The association between omega-3 fatty acid biomarkers and inflammatory arthritis in an anti-citrullinated protein antibody positive population

Abstract

Introduction

Seropositive RA exhibits a preclinical period characterized by circulating autoantibodies (e.g. RF and antibodies to ACPA). On average, these autoantibodies are present 3–5 years prior to the development of clinically apparent inflammatory arthritis (IA) and formal classification of RA [1, 2]. ACPAs, measured through anti-CCP antibody tests, provide a powerful tool to study the natural history of seropositive RA due to their high specificity for future disease [3–7]. Environmental factors may play a role in the development of the initial systemic autoimmunity, and in the progression from autoimmunity to clinical symptoms (which include joint swelling and pain) and finally to classified RA [1, 2]. Investigating the role environmental factors may have on RA pathogenesis during these stages is an important step towards prevention [1].

The long chain omega-3 fatty acids (n-3 FAs) eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) are environmental factors that may alter disease progression as they reduce proinflammatory cytokine production (i.e. IL-6), and can influence the innate and adaptive immune system (i.e. reduction in neutrophil recruitment, inhibition of T cell proliferation) through eicosanoid lipid derivatives (lipoxin, resolvins, protectins) [8–11]. The long-chain n-3 FA docosapentaenoic acid (DCPA) has not been as extensively studied, but evidence suggests it has similar properties to EPA and DHA [12].

As n-3 FAs cannot be endogenously synthesized, dietary sources are necessary. The most commonly consumed n-3 FA is α-linoleic acid (ALA), which is found in leafy green vegetables, nuts and seed oils. While ALA can be sequentially elongated in the body to form EPA, DCPA and DHA, this process is relatively inefficient in humans, where only approximately 8–10% of the longer chain FAs are converted from ALA [13, 14]. Therefore, dietary consumption of longer chain n-3 FAs via fatty fish (e.g. salmon and sardines), fortified foods and dietary supplements is necessary for physiological processes [15].

As for the relationship between n-3 FAs and RA, compounds including longer chain n-3 FAs have been shown to improve prognosis when paired with standard treatment in RA patients [16, 17]. Evidence suggests that consumption of foods rich in longer chain n-3 FAs is associated with a lower risk of RA [18–20]. Our previous work found n-3 FAs to be inversely associated with the presence of anti-CCP antibody and RF in a population at increased genetic and familial risk for developing RA [21, 22], suggesting that n-3 FAs could be protective against the initial development of preclinical autoimmunity. However, any potential preventive role n-3 FAs may play in the later transition from preclinical autoimmunity to IA, the primary clinical manifestation of RA, has yet to be elucidated.

The objective of our study was to explore the association between n-3 FAs and the presence or development of IA in a population that is seropositive for anti-CCP antibody.

Methods

Study population

The study population was selected as a subset of the longitudinal Studies of the aEtiology of Rheumatoid Arthritis cohort, described previously [23]. Study subjects were recruited from a Colorado-based free screening for undiagnosed RA provided at local health fairs from 2008 through 2014, described in detail elsewhere [24]. Screened health fair participants received a demographic and current joint symptom questionnaire, and a blood test for ACPA using the anti-CCP3 IgG ELISA (Inova Diagnostics, Inc., San Diego, CA, USA).

Health fair participants were eligible to be enrolled in a follow-up study if they were ⩾18 years old, had no self-reported prior diagnosis of RA and tested anti-CCP3 positive at the health fair. In total, 9329 participants were screened at the health fair; 250 were anti-CCP3 positive and did not have a self-report of diagnosed RA at the time of the health fair screening (Fig. 1). Fifty-six subjects agreed to participate in the follow-up study and completed a research visit (i.e. baseline visit), which occurred within ∼3 months of their health fair visit. This baseline visit consisted of a blood draw, and a 68-joint examination conducted by a study rheumatologist to evaluate joint tenderness and swelling consistent with IA. Subjects also completed questionnaires providing information on demographic characteristics, current and past joint symptoms and environmental exposures. Anti-CCP3 (IgG) was measured in serum obtained at the baseline visit using ELISA kits (Inova Diagnostics, Inc.) and was considered positive if greater than the kit cut-off of ⩾20 U. Subjects were excluded from the current study if they tested negative for CCP3 at the baseline visit (n = 9). Forty-seven anti-CCP3 positive subjects were included in the current study (Fig. 1). These individuals were invited for subsequent follow-up visits at 6-month intervals, where they were evaluated in a similar manner to the baseline visit.

Fig. 1

Study population recruitment and comparison groups IA: inflammatory arthritis.

