DOI 10.1002/art.27304 2010, American College of Rheumatology Fluoxetine and Citalopram Exhibit Potent Antiinflammatory Activity in Human and Murine Models of Rheumatoid Arthritis and Inhibit Toll-like Receptors Sandra Sacre, Mino Medghalchi, Bernard Gregory, Fionula Brennan, and Richard Williams Objective. Selective serotonin reuptake inhibitors (SSRIs), in addition to their antidepressant effects, have been reported to have antiinflammatory effects. The aim of this study was to assess the antiarthritic potential of 2 SSRIs, fluoxetine and citalopram, in murine collagen-induced arthritis (CIA) and in a hu- man ex vivo disease model of rheumatoid arthritis (RA). Methods. Following therapeutic administration of SSRIs, paw swelling was assessed and clinical scores were determined daily in DBA/1 mice with CIA. Joint architecture was examined histologically at the end of the treatment period. Cultures of human RA synovial membranes were treated with SSRIs, and cytokine pro- duction was measured. Toll-like receptor (TLR) func- tion was examined in murine and human macrophages, human B cells, and human fibroblast-like synovial cells treated with SSRIs. Results. Both SSRIs significantly inhibited disease progression in mice with CIA, with fluoxetine showing the greatest degree of efficacy at the clinical and histo- logic levels. In addition, both drugs significantly inhib- ited the spontaneous production of tumor necrosis factor, interleukin-6, and interferon-inducible pro- tein 10 in human RA synovial membrane cultures. Fluoxetine and citalopram treatment also inhibited the signaling of TLRs 3, 7, 8, and 9, providing a potential mechanism for their antiinflammatory action. Conclusion. Fluoxetine and citalopram treatment selectively inhibit endosomal TLR signaling, ameliorate disease in CIA, and suppress inflammatory cytokine production in human RA tissue. These data highlight the antiarthritic potential of the SSRI drug family and provide further evidence of the involvement of TLRs in the pathogenesis of RA. The SSRIs may provide a template for potential antiarthritic drug development. Rheumatoid arthritis (RA) is a chronic auto- immune inflammatory disease affecting 1% of the pop- ulation worldwide. It is characterized by a destructive inflammation of the joints, leading to progressive dis- ability and reduced life expectancy. The synovial mem- brane is infiltrated by immune cells, including macro- phages and T cells, resulting in the chronic production of proinflammatory cytokines and matrix metalloprotein- ases (MMPs), leading to cartilage and bone degradation (1). The mechanisms responsible for the perpetuation of the chronic inflammation associated within the RA joint are currently unknown. However, a family of evolution- ary conserved innate immune receptors, the Toll-like receptors (TLRs), has been suggested to contribute to this process (2). TLRs recognize and respond to pathogens and endogenous molecules that are potentially released during tissue inflammation and damage (3,4). They are type I membrane receptors that are mainly found at the cell surface, except for TLRs 3, 7, 8, and 9, which are localized to the endosome (5). Most data have indicated a role of TLR-4 in murine models of experimental arthritis (68). In human studies, most TLRs have been detected in rheumatoid tissue (913) in addition to potential TLR ligands (14,15). RA synovial tissue re- Supported by the Arthritis Research Campaign, the Kennedy Institute of Rheumatology Trustees, the European Commission Col- laborative Masterswitch grant, and by the Biomedical Research Centre funding program of the National Institute for Health Research. Sandra Sacre, PhD (current address: Brighton and Sussex Medical School, University of Sussex, Brighton, UK), Mino Medghal- chi, MSc, Bernard Gregory, PhD, Fionula Brennan, PhD, Richard Williams, PhD: Kennedy Institute of Rheumatology, Imperial College of Science, Technology, and Medicine, London, UK. Dr. Sacre has filed a patent application for the use of selective serotonin reuptake inhibitors in the treatment of arthritis (no financial benefit to declare). Dr. Brennan has received consulting fees, speaking fees, and/or honoraria from Wyeth (less than $10,000). Address correspondence and reprint requests to Sandra Sacre, PhD, Brighton and Sussex Medical School, University of Sussex, Falmer, Brighton BN1 9RY, UK. E-mail: s.sacre@bsms.ac.uk. Submitted for publication August 11, 2009; accepted in revised form November 15, 2009. 683 moved during elective surgery produces high levels of multiple cytokines and MMPs in cultures (11). Using this tissue, we have previously demonstrated that the spontaneous production of cytokines is partly dependent on TLR signaling (11). In subsequent studies, we iden- tified a role of TLR-8 in inducing the production of tumor necrosis factor (TNF). That study also showed that a serotonin receptor antagonist, mianserin, inhib- ited TLRs 3, 7, 8, and 9 in primary human cells and significantly decreased the production of TNF and interleukin-6 (IL-6) in human RA synovial cell cultures (10), suggesting a potential role of 1 or more of these TLRs in the pathogenesis of RA. We had initially become interested in selective serotonin reuptake inhibitors (SSRIs) because of their reported antiinflammatory effects (16). Increasing evi- dence has highlighted a link between the immune system and the symptoms of depression. This was originally referred to as the macrophage theory of depression, suggesting an association with inflammatory cytokines (17). It has since been shown that patients with depres- sion have elevated blood levels of cytokines, as com- pared with healthy controls, and that these levels are reduced upon treatment with SSRIs (18,19). In patients who failed to respond to SSRIs, no reduction in cytokine levels was observed (20), which suggests a connection between SSRIs, depression, and the immune system. In the present study, our aim was to investigate whether the antiinflammatory properties of 2 SSRIs, fluoxetine and citalopram, would be beneficial in murine and human disease models of RA and to determine whether these drugs could potentially work by a mech- anism similar to that of mianserin, through inhibition of TLR-induced cytokine production. MATERIALS AND METHODS Reagents. Cell culture reagents used were penicillin/ streptomycin, RPMI 1640, and Dulbeccos modified Eagles medium (DMEM) obtained from Cambrex (Verviers, Bel- gium), indomethacin from Sigma (St. Louis, MO), and fetal bovine serum (FBS) from PAA Laboratories (Linz, Austria). The TLR ligands used were chloroform-extracted Escherichia coli lipopolysaccharide (LPS), resiquimod (R-848), poly(I-C), CpG-containing oligonucleotide (ODN; 2006 and 1668), and imiquimod from InvivoGen (San Diego, CA). Flagellin (puri- fied) and Pam 3 Cys-Ser(Lys) 4 3HCl (Pam 3 CSK 4 ) were from Alexis (Nottingham, UK). Fluoxetine hydrochloride and cita- lopram hydrobromide were purchased from Sigma (Poole, UK). Macrophage colony-stimulating factor (M-CSF) was purchased from PeproTech (London, UK). Human CD19 microbeads were purchased from MACS Miltenyi Biotec (Bisley, UK). All reagents were tested for LPS using the Limulus amebocyte lysate assay from Cambrex (Walkersville, MD) (21) and found to have no detectable levels of LPS. Cell culture. RA synovial membrane cells were iso- lated from samples obtained from patients undergoing joint replacement surgery, as previously described (22,23). Immedi- ately after isolation, cells were cultured at 1 10 5 cells/well in 96-well tissue culture plates (Falcon, Oxford, UK) in RPMI 1640 containing 10% (volume/volume) FBS and 100 units/ml of penicillin/streptomycin. All patients gave written informed consent, and the study was approved by the Local Ethics Committee. Human macrophages were preincubated for 30 min- utes with 1 or 5 g/ml of fluoxetine or with 10, 20, or 30 g/ml of citalopram and then either left unstimulated or stimulated with 10 ng/ml of flagellin, 10 ng/ml of Pam 3 CSK 4 , 1 g/ml of R-848 or 10 ng/ml of LPS for 6 hours. Human RA synovial fibroblasts were preincubated with media alone or containing 1 or 5 g/ml of fluoxetine or with 10, 20, or 30 g/ml of citalopram and then either left unstimulated or stimulated with 20 g/ml of poly(I-C) for 24 hours. Primary human B cells were preincubated for 30 minutes with 1 or 5 g/ml of fluoxetine or with 10, 20, or 30 g/ml of citalopram and then either left unstimulated or stimulated with 10 g/ml of imi- quimod or 2.5 M ODN2006 for 24 hours. Human RA synovial membrane cells were cultured for 24 hours in the presence of media alone or 5 g/ml of fluoxetine, 30 g/ml of citalopram, or 10 or 100 M serotonin. Primary human synovial fibroblasts and peripheral blood monocytes were isolated and cultured as previously described (2426). Macrophages were derived from monocytes after differ- entiation for 4 days in the presence of 100 ng/ml of M-CSF. B lymphocytes were obtained by using CD19 microbeads accord- ing to the manufacturers instructions. Immediately after iso- lation, cells were cultured at 2 10 5 cells/well in 96-well tissue culture plates in RPMI 1640 containing 5% (v/v) FBS and 100 units/ml of penicillin/streptomycin. Murine bone marrowderived macrophages were de- rived from the femurs of male DBA/1 mice. Macrophages were cultured for 6 days with DMEM containing FCS (20% v/v), 100 units/ml of penicillin/streptomycin, 2-mercaptoethanol (50 M), and 10 ng/ml of M-CSF. Enzyme-linked immunosorbent assay (ELISA). Sand- wich ELISAs were used to measure human TNF, interferon- inducible protein 10 (IP-10), and IL-6 (R&D Systems, London, UK). Sandwich ELISAs were also used to measure murine TNF, RANTES, IL-12, and anticollagen IgG1 and IgG2a (BD PharMingen, San Diego, CA). Absorbance was read on a spectrophotometric ELISA plate reader (Labsystems Multiscan Biochromic) and was analyzed using Ascent soft- ware V2.6 (both from Thermo Labsystems, Cambridge, UK). Cell viability was not significantly affected over this time period, as evaluated by MTT assay (Sigma, Poole, UK) (27). CIA model. Male DBA/1 mice (812 weeks of age) were immunized subcutaneously at the base of the tail with 200 g of type II collagen emulsified in Freunds complete adjuvant (Difco, West Molesey, UK). The onset of arthritis was considered to be the day that erythema and/or swelling was first observed, and each limb of the arthritic mice was given a clinical score on a scale of 03, where 0 normal, 1 slight erythema and/or swelling, 2 pronounced edematous swell- 684 SACRE ET AL ing, and 3 joint deformity with ankylosis (maximum score 12 per animal). Fluoxetine (10 or 25 mg/kg) and citalopram (25 mg/kg) treatments were administered daily for 7 days starting on the day of arthritis onset. Paw swelling was assessed daily by measuring hindpaw thickness using calipers. All measurements were recorded in a blinded manner. Histologic assessment of arthritis. On completion of the experiment, the first mouse limb that was observed to show evidence of arthritis was processed for histologic assessment. The limb was removed, fixed, decalcified, and paraffin- embedded before sectioning and staining with hematoxylin and eosin. Sagittal sections of the proximal interphalangeal joint of the middle digit were examined by microscopy in a blinded manner. The histologic severity of arthritis was graded on a scale of 03, where 0 normal, 1 minimal synovitis, cartilage loss, and bone erosions limited to discrete foci, 2 synovitis and erosions present, but normal joint architecture intact, and 3 synovitis, extensive erosions, and disrupted joint architecture. Statistical analysis. The mean, SD, SEM, and