Hierarchy in the Avidity and Cross-reactivity of Anti-citrullinated Protein Antibodies in the Serum of Patients With Rheumatoid Arthritis

Fine specicity of anti-citrullinated protein antibodies (ACPAs), in which cross-reactivity exists, varies among patients with rheumatoid arthritis (RA), but it is unclear whether the mechanism of ACPA production is same or different among individuals. Since avidity of serum antibody reects the direction of immune response, we compared the levels of avidity and cross-reactivity between various ACPAs in a cohort of RA patient. Sera from 180 RA patients positive for anti-cyclic citrullinated peptide (CCP) 2 antibody were screened for positivity of antibodies against CCP1, and citrullnated brinogen (cFib), enolase (cEno), and vimentin (cVim) peptides. Avidity of the four ACPAs, and some autoantibodies and antibodies against foreign antigens was determined by an elution assay using sodium thiocyanate solution. Cross-reactivity between different ACPAs was estimated by measuring the inhibition of binding by competitor peptides.

varies among patients. This might re ect inter-patient variation in the repertoire of citrullinated antigens generated during pre-RA stage. Alternatively, a common antigen induces cross-reactive ACPA response with different e ciency for each citrullinated antigen, which might be in uenced by genetic, environmental as well as stochastic factors, causing individual variations. In fact, at poly-and monoclonal levels, varying extent of cross-reactivity among different ACPAs has been demonstrated (7)(8)(9)(10)(11)(12)(13). Thus, it is unclear whether the mechanism of ACPA production is same or different among RA patients.
In the germinal center (GC) of lymphoid follicles, only B cells expressing BCR with binding a nity above the selection threshold can survive (14). As a result, along with the progression of immune response, avidity of the antibody to the target antigen increases, the a nity maturation. Thus, the avidity of antibody indicates the direction of immune response, and therefore measuring avidity of ACPA could be a clue to identify the initiating antigen. In fact, analysis of ACPA-producing B cell clones revealed highly mutated heavy-and light-Ig chain genes (8)(9)(10)(11)(12). Reverting the mutation to the germ-line sequence lost the binding to the citrullinated peptides, further indicating the importance of a nity maturation (9,10,13,15). On the other hand, it has been shown that the avidity of ACPA is lower than antibodies against recall antigens (16). This might be because the immune response producing ACPA is ongoing, and newly recruited B cells are continuously involved in the antibody production. Alternatively, self-tolerance mechanisms prevent the production of high-avidity ACPA, although it is unknown whether low-avidity binding is unique to ACPA or common to autoantibodies in general.
Cross-reactivity of ACPA might be another explanation for the low avidity as well as the diverse speci city. In this regard, if a dominant antigen, which is common to all patients, induces the whole ACPA response, there should be a correlation or a xed hierarchy in the avidity of different ACPAs in the serum.
On the other hand, if each ACPA response, which cross-reacts with other citrullinated antigens to some extent, is induced by different antigen or if the antigen is different among patients, the hierarchy in the avidity of ACPAs may vary according to the maturation level of each immune response. Furthermore, in this case, avidity might be higher in the antibody with lower cross-reactivity, i.e. higher speci city.
In order to understand the mechanism of ACPA production, we compared the avidity of various ACPAs in the serum of a cohort of RA patients, based on the assumption that avidity of antibodies in the serum re ects the direction of immune response. We also examined the relationship between avidity and cross-reactivity of different ACPAs.

Patients
Two hundred and thirty six consecutive serum samples from RA patients, who ful lled the 1987 American College of Rheumatology (ACR) classi cation criteria (17), were screened, and those positive for CCP2 were used in the subsequent analysis (n=180). Subjects with other autoimmune diseases (Sjogren's syndrome, n=11; systemic lupus erythematosus (SLE), n=8; chronic thyroiditis, n=4, mixed connective tissue disease (MCTD), n= 7; CREST syndrome, n=12) were also included for the analysis. To de ne the cutoff levels of the antibodies, 40 healthy control subjects were recruited. Serum was isolated and stored at −20°C until use. This study complies with the Declaration of Helsinki, and the study protocol was approved by the Regional Committee of Ethics for Human Research at the Faculty of Medicine of the Kyushu University . All subjects provided written informed consent before participating in the study.