Open in new tabDownload slide

In addition, we tested subjects for RF as measured by nephelometry (Dade Behring, Newark, DE, USA), reported in IU/ml. Positivity for RF was based on 1987 ACR recommendations [25], using a cut-off level higher than that observed in 95% of 490 randomly selected blood donor controls from the Denver area. High-sensitivity CRP, a marker of general inflammation, was also measured by a nephelometric assay (Dade Behring, Deerfield, IL, USA), where positivity was defined by the manufacturer as >3 mg/l.

Genotyping

Subtyping for the shared epitope (SE) HLA-DR4 and HLA-DR1 alleles was done via a modification of a real-time PCR approach, as described previously [24]. DR4 subtypes that are considered SE positive include DRB1*0401, 0404, 0405, 0408, 0409, 0410, 0413, 0416, 0419 and 0421. DR1 subtypes that are considered SE positive include DRB1*0101, 0102, 0104, 0105, 0107, 0108 and 0111. A subject was considered to be SE positive if one or more allele contained the SE.

n-3 FA biomarker measurement

Percentages of n-3 FAs in red blood cell membranes (n-3 FA% in red blood cells; RBCs) were measured as a biomarker of fatty acid status/exposure at each visit. The protocol for testing n-3 FA% in RBCs has been described previously [21, 22]. Samples were sent to the University of Florida Analytical Toxicology Core Laboratory, where lipids were extracted for measurement of the FAs. FAs were measured as a percentage of the total lipid weight in the RBC sample [(µg FA/µg total lipid) × 100]. The following n-3 FAs were analysed: 18:3 n-3 (ALA), 20:5 n-3 (EPA), 22:5 n-3 (DCPA) and 22:6 n-3 (DHA). We combined the values of ALA, EPA, DCPA and DHA percentage lipid weight to estimate the total n-3 FA% in RBCs. Additionally, we created a variable that summed the two most common types of n-3 FAs found in n-3 FA supplements and fatty fish, EPA and DHA.

Demographic and exposure variables

Self-reported supplement use was collected by a Studies of the aEtiology of Rheumatoid Arthritis questionnaire obtained at each study visit. Participants were asked demographic questions about estimated annual income level, highest educational attainment, sex, race, age and smoking history. For this study, we focused on current smoker status (yes or no) in the analyses as we hypothesized current smoker status would be most relevant to the research questions. We also asked about any supplement use within the past year. If a participant reported yes to use of any supplement, they were then asked about specific supplements. Responses regarding the use of fish oil, fish liver oil or n-3 FA or use of multivitamins that contained n-3 FAs were used to create a binary (yes or no) n-3 FA supplement use variable. We categorized age at baseline as ⩾50 years and <50 years based on our previously reported findings that age ⩾50 years was significantly associated with inflammatory joint signs in RA-free first-degree relatives of RA probands [26].

Determination of IA

The presence of IA at the baseline visit or the subsequent development of IA was determined using data collected at each study visit, and with medical chart review if a subject reported having an examination by a health-care provider outside of the follow-up study. In this study, subjects were considered to have IA if examination demonstrated one or more joints with swelling consistent with RA-like synovitis, which is in line with the number of required joints necessary to classify RA using the 2010 ACR/EULAR criteria [27]. All joints in the 66/68 count exam are included in this definition, except the distal interphalangeals, first carpometacarpals and first metatarsolphalangeals. When an individual presents with the clinical signs and symptoms of IA, they can be further classified as having RA if in addition to IA they meet the 2010 ACR/EULAR criteria [27]. The date of IA or RA classification was determined by the date of the study visit at which these were observed, or the date IA or RA classification was determined by a health-care provider outside of the study, based on medical chart review.

Because the primary outcome of this study is IA, even though a subset of the subjects met RA classification criteria, we refer to subjects with either IA or RA as cases of IA.

Analysis of n-3 FA and prevalent IA at baseline visit

The association between n-3 FA% in RBCs and prevalent IA at baseline was tested cross-sectionally using logistic regression in R [28]. Specifically, we assessed each n-3 FA separately (ALA, EPA, DCPA, DHA), followed by assessment of the sum of EPA + DHA, and all n-3 FAs summed (total n-3 FA%). Each logistic regression model was adjusted for current smoking status (yes vs no), RF⁺, CRP⁺ and self-reported n-3 FA supplement use (yes vs no), as these variables met the operational definition of confounding. Here we define operational confounding as when the adjustment for a covariate of interest resulted in a change in the point estimate of the n-3 FA% variable of at least 10% relative to the crude model with only the n-3 FA% variable. In addition, we adjusted for SE⁺ as it is a well-known risk factor for RA. ORs and 95% CI were estimated for the odds of IA for a s.d. difference in n-3 FA% in RBCs. The s.d. for each n-3 FA% was calculated using the total cohort including all values measured at each visit (i.e. not just the baseline values). s.d. values are listed in the footnote of Table 2.