statisti- cal significance were calculated using GraphPad version 3 software (GraphPad Software, San Diego, CA). For statistical analysis, a 1-tailed t-test of paired data was used with a 95% confidence interval. The SEM was used for pooled experimen- tal data, whereas the SD was used in graphs showing repre- sentative experiments. A 1-tailed Mann-Whitney test was used with a 95% confidence interval for the CIA data. P values less than 0.05 were considered significant. RESULTS Fluoxetine halts disease progression in the mu- rine CIA model. To investigate whether the reported antiinflammatory effect of fluoxetine would be benefi- cial in a murine model of experimental arthritis, we decided to use the CIA model. This model was chosen because of its histologic similarities to human RA, with comparable synovitis, bone erosion, and pannus forma- tion, and because it responds similarly to anti-TNF therapy (28). Administration of fluoxetine was given therapeutically after the onset of clinical arthritis (day 1 of arthritis). Mice given 10 mg/kg of fluoxetine showed a small reduction in the clinical score and a slower increase in paw swelling (Figures 1A and B). At the higher dose Figure 1. Fluoxetine inhibition of disease progression and paw swelling in the mouse model of collagen-induced arthritis (CIA). Mice with CIA were given an intraperitoneal injection of vehicle (phosphate buffered saline [PBS]; controls) or fluoxetine (10 mg/kg or 25 mg/kg) once a day for 7 days. A and B, Mice were assessed for clinical score (n 1415 mice per group) (A) and paw thickness (n 1113 mice per group) (B) on a daily basis. C, Mice were bled on day 7, and serum levels of IgG1 and IgG2a anticollagen antibodies (n 11 mice per group) (left) and interleukin-12 (IL-12) (n 67 mice per group) (right) were measured by enzyme-linked immunosorbent assay. D, Arthritis was scored histologically as described in Materials and Methods (n 11 mice per group). Values in AD are the mean SEM of pooled data from 2 separate experiments. Values for IL-12 represent individual mice; horizontal lines show the mean. P 0.05; P 0.01; P 0.001 for fluoxetine 25 mg/kg versus controls. E, Histologic features of arthritic joints from control mice and mice treated with 25 mg/kg fluoxetine on day 7 of treatment. Photomicrographs of tarsometatarsal joints (top) and proximal interphalangeal joints (bottom) show protection from inflammation and joint erosion in fluoxetine-treated mice versus control mice (original magnification 5 in top panels; 10 in bottom panels). ANTIARTHRITIC POTENTIAL OF TWO SSRIs IN HUMAN AND MURINE MODELS OF RA 685 (25 mg/kg), fluoxetine profoundly halted disease pro- gression, with no further elevation in the clinical score or paw swelling (Figures 1A and B). Neither dose had a significant effect on the serum anticollagen IgG1 and IgG2a levels, but at the 25 mg/kg dose, serum IL-12 was inhibited by 83.5 9.9% (P 0.0256) (Figure 1C). The higher dose of fluoxetine (25 mg/kg) showed a signifi- cant reduction in the mean histology score, inhibiting it by 67 11.4% (P 0.0011) (Figure 1D). Vehicle control sections showed extensive inflammatory cell infiltration, bone erosion accompanied by pannus for- mation, and degradation of cartilage. In contrast, in mice treated with fluoxetine, particularly at the 25 mg/kg dose, there was reduced inflammation, reduced cartilage and bone erosion, and the joint architecture was largely maintained (Figure 1E). Citalopram treatment confers partial protection from CIA. Citalopram also reduced the severity of CIA but was not as effective as fluoxetine. Thus, mice given 25 mg/kg of citalopram showed a significant benefit in the clinical score from day 2 onward, but there was no improvement in paw swelling (Figures 2A and B). Citalopram had no appreciable effect on serum anti- collagen IgG1 and IgG2a levels (Figure 2C). Examina- tion of histology sections revealed a tendency toward reduced inflammatory cell infiltration, pannus forma- tion, and loss of joint architecture in the citalopram- treated group as compared with the control group. However, the reduction in the mean histology score (by 46.2 11.9%) did not reach statistical significance (P 0.0857) (Figure 2D), largely due to the low numbers of animals treated. Fluoxetine and citalopram inhibit cytokine pro- duction in murine bone marrowderived macrophages induced by TLRs 3, 7, and 9. Next, we sought to identify the mechanism by which SSRIs could suppress inflam- mation. We had previously shown that mianserin, a serotonin receptor antagonist, inhibited TLRs 3, 7, 8, and 9, but not TLRs 2, 4, and 5, in primary human cells (10). Given the structural and functional similarity of this class of molecule with the SSRIs, we decided to investigate whether fluoxetine and citalopram could also inhibit these TLRs. TLR-8 is not thought to be func- tional in the murine system since it does not respond to single-stranded RNA (ssRNA) or other TLR-8 ligands (29). For this reason, we chose to investigate TLRs 3, 7, and 9 using activation of TLR-4 as a control. In murine bone marrowderived macrophages, Figure 2. Citalopram inhibition of disease progression and paw swelling in the mouse model of collagen-induced arthritis (CIA). Mice with CIA were given an intraperitoneal injection of vehicle (phosphate buffered saline [PBS]; controls) or citalopram (25 mg/kg) once a day for 7 days. A and B, Mice were assessed for clinical score (n 7 mice per group) (A) and paw thickness (n 7 mice per group) (B) on a daily basis. P 0.05; P 0.01 for citalopram 25 mg/kg versus controls. C, Mice were bled on day 7, and serum levels of IgG1 and IgG2a anticollagen antibodies (n 5 mice per group) were measured by enzyme-linked immunosorbent assay. D, Arthritis was scored histologically as described in Materials and Methods (n 46 mice per group). Values are the