated with adalimumab, etanercept or rTNFα before stimulation with LPS. In agreement with our findings in mouse dendritic cells (not shown), treatment of MoDCs with adalimumab or etanercept resulted in a significant increase in p40 production, while treatment with rTNFα caused a significant reduction (online supplementary figure 1).
Having established that TNFα blockade increases p40 in human antigen presenting cells (APCs) in vitro, we next asked whether exposure of PBLs from patients with RA to TNFα inhibitors in vivo influenced their capacity to produce p40 upon LPS stimulation. Hence, p40 production before and after anti-TNFα was assessed in three different cell preparations: whole blood, PBMCs and purified monocytes. There were significant increases in p40 in whole blood after 4 and 8–12 weeks of therapy (figure 2A). In PBMCs, there was a progressive trend towards increased p40 levels, which reached significance after 8–12 weeks of therapy (figure 2B). Most strikingly, there was a fourfold increase in p40 production in purified monocytes after 8–12 weeks of therapy (figure 2C). These results confirm that TNFα blockade in RA results in upregulation of IL-12/23p40 production in vitro and in vivo.
Evidence for increased IL-17 and p40 production in anti-TNFα non-responder patients
The degree of response to treatment with TNFα inhibitors is variable, with ∼30% of patients with RA failing to exhibit a clinically meaningful response. We therefore hypothesised that enhanced IL-17 production is a contributory factor to suboptimal responses. In order to test this hypothesis, patients were divided into responders and non-responders, and IL-17 production by PBMCs was measured after stimulation with anti-CD3+anti-CD28. A trend towards increased IL-17 production was observed in non-responders after 8–12 weeks of therapy (figure 3A). In a separate study, we measured p40 production by non-stimulated (figure 3B) or LPS-stimulated (figure 3C) whole blood cultures and detected a significant increase in p40 expression in LPS-stimulated blood from non-responders after 8–12 weeks of therapy.
We also addressed the question of whether the number of Th17 cells at the start of therapy would predict the degree of response to anti-TNFα. A significantly negative correlation was found between percentage Th17 cells in peripheral blood before the start of anti-TNFα therapy and Disease Activity Score for 28 joints (DAS28)–erythrocyte sedimentation rate (ESR) score at 4 weeks after therapy, although this was lost by 8–12 weeks after therapy (data not shown).
Response to anti-TNFα in CIA correlates inversely with antigen-specific IL-17 production
The high variability between patients with RA led us to investigate the correlation between the response to anti-TNFα and the levels of IL-17 in the CIA model, which also shows variability in the response to anti-TNFα in the setting of a genetically homogeneous population. C57BL/6 mice were used for this study because this strain exhibits a more robust and sustained T cell response to type II collagen than DBA/1 mice.8 Mice with established CIA were treated for 10 days with monoclonal antibody against TNFα (250 µg/mouse every second day) or with isotype control antibody. The results showed that anti-TNFα-treated mice fell into one of two distinct groups: those that showed a reduction in clinical score after therapy (defined as responders) and those that showed no change or an increase in the clinical score (non-responders) (figure 4A). On day 10, IL-17 production by type II collagen-stimulated PBMCs was measured. In agreement with the data from patients with RA, IL-17 production was significantly higher in non-responder than in responder mice (figure 4B). More importantly, among the anti-TNFα-treated mice, there was a strong correlation (R2=0.67; p<0.0019) between IL-17 production and clinical score on day 10 of arthritis (figure 4C). We also observed an increase in the percentage of CD4+IL-17+ cells in the joints of non-responder mice (figure 4D), and a moderate correlation between the clinical score at day 10 and the number of CD4+IL-17+ cells in both the joints and blood of arthritic mice (not shown).
Synergistic effects of anti-TNFα and anti-IL-17 in CIA
The findings described above suggest that the anti-TNFα-mediated induction of Th17/IL-17 detracts from its therapeutic efficacy and that combined blockade of TNFα and IL-17 would have an additive or synergistic effect. We therefore assessed the outcome of combination therapy in CIA using suboptimal doses of anti-TNFα and anti-IL-17. We have already established that a dose of 300 µg/mouse anti-TNFα was optimal, whereas a dose of 50 µg/mouse was suboptimal in CIA.9 DBA/1 mice with established CIA were treated for 10 days with anti-TNFα (50 µg/mouse) alone, anti-IL-17 alone (50 µg/mouse), anti-TNFα plus anti-IL-17 or control antibodies. Anti-TNFα alone or anti-IL-17 alone failed to significantly ameliorate the disease, whereas combined anti-TNFα/anti-IL-17 therapy profoundly reduced clinical score and paw swelling (online supplementary figure 2). On this basis, it was concluded that anti-TNFα and anti-IL-17 acted synergistically at doses that were ineffective when used as monotherapy.
Finally, the curative potential of combination therapy was studied in chronic (non-resolving) CIA induced by immunising C57BL/6 mice with type II collagen.7 ,8 Anti-TNFα alone and/or anti-IL-17 was administered to arthritic mice for 10 days and then stopped. The mice were then monitored for an additional 30 days. After therapy had been stopped, arthritis severity increased rapidly in mice treated with anti-TNFα alone (figure 5A), probably because of the ability of anti-TNFα to increase pathogenic Th17 cells in the draining lymph nodes.1 Similarly, in some of the mice treated with anti-IL-17 alone, arthritis severity increased gradually from around 10 days after cessation of therapy and was still increasing at the end of the study, whereas the arthritis in some anti-IL17-treated mice was fairly stable in the long term. However, in all but one mouse treated with anti-TNFα plus anti-IL-17, disease was almost completely eliminated until day 40 after onset. Histological analysis of affected paws on day 40 confirmed the protective effect of combination therapy (figure 5B,D), which was associated with reduced antigen-specific responses (figure 5C). We conclude that combined anti-TNFα/anti-IL-17 therapy has a sustained therapeutic effect in chronic CIA in C57BL/6 mice.
