Objective: To determine whether synovial expression of triggering receptor expressed on myeloid cells 1 (TREM-1) is upregulated in patients with distinct types of inflammatory or non-inflammatory arthritis.
Methods: Synovial fluid (SF) samples were analysed for levels of soluble TREM-1 (sTREM; n = 132), tumour necrosis factor alpha (TNFα, n = 78) and leucocyte TREM-1 messenger RNA (n = 48). Synovial tissue from four rheumatoid arthritis (RA) patients, two patients with Crohn’s-associated arthritis, one patient with ankylosing spondylitis and one patient with osteoarthritis were examined for TREM-1 expression by immunohistology, and three of the RA samples were also analysed by Western blotting.
Results: Synovial fluid sTREM-1 levels in septic arthritis and RA were similar to each other and were each greater than those in gouty arthritis, non-septic/non-RA inflammatory arthritis and non-inflammatory arthritis. Synovial fluid TNFα and sTREM-1 levels correlated with each other, and sTREM-1 and leucocyte TREM-1 mRNA levels each correlated with SF leucocyte counts. TREM-1 in RA was expressed in situ in synovial tissue by cells of myelomonocytic lineage but was not detectably expressed in control osteoarthritis synovial tissue.
Conclusions: Synovial TREM-1 expression is increased in septic arthritis and RA. In patients with acute inflammatory arthritis, elevated SF sTREM-1 levels may point the clinician to a diagnosis of septic arthritis or RA. In RA patients, targeting TREM-1 may have therapeutic benefits by reducing local proinflammatory cytokine and chemokine release.
Statistics from Altmetric.com
Many conditions of widely differing aetiologies give rise to effusions within joint cavities. Inflammatory synovial effusions, regardless of aetiology, are characterised by the recruitment of numerous inflammatory cells into the synovial space mediated by, and resulting in, elevated levels of multiple cytokines and/or chemokines. Differences in synovial fluid (SF) cytokine concentrations among patients with discrete inflammatory and non-inflammatory arthritides have been described but have not been universally appreciated.1 2 Identification of some cytokines that could reliably distinguish among the various arthritides could potentially have great diagnostic utility and might afford new insights into disease pathogenesis.
Triggering receptor expressed on myeloid cells 1 (TREM-1) is a recently described member of the immunoglobulin superfamily expressed on neutrophils and a subset of monocytes and tissue macrophages.3 4 5 When bound to its as yet unidentified ligand(s), the transmembrane domain of TREM-1 associates with DAP12, and in turn triggers the secretion of inflammatory mediators such as tumour necrosis factor alpha (TNFα), IL-8, granulocyte macrophage colony-stimulating factor, monocyte chemotactic protein 1 and myeloperoxidase.3 4 5 6 Whereas crosslinking of TREM-1 alone induces modest cellular activation and proinflammatory cytokine secretion, TREM-1 acts synergistically with receptors for pathogen-associated molecular patterns.3 4 5 7 This results in considerably greater cytokine secretion than the sum of the responses induced by either TREM-1 or pathogen-associated molecular pattern-mediated activation alone. Although early studies did not reveal increased TREM-1 expression in tissue samples from subjects with non-infectious inflammatory conditions,4 5 more recent publications have implicated TREM-1-mediated inflammation in both acute and chronic non-infectious inflammatory disorders.8 9 10
In addition to the membrane-bound form, a soluble form of TREM-1 (sTREM-1) has been detected in mouse and human sera,11 12 and is suggested to arise from the proteolytic cleavage of membrane-bound TREM-1 by matrix metalloproteinases.13 In humans, sTREM-1 has been detected in plasma from volunteers injected with lipopolysaccharide and from patients with sepsis.14 15 Moreover, sTREM-1 has been advocated as a diagnostic marker of infection for patients with pneumonia and as a prognostic marker for patients with septic shock.16 17
In this study, we examined SF sTREM-1 levels in patients with infectious or distinct non-infectious inflammatory arthritis and in patients with non-inflammatory synovial effusions. As expected, we detected elevated SF sTREM-1 levels in septic arthritis. Unexpectedly, we also detected similarly elevated SF sTREM-1 levels in rheumatoid arthritis (RA) patients along with substantial in-situ TREM-1 expression on myelomonocytic cells in synovial tissue.