Analysis of n-3 FA and risk of incident IA

We assessed the association between n-3 FA% in RBCs and risk of incident IA using Cox proportional hazards regression models of the longitudinally collected data on both exposure and the IA outcome, using the R survival package [29]. In the CCP3⁺ subjects without IA at baseline (n = 35), follow-up time was calculated starting at the baseline visit and continued to either date of development of IA or date of last study visit in those subjects who did not develop IA. Our primary predictor variable of interest, n-3 FA% in RBCs, was treated as a time-varying covariate, which allowed it to vary at each study visit so that the n-3 FA% in RBC value at the previous visit contributed to the analysis of the outcome in those who were still at risk for IA at each visit.

Time-varying models were adjusted for n-3 FA supplement use, RF⁺ and CRP⁺ as time-varying covariates as these variables met the operational definition of confounding. Age ⩾50 years at baseline was treated as a fixed variable and met the operational definition of confounding. We also adjusted for SE⁺ as a fixed variable due to its association with RA, and as it was associated with incident IA and met the definition of a precision variable. We were unable to adjust for current smoking in the time-varying models as zero participants who eventually developed IA reported as current smokers. Hazard ratios (HRs) and 95% CI were estimated for the risk of IA for a s.d. difference in n-3 FA% in RBCs; s.d. values for the n-3 FA% in RBC variables are listed in the footnote of Table 4.

Ethical considerations

All study related recruitment and procedures were approved by the Colorado Multiple Institutional Review Board. Study subjects provided informed, written consent.

Results

Association between n-3 FA% in RBCs and IA at baseline visit

Ten of the 47 CCP3⁺ subjects recruited at the health fairs were found to have prevalent IA at their baseline study visit (Fig. 1); of these, eight were classified as RA by 2010 Criteria. Of the 37 CCP3⁺ subjects who were free of IA at baseline, 35 had subsequent follow-up visits and were included in this study. After an average follow-up of 2.56 years, 14 developed incident IA (and 11 of these were classified as RA by 2010 criteria).

Demographic and descriptive characteristics at the baseline visit are shown in Table 1. CCP3⁺ subjects with IA at baseline visit were more likely to be positive for RF and CRP compared with subjects without IA at baseline (Table 1).

Results from our assessment of the association between n-3 FAs and IA at baseline are presented in Table 2. We found that increasing n-3 FA% in RBCs was associated with a lower odds of prevalent IA at the baseline visit, adjusting for n-3 fatty acid supplement use, current smoking status, RF⁺, CRP⁺ and SE⁺. Specifically, a 1 s.d. increase in DCPA% in RBCs was associated with an 81% decrease in the odds of prevalent IA. A 1 s.d. increase in DHA% in RBCs was associated with a 76% decrease in the odds of prevalent IA. A 1 s.d. increase in EPA + DHA% in RBCs was associated with an 80% decrease in the odds of prevalent IA. A 1 s.d. increase in total n-3 FA% in RBCs was associated with a 91% decrease in the odds of prevalent IA.

Association between n-3 FA% in RBCs and incident IA

Baseline demographic and descriptive characteristics by incident IA status are presented in Table 3. Subjects who developed incident IA were more likely to be over the age of 50 years and SE positive.

The results from our time-varying proportional hazards models of the association between n-3 FA% in RBCs and incident IA in CCP3⁺ subjects are presented in Table 4. We found a significant association where a 1 s.d. increase in DCPA% in RBCs was associated with a 48% decreased risk of incident IA, adjusting for age ⩾50 at baseline, SE, RF⁺, CRP⁺ and n-3 FA supplement use (Table 4).

Discussion

In this study of CCP3⁺ individuals without a prior diagnosis of RA, higher levels of longer chain n-3 FAs were associated with a lower prevalence of IA. In addition, we found a significant association between increasing DCPA and decreased risk for incident IA. Our most interesting finding was the consistent association between DCPA and both prevalent IA and risk of incident IA. Our previous study also found that increased DCPA was inversely associated with presence of RF in SE positive subjects without RA [21]. These observations may suggest the potential for an undiscovered role for DCPA in RA pathogenesis.

The long-chain n-3 FAs EPA and DHA are known to modulate the adaptive immune system through several possible pathways. n-3 FA-derivative resolvins and protectins have been shown to actively resolve inflammatory processes [8]. Recent studies have shown DHA-derived resolvins to be protective against leucocyte infiltration, swelling and severity in murine models of IA [30]. EPA-derived resolvins have been shown to be associated with a reduced pain score in persons with IA [31]. In addition to resolvins, modulation of T cells is another pathway of interest, given the important pathogenic role they play in RA [32]. Long-chain n-3 FAs have been shown to alter T cell lipid membrane composition, promote differentiation from Th1 to Th2 and alter signalling pathways [33, 34]. Further research into these pathways could lead to preventive measures to protect against RA development.

