Autoantibody clusters in rheumatoid arthritis are not driven by antigen specificity or isotype

Abstract

Objective

Autoantibodies are a key feature of rheumatoid arthritis (RA). They can be detected years before disease onset, but it is unknown if there is any pattern in the co-occurrence of antigen recognition or isotype profiles. A common signature could point to a unique initial trigger for autoantibody development. Therefore, we sought to determine if there is a pattern in antigen or isotype reactivity in pre-symptomatic cases and established RA.

Methods

One pre-symptomatic cohort and one RA cohort were analysed for the co-occurrence of different isotypes of anti-modified protein antibodies (AMPA) and rheumatoid factor (RF). Patterns in autoantibody levels were investigated with clustering. Additionally, total IgG was measured in 1- year follow-up sera of a representative subgroup of the RA cohort.

Results

While especially anti-citrullinated protein antibodies (ACPA) IgG and RF IgA co-occurred with other autoantibodies, no specific patterns emerged. In both cohorts, clusters of autoantibody levels were not determined by particular antigen reactivities or isotype. However, clusters were driven by elevated levels of several different AMPA, with distinct AMPA high- and low-level clusters. A broad IgG autoantibody profile was not accompanied by high total IgG levels.

Conclusion

Autoantibody clusters are most likely not driven by AMPA specificity or isotype profile, neither before nor at RA onset, but are instead determined by a broad variety of autoantibodies. This indicates that the triggers for autoantibody development in RA do not skew the response towards certain autoreactivities or isotypes but rather lead to a broad and diverse autoantibody repertoire reflecting continuous and ongoing immune activation.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Anti-modified protein antibodies are arguably the best markers of the underlying disease process in rheumatoid arthritis (RA). They can be detected years before disease onset and are associated with a more severe disease course.

• It is unknown if there is any specific pattern in the co-occurrence of antigen-directed reactivities or isotype profiles.

WHAT THIS STUDY ADDS

• Triggers for autoantibody development in RA do not skew the response towards a preferred reactivity or a shared isotype profile but rather lead to a broad and diverse autoantibody repertoire reflecting continuous and ongoing immune activation.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Autoantibody profiles in RA are useful as a marker of disease progression but might not lead us to the original trigger of the disease.

Introduction

Autoantibodies are a key feature of rheumatoid arthritis (RA). These antibodies target various components, including the Fc tail of IgG (rheumatoid factor or RF) and post-translational modifications (PTMs) (anti-modified protein antibodies or AMPA). AMPA, anti-citrullinated (ACPA), anti-carbamylated (anti-CarP) and anti-acetylated protein antibodies (AAPA) are arguably the best markers of the underlying disease process. They can be detected years before disease onset, and ACPAs are even associated with a more severe disease course and linked to the most prominent genetic risk factors for RA.^{1 2}

Previous research has demonstrated that, in the period before disease onset, the autoimmune response matures through processes such as epitope spreading and an increase in AMPA levels.^{3 4} This maturation results in the presence of RF, AAPA, ACPA and anti-CarP with various isotype combinations and a stable composition after disease onset.⁵ Some individuals develop a broad antibody profile, which is associated with better treatment responses but not with more favourable long-term outcomes.

Despite prior research into the presence of AMPA and RF, it remains unknown if there is any specific pattern in the co-occurrence of antigen-directed reactivities or isotype profiles. A common signature might provide more insight into the initial triggers shaping the autoimmune response. For example, co-occurrence of AMPA of the IgA isotype could support a mucosal origin since IgA is the most prevalent antibody at mucosal sites.⁶ Furthermore, if IgA antibodies have preferred antigenic targets, this would indicate that mucosal exposure to these targets plays a role in disease development. Therefore, it is important to elucidate if distinct autoantibody signatures exist in RA, as well as in pre-symptomatic individuals.

For this reason, we sought to determine patterns in isotype usage or reactivities of RA-associated autoantibodies, using a unique cohort of pre-symptomatic cases and a cohort of established RA. Both cohorts underwent analysis for the co-occurrence of autoantibodies regarding isotypes and autoantigen-directed reactivity and for patterns in autoantibody levels. In this manner, we hope to obtain more insights into the development and origin of the autoimmune response in RA.