Materials and methods
The study protocol was approved by the Institutional Review Board of the University of Southern California Keck School of Medicine, and all subjects provided informed consent.
Patients requiring diagnostic and/or therapeutic arthrocentesis at the Los Angeles County + University of Southern California Medical Center were recruited for this study. A joint effusion was considered to be “inflammatory” if the SF leucocyte count was greater than 2000 cells/mm3. Underlying causes of joint effusions were determined by the medical history, physical examination and SF analyses (cell count, Gram stain, culture and crystal examination). Each patient’s demographics at the time of the arthrocentesis and the final diagnoses are listed in table 1. Four patients with osteoarthritis and one patient with a traumatic effusion had SF leucocyte counts modestly greater than 2000 cells/mm3 (“minimally inflammatory”) but were, nonetheless, included in the non-inflammatory cohort due to the underlying clinical nature of the joint effusion. Of the 18 SF samples from patients diagnosed as having septic arthritis, 14 were caused by Staphylococcus aureus, two by group A Streptococcus, one by Escherichia coli and one by Propionibacterium acnes. A definitive diagnosis could not be made in five patients with inflammatory arthritis and were labelled as unknown.
SF sTREM-1 level determination
SF was centrifuged within 1–2 h following arthrocentesis to pellet the cells. Cell-free SF was analysed for sTREM-1 levels by a recently developed sandwich ELISA, the methods for which have previously been described in detail.17 All measurements were performed in duplicate and in a blinded fashion.
Western blotting for SF sTREM-1
Protein samples were resolved by electrophoresis on a denaturating 4–12% acrylamide gel, followed by transfer to a PVDF membrane and probing with an anti-TREM-1 primary antibody (monoclonal mouse IgG1, 0.5 μg/ml, clone 193015; R&D, Minneapolis, Minnesota, USA). Polyclonal horseradish peroxidase (HRP)-labelled goat anti-mouse Ig (1/2000; Dako, Carpinteria, California, USA) was used as a secondary antibody. Binding was detected by chemoluminescence using a SuperSignal West Pico Kit (Pierce, Rockford, Illinois, USA).
SF leucocyte TREM-1 mRNA level determination
Total RNA/messenger RNA was extracted from the SF pellet using RNA STAT-60 isolation reagent (Tel-Test, Friendswood, Texas, USA) and complementary DNA was generated using the First Strand cDNA Synthesis Kit (Fermentas, Hanover, Maryland, USA). Real-time PCR analysis was performed with the Cyber Green-based real-time quantitative PCR technique using SuperArray (Frederick, Maryland, USA) TREM-1 and β-actin (control) primer sets and assay conditions and was analysed on the ABI PRISM 7900HT system (Applied Biosystems, Foster City, California, USA). Results are presented as the ratio of TREM-1 mRNA to β-actin mRNA, which was calculated by the formula: (2 exp (Ctβ-actin − CtTREM-1)).
SF TNFα level determination
Cell-free SF was analysed for TNFα levels by the solid-phase ELISA method (BD Biosciences, Le Pont de Claix, France). The sensitivity of the technique allows the detection of levels as low as 2 pg/ml. All measurements were performed in duplicate and in a blinded fashion.
Synovial tissue from four RA patients, two patients with Crohn’s-associated arthritis, one patient with ankylosing spondylitis and one patient with knee osteoarthritis were collected during joint surgery, snap-frozen in liquid nitrogen, and stored at −80°C. Five micrometre-thick frozen serial sections were cut at −30°C, collected on clean glass slides and fixed for 30 s in ethanol. Immunohistochemistry was performed with the Dako Immunocytomaton (Envision HRP kit; Dako, Glostrup, Denmark) protocol and reagents. After rehydration in phosphate buffered saline (PBS), endogenous peroxidase was blocked for 15 minutes at room temperature. The sections were then incubated with unconjugated monoclonal antibodies to CD33, CD14 and HLA-DR (Beckman Coulter, Miami, Florida, USA) or TREM-1 (R&D) for 45 minutes at room temperature in a moist chamber. After three washes in PBS, a second incubation was performed with a polyclonal HRP anti-mouse antibody for 45 minutes at room temperature. After another three washes in PBS, the sections were developed with a diaminobenzidine solution in substrate buffer for 5 minutes at room temperature, washed in distilled water and ammonia and counterstained with haematoxylin.