This study has revealed that circulating Th17 cells are increased in patients with RA compared with healthy controls. Furthermore, we report the important observation that Th17 percentage and numbers increase following anti-TNFα therapy despite favourable clinical responses in the majority of patients. These findings are in agreement with published reports showing increased production of IL-17 in patients with RA treated with anti-TNFα agents and may also explain why reductions in disease activity are generally incomplete and that some patients relapse on cessation of treatment.10 ,11 While this may question the role of Th17 cells in initiating arthritis, it is possible that their pathogenicity only reaches its full impact in the presence of TNFα. This is supported by the observation that IL-17 induces TNFα production by myeloid cells12 and synergises with TNFα in a number of cellular systems.13,–,15
Consistent with the findings in CIA, the results of the present study also reveal an increase in p40 production by myeloid cells from patients with RA after anti-TNFα therapy.1 However, the chief difference between the mouse and human data is that Th1 cells were induced by anti-TNFα in the lymph nodes of mice with CIA but not in the blood of patients with RA, although we did observe a statistically insignificant increase in IFNγ production (data not shown). One possible explanation for this is provided by the finding that IL-12 receptors are downregulated on T cells after anti-TNFα therapy.16
A potentially important observation in patients with RA to emerge from this study is the significant inverse correlation between numbers of Th17 cells at the start of anti-TNFα therapy and response to therapy at 4 weeks (figure 3D), which raises the possibility that numbers of circulating Th17 cells could predict poor response to therapy. However, further studies will be needed to confirm this finding in a larger cohort of patients.
The inverse relationship between Th17 cell activity and response to anti-TNFα therapy was more clearly confirmed in mice with CIA in which we detected a positive correlation between IL-17 production and clinical score (figure 4C), although further studies will be required to confirm these findings and establish the mechanisms involved. While we did detect a weak to moderate correlation between Th17 cells and clinical score, the strongest relationships emerged when we looked at antigen-specific responses. In contrast with CIA, which occurs in response to a known antigenic stimulus in genetically identical mice, RA is a syndrome with a range of gene associations and heterogeneous autoantibody and clinical expression. In RA blood the induction of Th17 cells and IL-17 production ex vivo required mitogenic and T cell receptor stimulation, respectively. Thus, it is possible that the lack of a strong correlation between the response to anti-TNFα and Th17/IL-17 in patients with RA is due to the use of a non-specific stimulus to activate T cells.
Anti-TNFα treatment of CIA did not, however, lead to exacerbation of arthritis beyond the level observed in control mice and generally led to a reduction in the severity of disease, indicating that TNFα continues to play a proinflammatory role in spite of the enhancement in Th17 responses. Nevertheless, our results in mouse and man suggest that a full and durable response to TNFα inhibitors in RA may be compromised by their ability to induce pathogenic Th17 cells. This led us to test the therapeutic outcome of using anti-TNFα/anti-IL-17 as combination therapy in CIA. When administered at suboptimal doses, the combination was found to act in a synergistic fashion (online supplementary figure 2), and there was evidence of a long-term therapeutic benefit that lasted for at least 30 days after cessation of therapy (figure 5A). These findings are in agreement with the study of Koenders et al,17 who showed that overexpression of soluble IL-17 receptor and TNF binding protein in CIA by gene delivery resulted in more joint destruction in combination than the IL-17 overexpression alone, supporting the therapeutic advantage of targeting both IL-17 and TNFα.
Early data suggest that anti-IL-17 therapy is effective in human RA,18 and our study supports further evaluation of this form of therapy, particularly for patients who show a poor response to anti-TNFα. Our demonstration of enhanced therapeutic effects from combined anti-TNFα/anti-IL-17 therapy in CIA raises the possibility of combination biologic therapy for RA, a suggestion that has been supported by at least two human studies19 ,20 —although the potential benefits would need to be weighed carefully against increased infectious risk. However, if it can be shown that combination therapy results in reduced frequency of re-administration, then this would have clear advantages over monotherapy. Another important finding from the mouse studies was that combination therapy allows dose reduction. Thus, it may be possible to inhibit inflammation by administering anti-TNFα with anti-IL-17 at doses that avoid total blockade of either cytokine, allowing their physiological functions to be maintained.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
RW and RM are joint senior authors.
Correction notice This article has been updated since it was published Online First. The statement ‘RW and RM are joint senior authors’ has been added.
Acknowledgements This project was funded by the Medical Research Council. The authors are grateful for support from the NIHR Biomedical Research Centre funding scheme, Arthritis Research UK, and the Kennedy Institute of Rheumatology Trustees, and a research grant from Wyeth Pharmaceuticals (Pfizer).
Contributors SA designed and wrote the manuscript. SA, AP and KM performed experiments in cohort 1 and all animal work. SMA recruited patients to cohort 1 and provided intellectual input. TET performed experiments for cohort 2. DH, EP and AK helped recruit patients for cohort 1. AJ recruited patients to cohort 2. PT, RW and RM are principal investigators on the project and provided intellectual input.
Funding Medical Research Council, UK.
Competing interests None.
Ethics approval Ethics approval was given by independent ethics committees before recruitment of patients (RREC 07/8070681 and REC 06/Q0605/8).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.