Enzyme-linked immunosorbent assay(ELISA)
To detect antibodies speci cally bound to CCP1, citrullinated brinogen (cFib), citrullinated a-enolase (cEno), and citrullinated vimentin (cVim) peptides, each pairs of citrullinated and control non-citrullinated peptides were used (Supplementary Table 1). All peptides were produced through amino acid synthesis and conjugated with biotin at the C-terminal. Ninety-six well avidin-coated plates (Avidin Plate, Sumitomo Bakelite Co., Ltd., Tokyo, Japan) were incubated with the peptides (2 mg/mL) overnight at 4°C, and after blocking non-speci c binding by an incubation with Blocking One solution (Nacalai Tesque, Kyoto, Japan), the plates were incubated with the serum diluted at 1:30 by LowCross-Buffer (CANDOR Bioscience GmbH, Wangen, Germany), for 2 h at 37°C. Peroxidase-conjugated F(ab) 2 goat anti-human IgG (Rockland Immunochemicals, Limerick, PA, USA) was used for the detection. Antibody binding was visualized using tetramethylbenzidine substrate solution (Interchim, Montluçon, France), and the absorbance was measured at 450 nm. The antibody level was expressed as DOD between citrullinated and native peptides. Sera with DOD higher than the mean plus 2SD of the healthy control sera were considered positive. To measure the levels of IgG against CCP2, we used a commercially available ELISA kit (Euro-Diagnostica, Malmö, Sweden). For the detection of anti-in uenza, and anti-diphtheria toxin antibodies, 96-well plates were coated with 1 μg/ml of A/H1N1 subunit (Bio-Rad, Hercules, CA, USA) and diphtheria toxin (List Biological Laboratories, Campbell, CA, USA), respectively. We utilized precoated commercial kits to measure the avidity of anti-SS-A/Ro, anti-dsDNA, anti-centromere, and anti-U1-RNP antibodies (all from IBL INTERNATIONAL, Hamburg, Germany).

Measuring avidity of antibodies
Avidity of antibodies was determined by an elution assay using sodium thiocyanate (NaSCN) according to a previous report by Suwannalai et al. (16) with minor modi cations. Serum was serially diluted (1:25, 1:50, 1:100, 1:200, 1:400) with Low cross buffer (CANDOR Bioscience GmbH) to de ne the dilution at which the antibody response was in the linear part of the curve. Sera with OD value of native peptide higher than 0.5 were excluded from the analysis in order to ensure data quality. Antigen-coated plates were incubated with the diluted serum for 2 h at 37°C. After washing, the plate wells were incubated with NaSCN at the concentration of 0, 0.125, 0.25, 0.5, 1.0, 2.0 or 4.0 M for 15 min at 37°C. The plates were washed, and remaining bound antibodies were detected using peroxidase-conjugated goat anti-human IgG. We calculated the avidity index (AI) at each concentration of NaSCN as follows: (OD value of serum treated with given M of NaSCN -OD value of serum treated with 4M of NaSCN) / (OD value of serum treated with 0 M of NaSCN -OD value of serum treated with 4 M of NaSCN) × 100.
The strength of avidity is depicted as AI AUC, the area under the curve of AI at seven reference points, in order to minimize the bias in the avidity of antibodies with different concentrations (dilutions). AI AUC was calculated according to Perciani

Inhibition assay
The extent of cross-reactivity between different ACPAs was examined by a peptide inhibition assay as reported previously with minor modi cation (7). In brief, serum in appropriate dilution at which the antibody response was in the linear part of the curve as de ned above, were incubated with 100 mg/ml of competitor peptide (non-biotinylated peptide) overnight at 4°C with continuous mixing. Then the serum samples were incubated in the plates precoated with the biotinylated test peptides, and subsequent ELISA steps were performed as described above. Only sera with OD value lower than 0.5 for the native peptides and those with DOD value larger than 0.2 were analyzed. The inhibition rate of antibody binding was calculated as follows: (OD value of untreated serum -OD value of serum pretreated with competitor peptide / OD value of untreated serum) × 100.

Statistics
All statistical analyses were performed with EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan) (19), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria). Difference between groups was analyzed by Steel-Dwass test, and correlation was determined by Spearman's correlation coe cient. P values less than 0.05 were considered signi cant.