Western blotting of synovial tissue extracts
Synovial tissue from three RA patients was extracted with the RIPA lysis buffer (Santa Cruz Biotechnology, Santa Cruz, California, USA) supplemented with protease inhibitors. The clear protein extract was subjected to sodium dodecylsulphate–polyacrylamide gel electrophoresis and Western blotting by blocking the membrane with 5% bovine serum albumin and incubating for 1 h with anti-TREM-1 (R&D) in PBS-Tween followed by HRP anti-mouse (Dako). Chemoluminescence substrate was used to reveal the bands (Chemoluminescence reagent Plus; NEN Life Science, Boston, Massachusetts, USA) for 5 minutes, which were then impressed in the dark on X-Omat blue film (Sigma-Aldrich, St Louis, Missouri, USA). After developing, the film was read with a Biorad Gel Doc XR (Biorad, Hercules, California, USA).
All analyses were performed using SigmaStat software (SPSS). Results that did not follow a normal distribution were log-transformed to achieve normality. Parametric testing between two matched or unmatched groups was performed by the paired or unpaired t test, respectively. When log transformation failed to generate normally distributed data or the equal variance test was not satisfied, non-parametric testing was performed between two groups by the Mann–Whitney rank sum test. Correlations were determined by Pearson product moment correlation for interval data and by Spearman rank order correlation for ordinal data or for interval data that did not follow a normal distribution. Analyses are presented for each cohort of patients described in table 1. Additional analyses were also performed with exclusion of the five “non-inflammatory” patients who had “minimally inflammatory” effusions and with re-assignment of these patients to the “other inflammatory” cohort. In neither case did changes in statistical significance emerge.
Elevated SF sTREM-1 levels in septic arthritis and RA
As illustrated in fig 1A, SF sTREM-1 levels in septic arthritis were greater than those in gout, non-septic/non-RA inflammatory arthritis and non-inflammatory arthritis (p = 0.003, p<0.001 and p<0.001, respectively). These results are consistent with previously published observations of increased sTREM-1 in infectious conditions compared with those in similar non-infectious states.14 17 Unexpectedly, SF sTREM-1 levels in RA were also greater than those in gout, non-septic/non-RA inflammatory arthritis and non-inflammatory arthritis (p = 0.002, p = 0.001 and p<0.001, respectively) and were very similar to SF sTREM-1 levels in septic arthritis.
As rheumatoid factor in biological fluids (including SF) can confound determinations by ELISA by promoting multimeric complexes of unpredictable sizes in the presence of the assay’s detecting antibody, we serially diluted RA SF and measured sTREM-1 levels in the diluted samples. Linear responses with serial dilutions of RA SF were obtained (data not shown), suggesting little interference from rheumatoid factor in these samples. To confirm further that bona fide sTREM-1 was being detected, Western blotting of SF was performed and verified the presence of sTREM-1 (fig 1B). Its molecular weight of approximately 28 kDa strongly suggests that the measured sTREM-1 in our ELISA is a product of cleaved membrane-bound TREM-1 rather than being an alternative splice variant lacking exon 3, which would have a lower molecular mass of approximately 15 kDa.13
SF TNFα levels
One of the inflammatory cytokines secreted in response to TREM-1 ligation is TNFα.3 4 In our cohort of patients, SF TNFα levels were markedly greater in septic arthritis than in gout, non-septic/non-RA inflammatory arthritis, or non-inflammatory arthritis (p = 0.005, p = 0.040 and p<0.001, respectively) and trended towards greater levels than in RA (p = 0.079; fig 1C). SF TNFα levels in RA were similar to those in gout and non-septic/non-RA inflammatory arthritis (p = 0.243 and p = 0.547, respectively) and trended towards greater levels than in non-inflammatory arthritis (p = 0.052). These observations agree well with other studies of SF TNFα levels in patients with various types of arthritis.2 18 Of the 39 RA SF samples, 14 were from patients on anti-TNFα therapy at the time of arthrocentesis. SF TNFα levels were actually higher in these patients than in RA patients who were not being treated with an anti-TNFα agent (p = 0.031). Although no statistically significant difference in SF sTREM-1 levels between these two sub-cohorts of RA could be detected (p = 0.783, data not shown), there was a positive correlation among all patients tested between SF sTREM-1 and SF TNFα levels (r = 0.366, p = 0.001, fig 1D), suggesting a relationship between the two.
