N-Linked Glycans in the Variable Domain of IgG Anti-Citrullinated Protein Antibodies Predict the Development of Rheumatoid Arthritis.

OBJECTIVE: Anti-citrullinated protein antibodies (ACPAs) are disease-specific biomarkers in rheumatoid arthritis (RA). More than 90% of IgG ACPAs harbor N-linked glycans in the antibody variable (V) domain. The corresponding N-glycosylation sites in ACPA V-region sequences result from somatic hypermutation, a T cell-dependent process. As ample evidence indicates that T cells drive the maturation of the ACPA response prior to arthritis onset, we undertook this study to investigate whether the presence of glycans in IgG ACPA V domains predicts the transition from predisease autoimmunity to overt RA.
METHODS: We analyzed 2 independent sets of serum samples obtained from 126 ACPA-positive first-degree relatives (FDRs) of RA patients. Both sets originated from an Indigenous North American population and comprised cross-sectional and longitudinal samples of individuals who did or did not develop inflammatory arthritis. Serum IgG ACPAs were affinity-purified and subjected to ultra high-performance liquid chromatography-based glycan analysis.
RESULTS: In both data sets, FDR-derived IgG ACPA displayed markedly lower levels of V domain glycans (<50%) compared to IgG ACPA from RA patients. Notably, FDRs who later developed RA showed extensive V-domain glycosylation before the onset of arthritis. Moreover, IgG ACPA V-domain glycosylation was strongly associated with future development of RA (hazard ratio 6.07 [95% confidence interval 1.46-25.2]; P = 0.013).
CONCLUSION: Extensive glycosylation of the IgG ACPA V domain is present in a subset of predisposed FDRs of Indigenous North American RA patients. The presence of this feature substantially increases the risk of RA development. Based on these findings, we propose that glycosylation of the IgG ACPA V domain represents a predictive marker for RA development in ACPA-positive individuals and may serve to better target prevention measures.

Introduction

Rheumatoid arthritis (RA) is a systemic autoimmune disease. Approximately 1% of the global population is affected, but higher prevalence rates have been observed in certain defined populations, such as Indigenous North Americans 1, 2. Indigenous North Americans develop RA at a younger age, experience higher disease burden, have a remarkably high prevalence of the major genetic risk factor for RA (HLA class II shared epitope [SE] alleles) 3, and develop RA that is primarily seropositive for RA‐associated autoantibodies, particularly anti–citrullinated protein antibodies (ACPAs) 4.

It is now well established that ACPAs can be present for many years without evidence of clinical symptoms of RA 5. Notably, ACPA levels, isotype usage, and the citrullinated antigen recognition profile broaden relatively close to the onset of arthritis 6. Thus, it has been postulated that the development of ACPA‐positive disease is a multistep process 7, 8, 9, in which tolerance to citrullinated antigens is initially broken, followed by a putative “second hit” that leads to the expansion of the ACPA response and, ultimately, to development of clinically detectable disease. Hence, it has become of considerable interest to understand the drivers of this predisease expansion of the ACPA response and to identify markers that predict the transition from asymptomatic autoimmunity to ACPA‐positive inflammatory arthritis (IA). Interestingly, recent immunogenetic evidence indicates that HLA SE alleles might contribute to the expansion of the ACPA response by facilitating the provision of T helper cell activity to ACPA‐expressing B cells. This is based on the observation that HLA SE alleles are risk factors for ACPA‐positive disease but do not predispose to the development of ACPA positivity in healthy individuals 10, 11. Thus, it can be hypothesized that in individuals destined to develop RA, ACPA‐expressing B cells receive predisease T cell “help” that initiates and drives B cell maturation, including isotype switching and somatic hypermutation (SHM) 6.

Recently, we found that almost all IgG ACPA molecules carry N‐glycans in their B cell receptor variable (V) domains 12, 13. Notably, all of the consensus N‐glycosylation sequences (Asn‐X [≠Pro]–Ser/Thr) found in these V regions were generated upon SHM and not encoded in the germline genetic repertoire 14. Moreover, we found evidence that the generation of such N‐glycosylation sites offers an advantage to ACPA‐expressing B cells that helps with their escape from selection checkpoints, as these cells acquire extensive somatic mutations despite a lack of avidity maturation. Together, these observations are consistent with the notion that T cells have a pivotal role in the selection and expansion of ACPA‐expressing B cells, possibly by facilitating the introduction of N‐glycosylation sites in IgG ACPA V domains.