mean SEM. 686 SACRE ET AL both fluoxetine and citalopram were able to inhibit in a dose-dependent manner cytokine production induced by poly(I-C) (TLR-3), R-848 (TLR-7), and CpG (TLR-9) (Figure 3), whereas LPS (TLR-4) activation was unaf- fected (Figures 3A and B). Fluoxetine inhibited TLR-7 and TLR-9induced TNF production in murine macro- phages by 83 2.894% (P 0.0006) and 53.6 16.7% (P 0.0425), respectively (Figure 3A). Citalopram reduced the amount of TNF released by 70 19% (P 0.0333) for TLR-7 and by 97 0.7% (P 0.0001) for TLR-9 (Figure 3B). TLR-3 activation with poly(I-C) did not induce TNF production, but it did stimulate RANTES production in these cells. This was inhibited by 84 5% (P 0.0002) with citalopram treatment and by 96.4 2% with fluoxetine treatment (P 0.0001) (Figure 3C). No reduction in cell viability was observed. Fluoxetine and citalopram inhibit cytokine pro- duction induced by TLR-3 and TLR-8 in primary hu- man M-CSF macrophages and RA synovial fibroblasts. To establish the relevance of these findings in human disease, we further examined whether fluoxetine and citalopram were also able to inhibit endosomal TLR signaling in primary human cells. Neither fluoxetine nor citalopram had any appreciable effect on Pam 3 CSK 4 (TLR-1/2), LPS (TLR-4), or flagellin (TLR-5) activity (Figures 4A and B), but both treatments selectively inhibited (by 55.7 4.8% [P 0.0007] for fluoxetine and by 80.2 7.3% [P 0.0041] for citalopram) the R-848 (TLR-7/8 ligand)induced production of TNF in human macrophages in a dose-dependent manner (Fig- ures 4A and B). Macrophages do not induce cytokines in response Figure 3. Fluoxetine and citalopram inhibition of Toll-like receptor 3 (TLR-3), TLR-7, and TLR-9 induction of cytokines from murine bone marrowderived macrophages. A and B, Primary murine macrophage colony- stimulating factor (M-CSF)derived macrophages were preincubated for 30 minutes with 2.5, 5, or 10 g/ml of fluoxetine (A) or with 10, 20, or 30 g/ml of citalopram (B) and were then either left unstimulated (US) or were stimulated for 6 hours with 1 g/ml of R-848, 2 M CpG-containing oligonucleotide (ODN; 1668), or 100 ng/ml of lipopolysaccharide (LPS). The dose reponse of tumor necrosis factor (TNF) inhibition (left and middle) was measured, and the data from the highest dose of each selective serotonin reuptake inhibitor (SSRI) were pooled from 3 separate experiments (right). C, Primary murine M-CSFderived macrophages were preincubated for 30 minutes with 10, 20, or 30 g/ml of citalopram or with 2.5, 5, or 10 g/ml of fluoxetine and were then either left unstimulated or were stimulated for 24 hours with 20 g/ml of poly(I-C). The dose response of RANTES inhibition (left and middle) was measured, and the data from the highest dose of each SSRI were pooled from 3 separate experiments (right). Dose response data are the mean and SD of triplicate cultures and are representative of 3 separate experiments. Pooled data are the mean and SEM percentage of the ligand-only response. P 0.05; P 0.001 versus ligand-only response. ANTIARTHRITIC POTENTIAL OF TWO SSRIs IN HUMAN AND MURINE MODELS OF RA 687 to TLR-7 ligands; thus, in these cells, R-848 is activating TLR-8 signaling. Human macrophages do not produce TNF in response to activation of TLR-3; however, poly(I-C) activation of TLR-3 on RA synovial fibroblasts induces a strong IL-6 response (30). Fluoxetine and cita- lopram inhibited TLR-3induced IL-6 production in RA synovial fibroblasts by 56.8 6% (P 0.0012) and by 92.7 4% (P 0.0009), respectively (Figure 4C). Primary human macrophages and RA synovial fibroblasts do not respond to TLR-7 or TLR-9 ligands (31); Figure 4. Fluoxetine and citalopram inhibition of Toll-like receptor 3 (TLR-3) and TLR-8 induction of cytokines from primary human macrophages. A and B, Primary human macrophages were preincubated for 30 minutes with 1 or 5 g/ml of fluoxetine (A) or with 10, 20, or 30 g/ml of citalopram (B) and were then either left unstimulated (US) or were stimulated for 6 hours with 1 g/ml of R-848, 10 ng/ml of Pam 3 Cys-Ser(Lys) 4 3HCl (Pam 3 CSK 4 ), 10 ng/ml of flagellin, or 10 ng/ml of lipopolysaccharide (LPS). The dose response of tumor necrosis factor (TNF) inhibition (left) was measured, and the data from the highest dose of each selective serotonin reuptake inhibitor (SSRI) were pooled from 3 separate experiments (right). C, Rheumatoid arthritis synovial fibroblasts were preincu- bated with media alone or with media containing 1 or 5 g/ml of fluoxetine or containing 10, 20, or 30 g/ml of citalopram and were then either left unstimulated or were stimulated for 24 hours with 20 g/ml of poly(I-C). The dose response of interleukin-6 (IL-6) inhibition (left) was measured, and the data from the highest dose of each SSRI were pooled from 3 separate experiments (right). Dose response data are the mean and SD of triplicate cultures and are representative of 3 separate experiments using cells from 3 unrelated donors. Pooled data from these donors are the mean and SEM percentage of the ligand-only response. P 0.01; P 0.001 versus ligand-only response. 