Studies of Preclinical Rheumatoid Arthritis

Multiple studies have shown that RA-related autoantibodies are present years before the diagnosis of RA ( Table 1 ). del Puente and colleagues, who investigated RA in the Pima Indians in the Southwestern United States, showed that rheumatoid factor was present before the onset of clinically apparent RA. Aho and colleagues, who investigated preclinical RA in Finland using a biobank of stored prediagnosis samples, and Jonsson and colleagues, who used Icelandic biobank samples, also demonstrated that rheumatoid factor (RF) (by various methodologies) was present prior to the onset of clinically apparent RA. Also, in a prospective study of initially healthy family members of patients with RA, Silman and colleagues showed that the presence of RF preceded the onset of clinically apparent RA.

Later studies also showed preclinical RA positivity for RF, as well as positivity for the highly RA-specific a ntibodies to c itrullinated p rotein a ntigens (ACPAs). Again using a Finnish biobank, Aho and colleagues demonstrated preclinical RA positivity of antikeratin (AKA) and antiperinuclear factor (APF) antibodies, and antifillagrin antibodies (AFA) —autoantibody targets that, based on later findings, represented citrullinated antigens. Using stored blood samples available through the Medical Biobank of Northern Sweden, Rantapaa-Dahlqvist and colleagues evaluated for elevations of RF isotypes (immunoglobulins [Ig] M, G and A) and the anticyclic citrullinated peptide (anti-CCP) antibody in 98 preclinical samples from 83 RA patients, collected a median of 2.5 years prior to symptomatic disease onset. In this study, approximately 34% of patients were positive for anti-CCP within 1.5 years prior to diagnosis of RA, and the prevalence of RF isotype positivity ranged from about 17% to 34% during this same period. In addition, in comparison to controls, a combination of both anti-CCP and any RF isotype was highly specific (99%) for the future development of classifiable RA. This report was followed by a similarly designed retrospective cohort study by Nielen and colleagues that evaluated preclinical RA RF and anti-CCP positivity using pre-RA diagnosis samples stored in a Dutch blood donor biobank. In this study, 79 RA patients with stored prediagnosis samples were identified, with a median of 13 preclinical samples per case. Of these 79 cases, approximately 28% and 41% had pre-RA diagnosis elevation of RF (IgM isotype) or anti-CCP, respectively. RF was positive a median of 2.0 years prior to RA diagnosis (range 0.3–10.3 years), and anti-CCP was positive a median of 4.5 years prior to RA diagnosis (range 0.1–13.8 years).

Additional studies using stored biobank samples have also demonstrated preclinical RA positivity for autoantibodies. Using stored pre-RA samples from 83 United States military subjects with RA, Majka and colleagues demonstrated pre-RA diagnosis positivity for RF and anti-CCP. In addition, they demonstrated that individuals that were older at the time of diagnosis of RA had longer duration of preclinical positivity of autoantibodies, a finding that was supported by work by Bos and colleagues. Chibnik and colleagues utilized the Nurses’ Health Study (NHS) biobank to demonstrate anti-CCP positivity prior to RA diagnosis, and showed that a lower cut-off value for anti-CCP than the kit suggested was also highly specific for future RA.

Multiple studies have also evaluated elevations of inflammatory markers prior to diagnosis of RA, with varying results (see Table 1 ). In their Finnish biobank study, Aho and colleagues found no significant elevation of C-reactive protein (CRP) in pre-RA diagnosis samples, although the time of sample collection prediagnosis was not reported. Using the same Medical Biobank of Northern Sweden as in the study of pre-RA RF and anti-CCP positivity, Rantapaa-Dahlqvist and colleagues showed a significant elevation of monocyte chemoattractant protein-1 (MCP-1) in 92 pre-RA cases versus controls, but not secretory phospholipase A2 (sPLA2), high-sensitivity CRP (hsCRP) or interleukin (IL)-6. Nielen and colleagues demonstrated, in the Dutch pre-RA sample cohort described above, that CRP was elevated in the pre-RA period, most commonly about 2 years before diagnosis, regardless of prediagnosis autoantibody positivity, although these investigators were unable in this study to demonstrate the chronologic sequence of appearance of CRP versus autoantibodies. Nielen and colleagues later additionally demonstrated pre-RA elevation of sPLA2 in this cohort, but were unable to determine whether CRP or sPLA2 preceded autoantibodies in the preclinical period, and they concluded that the temporal development of autoantibodies and these 2 markers were probably similar. Jorgensen and colleagues utilized samples from a Norwegian biobank to examine in 49 cases with RA prediagnosis elevations of autoantibodies (RF and anti-CCP) and multiple cytokines. These investigators found that RF and anti-CCP were elevated in RA cases prediagnosis; however, they could not demonstrate any cytokine elevations prior to 5 years before disease onset, and only tumor necrosis factor α (TNFα) was statistically significantly elevated in cases (vs controls) during the 5-year period just before diagnosis of RA. Using a single pre-RA diagnosis blood sample from 90 incident RA cases (case status confirmed by chart review) identified in the Women’s Health Study (WHS), Shadick and colleagues were unable to demonstrate significant elevations in CRP levels in cases versus controls (with case samples available a mean of 6.6 years prior to diagnosis), and they were unable to use a single CRP level to predict future RA. However, in a later study using 170 RA cases from a combination of the NHS and WHS cohorts, the same group was able to demonstrate statistically significant elevations of soluble tumor necrosis factor receptor II (sTNFRII) prior to RA diagnosis, but not IL-6 or hsCRP. Most recently, Rantaapa-Dahlqvist and colleagues used an expanded sample set from their Swedish Biobank, to demonstrate elevations of multiple cytokines and chemokines prior to the diagnosis of RA, with these elevations likely indicating pre-clinical RA activity of Th1, Th2, and T regulatory processes.