As SF from septic arthritis patients harboured the highest levels of both TNFα and sTREM-1, we excluded these patients from a secondary analysis. Among the non-septic arthritis patients, the correlation between SF sTREM-1 and TNFα levels persisted (r = 0.343, p = 0.005). When only samples with detectable levels of TNFα within this cohort were assessed, the correlation remained strong (r = 0.542, p = 0.024), demonstrating that even in the absence of infection, a relationship exists between sTREM-1 and TNFα in inflammatory effusions.
Relationship between SF leucocyte counts and SF sTREM-1 levels
As neutrophils and monocytes each express TREM-1,3 4 19 20 these inflammatory cells are candidate sources of SF sTREM-1. Given that SF leucocytes positively correlated with SF TNFα levels in our patients (r = 0.269, p = 0.018) and that TREM-1 ligation is associated with inflammatory cytokine release, we hypothesised that a relationship would exist between SF leucocytes and SF sTREM-1 levels. Indeed, among all patients, SF sTREM-1 levels did positively correlate with synovial leucocyte counts (r = 0.357, p<0.001, fig 2A). In addition, SF neutrophils, lymphocytes and monocytes individually also correlated significantly with SF sTREM-1 levels (r = 0.285, p = 0.001; r = 0.299, p<0.001; r = 0.206, p = 0.021, respectively; data not shown).
As the SF leucocyte counts in septic arthritis patients were markedly greater than those in any of the other patient cohorts (fig 2B) and may have skewed the overall results, we separately analysed the non-septic arthritis patients. The correlation between total SF leucocyte count and the sTREM-1 level persisted (r = 0.314, p<0.001; fig 2C) as well as the SF leucocyte differentials (neutrophils r = 0.227, p = 0.018; lymphocytes r = 0.345, p<0.001; and monocytes r = 0.218, p = 0.024; data not shown), suggesting that even in non-infectious effusions, recruited leucocytes contribute substantially to the elevation in SF sTREM-1 levels.
SF leucocyte TREM-1 mRNA levels
Despite the significant differences in both SF sTREM-1 levels and SF leucocyte counts among the individual patient cohorts, there were no significant differences in SF leucocyte TREM-1 mRNA levels among them (fig 3A). The median SF leucocyte TREM-1 mRNA levels, nevertheless, were greater in the septic arthritis and RA groups, so significant differences may have emerged with a greater number of samples. Indeed, SF sTREM-1 levels positively correlated with SF leucocyte TREM-1 mRNA levels (r = 0.344, p = 0.017; fig 3B) suggesting that the measured SF sTREM-1 arises at least partly from leucocytes that have migrated into the SF.
