IgG anti-hinge antibodies against IgG4 F(ab’)2 fragments generated using pepsin are useful diagnostic markers for rheumatoid arthritis: implications of the possible roles of metalloproteinases and IgG subclasses in generating immunogenic hinge epitopes

Background Pepsin agglutinators, discovered over 50 years ago, have been recently referred to as anti-hinge antibodies (AHAs) because of their reaction with the IgG hinge epitope. AHAs have different reactivity for each hinge epitope generated by each protease that cleaves the hinge region at different sites. Moreover, AHAs have different reactivity against different hinge epitopes derived from each IgG subclass even when the same protease is used. Since the expression of matrix metalloproteinase-3 (MMP-3) is enhanced in rheumatoid arthritis (RA), AHA production could also be increased. The purpose of this study was to determine whether the levels of AHAs against IgG hinge epitopes produced by MMP-3 are elevated in RA. Methods The serum levels of IgG or IgA AHAs against the IgG1/IgG4 F(ab’)2 fragments, generated by either MMP-3 or pepsin, were measured using ELISA in 111 patients with RA and 81 healthy controls (HC). Receiver operating characteristic (ROC) analysis was used for obtaining optimal cutoff values and cutoff values indicating high specificity (> 95%) of the AHA. The targeted epitope of a specific AHA was investigated through inhibition ELISA. Results Seven AHAs were statistically higher in RA patients than in HC, except IgG AHA against IgG1 F(ab’)2, which was generated by MMP-3 proteolytic cleavage. The areas under the ROC curve were 0.66–0.80, although the sensitivities at high specificity were low (5.4–24.3%). The cumulative number of positive AHAs in each individual was statistically higher in RA patients than in HC, suggesting the extreme extent of AHA repertoires in RA. Inhibition studies revealed that IgG AHAs against IgG4 F(ab’)2 fragments generated by pepsin cross-reacted with IgG1 F(ab’)2 fragments generated by pepsin. Multivariate logistic regression analysis identified the IgG AHA against IgG4 F(ab’)2 fragments generated by pepsin as an independent variable for RA diagnosis, even in RA patients who were negative for both RF and ACPA (odds ratio 1.18, 95% CI 1.06–1.32; P = 0.003). Additional experiments using non-RA patients finally strengthened the diagnostic utility. Conclusion In RA patients, we observed diversification and amplification of AHA repertoires and diagnostic utility of the specific AHA against IgG4 F(ab’)2 fragments generated by pepsin but not MMP-3.


Background
Rheumatoid arthritis (RA) is a chronic inflammatory disorder characterized by persistent synovitis, destruction of bone and cartilage in multiple joints, and disability [1]. RA is also a systemic autoimmune disease representing autoantibodies, such as rheumatoid factors (RFs; autoantibodies to the IgG Fc fragment), anti-citrullinated protein antibodies (ACPA), and anti-carbamylated protein antibodies [2]. Although RFs are the leading anti-IgG antibodies in RA, antibodies against the IgG hinge epitopes have been also of interest [2].
Multiple proteases cleave IgG at the lower hinge region and produce F(ab') 2 fragments [3,4]. Antibodies to IgG F(ab') 2 , once designated as pepsin agglutinators, have been detected not only in patients with autoimmune diseases but also in healthy subjects [5,6]. Lately, they have been termed as anti-hinge antibodies (AHAs), since the majority of the antibodies bind strongly to IgG F(ab') 2 and not to intact IgG [7]. Furthermore, AHAs usually target a C terminus of the amino acid sequence in the lower hinge region [8,9], implying they react specifically to the hinge neoepitope and no other epitopes of IgG F(ab') 2 .
In recent studies, the reaction of AHAs with human hinge epitopes of human IgG1 or IgG4 appearing after pepsin cleavage showed higher positivity rate and titer in RA patients than in healthy controls (HC) [10]. However, AHAs against IgG1 or IgG4 F(ab') 2 generated by other proteases, such as the IgG-degrading enzyme of Streptococcus pyogens (IdeS), did not show any difference, suggesting certain AHAs may be associated with a specific disorder.
Matrix metalloproteinase-3 (MMP-3) is produced in synovial cells and has been associated with RA disease activity [11,12]. It seems likely that excessive production of IgG F(ab') 2 fragment in RA is caused by MMP-3, followed by the overproduction of specific AHAs, since activated MMP-3 generates human IgG1 F(ab') 2 fragment in vitro and in vivo [3,13].
The purpose of the present study is to measure IgG/ IgA AHAs against IgG1, IgG2, and IgG4 monoclonal therapeutic biologics cleaved by MMP-3 or pepsin and to evaluate their characteristics in RA. To our knowledge, a clinical study of serum AHAs against IgG1 and IgG4 F(ab') 2 fragments generated by MMP-3 has never been attempted before.