There are caveats when interpreting the data regarding pre-clinical elevations of biomarkers. Certainly the prospective studies by del Puente and colleagues and Silman and colleagues suggest that autoantibodies are truly elevated before the onset of clinically apparent disease. However, in the studies of preclinical autoantibody and inflammatory marker elevations using stored samples from biobanks, there were not detailed joint-directed questionnaires or examinations performed at baseline or over time to ensure that no clinically apparent arthritis was present at the time of presumed preclinical RA blood collections. As such, it may be that the duration of pre-RA diagnosis autoantibody positivity is overestimated in the studies utilizing stored biobank samples, as subjects may have had mild or fluctuating inflammatory symptoms long before a confirmed diagnosis of RA. Also of importance, these pre-RA studies show that not all RA patients have detectable pre-RA diagnosis autoantibody positivity. Methodologic issues may in part explain these findings. For example, patients may not have a stored blood sample available for analysis from the correct time period to demonstrate their prediagnosis autoantibody positivity. Also, current assays for autoantibodies may not detect the earliest preclinical autoantibody specificities; perhaps an as-of-yet unknown citrullinated antigen is the earliest autoantibody target in preclinical RA? In addition, not all patients with RA may develop detectable circulating autoantibodies in the preclinical period. For example, some patients may develop circulating autoantibodies after clinically apparent disease develops. Of note, in the Nielen study, while only about 28% of patients with RA were RF-positive prediagnosis, about 67% of these same patients were positive for RF by 6 years post diagnosis, suggesting that circulating RF develops in some cases after symptomatic onset of RA. All of these issues will need to be considered in studies of preclinical RA going forward.

The Contribution of Genetic and Environmental Factors Associated with Rheumatoid Arthritis to Understanding Preclinical Development

Although the exact etiology of RA is unknown, there are multiple genetic and environmental factors that have been associated with disease. Estimates of the genetic contribution to RA have ranged between 30% and 60%. Of the known genetic factors associated with RA, a DRB1 allele containing the “shared epitope” is the most important genetic factor associated with RA. Of the HLA alleles, HLA-DRB1*0401 and HLA-DRB1*0404 (within the HLA-DR4 group) are the most strongly associated with RA, with an approximate relative risk for RA of almost 4 in Caucasians. In addition, several genes outside the HLA-DRB1 gene have recently been associated with RA. These genes include PTPN22 , TRAF1/C5 , CTLA4 , and STAT4 , and the list is rapidly growing.

Several potential environmental factors that could play important roles in modifying either susceptibility to RA or disease severity have been identified. High levels of coffee consumption, in particular decaffeinated coffee, as well as a positive smoking history, exposure to air pollution, and environmental exposure to silica-containing dust are also associated with increased risk for RA. In addition, infections have been associated with RA including pathogens such as the Epstein-Barr virus (EBV), Mycoplasma or Proteus species. Finally, several factors have been identified as possibly protective against development of RA including a history of successful pregnancy, oral contraceptive pill use, and higher vitamin D intake.

Several studies have examined the association of environmental factors with asymptomatic RA-related autoantibody positivity. These studies suggest that lack of oral contraceptive use and a history of heavy smoking are associated with an increased prevalence of RF, whereas 25,hydroxyvitamin D levels are not associated with RF positivity in individuals without RA. Active pulmonary infection with tuberculosis has also been associated with RF and anti-CCP positivity in individuals without clinically apparent synovitis. These studies suggest that environmental factors affect RA-related autoimmunity, although prospective evaluation of future risk for RA in these types of autoantibody positive subjects is lacking.