688 SACRE ET AL thus, to investigate these TLRs, we used B cells that produce IL-6 in response to imiquimod (TLR-7) and CpG DNA (TLR-9) (32). Fluoxetine and citalopram signifi- cantly inhibited TLR-7 and TLR-9induced production of IL-6 from B cells in a dose dependent manner (Figure 5). Fluoxetine inhibited TLR-7induced IL-6 production by 80 7.2% (P 0.004) and inhibited TLR-9induced IL-6 production by 92 4% (P 0.001) (Figure 5A). Citalopram inhibited TLR-7induced IL-6 production by 72 11.8 (P 0.0044) and inhibited TLR-9induced IL-6 production by 81 15.3% (P 0.017) (Figure 5B). Cell viability was measured after all experiments by MTT assay, and no toxicity was observed (data not shown). Fluoxetine and citalopram inhibit spontaneous cytokine production in human RA synovial membrane cultures. The success of citalopram and, in particular, fluoxetine in the murine CIA model, in addition to the ability of these drugs to inhibit endosomal TLR activity in both murine and human cells, made them attractive candidates for testing in human RA synovial membrane cultures. We have previously identified a potential role of the endosomal TLRs and, in particular, TLR-8 in the production of TNF and IL-6 in RA synovial membrane cultures (10). These cultures are produced from RA tissues removed during elective surgery and are com- posed of a mixed population of cells that spontaneously release cytokines without the need for exogenous stim- ulation (22). This is considered an accepted model of human disease, and this model was used for the initial studies that identified the importance of TNF in RA (22). Fluoxetine at 5 g/ml inhibited TNF by 50.23 9.7% (P 0.0032), IL-6 by 18.4 22.3% (P 0.0043), and IP-10 by 54 15.5% (P 0.0035) (Figure 6A). Citalopram at 30 g/ml was more effective at inhibiting TNF by 69.6 5.6% (P 0.0032), IL-6 by 33 21.2% (P 0.0067), and IP-10 by 63.6 7.7% (P 0.0043) (Figure 6B). Fluoxetine and citalopram are both SSRIs, and we therefore wished to ensure that any inhibition ob- served was not due to the effects of serotonin on the cells. In addition to the important function of serotonin in the central nervous system, there is also a role for this Figure 5. Fluoxetine and citalopram inhibition of Toll-like receptor 7 (TLR-7) and TLR-9 production of interleukin-6 (IL-6) from primary human B cells. A and B, Primary human B cells were preincubated for 30 minutes with 1 or 5 g/ml of fluoxetine (A) or with 10, 20, or 30 g/ml of citalopram (B) and were then either left unstimulated (US) or were stimulated for 24 hours with 10 g/ml of imiquimod or 2.5 M CpG-containing oligonucleotide (ODN; 2006). Dose responses (left) are the mean and SD of triplicate cultures and are represen- tative of 3 separate experiments using cells from 3 unrelated donors. Pooled data from these donors (right) are the mean and SEM percentage of the ligand-only response. Values are the mean and SD. P 0.05; P 0.01; P 0.001 versus ligand alone. ANTIARTHRITIC POTENTIAL OF TWO SSRIs IN HUMAN AND MURINE MODELS OF RA 689 molecule in the periphery. Serotonin receptors and serotonin reuptake receptors are expressed on many of the cells within the immune system, including macro- phages, B cells, and T cells (33). Addition of 10 M or 100 M serotonin to the RA synovial membrane cul- tures had no appreciable effect on the production of TNF or IL-6 (Figure 6C). Cell viability was measured after all experiments by MTT assay, and no toxicity was observed (Figure 6D). DISCUSSION In this study of 2 licensed SSRIs, fluoxetine and citalopram, we provide data demonstrating a therapeutic benefit of these drugs, both in the murine CIA model and in a validated human RA disease tissue model. In addition, we suggest a potential mechanism for at least part of the antiinflammatory action observed, perhaps through inhibition of the endosomal TLRs. Fluoxetine powerfully halted any further increase in the clinical and histopathologic severity of established disease in the CIA model, demonstrating a remarkable inhibition of inflammation. Citalopram, although effective, was not as successful as fluoxetine at inhibiting disease pathogene- sis in this model. However, this may be due to the reported, distinctly longer half-life of fluoxetine in vivo (34). Citalopram affected the clinical score to a greater extent than paw swelling. This is probably because the clinical score takes into account the number of paws. Therefore, it is a more sensitive measure of overall disease severity than is paw swelling. The precise mechanism for the antiinflammatory action of the SSRIs is unknown. However, there have been reports of fluoxetine suppressing T cell prolifera- tion (35) and inhibiting interferon- (IFN) production in human whole blood cultures (36). In the CIA model, the levels of serum IL-12 were significantly inhibited by fluoxetine treatment, suggesting that in addition to the reported ability to influence T cell proliferation, fluox- etine may also be inhibiting the response of antigen- presenting cells. TLRs are strong activators of inflam- matory cytokines and have been suggested to play a role in RA (37), with most studies indicating a role of TLR-4 in models of experimental arthritis (68). Upon investi- gation, fluoxetine and citalopram were found to inhibit Figure 6. Fluoxetine and citalopram inhibition of spontaneous cytokine production in human rheumatoid synovial membrane cultures. AC, Rheumatoid synovial membrane cells were cultured for 24 hours in the presence of media alone, 5 g/ml of fluoxetine (n 1012 donors) (A), 30 g/ml of citalopram (n 911 donors) (B), or 10 or 100 M serotonin (n 5 donors) (C), and the production of tumor necrosis factor (TNF), interleukin-6 (IL-6), and interferon-inducible protein 10 (IP-10) was determined. D, Cell viability was measured by MTT assay. Results are representative of all experiments performed in A and B. Each data point in A and B represents an individual donor. P 0.01 versus unstimulated controls. Values in C and D are the mean and SEM. OD optical density. 