Rheumatoid arthritis (RA) is a systemic autoimmune disease that causes chronic, persistent joint inflammation, leading to irreversible structural damage. The etiology remains unclear, but the prevailing disease model is that multiple genes, especially HLA-DRB1 alleles, interact with environmental risk factors, including tobacco smoking, culminating in an adaptive immune response to citrullinated autoantigens that perpetuates joint inflammation 1. T lymphocytes are key "conductors" of the inflammatory process, providing help to B cells and leading to the production of autoantibodies and the recruitment/activation of other immune effectors 23. Ultimately, the activation of plasma cells, macrophages, fibroblasts, chondrocytes and osteoclasts, and downstream production of inflammatory cytokines (that is, TNF, IL-1, and IL-6), reactive oxygen species, matrix metalloproteinases and other toxic molecules, mediates destruction of articular cartilage and bone 4.

Advances in therapeutics have dramatically improved the outlook for joint structure and function in patients with RA. Current treatment strategies retard joint destruction in the majority. Nonetheless, between 5 to 30% of patients, depending on the treatment strategy used, still experience rapid progression of joint damage 56. Models describing the probability of rapid radiographic progression have limited accuracy and reliability and do not fully account for joint damage 789. More patients suffer slow structural deterioration over time, which can occur despite intensive treatment 101112. The most likely explanation is that inflammation can persist despite treatment, even when the disease appears to be in clinical remission 13141516. Therefore, there is an urgent need to identify new pathogenic mechanisms of persistent, treatment-refractory inflammation, in hopes of developing novel targeted therapies that prevent structural deterioration over time.

We have devised an approach to discover immunological correlates of disease phenotypes based on profiles of ex vivo cytokine production from peripheral blood mononuclear cells (PBMC) in response to a number of non-specific stimuli. This approach was based on the concept that we could sample a broad array of activation pathways and, thereby, develop a general profile of immune responsiveness rather than define the responsiveness of specific innate or adaptive immune pathways. Our previous studies demonstrated the success of this approach by characterizing differences in the profiles of immune response between patients with early versus late RA and healthy controls, and also between patients with and without myocardial dysfunction 1718. The results informed our hypothesis that the broad responsiveness of one or more immune pathways is associated with the potential of the inflammatory process to mediate joint destruction. The objective of this study was to identify a profile of immune response based on ex vivo cytokine production that is associated with the severity of radiographic joint damage in patients with RA, after accounting for known predictors of severe disease.

The current study represents a translational effort to discover new immune pathways underpinning structural joint damage. Our approach was to induce cytokine response profiles using a panel of stimuli designed to activate both innate and adaptive immunity, then to screen these profiles for correlation with radiographic joint damage in patients with early RA. The singular finding of this study is the association between a profile of cytokine production in response to stimulation with purified CMV/EBV and radiographic joint damage. This finding assumes significance in view of the literature supporting a potential role of latent herpesvirus infection in the pathophysiology of RA.

Both CMV and EBV are in the family of human herpes viruses, which have large double-stranded DNA genomes enclosed in a protein nucleocapsid, surrounded by a protein tegument and also an outer glycoprotein-studded lipid membrane 2728. Infection with these viruses is ubiquitous in people worldwide. In the United States, 50 to 60% of adults have ever been infected with CMV, whereas 90 to 95% of adults have ever been infected with EBV 2729. Of crucial relevance to our findings is the knowledge that herpes viruses induce a robust, primary immune response capable of controlling viral replication, yet they are still able to establish a latent infection in cell reservoirs 3031. CMV primarily infects monocytes and macrophages but also infects dendritic cells, epithelial cells, endothelial cells and fibroblasts 28. EBV primarily infects B cells yet can also exist latently within nasopharyngeal epithelial cells 27. Remarkably, 10% of the entire memory T-cell repertoire in humans, on average, is devoted to defense against latent CMV infection, underscoring the impact of this virus on the immune system 32. Previous epidemiological studies report associations between infection with these herpes viruses (as defined by serological assays) and RA 333435363738. Indeed, latent infection with CMV and EBV is consistently detectable in the synovium 394041424344, and CMV has been isolated from synovial cells of a patient with RA 45. Studies also report the detection of viral replication in RA joints, suggesting active subclinical infection 444647.