In-situ studies of TREM-1 expression in RA synovial tissue
The likely contribution of SF leucocytes to the increased SF sTREM-1 levels in RA did not preclude other potential sources of sTREM-1. In addition to TREM-1 expression by SF leucocytes, TREM-1 was strongly expressed in RA synovial tissue in a fine punctate pattern along the surface membrane of large cells located among lymphoid infiltrates in the subintimal layer (fig 4A). That TREM-1-expressing cells were of myelomonocytic lineage was established by CD14, CD33 and HLA-DR staining of serial sections (data not shown). In contrast, non-inflammatory osteoarthritis synovial tissue demonstrated essentially no surface expression of TREM-1 (fig 4B), indicating that TREM-1 is not constitutively expressed in synovial tissue. Western blotting of synovial tissue protein extracts demonstrated in each tested RA sample a strong TREM-1 band with the expected molecular weight of 30 kDa for the full non-cleaved protein (fig 4C), confirming the immunostaining results. We also observed binding of anti-TREM-1 to synovial macrophages in each sample from the two patients with Crohn’s-associated arthritis and from the patient with ankylosing spondylitis (data not shown), suggesting that macrophages may contribute to synovial inflammation in other inflammatory arthritides.
In this study, we show that SF sTREM-1 levels are significantly elevated in septic arthritis patients compared with those in patients with non-septic/non-RA inflammatory arthritis or non-inflammatory arthritis (fig 1A). This preferential elevation of SF sTREM-1 levels in septic arthritis is reminiscent of the preferential elevations of sTREM-1 levels in bronchoalveolar lavage fluids from patients with infectious pneumonia and in sera from patients with sepsis.14 17
Unexpectedly, sTREM-1 levels were also elevated in SF from RA patients (fig 1A). In fact, the levels were similar to those seen in septic arthritis, and the degree to which SF sTREM-1 levels are elevated appears to be unique to RA among the various causes of non-infectious inflammatory arthritis. The sTREM-1 levels measured by ELISA in RA SF are not spuriously elevated, as evidenced by the detection of sTREM-1 bands by Western blotting (fig 1B) and the linear diminution of measured sTREM-1 levels in serially diluted SF samples (data not shown). Our finding adds to a recently growing list of non-infectious inflammatory conditions associated with increased sTREM-1 levels or surface TREM-1 expression, including inflammatory bowel disease (IBD),9 21 paraneoplastic pleural effusions,22 acute pancreatitis23 and Helicobacter pylori-negative peptic ulcer disease.24
The mechanistic underpinnings of increased SF sTREM-1 levels in RA remain uncertain. TNFα can modestly upregulate TREM-1 expression,3 raising the possibility that elevated SF sTREM-1 levels in RA were consequent to elevated SF TNFα levels. However, inasmuch as TNFα levels in RA were not significantly elevated compared with non-septic/non-RA inflammatory arthritis despite increased sTREM-1 concentrations in the former (fig 1A, C), the elevated SF sTREM-1 levels in RA cannot solely be passive reflections of SF TNFα concentrations. In addition, even though RA patients treated with an anti-TNFα agent had higher levels of SF TNFα compared with the other RA patients, the use of anti-TNFα therapy did not appear to influence the expression of SF sTREM-1.
Indeed, it is possible that heightened synovial expression of TREM-1 actually promotes increased SF TNFα levels in addition to being upregulated itself by TNFα. Ligation of TREM-1 leads to the release of several pro-inflammatory cytokines, including TNFα.3 4 5 6 The correlation between SF sTREM-1 and TNFα levels in our patients (fig 1D) suggests that local TNFα production is linked to levels of SF sTREM-1 and, by inference, to the upregulation of surface TREM-1 expression. Schenk et al21 demonstrated marked reductions in TNFα and TNFα mRNA in mice with colitis following treatment with a TREM-1 antagonist peptide, suggesting a direct link between TREM-1 expression and TNFα production. Additional studies are needed to test this possibility in RA.