690 SACRE ET AL the activation of endosomal TLRs 3, 7, and 9, but not TLR-4, in murine bone marrowderived macrophages. Production of type I IFN by activation of TLRs 3 and 7, however, has been reported to enhance TLR-4 cytokine production in monocyte-derived dendritic cells (38). Thus, inhibition of the endosomal TLRs in the CIA model by fluoxetine and citalopram treatment may, in turn, affect TLR-4induced cytokine production in vivo. Similar to the observation in the CIA model, fluoxetine was more effective than citalopram at inhibiting endo- somal TLR activation. This would be expected if inhibi- tion of endosomal TLRs forms part of the mechanism by which these drugs suppressed the disease activity in the CIA model. Inhibition of the endosomal TLRs was also evi- dent in primary human cells, where both drugs inhibited TLRs 3, 7, 8, and 9, but not TLRs 1/2, 4, or 5. The lack of inhibition of TLR-4 is consistent with a separate study performed in whole blood (35). In a previous study, we discovered that mianserin, a serotonin receptor antago- nist, is also an inhibitor of TLRs 3, 7, 8, and 9, and suppressed the spontaneous production of TNF and IL-6 in human RA synovial membrane cultures (10). Here, we demonstrated that fluoxetine and citalopram also inhibit TNF, IL-6, and IP-10 production in these cultures, supporting a concept of activation of 1 or more of these TLRs in the pathogenesis of RA. This inhibition of cytokines was not due to effects of increased seroto- nin, since the addition of serotonin to the RA cultures had no effect. Furthermore, the percentages of inhibi- tion of TNF and IL-6 were similar to the percentage inhibition observed in our previous study with mianserin. This may suggest a common mechanism that is indepen- dent of the serotonergic system, since SSRIs and sero- tonin receptor antagonists act on different receptors. Other studies have also suggested that the antiinflam- matory mechanism of SSRIs is independent of mono- amine transporter blockade (35). There is increasing evidence in support of a potential role of 1 or more of these TLRs in RA. It has recently been demonstrated that TLR-3 can be activated on synovial fibroblasts by necrotic cells, as would be found in the RA joint (39). More recently, activation of TLR-3 on RA fibroblasts was reported to promote osteoclastogenesis (40). We have previously identified TLR-8 as a contributor to the TNF production in human RA cultures, although TLR-8 may not play a role in the CIA model because murine TLR-8 is not considered to be functional (29). However, murine TLR-7 responds in a similar way to human TLR-8; both receptors are activated by the same natural ligand ssRNA (29) and both can activate NF-B and produce TNF in macro- phages. Human TLR-7 is not active on macrophages, but is instead expressed on dendritic cells, where it induces IFN production (41). It is therefore conceivable that TLR-7 may have more importance in the CIA model. It has recently been reported that cross-tolerance of TLRs 2, 7, and 9 induced by daily administration of low doses of a TLR-7 ligand has proved beneficial in a murine serum-transfer model of arthritis, reducing cell infiltration and bone erosion (42). A role of TLR-9 may be more predominant on B cells, where activation by DNA-containing immune complexes has been shown to stimulate rheumatoid factorpositive autoreactive B cells (43). More clinically relevant support for a role of TLRs in RA has come from the therapeutic use of chloroquine and from studies of patients deficient in endosomal TLR function. Chloroquine has been used for many years in combination with other therapies in a clinical setting as a treatment for RA (44). It is an inhibitor of endosomal TLR function, and we have previously demonstrated that chloroquine significantly inhibits cytokine production in human RA synovial membrane cultures (10). It has been suggested that the benefit observed clinically may be mediated through the inhibition of these TLRs (45). Additional support for a role of endosomal TLRs in autoimmunity has emerged from the identification of patients deficient in UNC93B1 (46), a protein required for TLR-3, TLR-7, TLR-8, and TLR-9 signaling (47). These patients show an unexpect- edly mild phenotype (48,49), but have an increased number of polyreactive and autoreactive B cells in the periphery (46), similar to that observed in RA patients (50). Intriguingly, none have progressed to develop autoimmunity (46), which suggests that activation of 1 or more of these TLRs may be required to induce an autoimmune phenotype. The mechanism by which fluoxetine and citalo- pram inhibit endosomal TLRs seems unlikely to be linked to the antidepressant mechanism of SSRIs. Flu- oxetine and citalopram have a high affinity for the serotonin transporter, with a K d of 0.81 nM and 1.16 nM, respectively (51). In the present study, TLR-induced cytokine production was inhibited in vitro by concentra- tions equivalent to 7.530 M fluoxetine (2.510 g/ml) and 2574 M citalopram (1030 g/ml), which are much higher than the concentrations required for SSRI activity or the concentrations that would be achieved in the serum of patients. Fluoxetine and citalopram are normally prescribed for depression at a maximum dos- age of 80 mg/day, which produces a serum concentration ANTIARTHRITIC POTENTIAL OF TWO SSRIs IN HUMAN AND MURINE MODELS OF RA 691 of 1 M and 250 nM for the 2 SSRIs, respectively (52). In the CIA model, suppression of disease progres- sion also required much higher doses (25 mg/kg, as compared with 1 mg/kg used to treat depression in humans). Moreover, the addition of serotonin to the human RA cultures had no effect on the spontaneous release of TNF. It would therefore seem more likely that the mechanism is an off-target effect of these drugs. A possible mechanism by which fluoxetine and citalopram inhibit endosomal TLRinduced cytokine production could be through the inhibition of cellular signaling molecules shared by the endosomal TLRs. All of the TLRs (except TLR-3) use the myeloid differenti- ation factor 88dependent pathway to activate NF-B. It is doubtful, however, that these drugs inhibit a shared signaling molecule on this pathway, since there was no inhibition of TLR-2, TLR-4, or TLR-5. In addition, TLR-3 signals through a different pathway, using TRIF to induce IFN (30). An alternative mechanism could be through direct