Considering herpes virus biology, our findings suggest that the immune response to latent infection with CMV or EBV may play a fundamental role in the pathophysiology of chronic inflammation and joint destruction in RA. Specifically, the key observation is that a profile of cytokine production in response to stimulation of PBMC with purified CMV and EBV lysates is associated with the severity of radiographic joint damage in patients with RA. Several pathways could give explanation for our findings. Innate immune receptors could recognize various non-self, pathogen-associated molecular patterns of the viruses 48. Toll-like receptors (TLR) 2, 3, 7 and 9 likely have important roles in recognizing and activating responses to both CMV 495051525354 and EBV 5556575859. Additionally, the protein DAI (ZBP1), and the AIM2 inflammasome are crucial sensors of viral double-stranded DNA in defense against CMV infection 4860616263. However, our results do not clearly suggest activation of innate immune cells by such pathways. In this case, one would expect to see production of IL-1β, IL-6, IL-8, IL-12 or TNF in response to the viruses 50545556, but this is not what we observed. The lack of correlation between the immune response to CpG oligonucleotide, a TLR-9 ligand and radiographic joint damage further suggests that recognition of CMV/EBV by pathogen recognition receptors and innate immune activation may not be the main explanation for our findings.

Another possible explanation for our findings is activation of cytokine production in memory T cells specific for CMV or EBV. The composition of the CMV/EBV immune profile suggests the main source of cytokine production is CD4⁺ or CD8⁺ T cells. These should be memory T cells, considering the high prevalence of CMV and EBV infection in our sample, in addition to the short time-period for the development of an immune response to the viruses. A key result is the significant correlation of the CMV/EBV immune response to CMV IgG⁺. The significance is that patients with RA and other immune-mediated disorders have large, clonally expanded populations of 'CD28^null' T cells in the blood 64656667. It happens that these CD28^null T cells are actually highly differentiated, effector-memory T cells, which are largely antigen-specific for CMV 646566686970. The down-regulation of the co-stimulatory molecule CD28 signifies chronic activation of the cells, which corresponds to progressive attenuation of proliferative capacity and increase in cytokine production 697172. In contrast, EBV-specific effector-memory T cells are typically CD28⁺ 6973. Of further note, clonally expanded populations of CD8⁺ T cells specific for CMV are present in RA synovial fluid 747576. Their frequencies are higher in synovial fluid than peripheral blood, suggesting selective enrichment of CD8⁺ effector-memory T cells in joints 74. Using HLA class I tetramers, CD8⁺ T cells have been shown to react specifically with CMV pp65 epitopes 74. Further, CMV-specific CD28^null T cells have a known association with the systemic complications of RA, including lung disease, vasculitis, cardiovascular disease and mortality, highlighting their potential role in the pathophysiology of RA 65677778. In view of the evidence, CMV appears to be the more compelling of the two viruses for further study.

CMV-specific effector-memory T cells could possibly mediate inflammation in RA in many, complementary ways. They could do so in an antigen-specific manner, by targeting cells that harbor latent virus, and thus amplify the autoimmune joint disease 79. Alternatively, CMV-specific effectors could mediate inflammation non-specifically, by migrating into inflamed joints according to chemokine gradients and becoming activated through cell-cell contact, or, by binding other cytokines in the inflammatory milieu, leading to activation of macrophage TNF production 80. This type of 'bystander activation' of effector-memory T cells is considered a critical mechanism in RA pathophysiology 798081. CMV-specific effector-memory CD4⁺ T cells are equipped to kill target cells rapidly and effectively by production of inflammatory cytokines (that is, IFN-γ or TNF) as well as perforin- and granzyme-dependent cytotoxic function 8283. Additionally, the expansion of CMV-specific memory T cells can overwhelm immunological niches, leading to the loss of both naïve T cells and smaller populations of memory T cells 84. Indeed, latent CMV infection causes telomere shortening in T cells, indicative of immune senescence 8586. One intriguing possibility relevant to RA is that latent CMV infection is associated with increased differentiation of diverse T-cell specificities, suggesting that this could augment the function of autoreactive T cells in the joints of patients with RA 87. Further research is necessary to elucidate the contribution of CMV-specific T-cell immunity to the progression of joint destruction in patients with RA.