Methods

This study used two separate cohorts: (1) 333 patients fulfilling the American College of Rheumatology 1987 criteria for RA from the Leiden Early Arthritis Clinic (EAC).^{7 8} (2) 68 pre-symptomatic individuals, all of whom developed RA in time, described in the online supplemental file 1.8,10 Co-occurrence of autoantibodies was investigated using data from ELISA measurement determining ACPA and RF (IgG, IgA and IgM) in the pre-symptomatic individuals and IgG, IgA and IgM for AMPA (ACPA, anti-CarP and AAPA) and RF IgA and IgM in the RA cohort. In addition, autoantibody clusters were assessed using the Louvain method. Total IgG was measured in 1-year follow-up sera of a representative subgroup of the RA cohort.¹¹

Results

Co-occurrence of epitope specificity or isotype

Using the data of all ELISA measurements, we investigated the co-occurrence of AMPA and RF. The percentages of autoantibody positivity are listed in online supplemental table S1. Figure 1 shows the pattern for the presence of RF, AMPA and the isotype usage. Depicted are the percentages of participants positive for the autoantibody on the vertical axis who are also positive for the autoantibody on the horizontal axis; for example, 25% of ACPA IgA-positive pre-symptomatic participants are also positive for ACPA IgM. Black frames indicate antibodies belonging to the same (PTM-directed) reactivity family. Surprisingly, most autoantibodies or isotypes did not co-occur more frequently within their family (i.e., the same reactivity or isotype) than outside their respective family (i.e., another reactivity or isotype) in either cohort. In contrast, co-occurrence rather seemed to depend on the underlying autoantibody prevalence and was thus most frequently observed with the most prevalent autoantibodies (i.e., ACPA IgG and RF IgA in both cohorts and RF IgM and anti-CarP IgG in the RA cohort).

Figure 1. Autoantibody profiles in the pre-symptomatic and RA cohort. Depicted is the percentage of participants positive for the autoantibody on the vertical axis who are also positive for the autoantibody on the horizontal axis; for example, 25% of ACPA IgA-positive pre-symptomatic individuals are also positive for ACPA IgM. AAPA, anti-acetylated protein antibodies; ACPA, anti-citrullinated protein antibodies; anti-CarP, anti-carbamylated protein antibodies; RA, rheumatoid arthritis; RF, rheumatoid factor.

Cluster analysis

Since dichotomising autoantibody levels into positive and negative may lead to loss of information, we also explored the levels of these autoantibodies with a clustering analysis. This revealed three clusters in the pre-symptomatic and six clusters in the RA cohort shown in a Uniform Manifold Approximation and Projection (UMAP) in figure 2A. Our clustering method repeatedly showed three and six clusters, respectively, which argues in favour of the robustness of the methodology. The orange and blue clusters in the UMAPs differ in their autoantibody levels, which are depicted in figure 2B. The cluster formations were again not determined by any particular family of antigen reactivity or isotype. Instead, these clusters were driven by elevated levels of several different AMPA and RF, with distinct high levels, blue clusters, and low levels, orange clusters. Because the clearest differences existed between low-level versus high-level clusters, we have focused on this distinction. Online supplemental figure 1 illustrates the significant difference between these clusters for ACPA IgG levels. These level clusters emerged in pre-symptomatic cases as well as in RA. For the RA cohort, we also investigated clustering of only the seropositive individuals, which resulted in similar clusters (online supplemental figure S2). To investigate whether patient characteristics differed among high-level versus low-level clusters, we compared age, body mass index (BMI), female (%), smoking (ever), shared epitope (SE), DAS28, erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) between these groups (online supplemental table S2). While there were no differences in the pre-symptomatic cohort, in the RA cohort, patients in high-level clusters were more often HLA SE-positive, smokers and had higher ESR levels.

Figure 2. A clustering analysis and UMAP visualisation of autoantibody-level data. (A) Three clusters are formed in the pre-symptomatic cohort (left), and six clusters are formed in the RA cohort (right). The blue clusters have high autoantibody levels, and the orange clusters have low autoantibody levels. (B) The level data in the various clusters are shown for all autoantibodies separately (orange=ACPA, pink=RF, blue=anti-CarP and grey=AAPA). The y-axis is normalised with the highest occurrence set at 100%. The x-axis depicts the different AMPA isotypes. AAPA, anti-acetylated protein antibodies; ACPA, anti-citrullinated protein antibodies; AMPA, anti-modified protein antibodies; anti-CarP, anti-carbamylated protein antibodies; RA, rheumatoid arthritis; RF, rheumatoid factor; UMAP, Uniform Manifold Approximation and Projection.