Based on these considerations, we hypothesized that the detection of IgG ACPA V‐domain glycosylation in ACPA‐positive individuals is indicative of maturation of this autoimmune response and could potentially serve as a predictor for the development of ACPA‐positive RA. To address this hypothesis, we analyzed IgG ACPA V‐domain glycosylation in a longitudinal manner in unaffected ACPA‐positive first‐degree relatives (FDRs) of RA patients in an Indigenous North American population predisposed to the disease. We determined that there is a heterogeneity of IgG ACPA V‐domain glycosylation levels in unaffected FDRs and that high levels, comparable to those seen in most RA patients, serve to substantially increase the risk of future RA development.

Study population

The current study is nested within a longitudinal research project initiated in 2005 entitled “Early Identification of Rheumatoid Arthritis in First Nations,” that was based at the Arthritis Centre at the University of Manitoba. The study and its consenting process were fully approved by the University of Manitoba Research Ethics Board and the Band Councils of the study's participating communities, the latter based on specific study agreements with the communities 15. Sera were sent in 2 sets, in a blinded manner, to Leiden University Medical Center. The first set was collected from a cross‐sectional cohort consisting of 10 RA patients and 84 ACPA‐positive FDRs. A second set of samples represented a longitudinal cohort of FDRs who had serial visits. As a quality control measure, there was intentional overlap, with 11 FDRs being in both sets (Table 1 and Supplementary Figure 1, on the Arthritis & Rheumatology web site at http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract). Additional information on study design and IA case definition are available in the Supplementary Methods.

Table 1

Patient characteristics*

	Age, mean ± SD years	Female sex, no. (%)	CCP‐3, mean ± SD titer	Follow‐up duration, mean ± SD months
Cross‐sectional cohort
RA patients (n = 10)†	50 ± 10	9 (90)	148 ± 71	–
FDRs (n = 84)	38 ± 14	55 (65)	66 ± 68	–
FDRs included based on IgG ACPA QC (n = 15)	36 ± 12	11 (73)	131 ± 124	–
Longitudinal cohort
Transitioned FDRs (n = 13)‡	34 ± 14	9 (69)	122 ± 120	50 ± 28
Nontransitioned FDRs (n = 19)§	39 ± 15	12 (63)	49 ± 85	64 ± 24
Nontransitioned FDRs included based on IgG ACPA QC (n = 15)	40 ± 18	11 (73)	62 ± 90	45 ± 29

For relationships between donors, quality control, and number of samples, see Supplementary Figure 1 and Supplementary Methods (http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract). CCP‐3 = cyclic citrullinated peptide 3.

Median disease duration of rheumatoid arthritis (RA) patients was 6.8 years (interquartile range 2.9–18.7). After quality control (QC) measures, samples from all 10 RA patients were included for analysis of IgG anti–citrullinated protein antibodies (ACPAs).

After QC measures, samples from all 13 transitioned first‐degree relatives (FDRs) were included for analysis of IgG and analysis of IgG ACPA. The mean ± SD follow‐up duration for samples included for analysis of IgG ACPA was 46 ± 24 months.

After QC measures, samples from all 19 nontransitioned FDRs were included for analysis of IgG.

Isolation of IgG ACPA and of total IgG depleted of ACPA

IgG ACPA and IgG were captured as previously described 16. Briefly, ACPA were purified from 25 μl serum by antigen affinity chromatography using NeutrAvidin Plus UltraLink beads (Thermo Scientific) coated with cyclic citrullinated peptide 2 (CCP‐2). Incubation was performed by shaking the plate for 2 hours at room temperature. The flowthrough was collected by centrifugation. Beads were washed with phosphate buffered saline, and ACPAs were eluted with formic acid and immediately neutralized 16. IgG and IgG ACPAs were subsequently isolated in a similar manner using 20 μl Protein G Sepharose (GE Healthcare Life Sciences). Two hundred microliters of ACPA elution or 15 μl of the flowthrough was loaded onto protein G beads and incubated by shaking at room temperature. Elution was performed with formic acid and centrifugation at 50g for 1 minute.