interaction with the receptors, or possi- bly, a shared accessory molecule required for their activation, for example UNC93B1. This protein has been shown to be required for trafficking of the endosomal TLRs from the endoplasmic reticulum to endosomes, where they are then activated (53). The data presented herein demonstrate the abil- ity of fluoxetine and citalopram to selectively inhibit endosomal TLRs, to decrease inflammation in the mu- rine CIA model, and to decrease inflammatory cytokine production in human RA tissue. Although these drugs will undoubtedly have multiple effects in vivo, the data suggest that targeting the endosomal TLRs may provide therapeutic benefit in RA. Fluoxetine and citalopram are not ideal candidates to be progressed into clinical trials, since our in vitro data suggest that effective inhibition would require levels above their safe thera- peutic dosages. Accordingly, there does not appear to be any published anecdotal evidence of a clinical benefit in disease pathogenesis in RA patients prescribed SSRIs for depression. Nevertheless, these data in conjunction with our previous study support the concept of a contri- bution from endosomal TLRs to the chronic inflamma- tion associated with the pathogenesis of RA and dem- onstrate the potential to selectively target these receptors with small molecules in the future for the effective treatment of RA. ACKNOWLEDGMENTS We would like to thank Renee Best for preparation of the human RA synovial tissue and Marc Feldmann for critical reading of the manuscript. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Sacre had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Sacre, Medghalchi, Gregory, Brennan, Williams. Acquisition of data. Sacre, Medghalchi. Analysis and interpretation of data. Sacre, Medghalchi, Gregory, Williams. REFERENCES 1. Feldmann M, Brennan FM, Maini RN. Role of cytokines in rheumatoid arthritis. Annu Rev Immunol 1996;14:397440. 2. Drexler S, Sacre SM, Foxwell BM. Toll-like receptors: a new target in rheumatoid arthritis? Expert Rev Clin Immunol 2006;2:58599. 3. Kawai T, Akira S. The roles of TLRs, RLRs and NLRs in pathogen recognition. Int Immunol 2009;21:31737. 4. Wagner H. Endogenous TLR ligands and autoimmunity. Adv Immunol 2006;91:15973. 5. Uematsu S, Akira S. Toll-like receptors and innate immunity. J Mol Med 2006;84:71225. 6. Popa C, Abdollahi-Roodsaz S, Joosten LA, Takahashi N, Sprong T, Matera G, et al. Bartonella quintana lipopolysaccharide is a natural antagonist of Toll-like receptor 4. Infect Immun 2007;75: 48317. 7. Abdollahi-Roodsaz S, Joosten LA, Roelofs MF, Radstake TR, Matera G, Popa C, et al. Inhibition of Toll-like receptor 4 breaks the inflammatory loop in autoimmune destructive arthritis. Arthri- tis Rheum 2007;56:295767. 8. Choe JY, Crain B, Wu SR, Corr M. Interleukin 1 receptor dependence of serum transferred arthritis can be circumvented by toll-like receptor 4 signaling. J Exp Med 2003;197:53742. 9. Iwahashi M, Yamamura M, Aita T, Okamoto A, Ueno A, Ogawa N, et al. Expression of Toll-like receptor 2 on CD16 blood monocytes and synovial tissue macrophages in rheumatoid arthri- tis. Arthritis Rheum 2004;50:145767. 10. Sacre SM, Lo A, Gregory B, Simmonds RE, Williams L, Feldmann M, et al. Inhibitors of TLR8 reduce TNF production from human rheumatoid synovial membrane cultures. J Immunol 2008;181: 80029. 11. Sacre SM, Andreakos E, Kiriakidis S, Amjadi P, Lundberg A, Giddins G, et al. The Toll-like receptor adaptor proteins MyD88 and Mal/TIRAP contribute to the inflammatory and destructive processes in a human model of rheumatoid arthritis. Am J Pathol 2007;170:51825. 12. Radstake TR, Roelofs MF, Jenniskens YM, Oppers-Walgreen B, van Riel PL, Barrera P, et al. Expression of Toll-like receptors 2 and 4 in rheumatoid synovial tissue and regulation by pro- inflammatory cytokines interleukin-12 and interleukin-18 via interferon-. Arthritis Rheum 2004;50:385665. 13. Seibl R, Birchler T, Loeliger S, Hossle JP, Gay RE, Saurenmann T, et al. Expression and regulation of Toll-like receptor 2 in rheumatoid arthritis synovium. Am J Pathol 2003;162:12217. 14. Foell D, Wittkowski H, Roth J. Mechanisms of disease: a DAMP view of inflammatory arthritis. Nat Clin Pract Rheumatol 2007;3: 38290. 15. Midwood K, Sacre S, Piccinini AM, Inglis J, Trebaul A, Chan E, et al. Tenascin-C is an endogenous activator of Toll-like receptor 4 that is essential for maintaining inflammation in arthritic joint disease. Nat Med 2009;15:77480. 16. Maes M. The immunoregulatory effects of antidepressants. Hum Psychopharmacol 2001;16:95103. 692 SACRE ET AL 17. Smith RS. The macrophage theory of depression. Med Hypothe- ses 1991;35:298306. 18. Tsao CW, Lin YS, Chen CC, Bai CH, Wu SR. Cytokines and serotonin transporter in patients with major depression. Prog Neuropsychopharmacol Biol Psychiatry 2006;30:899905. 19. Basterzi AD, Aydemir C, Kisa C, Aksaray S, Tuzer V, Yazici K, et al. IL-6 levels decrease with SSRI treatment in patients with major depression. Hum Psychopharmacol 2005;20:4736. 20. OBrien SM, Scully P, Fitzgerald P, Scott LV, Dinan TG. Plasma cytokine profiles in depressed patients who fail to respond to selective serotonin reuptake inhibitor therapy. J Psychiatr Res 2007;41:32631. 21. Levin J, Bang FB. The role of endotoxin in the extracellular coagulation of Limulus blood. Bull Johns Hopkins Hosp 1964;115: 26574. 22. Brennan FM, Chantry D, Jackson A, Maini R, Feldmann M. Inhibitory effect of TNF antibodies on synovial cell interleukin-1 production in rheumatoid arthritis. Lancet 1989;2:2447. 23. Brennan FM, Chantry D, Jackson AM, Maini RN, Feldmann M. Cytokine production in culture by cells isolated from the synovial membrane. J Autoimmun 1989;2 Suppl:17786. 24. Butler DM, Feldmann M, Di Padova F, Brennan FM. p55 and p75 tumor necrosis factor receptors are expressed and mediate com- mon functions in synovial fibroblasts and other fibroblasts. Eur Cytokine Netw 1994;5:4418. 