Given the role that TREM-1 ligation plays in the upregulation of inflammatory cytokines and chemokines, the individual cells that express TREM-1 and release sTREM-1 in each of the discrete inflammatory arthritis conditions could have important consequences for disease pathogenesis and/or maintenance. Whereas the exact cellular source of sTREM-1 in the SF of patients with inflammatory arthritis is unknown, the positive correlation between SF leucocytes and both SF sTREM-1 levels and TREM-1 mRNA expression (figs 2A and 3B) suggests that inflammatory leucocytes recruited into the SF are, at least partly, responsible for sTREM-1 production. Whereas the significantly elevated SF sTREM-1 levels were expected in septic arthritis, the similarly increased levels observed in RA SF may reflect additional sources of sTREM-1 independent of recruited leucocytes. In our study, the mean SF leucocyte differentials for each cohort were as follows: RA 67% neutrophils (N), 21% lymphocytes (L), 12% monocytes (M); septic 92% N, 3% L, 5% M; gout 87% N, 4% L, 9% M; other inflammatory 73% N, 14% L, 13% M; and non-inflammatory 34% N, 43% L, 23% M. The differences seen in leucocyte differentials within each cohort do not appear to explain the differences seen in SF sTREM-1 levels. For example, whereas gout had a total SF leucocyte count similar to that of RA, the SF leucocyte differential in gout (relatively rich in neutrophils with the potential to express high levels of TREM-1) would seemingly have favoured higher sTREM-1 levels compared with RA (relatively rich in lymphocytes, which are relatively poor expressers of TREM-1). This, however, was empirically not observed, suggesting that alternative cellular sources of synovial sTREM-1 substantially contributed to the marked elevation seen in RA.
One possibility is that a significant contributor of sTREM-1 in RA SF arises from myelomonocyte cells within synovial tissue (fig 4A). In RA, resident tissue macrophages and fibroblasts produce the majority of cytokines and chemokines found in the affected joints, and the observed spectrum is similar to that which follows TREM-1 ligation.3 4 Supportive of this argument is the dramatic increase in the frequency and number of TREM-1-expressing intestinal macrophages in humans with ulcerative colitis or Crohn’s disease and in experimental mouse models of IBD.21 Increased TREM-1 expression correlated with disease activity in both clinical and experimental IBD, and TREM-1 crosslinking resulted in marked increases in proinflammatory cytokines, including TNFα, IL-8 and monocyte chemotactic protein 1.21 Indeed, in samples from two patients with inflammatory arthritis associated with Crohn’s disease, immunoflorescence staining revealed the presence TREM-1-expressing synovial macrophages (data not shown). Furthermore, the lack of TREM-1 expression in non-inflammatory synovial tissue from osteoarthritis (fig 4B) suggests that TREM-1 is not constitutively expressed in synovial tissue. TREM-1 expression is thus probably an important mediator of inflammation in several inflammatory arthritis conditions, including IBD-associated arthritis and gout. The abundance of sTREM-1 in RA may merely be a reflection of the hypercellular synovium of RA patients relative to that in patients with other types of inflammatory arthritis. Accordingly, increased TREM-1 expression in RA synovial tissue may play an important contributory role in the ongoing inflammatory response. Additional experiments utilising increased numbers of surgical samples are needed to determine if relationships exist between synovial TREM-1 expression, quantitative synovial monomyelocytic assessments and levels of SF sTREM-1.
In conclusion, sTREM-1 levels are increased in the SF from patients with septic arthritis compared with those in SF from patients with other types of arthritis, with the noteworthy exception of RA. The realisation of increased SF TREM-1 expression in RA may not only provide additional insight into the pathogenesis of RA, but may potentially suggest an additional therapeutic target. Although the endogenous ligand of TREM-1 has yet to be determined, blocking TREM-1 with a monoclonal antibody in a murine model of lipopolysaccharide-induced shock and peritonitis reduced hyperresponsiveness and lethality.6 In addition, interfering with TREM-1 engagement in murine colitis resulted in reductions of several proinflammatory cytokines and amelioration of disease.21 Accordingly, blockade of TREM-1 in an active RA patient could potentially result in the downstream reduction of excessive chemokine and cytokine release. Further studies assessing sTREM-1 and TREM-1 expression from various sites are needed in RA as well as the determination of TREM-1 expression in other autoimmune inflammatory diseases.
The authors thank all the subjects for their participation in this study, Professors D Mainard and F Sirveaux for providing RA synovial tissue samples, and Anne Sophie Didier for her technical help with the studies on the synovial tissue samples.
Competing interests None.
Ethics approval The study protocol was approved by the Institutional Review Board of the University of Southern California Keck School of Medicine.
Patient consent Obtained.
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.