Glycan analysis

Glycans were isolated and analyzed as previously described 12. Briefly, IgG and IgG ACPA eluates were dried by vacuum centrifugation. Glycans were released using PNGase F (Roche) and labeled with 2‐aminobenzoic acid (2‐AA) and 2‐picoline borane (Sigma‐Aldrich). The 2‐AA–labeled glycans were purified via hydrophilic interaction liquid chromatography–solid phase extraction (HILIC‐SPE) using multiwell filter plates, as previously described 17, 18. HILIC‐SPE–purified 2‐AA–labeled glycans were diluted in 100% acetonitrile and injected into a UHPLC Dionex Ultimate 3000 (ThermoFisher Scientific) equipped with an Acquity UHPLC BEH Glycan column and a fluorescence detector.

Data analysis

HappyTools was used to align, calibrate, and integrate the raw emissions of the chromatograms exported from Chromeleon, version 7.1.2.1713 (ThermoFisher Scientific) 19. The calibrations list, settings, and quality control measurements are provided in the Supplementary Methods and Supplementary Tables [Link], [Link] (http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract). V‐domain glycosylation was calculated using the following formula: percentage V‐domain glycosylation = (V‐domain glycans/Fc glycans) × 100, where the V‐domain glycans were GP19 plus GP23 plus GP24, and the Fc glycans were GP4 plus GP8 plus GP14. We selected the glycans used for the calculation based on our previous observation of their respective, exclusive presence in either the Fc portion or in the V domain 12, 13 (Supplementary Figure 2). IgG ACPA glycan profiles could be obtained for all 10 RA samples from the cross‐sectional cohort; 15 of the 84 FDR samples showed IgG ACPA glycan profiles that met the criteria. In the longitudinal cohort, 67 of the 117 samples from the FDRs yielded sufficient signal to determine the IgG ACPA glycan profile (Table 1 and Supplementary Figure 1, http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract).

Statistical analysis

In the cross‐sectional cohort, the percentage of V‐domain glycosylation between RA patients and FDRs was compared by Mann‐Whitney U test. In the longitudinal cohort, levels of V‐domain glycosylation over time were compared between FDRs who developed RA and those who did not, using linear mixed models with a random intercept and slope. A multivariable Cox proportional hazards regression analysis was performed, with RA diagnosis as outcome and V‐domain glycosylation level as predictor. We used V‐domain glycosylation at the first moment of sampling, dichotomized as above or below the group median to draw Kaplan‐Meier survival curves and estimate risk of developing RA. Receiver operating characteristic curve (ROC) regression was used to calculate diagnostic properties. All analyses in the longitudinal cohort included adjustment for age and sex and were conducted using Stata SE 14.1. All reported percentage values refer to IgG ACPA V‐domain glycosylation levels and not to relative changes. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were calculated.

Lower levels of ACPA V‐domain glycosylation in ACPA‐positive healthy FDRs compared to RA patients

Previously, we demonstrated that IgG ACPAs, in contrast to other antigen‐specific IgG molecules (e.g., IgG against tetanus toxoid), are hyperglycosylated in the V domain in patients with established RA. These glycans are biantennary N‐linked glycans that are fully galactosylated and sialylated 12, 13. Although ACPA can be present before disease onset in individuals who will eventually develop RA, the V‐domain glycosylation state of ACPA has not been studied in the predisease phase. To address this question, IgG ACPAs were isolated from the serum of ACPA‐positive Indigenous North American patients with RA and their unaffected ACPA‐positive FDRs. By analyzing an initial cross‐sectional set of samples, we observed that FDR‐derived IgG ACPA showed, on average, substantially lower levels of V‐domain glycosylation than IgG ACPA from patients with established RA (Supplementary Figure 3A, http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract). Notably, the levels of V‐domain glycosylation of IgG ACPA isolated from Indigenous North American RA patients were comparable to the levels we previously reported in Dutch RA patients (112% versus 93%, respectively) 12. Additionally, tetanus toxoid–specific IgG in Indigenous North American RA patients did not show the enhanced V‐domain N‐glycosylation observed in IgG ACPA, which is consistent with our previous observations in Dutch RA patients (data not shown) 13, 14.

Analysis of a second set of samples derived from longitudinally followed up, unaffected, ACPA‐positive FDRs essentially replicated the cross‐sectional results, which demonstrated low levels of IgG ACPA V‐domain glycosylation compared to those in RA patients (Supplementary Figure 3B, http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract). For subsequent group‐level analyses, we combined the IgG ACPA V‐domain glycosylation data from the initial cross‐sectional cohort with the data from baseline samples from the FDRs included in the longitudinal cohort (Figure 1A). Thus, each individual FDR contributed only once to the analysis, but overall this approach served to increase sample size and statistical power of the analysis.