Total IgG levels in autoantibody positivity

To investigate whether the presence of high-level autoantibodies and broad autoantibody profiles may be due to overall high immunoglobulin production, we examined this for IgG production in a subpopulation of the RA cohort. As shown in table 1, a broad IgG autoantibody profile was not accompanied by high total IgG levels.

Table 1. Number of participants from a 1-year follow-up in the EAC subpopulation who are positive for 0, 1, 2 or 3 IgG autoantibodies and the mean (M ± SD) total IgG serum levels for every group. EAC, Early Artritis Clinic.

Number of IgG autoantibodies	Participants positive	Total serum IgG levels (M ± SD)
0	5	16 ± 9.1
1	44	12 ± 4.4
2	33	14 ± 5.7
3	19	15 ± 5.6

Discussion

The current study assessed the presence of patterns in autoantibody reactivity in two defined stages of disease by investigating autoantigen reactivity and isotype specificities. We discovered that autoantibody clusters were not driven by RF, AMPA specificity or isotype profile, neither before nor at disease onset. Rather, autoantibody clusters appear to be determined by overall high versus low levels of various different autoantibody specificities and isotypes. Differences between high- and low-level antibody groups were corroborated by the association of the high antibody-level groups with known risk factors (shared epitope and smoking) and ESR, as described previously.¹² Broad IgG autoantibody profiles with high autoantibody levels were not accompanied by, or attributable to, high total IgG levels. This suggests that the presence of autoantibodies is independent of the total level of antibody production.

Our results regarding the co-occurrence of autoantibody reactivities align with previous studies, indicating the absence of one dominant antigen or isotype driving the autoantibody response in RA.^{3 4} This is in contrast to other autoimmune diseases where immunodominant epitopes have been described and epitope shifting is known to play a key role. For example, in Fogo Selvagem, an autoimmune disease characterised by skin blistering and anti-demoglein-1 antibodies, individuals pre-disease demonstrated reactivity against only one individual epitope of anti-demoglein-1.¹³ However, epitope shifting to other domains of the same protein led to disease onset, illustrating the potential importance of epitope shifting. Preferential epitope recognition before disease was not observed in RA, indicating that the development of the autoimmune response appears to work differently in other autoimmune diseases.

An additional aim of this study was to explore isotype co-occurrence. Similar to our findings regarding reactivity to antigenic determinants, there was no predominant presence of an isotype. Nonetheless, the absence of isotype dominance does not necessarily indicate the absence of a site at which the autoimmune response is initiated, since we cannot exclude that quick maturation and isotype switching could have obscured a distinctive initial autoantibody signature. Yet, considering that we also could not detect an isotype dominating signature in the pre-disease samples, the data overall do suggest simultaneous development of multiple autoantibodies.

Intriguingly, our data indicate that approximately half of the RA patients have an ongoing and active autoimmune response marked by elevated levels of a broad variety of autoantibodies. An explanation for the presence of this broad antibody profile in the RA cohort could be the known AMPA cross-reactivity, whereby antibodies reacting to citrullinated proteins could display anti-carbamylated and anti-acetylated protein reactivity.¹⁴ Thus, the initial reactivity to one PTM could be obscured by the promiscuous nature of the AMPA response.

Our study has some limitations. First, as discussed above, a clear pattern might have partly been precluded by the known cross-reactivity of AMPA and the low avidity of this response.¹⁵ However, it would not be expected to have concealed a dominant isotype co-occurrence. Another limitation is that not all the autoantibodies were measured in both cohorts. We cannot exclude that patterns would have emerged in the presymptomatic cohort if all AMPA had been measured, although this appears unlikely based on the present data. Strengths of the current study include the measurement of many different autoantibodies and isotypes, the use of binary data as well as level data for clustering analysis, and the inclusion of both pre-symptomatic individuals as well as RA, providing important insights into immunological patterns during disease development.

In summary, autoantibody clusters in RA are most likely not driven by reactivity or isotype. Instead, they can rather be distinguished by a broad autoantibody profile, indicating an ongoing immune response. These data indicate that the triggers for autoantibody development in RA do not skew the response towards a preferred reactivity or a shared isotype profile but rather lead to a broad and diverse autoantibody repertoire reflecting continuous and ongoing immune activation.