Results
Prevalence of various ACPAs in a cohort of Japanese RA patients We rst examined the prevalence of various ACPAs in a cohort of Japanese RA patients who were positive for anti-CCP2 antibody (n=180). Baseline characteristics of these patients are shown in Table 1.
To exclude false positive samples from the analysis, we expressed the antibody levels after subtracting the values of non-citrullinated peptides. The prevalence of anti-CCP1, cFib, cEno, and cVim antibodies were 37.7%, 38.3%, 15.6%, and 23.9%, respectively ( Figure 1A). Sixty one percent (n=110) of the serum samples reacted at least one peptide, in which 44.5% (n=49) reacted to single peptide only, while 7.3% (n=8) reacted to all four peptides ( Table 2). There is a positive trend between the number and the level of each ACPA, as reported previously ( Figure 1B) (20). However, there was no correlation in the levels between different ACPAs in any combination ( Figure 1C). Avidity of different ACPAs and other autoantibodies We next measured the avidity of serum antibodies to the four citrullinated peptides by an elution assay using gradients of NaSCN. In order to minimize the bias in comparing the avidity of antibodies with different concentrations (dilutions), the values are expressed as AI AUC (Supplementary Figure 1). In addition to ACPA, we also measured avidity of some autoantibodies, including anti-dsDNA, anti-SS-A, anti-U1-RNP, and anti-TPO, and antibodies against foreign antigens, including, in uenza and diphtheria toxin ( Figure 2). We found that the avidity of ACPA, except for anti-cVim antibody, was signi cantly lower than that of antibodies against foreign antigens. There was no difference in the avidity level between anti-CCP, anti-cFib, and anti-cEno antibodies. Among the autoantibodies examined, anti-dsDNA antibody showed the lowest avidity, even lower than any ACPA. Anti-TPO and anti-centromere antibodies, on the contrary, tended to show higher avidity than antibody against foreign antigens. Thus, low avidity is not a general feature of autoantibodies but might be unique to ACPA, in which anti-cVim antibody showed higher avidity than others.

The relationship in the avidity of different ACPAs
We analyzed the relationship in the avidity of different ACPAs at an individual level using serum samples reactive to more than two citrullinated antigens. Although, no correlation was detected in the levels ( Figure 1C), there was a signi cant correlation in the avidity between anti-CCP1 and anti-cFib antibody ( Figure 3). A trend of correlation was observed in the avidity between anti-CCP1 and anti-cEno or anti-cVim antibodies, but it was not statistically signi cant. Furthermore, we con rmed that the avidity of anti-cVim antibody was higher than anti-CCP1, cFib, and cEno antibodies at individual levels. Thus, there is a correlation or xed hierarchy in the avidity of different ACPAs. We observed no correlation or xed hierarchy in the avidity of different antibodies against recall antigens (Supplementary Figure 2).

The extent of cross-reactivity between different ACPAs
We lastly examined the relationship between the levels of avidity and cross-reactivity. The extent of crossreactivity was examined by an inhibition assay, in which reduction rate of binding to an antigen by preincubation with another antigen was measured. Although there was a substantial variation in the extent of cross-reactivity, it exceeded 50% in most cases ( Figure 4). Interestingly, there was a clear hierarchy in the cross-reactivity in some combinations of peptides. Thus, preincubation with CCP1, cFib or cEno inhibited binding to cVim more e ciently than the opposite, indicating anti-cVim antibody is more cross-reactive than the others. Similarly, serum binding to CCP1 was inhibited by preincubation with cEno more e ciently than did the preincubation with CCP1 to cEno.