Figure 1

Anti–citrullinated protein antibody (ACPA)–positive first‐degree relatives (FDRs) have lower levels of IgG ACPA variable (V)–domain glycosylation than do patients with clinically evident rheumatoid arthritis (RA). Percentages of IgG ACPA V‐domain glycosylation in RA patients versus unaffected FDRs (A) and in FDRs who transitioned to inflammatory arthritis versus those who did not (B) are shown. Symbols represent individual subjects; bars show the median. *** = P < 0.001; **** = P < 0.0001.

FDRs displayed heterogeneous levels of IgG ACPA V‐domain glycosylation. This observation prompted us to analyze whether FDRs with high levels subsequently developed clinically detectable IA. As depicted in Figure 1B (and in Supplementary Figures 3A and B, http://onlinelibrary.wiley.com/doi/10.1002/art.40920/abstract), FDRs who transitioned to a state of clinically detectable IA later in life had substantially higher levels of IgG ACPA V‐domain glycosylation compared to individuals who did not transition during follow‐up (89% versus 20%; P < 0.0001). Furthermore, a small portion of FDRs reported new arthralgia; these individuals demonstrated higher IgG ACPA V‐domain glycosylation levels compared to FDRs who remained asymptomatic (Supplementary Figure 3). Therefore, the initially unaffected FDRs who subsequently developed clinically detectable IA or new‐onset arthralgia displayed high levels compared to FDRs who remained in an unaffected or asymptomatic state.

Association of high IgG ACPA V‐domain glycosylation levels in FDRs and development of RA

The observation that unaffected FDRs who developed IA later in life had significantly higher IgG ACPA V‐domain glycosylation levels than FDRs who did not develop IA (Figure 1B) prompted us to determine the time course of glycosylation in these individuals. Intriguingly, we observed that V‐domain glycosylation levels were already increased in some FDRs up to 7 years before the onset of arthritis. Other individuals who developed IA showed an increase in glycosylation ~1–2 years before disease onset (Figure 2A). When all available time points were considered, IgG ACPA V‐domain glycosylation was 34% higher (95% CI 7–62%) in FDRs who developed IA compared to FDRs who did not (P = 0.015) (Figure 2B). Notably, FDRs with new‐onset arthralgia were classified as nontransitioned. Taken together, these data indicate that ACPA‐positive individuals who subsequently developed IA displayed a high degree of IgG ACPA V‐domain glycosylation in the preclinical phase. The time course for development of IA, however, varied considerably between individuals.

Figure 2

IgG ACPA V‐domain glycosylation is elevated in FDRs who transitioned to RA during follow‐up. A, Longitudinal data on IgG ACPA V‐domain glycosylation. Time of RA diagnosis is shown with a red cross; lines without crosses indicate that the patient transitioned to RA but did not have a sample obtained at the time of diagnosis. B, Median IgG ACPA V‐domain glycosylation levels in FDRs who transitioned to RA versus those who did not. Bars show the median and range. Dotted lines indicate the range of IgG V‐domain glycosylation observed in healthy individuals. See Figure 1 for definitions.

Absence of elevated V‐domain glycosylation in IgG ACPA

The degree of V‐domain glycosylation of IgG ACPA is remarkably high compared to other IgG molecules. To date, we have been unable to identify another autoreactive B cell response with comparably high levels of V‐domain glycans 12, 13. Furthermore, among the Dutch RA population, non‐ACPA IgG V‐domain glycosylation levels were only 17%. To confirm that non‐ACPA IgG V‐domain glycosylation is also low in the Indigenous North American population and to investigate a potential association with clinical outcome, we used the flowthrough of the ACPA isolation columns to further purify non‐ACPA IgG from FDR‐derived sera, followed by glycosylation analysis as it was performed for IgG ACPA. V‐domain glycosylation of non‐ACPA IgG was as low as 15–25% in these samples, with a mean of 17% for FDRs who did not develop IA and 20% for those who did (P not significant) (Figure 3). These levels are consistent with published data on IgG in healthy individuals 12, 20. Moreover, the degree of V‐domain glycosylation remained low over time and showed no differences in mean levels between transitioning and nontransitioning FDRs (P = 0.33). These data suggest that increased V‐domain glycosylation is not a feature of non‐ACPA IgG and that this phenomenon is inherent in the citrulline‐directed immune response.