Of recent interest regarding the pathogenesis of RA are interactions between genetic and environmental factors that may lead to RA-related autoimmunity and disease. For example, several studies have reported that smoking in individuals with specific HLA alleles is associated with ACPA-positive RA, suggesting that a gene-environment interaction leads to RA-related autoimmunity.

However, while genetic and environmental factors have been associated with RA, the exact role that these factors play in the development of RA-related autoimmunity is not yet clear. Also, it is unclear whether these factors may lead to initial RA-related immune dysregulation or transition from asymptomatic autoimmunity to clinically apparent RA. It is important that, as discussed later, prospective study of the early phases of RA development should help to clarify these issues.

Preclinical Studies in Other Autoimmune Diseases

In addition to the information about preclinical RA available from RA-specific studies, much about the evolution of autoimmunity can be learned by examining the preclinical natural history of other autoimmune diseases including T1DM and systemic lupus erythematosus (SLE). In particular, prospective studies of T1DM have led to the development of a model of disease evolution that may be of importance to RA.

T1DM results from autoimmune-mediated destruction of the pancreatic β cells, and it affects approximately 1 in 300 children. Numerous studies of T1DM have established that there are specific high-risk HLA genotypes that predispose to disease. Moreover, the autoimmune attack on the pancreatic β cells can be detected years before clinical onset of T1DM via the presence of any of the T1DM-related autoantibodies in the blood including antibodies to the following antigens: islet cell (ICA), insulin (IAA), protein tyrosine phosphatase (IA2), and glutamic acid decarboxylase (GAD65).

Prospective studies of T1DM have established that T1DM exhibits several phases during its development. The first phase is defined as the presence of genetic risk factors that, either alone or in combination, may predispose to the loss of self tolerance. The second phase is reached when transformation from the risk state to a state of immunologic autoreactivity occurs; this second phase of “asymptomatic autoimmunity” is measurable by testing for T1DM-related biomarkers, but this phase is not yet associated with the presence of clinically apparent hyperglycemia. This transition from genetic risk to asymptomatic autoimmunity occurs either because of the introduction of an environmental factor, or perhaps because of the stochastic nature of the immune response. In T1DM, this preclinical autoimmune phase is marked by initial immune reactivity to only a small number of autoantigens. The third phase of T1DM is the development of clinically apparent hyperglycemia; this final phase is characterized immunologically by autoreactivity to numerous antigens and extensive immune-mediated tissue destruction of islet cells caused by many proinflammatory pathways.

The presence of these 3 phases of diseases in T1DM is well established, as is the value of prospectively studying children with highly predictive autoantibodies in the preclinical phases of disease for epidemiologic and genetic associations. Of importance, due to the strong association of T1DM with β-cell (or islet) autoimmunity (IA), defined as the presence of autoantibodies specific to β-cell autoantigens, IA has become an alternative end point to clinically apparent hyperglycemia in T1DM research. Advantages of this approach include the opportunity to study the pathologic process underlying T1DM in a preclinical state, and to verify that a candidate risk factor is not only related to diabetes but also to the alteration of immune function preceding clinical onset of disease. Studies using as end points both autoimmunity and clinically apparent diabetes have allowed for differentiation of the risk factors that initiate humoral autoimmunity from those that promote progression from subclinical autoimmunity to diabetes. Of note, prospective studies using asymptomatic IA as an outcome measure have identified important findings regarding dietary and other factors in T1DM, including the lack of association of islet-cell autoimmunity with childhood vaccines, and the important relationship of timing of cereal exposure during the first year of life to the later development of IA. This latter finding has been of importance in terms of potentially reducing risk for T1DM based on dietary considerations.

The presence of autoantibodies can be also used in the prediction of future T1DM. In first-degree relatives (FDRs) of diabetic individuals, the presence of a single autoantibody on one occasion has been found to have a sensitivity and specificity of 95% and 99%, respectively, and a positive predictive value (PPV) of 46% for type 1 diabetes with 5 years of follow-up. Of note, the number of unique T1DM-related autoantibodies that are positive also predicts risk for future T1DM. For example, a person who tests positive for multiple autoantibodies is at a much higher risk of developing T1DM than a person with a single detectable autoantibody. However, T1DM-related autoantibodies may be transiently positive, and there are not yet enough longitudinal data to determine if all people that develop autoantibodies will eventually develop T1DM. Although not 100% predictive of future T1DM development, IA still serves as a very useful intermediate end point when studying the autoimmune disease process and potential risk factors for T1DM. In terms of RA, given the high specificity of certain autoantibodies (notably ACPAs) for disease, using autoantibody positivity as a surrogate end point for autoimmunity in preclinical RA is likely to be of similar great value.