RA patients, healthy controls (HC), and non-RA patients
In this cross-sectional and case-control study, serum samples were collected from 111 patients with RA who met the ACR classification criteria [14]. As HC, we obtained sera from 81 healthy staffs in our hospital. All sera were leftovers, and there was a shortage of five sera for testing IgA AHAs. All the samples were stored at − 80°C until use.

Generation of F(ab') 2 fragments by proteolytic cleavage
The following biologics were used: tocilizumab (TCZ; IgG1), infliximab (IFX;IgG1), panitumumab (PAN;IgG2), and natalizumab (NTZ;IgG4). All biologics, except IFX, were humanized monoclonal IgG. Although IFX is chimeric, the hinge region and CH2 domain were derived from human IgG1 according to the database from the National Center for Biotechnology Information (NCBI, USA). We used pepsin (Sigma-Aldrich, St. Louis, USA) and human MMP-3 as proteases to cleave the biologics. The pro-MMP-3 was a kind gift from Daiichi-Fine Chemical, Toyama, Japan.
The biologics were dialyzed with 0.1 M sodium citrate buffer (SCB, pH 3.5), and then 100 μg of pepsin in SCB was incubated at 37°C with 10 mg of each biologic. Incubation time was overnight except 2 h for PAN. Tris 1 M was added to the IgG solution until the pH increased to 7.4 to stop the digestion. Proteolysis by MMP-3 was performed after the activation of pro-MMP-3 by incubation at 55°C for 25 min [15]. The activated MMP-3 (50 μg) in 50 mM Tris-HCl (pH 7.5) containing 150 mM NaCl and 10 mM CaCl 2 was mixed with 5 mg of each biologic at 37°C. After 2 h of incubation, each small amount of the reaction mixture (20 μL) was removed, and the reaction was stopped by rapid freezing. The remaining reaction mixture was continued for 24 h and stopped by adjustment to 20 mM ethylenediaminetetraacetic acid (EDTA).

Detection of human IgG fragments
Cleaved human IgG fragments were analyzed by SDS-PAGE in Tris-glycine buffer using 10% gels under nonreducing conditions. Samples were heated at 100°C for 2 min in 50 mM Tris-HCl buffer (pH 6.8) with a final concentration of 20% glycerin and 1% (w/v) SDS. Protein bands were visualized by staining with 0.25% (w/v) Coomassie Brilliant Blue R-250 (Nacalai Tesque, Kyoto, Japan) in 50% (v/v) methanol and 10% (v/v) acetic acid.

Purification of IgG F(ab') 2 fragments
At the outset, human IgG digested by pepsin or MMP-3 was separated by gel filtration on a Sephadex G-150 column (Pharmacia Fine Chemicals, Uppsala, Sweden). The estimated IgG F(ab') 2 fractions were concentrated by Vivaspin 20 (Sartorius Stedium, Goettingen, Germany). Finally, to remove IgG possessing Fc, the concentrated crude IgG F(ab') 2 were applied to a Protein G Mag Sepharose column (GE Healthcare, Uppsala, Sweden) using 50 mM Tris buffer with 150 mM NaCl at pH 7.5. Purification of the IgG F(ab') 2 was confirmed by the SDS-PAGE described above. Each purified F(ab') 2 fragment was denoted by the addition of an italic subscript to specify the protease responsible for its cleavage, e.g., IgG1 F(ab') 2MMP-3 .