Figure 3

IgG depleted of ACPA does not display enhanced levels of V‐domain glycosylation. A, Longitudinal data on IgG V‐domain glycosylation. Time of RA diagnosis is shown with a red cross; lines without crosses indicate that the patient transitioned to RA but did not have a sample obtained at the time of diagnosis. B, Median IgG V‐domain glycosylation levels in FDRs who transitioned to RA versus those who did not. Bars show the median and range. Dotted lines indicate the range of IgG V‐domain glycosylation observed in healthy individuals. See Figure 1 for definitions.

Elevated IgG ACPA V‐domain glycosylation level as a predictive marker for disease development in ACPA‐positive at‐risk subjects

Given that FDRs who developed IA had higher levels of IgG ACPA V‐domain glycosylation over time compared to those who did not, we attempted to create models of glycosylation as a predictor of IA development. To account for the variance between time points and duration of follow‐up of available study subjects, 2 statistical models were established. In the first model, we used the median IgG ACPA V‐domain glycosylation level in the cross‐sectional analysis (58%) as a cutoff point, thus generating a binary variable. Cox regression analysis, corrected for age and sex, revealed that IgG ACPA V‐domain glycosylation levels above the median were associated with development of IA (HR 6.07 [95% CI 1.46–25.2]; P = 0.013) (Figure 4). A similar cutoff point was created when ROC regression was used to determine the predictive value of high IgG ACPA V‐domain glycosylation levels. At this cutoff point, IgG ACPA V‐domain glycosylation had a sensitivity of 76.9% (95% CI 46.2–95.0%), a specificity of 78.6% (95% CI 49.2–95.3%), a positive predictive value (PPV) of 76.9% (95% CI 46.2–95.0%), and a negative predictive value (NPV) of 78.6% (95% CI 49.2–95.3%) for IA development.

Figure 4

Development of RA in ACPA‐positive FDRs, based on degree of IgG ACPA V‐domain glycosylation at the first instance of sampling. IgG ACPA V‐domain glycosylation levels below and above the 58.5% median are shown. HR = hazard ratio; 95% CI = 95% confidence interval (see Figure 1 for other definitions).

Discussion

IgG ACPA represents the most relevant prognostic and diagnostic biomarker in RA and is associated with poor prognosis and progressive joint destruction 21. A large body of evidence indicates that early intervention in RA improves clinical outcomes 22, 23. Based on these observations, it is hypothesized that intervention at a stage prior to the onset of clinically detectable arthritis may result in better clinical outcomes and possibly even prevent disease 22. Currently, this preclinical phase is identified based on the presence of suggestive joint symptoms such as arthralgia, the detection of RA‐associated autoantibodies, especially ACPA, or both of these factors. Imaging studies indicate that this clinical phenotype may also feature subclinical synovitis 22, 24. Although arthralgia in conjunction with a broad autoantibody response is a strong predictor of imminent RA onset, a linear progression toward these features has been difficult to demonstrate 25. In particular, the existing longitudinal data suggest that a considerable proportion of individuals with ACPA do not develop RA. Thus, it has become increasingly important to delineate biomarkers that can be used to improve the risk model and that in turn provide actionable clinical information on which interventions can be based 24, 26. Our current study demonstrates that the glycosylation state of the IgG ACPA V domain may serve such a purpose. We found that the presence of extensive IgG ACPA V‐domain glycosylation in unaffected ACPA‐positive individuals is a strong predictor of progression toward disease. Indeed, a high level of IgG ACPA V‐domain glycosylation had a PPV of 76.9% and an NPV of 78.6% for predicting the development of IA.

Considering these clinical implications, the present data refine our understanding of the ACPA immune response and its predisease evolution. Conceptually, the development of ACPA‐positive RA has been proposed as a multistep process 7, 8 in which the break in immune tolerance to citrullinated antigens develops first, followed by expansion and maturation of the autoimmune response, a so‐called second hit. Recent evidence suggests that this putative second hit is driven by T helper cells that provide help to ACPA‐expressing B cells. This T cell–B cell interaction, which likely mediates the increased usage of isotypes and epitope spreading observed before disease onset, may also be responsible for the introduction of N‐glycosylation sites in the V domain of IgG ACPA and, therefore, the presence of glycans in this region. This hypothesis is further supported by the observation that the consensus sequences for N‐glycosylation sites in IgG ACPA V domains are not encoded in their respective germline genes but introduced upon SHM 14. In fact, the presence, frequency, and distribution of these sites suggest a role of the glycans in the selection, development, and activation of the B and plasma cells that express/secrete IgG ACPA.