This study has a number of limitations. The first is its discovery-oriented nature; our findings must be replicated in other patient populations. The study included relatively few patients currently receiving biologic therapies, limiting generalization of our findings to patients mainly on oral disease-modifying agents. Because we only collected data on current use, the results could underestimate the prevalence of ever having taken a biologic DMARD. Another limitation is the study of mixed populations of cells, precluding determination of the specific populations producing our results. We acknowledge that differences in the frequencies of cell types could contribute to our results, but in previous studies we could not explain the associations of ex vivo cytokine production with disease characteristics on the basis of such differences 1718. The use of a mixture of both CMV and EBV lysates creates uncertainties. We cannot be sure that CMV is more important then EBV in regard to RA, but overall the serological findings point to CMV as the most promising candidate for further study. The specific viral moieties that mediate the observed cytokine production are unknown. It certainly remains possible that activation of TLR pathways in innate immune cells with downstream activation of T cells could contribute to our results. Finally, this is a correlative study, so future prospective longitudinal studies are imperative to determine the causal relationship between CMV immunity and joint destruction in RA.

We report the identification of a profile of ex vivo immune response to CMV/EBV stimulation that correlates with radiographic joint damage in patients with RA. The correlation of this immune response to CMV serology suggests the underlying drive is latent CMV infection. The nature of the response suggests the involvement of T-cell immunity. The findings assume significance in the knowledge that the magnitude of CMV-specific CD28^null T cells predicts severe manifestations of RA and other immune-mediated diseases. Based on our findings and the literature, we hypothesize that the immune response to latent CMV infection contributes to the propagation of inflammation and progression of structural joint damage in patients with RA. Further research is necessary to elucidate the determinants of the immune response to CMV and/or EBV that aggravate joint destruction in this disease.

ACPA: anti-citrullinated protein antibodies; anti-CD3/anti-CD28: anti-CD3 and anti-CD28 monoclonal antibodies; CMV: cytomegalovirus; CMV/EBV-1: CMV/EBV first principal component score; CpG: bacterial CpG oligonucleotides; CRP: C-reactive protein; DAS28: Disease Activity Score in 28 joints; DMARDs: disease-modifying antirheumatic drugs; EBV: Epstein-Barr virus; HAQ: Health Assessment Questionnaire; HLA-DRB1: human leukocyte antigen DRB1; IgG: immunoglobulin G; IFN-γ: interferon gamma; IL: interleukin; IQR: interquartile range; JSN: joint space narrowing; MCP-1: monocyte chemoattractant protein-1; MHV: murine gamma-herpesvirus; MIP-1β: monocyte inflammatory protein-1β; MSD: Meso Scale Discovery; PBMC: peripheral blood mononuclear cells; PCA: principal components analysis; PMA: phorbol myristate acetate; PMA/ionomycin-2: PMA/ionomycin stimulation second principal component score; RA: rheumatoid arthritis; RF: rheumatoid factor; SEA/SEB: Staphylococcal enterotoxins A and B; SF-36: Medical Outcomes Study short form 36; SHS: Sharp-van der Heijde score; TLR: Toll-like receptor; TNF: tumor necrosis factor; VAS: visual analog scale

JMD, KLK and SEG are inventors of technology referenced in this article. That technology is entitled "Cytokine Response Profiling in Rheumatoid Arthritis." The Mayo Clinic has filed a provisional patent application for this technology.

JMD, KLK and SEG designed the study. JMD recruited and clinically evaluated the patients. MSS performed the PBMC assays and multiplexed cytokine analysis. JAS interpreted and scored all joint radiographs. JMD, CSC, KLK and TMT performed the statistical analysis. All authors contributed to analysis and critical interpretation of the results as well as manuscript preparation. All authors have read and approved the final manuscript for publication.