25. Foxwell B, Browne K, Bondeson J, Clarke C, de Martin R, Brennan F, et al. Efficient adenoviral infection with IB reveals that macrophage tumor necrosis factor production in rheuma- toid arthritis is NF-B dependent. Proc Natl Acad Sci U S A 1998;95:82115. 26. Buchan G, Barrett K, Turner M, Chantry D, Maini RN, Feldmann M. Interleukin-1 and tumour necrosis factor mRNA expression in rheumatoid arthritis: prolonged production of IL-1. Clin Exp Immunol 1988;73:44955. 27. Mosmann T. Rapid colorimetric assay for cellular growth and survival: application to proliferation and cytotoxicity assays. J Im- munol Methods 1983;65:5563. 28. Holmdahl R, Bockermann R, Backlund J, Yamada H. The molec- ular pathogenesis of collagen-induced arthritis in mice: a model for rheumatoid arthritis. Ageing Res Rev 2002;1:13547. 29. Heil F, Hemmi H, Hochrein H, Ampenberger F, Kirschning C, Akira S, et al. Species-specific recognition of single-stranded RNA via toll-like receptor 7 and 8. Science 2004;303:15269. 30. Lundberg A, Drexler SK, Monaco C, Williams LM, Sacre SM, Feldmann M, et al. Key differences in TLR3/poly I:C signaling and cytokine induction by human primary cells: a phenomenon absent from murine cell systems. Blood 2007;110:324552. 31. Kyburz D, Rethage J, Seibl R, Lauener R, Gay RE, Carson DA, et al. Bacterial peptidoglycans but not CpG oligodeoxynucleotides activate synovial fibroblasts by Toll-like receptor signaling. Arthri- tis Rheum 2003;48:64250. 32. Poeck H, Wagner M, Battiany J, Rothenfusser S, Wellisch D, Hornung V, et al. Plasmacytoid dendritic cells, antigen, and CpG-C license human B cells for plasma cell differentiation and immunoglobulin production in the absence of T-cell help. Blood 2004;103:305864. 33. Mossner R, Lesch KP. Role of serotonin in the immune system and in neuroimmune interactions. Brain Behav Immun 1998;12:24971. 34. Van Harten J. Clinical pharmacokinetics of selective serotonin reuptake inhibitors. Clin Pharmacokinet 1993;24:20320. 35. Diamond M, Kelly JP, Connor TJ. Antidepressants suppress production of the Th1 cytokine interferon-, independent of monoamine transporter blockade. Eur Neuropsychopharmacol 2006;16:48190. 36. Kubera M, Lin AH, Kenis G, Bosmans E, van Bockstaele D, Maes M. Anti-inflammatory effects of antidepressants through suppres- sion of the interferon-/interleukin-10 production ratio. J Clin Psychopharmacol 2001;21:199206. 37. ONeill LA. Primer: Toll-like receptor signaling pathways: What do rheumatologists need to know? Nat Clin Pract Rheumatol 2008;4:31927. 38. Roelofs MF, Wenink MH, Brentano F, Abdollahi-Roodsaz S, Oppers-Walgreen B, Barrera P, et al. Type I interferons might form the link between Toll-like receptor (TLR) 3/7 and TLR4- mediated synovial inflammation in rheumatoid arthritis (RA). Ann Rheum Dis 2009;68:148693. 39. Brentano F, Schorr O, Gay RE, Gay S, Kyburz D. RNA released from necrotic synovial fluid cells activates rheumatoid arthritis synovial fibroblasts via Toll-like receptor 3. Arthritis Rheum 2005;52:265665. 40. Kim KW, Cho ML, Oh HJ, Kim HR, Kang CM, Heo YM, et al. TLR-3 enhances osteoclastogenesis through upregulation of RANKL expression from fibroblast-like synoviocytes in patients with rheumatoid arthritis. Immunol Lett 2009;124:917. 41. Kadowaki N, Ho S, Antonenko S, Malefyt RW, Kastelein RA, Bazan F, et al. Subsets of human dendritic cell precursors express different toll-like receptors and respond to different microbial antigens. J Exp Med 2001;194:8639. 42. Hayashi T, Gray CS, Chan M, Tawatao RI, Ronacher L, Mc- Gargill MA, et al. Prevention of autoimmune disease by induction of tolerance to Toll-like receptor 7. Proc Natl Acad Sci U S A 2009;106:27649. 43. Viglianti GA, Lau CM, Hanley TM, Miko BA, Shlomchik MJ, Marshak-Rothstein A. Activation of autoreactive B cells by CpG dsDNA. Immunity 2003;19:83747. 44. Suarez-Almazor ME, Belseck E, Shea B, Homik J, Wells G, Tugwell P. Antimalarials for treating rheumatoid arthritis. Coch- rane Database Syst Rev 2000;4:CD000959. 45. Kyburz D, Brentano F, Gay S. Mode of action of hydroxychloro- quine in RA: evidence of an inhibitory effect on toll-like receptor signaling. Nat Clin Pract Rheumatol 2006;2:4589. 46. Isnardi I, Ng YS, Srdanovic I, Motaghedi R, Rudchenko S, von Bernuth H, et al. IRAK-4-and MyD88-dependent pathways are essential for the removal of developing autoreactive B cells in humans. Immunity 2008;29:74657. 47. Tabeta K, Hoebe K, Janssen EM, Du X, Georgel P, Crozat K, et al. The Unc93b1 mutation 3d disrupts exogenous antigen presentation and signaling via Toll-like receptors 3, 7 and 9. Nat Immunol 2006;7:15664. 48. Casrouge A, Zhang SY, Eidenschenk C, Jouanguy E, Puel A, Yang K, et al. Herpes simplex virus encephalitis in human UNC-93B deficiency. Science 2006;314:30812. 49. Sancho-Shimizu V, Zhang SY, Abel L, Tardieu M, Rozenberg F, Jouanguy E, et al. Genetic susceptibility to herpes simplex virus 1 encephalitis in mice and humans. Curr Opin Allergy Clin Immunol 2007;7:495505. 50. Samuels J, Ng YS, Coupillaud C, Paget D, Meffre E. Impaired early B cell tolerance in patients with rheumatoid arthritis. J Exp Med 2005;201:165967. 51. Tatsumi M, Groshan K, Blakely RD, Richelson E. Pharmacolog- ical profile of antidepressants and related compounds at human monoamine transporters. Eur J Pharmacol 1997;340:24958. 52. Reis M, Aamo T, Spigset O, Ahlner J. Serum concentrations of antidepressant drugs in a naturalistic setting: compilation based on a large therapeutic drug monitoring database. Ther Drug Monit 2009;31:4256. 53. Kim YM, Brinkmann MM, Paquet ME, Ploegh HL. UNC93B1 delivers nucleotide-sensing toll-like receptors to endolysosomes. Nature 2008;452:2348. ANTIARTHRITIC POTENTIAL OF TWO SSRIs IN HUMAN AND MURINE MODELS OF RA 693