Notably, the ACPA‐positive FDRs examined here exhibited lower levels of IgG ACPA V‐domain glycosylation than their relatives with RA, but there was considerable heterogeneity in the population. As a matter of fact, a subset of FDRs had high levels of V‐domain glycosylation, which in some cases was detectable years before disease onset. Individuals who exhibited fluctuation between low and normal levels over time (15–25%) rarely transitioned to disease. Given this heterogeneity, we used the available longitudinal samples to develop a kinetic understanding of how IgG ACPA V‐domain glycosylation evolves in individuals who ultimately develop IA. Although the data are limited by the small number of those who transitioned to disease, it is clear that there is more than one trajectory and time course for disease development, with some individuals demonstrating a rapid increase in glycosylation and others showing high levels for an extended period of time. These observations suggest that the acquisition of N‐glycans in the ACPA V domain is a process that requires repeated T cell–dependent B cell hypermutation events, potentially as a result of multiple hits that occur with varying kinetics. Ultimately, these may lead to the persistence, if not the pathogenicity, of the ACPA response.

Our study is limited by the relatively small number of samples with longitudinal follow‐up. Also, the number of samples with low‐level ACPA that did not pass quality control may have introduced bias. However, except for ACPA levels, the patients who could be included in the analysis had similar baseline characteristics compared to those who had to be excluded because of technical limitations (Table 1). Furthermore, in samples that passed our strict controls, estimates of the predictive value of IgG ACPA V‐domain glycosylation were substantial and hold promise for clinical application. Nonetheless, additional studies are necessary to validate and extend these findings.

Finally, our study was conducted exclusively in an Indigenous North American population. Seropositive RA in this population was associated with HLA–DRB1 SE–encoding alleles, as it is in most other populations worldwide. The primary SE allele in the Indigenous North American population is HLA–DRB1*14:02, which is prevalent in the background population and almost unique to Indigenous North Americans. In contrast to other RA‐predisposing SE alleles such as HLA–DRB1*04:01, which is seen primarily in white populations, HLA–DRB1*14:02 can accommodate both citrulline and arginine peptides with comparable affinity, but the orientation of citrulline‐containing peptides is upright and directly interfaces with the T cell receptor 27, 28. It is unclear whether these differences in peptide presentation to T cells impact their capacity to provide help for ACPA‐expressing B cells. However, the observation that ACPA V‐domain glycosylation patterns are comparable in Indigenous North American and Dutch RA patients suggests that, irrespective of how T cell autoimmunity develops, the T cell–dependent SHM of ACPA B cells is a final common mechanism in the evolution of the ACPA response. Notably, tetanus toxoid–specific IgG did not show enhanced V‐domain glycosylation in the Indigenous North American RA patients, in accordance with our previous observations in Dutch RA patients and with the absence of additional N‐glycosylation sites in B cell receptors of tetanus toxoid–specific B cells 13, 14.

In summary, we have shown that IgG ACPA V‐domain glycosylation is a strong predictive biomarker for the development of ACPA‐positive RA. Our findings have important implications for assessing the risk of future RA development in unaffected ACPA‐positive individuals and, in turn, for stratifying these individuals for intervention studies. The results also provide mechanistic information regarding the evolution of preclinical RA autoimmunity. Future studies in this and other populations will be needed to determine the ultimate clinical utility of these findings.

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. Scherer 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

Hafkenscheid, Tanner, Huizinga, Toes, El‐Gabalawy, Scherer.

Acquisition of data

Hafkenscheid, Smolik, Tanner, Meng, Wuhrer.

Analysis and interpretation of data

Hafkenscheid, de Moel, Jansen, Bondt, Wuhrer, Huizinga, Toes, El‐Gabalawy, Scherer.

Additional Disclosures

Author Jansen is an employee of Ludger Ltd.

Supplementary Figure S1

Click here for additional data file.

Supplementary Figure S2

Click here for additional data file.

Supplementary Figure S3

Click here for additional data file.

Supplementary Table S1

Click here for additional data file.

Supplementary Table S2

Click here for additional data file.

Supplementary Table S3

Click here for additional data file.

Click here for additional data file.