Discussion
In the present study, we measured the avidity of various ACPAs in the serum of RA patients in order to understand the mechanism of ACPA production, because avidity of antibody is supposed to re ect the direction of antigen speci city. Importantly, avidity can be compared among patients with varying antibody levels and is not likely reduced by therapeutic intervention once antibody is produced.
In line with the results of previous reports (16), we observed that the avidity of ACPAs, except for anti-cVim antibody, in a Japanese cohort of RA patients was lower than that of antibodies against recall antigens. This is not likely because those ACPAs are autoantibodies, as we found some autoantibodies even higher avidity than the antibodies against recall antigens. Another explanation is that ACPA immune response is ongoing and that a large portion of serum ACPA is produced by B cells that have not undergone extensive a nity maturation, which may not circulate in the peripheral blood. This can explain the discrepancy with the previous report showing high levels of SHM in peripheral blood B cells (8)(9)(10)(11)(12). However, reverting mutation of ACPA genes of those B cells to the germline sequence lost the binding to the citrullinated peptides (9,10,13,15), arguing against this possibility. Recently, an alternative hypothesis has been raised. Most ACPA has N-glycosylation sites in the Fab portion, which are generated via somatic hypermutation and are assumed to provide survival advantage in the GC (21). Hence the threshold of binding a nity to the target antigen could be lowered. Although the mechanism why Fab glycosylation sites are preferentially induced in ACPA is unclear, it indicates that recognition of citrulline residue, albeit at low avidity, is independently involved in the selection of GC B cells.
Although numerous citrullinated proteins/peptides have been demonstrated to be the target of ACPA, it is unlikely that all these antigens induce the corresponding speci c antibodies. Instead, crossreactivity generates the diversity of ACPA to a substantial extent, which might also be involved in the low avidity of ACPA. A question is whether the ACPA-inducing antigens are same or different among patients.
The variety in ACPA ne speci city ts better with the latter hypothesis. Sequential appearance of different antigens might be the mechanism of the epitope spreading. In this case, hierarchy in the avidity of different ACPAs might vary among patients. However, our data actually showed a xed hierarchy or positive correlation in the avidity of different ACPAs. It is thus hypothesized that a dominant and common antigen induces the whole ACPA response, in which cross-reactivity between different antibody clones may also exist. In support of this, an amino acid motif critical for ACPA binding has been identi ed (22). Crystal structure of monoclonal ACPAs demonstrated the conformation that binds different citrullinated peptides (23). Taken together these speculate that the avidity of ACPA is primarily determined by its binding strength to the core-epitope.
As reported previously (6, 7), we observed cross-reactivity between different ACPAs. We further found that anti-cVim antibody is more cross-reactive than the other three ACPAs, indicating that the high avidity of anti-cVim antibody is not owing to low levels of cross-reactivity, i.e. high speci city. Interestingly, it was reported that a nity maturation of ACPA increased cross-reactivity, rather than shaping up the speci city (13). However, it is premature to conclude that cVim is the dominant target antigen inducing whole ACPA response, because anti-cVim reactivity was not always detected in the serum, and, the cVim peptide used in the experiments is not a natural product. Further studies comparing the avidity of antibodies against a variety of natural antigens might identify such dominant antigen.
Our results might not be as clear-cut as those from studies analyzing monoclonal ACPAs, because whole serum contains polyclonal antibodies with different avidity in different amount. However an advantage of using serum samples is that it re ects the whole picture of in vivo immune response and avoids the risk of missing any antibody clone. Thus, it is important to integrate the notion obtained from studies using monoclonal antibodies and polyclonal serum. Another limitation of our study is the small number of citrullinated peptides used for the analysis. In addition, increasing the number of patient samples could more clearly demonstrate the statistical relationship.
In conclusion, we found in this study that there is a xed hierarchy in the avidity and crossreactivity of different ACPAs despite large variations in the speci city and levels. This emphasizes the importance of low-a nity cross-reactive binding in the generation of diverse antigen speci city of ACPA and suggests the presence of dominant antigen that induces ACPA production. Interestingly, crossreactivity of ACPA to carbamylated or acetylated peptides has also been revealed (12,24,25), raising a possibility that, not only ACPA, but all these RA-related antibodies are generated by the common mechanism.

Conclusions
We detected xed hierarchy in the avidity and cross-reactivity between different ACPAs in the serum of a cohort of RA patients, suggesting that the mechanism underlying ACPA production is common to RA patients. Presence of a dominant antigen that induces whole ACPA response is speculated. This study complies with the Declaration of Helsinki, and the study protocol was approved by the Regional Committee of Ethics for Human Research at the Faculty of Medicine of the Kyushu University (24-174). All subjects provided written informed consent before participating in the study.

Consent for publication:
Not applicable.
Availability of data and materials: All data generated or anayzed during this study are included in this published article and its supplementary information les.

Competing interests:
The authors declare that they have no competing interests.

Funding:
This work was supported in part by Grants-in-aid for Scienti c Research, The Japan Society for the Promotion of Science (24591449).

Authors' contributions:
AH and HY participated in the study design. TT and MK participated in the collection of serum samples. AH, TS and HY performed laboratory assays and led the interpretation of the study results with assistance from the remaining authors. HY and YN critically revised the manuscript and provided nal approval. All authors read and approved the nal manuscript.  Comparison of avidity of various antibodies in the serum. A, Avidity of various antibodies in the serum was measured by an elution assay using NaSCN solution and is expressed as AIAUC (see Materials and Methods). B, Statistical differences between the avidity of different antibodies are shown.

Figure 3
Correlation and hierarchy in the avidity of different ACPAs. The relationship between the avidity (AIAUC) of different ACPAs in serum samples reactive to more than two citrullinated antigens are analyzed.
Dashed diagonal lines indicate the points of equal value, and solid lines indicate the correlation curve.

Figure 4
The extent of cross-reactivity between different ACPAs. Cross-reactivity between different ACPAs was estimated by measuring reduction rate of peptide binding by preincubation with inhibitor peptide.
Combinations of inhibitor and test peptides are indicated in the panels. Dashed diagonal lines indicate the points of equal value.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.