DCPA could also be serving as a relatively stable estimate of n-3 FA status, as DCPA% in RBCs is a product of several different factors, including dietary intake, conversion and metabolic function. Conversion from dietary intake of the precursor fatty acids ALA and EPA is an important determinant of levels of DCPA% in RBCs [13, 14]. Also, DCPA makes up a small proportion of dietary long-chain n-3 FAs consumed in Western populations [35–38], and tends to make up only a small portion of the overall n-3 FA content in most fish oil supplements [38]. These factors could suggest that higher levels of DCPA require an n-3 FA dietary profile consistent with long-term and regular consumption of seed and nut oils and fatty fish to provide enough precursor fatty acids for conversion. Whether increased DCPA% in RBCs indicates a different dietary profile, or whether DCPA has any unique and beneficial biological process in RA pathogenesis warrants further investigation.

It is not clear why we did not find statistically significant associations with EPA in the baseline analysis and with EPA and DHA in the incident analysis, as we did for DCPA. However, for both EPA and DHA, the odds ratios were <1.0, suggesting a trend for an inverse/protective association that may require a larger sample size to detect.

We were not able to account for dietary intake of n-3 FAs in this study. Furthermore, we only had a crude measure of self-reported n-3 FA supplement use over the past year, to which nearly 60% of participants reported ‘yes’ at baseline (Table 1). n-3 FA supplements were also positively associated with levels of n-3 FA% in RBCs (data not shown), which suggested confounding bias. This influenced our decision to adjust for n-3 FA supplement use in both the baseline prevalent IA and incident IA study designs. This adjustment was important for the prevalent IA at baseline analysis, and was not as important in the incident IA analysis but was done to keep a harmonious analytic approach (data not shown). It also led us to focus our research question on the biomarker n-3 FA% in RBCs as a better reflection of a participant’s n-3 FA status that accounts for both dietary intake and physiological processes at the point in time in which joint symptoms were assessed.

There is also the possibility that our prevalent IA results are explained by IA status affecting n-3 FAs, where inflammation and a reactive immune system could be using up longer chain n-3 FAs in an attempt to resolve themselves, resulting in lower n-3 FA levels. We attempted to address this possibility through the adjustment of CRP⁺ and RF⁺. Also, the consistency with our incident IA analysis suggests that it is less likely that IA status is affecting n-3 FA levels.

More of the participants with prevalent IA at the baseline visit were current smokers (∼30%), compared with the incident IA group (0%) and the group that did not develop IA over the course of the study (∼9.5%). These data are in line with our knowledge of smoking as an environmental risk factor for RA [26, 39, 40]. In regard to SE⁺, 70% of participants with prevalent IA were SE⁺, which was similar in proportion to the incident IA group (64%); both these groups were higher than 24% in the group that did not develop IA over the course of the study. These findings are in line with our knowledge of SE being the strongest genetic predictor of RA [41, 42]. Our findings with smoking and SE could suggest certain factors are associated with a more rapid or aggressive progression of IA or RA. Similarly, the participants who did not develop IA over the course of the study could represent a group that has a slower transition from ACPA-positivity to clinical symptoms, or a group that does not convert at all. Continued follow-up and expansion of this unique population would allow for an in-depth evaluation of these hypotheses.

Identifying complex relationships between self-reported behaviours (n-3 FA supplement use), biomarkers (n-3 FA% in RBCs and autoantibodies) and clinical symptoms (joint signs consistent with IA) in observational studies is a difficult task. Therefore, experimental studies should be undertaken to understand the mechanistic relationships between n-3 FAs, systemic autoimmunity and IA. Randomized clinical trials should be performed to see if longer chain n-3 FAs might alter the autoimmune response or delay progression of those at increased risk for developing IA. However, our study offers valuable insight regarding the role of n-3 FAs on IA pathogenesis, and further suggests a possible protective effect of n-3 FAs on IA in ACPA-positive persons who do not yet have joint signs, supporting n-3 FAs as a possible preventive agent against development of IA.

Acknowledgements

Special thanks to the 9Health Fair organization for providing the resources to make the initial evaluations possible. We would like to thank all study participants.

Funding: This work was supported by National Institutes of Health [K23 AR051461, T32 AR007534, R01 AR051394 and U01 AI101990]. Additional support was from an Investigator-Initiated Grant from Abbvie to K.D.D. and the Walter S. and Lucienne Driskill Foundation to K.D.D.

Disclosure statement: The authors have declared no conflicts of interest.

References