SLE is a multisystem autoimmune disease of unknown etiology. More than 98% of SLE patients are positive for antinuclear antibodies (ANA) and many SLE patients have additional reactivity to specific nuclear antigens including double-stranded DNA (dsDNA), Smith, ribonuclear proteins, and Ro and La. Some of these autoantibodies are associated with specific clinical manifestations of disease. For example, autoantibodies to dsDNA are associated with nephritis, and levels of anti-dsDNA antibodies may fluctuate in conjunction with disease activity. In addition, anti-Ro antibodies have also been shown to be associated with specific manifestations of SLE including skin disease and neutropenia. Although these autoantibody systems have been extensively studied for more than 50 years, the initiating events for the development of these autoantibodies and the actual roles these antibody specificities play in clinical SLE are not known. However, these SLE-related autoantibodies are present in the serum of lupus patients many years prior to the first clinical evidence of disease. Furthermore, the spectrum/number of autoantibody specificities increases up to the time of diagnosis. The discovery of preclinical autoantibody positivity and evolution of antibody specificity has led to important investigations into potential etiologic agents such as EBV infection in SLE, investigations that could not be performed without preclinical studies.

Animal Models and Preclinical Rheumatoid Arthritis

While animal models of arthritis are not an exact match to human disease, they have provided important insights into the preclinical period of arthritis development. In particular, investigations into the relationships between the major histocompatibility complex (MHC) class II dependence of disease, a requirement in some instances for the specific exposure to citrullinated autoantigen, and the timing between the evolution of ACPA and clinically apparent arthritis have been helpful in understanding important factors in preclinical arthritis. An early example was provided by evidence that citrullinated proteins are found in the synovium after immunization with bovine collagen type II in DBA/1j mice, a strain that is commonly used to study the type II collagen-induced arthritis (CIA) model of RA. However, in this study no ACPAs were detected. In a second study of the same model, ACPAs were found to develop before clinically detectable joint inflammation. In this latter study, 2 methods were used to show that the ACPAs are pathogenic: first, treatment of mice with a “tolerogenic” form of a citrullinated peptide decreased the arthritis disease severity by approximately 60% and decreased the levels of ACPAs and epitope spreading to other citrullinated peptides; and second, a mouse monoclonal antibody that recognized citrullinated peptides, and could be used as a representative example of an ACPA, was found to greatly amplify joint inflammation. This process occurred only following the administration of low doses of antibodies specific for type II collagen, which induced the generation of citrullinated target antigens, presumably through the actions of macrophages and neutrophils containing peptidyl arginine deiminase (PAD) enzymes that participated in the inflammatory response. This second experiment suggested that, in the absence of citrullinated target antigens, there is no target injury; however, when these antigens are present, such as during treatment with monoclonal antibodies specific for type II collagen, ACPAs can greatly amplify inflammation and damage. Recently, additional monoclonal antibodies reactive with citrullinated type II collagen, with or without cross-reactivity to other epitopes, were also found to amplify the development of arthritis in mice in vivo. In addition, another study in rats has shown an increased incidence and severity of arthritis when collagen type II is citrullinated before immunization.

Other animal studies have focused on the relationship between cellular and humoral autoimmunity to citrullinated antigens and the shared epitope and class II molecules. For example, immunization with citrullinated vimentin peptide leads to CD4 ⁺ T-cell activation and proliferation in transgenic mice expressing human HLA-DRB1*0401. In addition, immunization of human HLA-DRB1*0401-transgenic C57BL/6 mice with citrullinated human fibrinogen, but not native human fibrinogen or either form of mouse fibrinogen, resulted in the development of arthritis in a substantial proportion of animals. In this setting, a relatively low level of inflammation was present, but this was accompanied by specific antibody responses to citrullinated human fibrinogen peptide, as well as T-cell responses to citrullinated protein. ACPA-positive arthritis was also found to develop in mice that are immunized with low doses of collagen type II and that have dysregulated MHC expression in the joints owing to transgenic expression of class II transactivator. Finally, a particularly intriguing study has recently shown that immunization of a subset of strains of mice with human fibrinogen alone resulted in destructive arthritis as well as T- and B-cell reactivity to fibrinogen. In this study, epitope spreading to citrullinated fibrinogen as well as many synovial citrullinated and noncitrullinated autoantigens developed. Remarkably, arthritis could be transferred to naïve mice with either serum or fibrinogen-reactive T cells from immunized mice.