Measurement of AHAs by enzyme-linked immunosorbent assay (ELISA)
ELISA plates (Sumitomo Bakelite, Tokyo, Japan) were coated overnight at 4°C with 100 μL/well of a solution of 0.5 μg/ml IgG F(ab') 2 in 0.1 M carbonate/bicarbonate buffer at pH 9.6. After washing with 10 mM Tris buffer containing 0.9% NaCl with 0.05% Tween-20 (TBST) at pH 7.4, serum samples diluted 1:200 with TBST were added to the plate (100 μL/well). These were incubated for 2 h at room temperature (RT). After washing, 100 μL/ well of alkaline phosphatase (ALP)-conjugated antihuman IgG Fc (Sigma-Aldrich) diluted 1:10,000 with TBST, or ALP-conjugated anti-human IgA (Sigma-Aldrich) diluted 1:10,000 with TBST was added and incubated for 1 h at RT. After washing, the AHAs were visualized with 1 mg/mL of p-nitrophenyl phosphate tablets (Sigma-Aldrich) in diethanolamine buffer at pH 9.8 for 30 min or for 2 h for IgA AHA measurement. Absorbance was measured at 405 nm using a microplate reader. Levels of IgG AHA were calculated by a calibration curve using pooled human IgG purified using 40% ammonium sulfate and DEAE sephadex (Pharmacia Fine Chemicals, Uppsala, Sweden). We arbitrarily defined 1 mg/mL of the pooled IgG as containing 800 arbitrary units (AU)/mL of IgG AHA against IgG F(ab') 2pepsin . We also used this calibration curve for measuring IgG or IgA AHA to other IgG F(ab') 2 fragments.

Inhibition study for specificities of IgG AHAs against IgG1/IgG4 F(ab') 2pepsin
An equal volume of inhibitors with various concentrations (5, 000, 1, 000, 200, 40, 8, 1.6, 0.32, 0 μg/mL) and 1:200 diluted IgG AHA-positive serum from RA patients were thoroughly mixed, followed by incubation for 2 h at RT. The mixtures were added to the ELISA plate (100 μL/well) coated with IgG1/IgG4 F(ab') 2pepsin and then allowed to react for 2 h at RT. The subsequent procedure was the same as the measurement of AHA described above. The extents of inhibition were expressed as percent inhibition of the AHA responses, calculated as follows: A is the absorbance in the presence of inhibitors and B absorbance in the absence of inhibitors.

Statistical analysis
We used the Mann-Whitney U test and the Kruskal-Wallis test to compare the differences between two groups and among multiple groups, respectively. We also used Fisher's exact and χ 2 test for nominal characteristic. To elucidate the independent variables associated with RA diagnosis, univariate logistic regression followed by multivariate logistic regression analysis was performed. Age and gender added to the model. A two-tailed P < 0.05 was considered statistically significant. Data were analyzed on a personal computer using SPSS version 19 (IBM Japan, Tokyo, Japan) and Statflex version 6 (Artech, Osaka, Japan).
AHA levels in RA patients were compared after stratification according to positive/negative for RF and anti-CCP2, namely double positive RA (DPRA), double negative RA (DNRA), RA with single positive RF (SPRA (RF)), and RA with single positive anti-CCP2 (SPRA (CCP)). As shown in Fig. 2, the Kruskal-Wallis test revealed significant AHA level differences among the four groups in AHA1, AHA2, and AHA6. Moreover, the Mann-Whitney U test was used to compare the differences between the groups, and the AHA levels of the DNRA in AHA, AHA2, and AHA6 were significantly lower than those of the DPRA (AHA1, P = 0.004; AHA2, P = 0.003; AHA6, P = 0.03).

Specificity of IgG AHA responses against the F(ab') 2pepsin of IgG1 or IgG4
To elucidate whether IgG AHA responses against the F(ab') 2pepsin of IgG1 or IgG4 possess a specificity for epitopes on each IgG subclass F(ab') 2pepsin , inhibition studies were implemented. The response of serum AHAs against IgG1 TCZ F(ab') 2pepsin from a patient with RA (S-47) was not inhibited by neither IgG2 nor IgG4 F(ab') 2pepsin , but it was inhibited in a dose-dependent manner by IgG1 F(ab') 2pepsin (Fig. 3a). Meanwhile, the AHA response against IgG4 F(ab') 2pepsin was inhibited by not only IgG4 F(ab') 2pepsin , but also IgG1 F(ab') 2pepsin to a certain extent (Fig. 3b) in another RA patient (S-212).

Use of AHAs for clinical RA testing
We assessed the clinical usefulness of AHAs for RA diagnosis by ROC curve analysis. The values of the area under the curve and their mean standard error (SE) ( Table 2) ranged from 0.57 (0.04; AHA2) to 0.80 (0.03; AHA6 and AHA7). After the optimal cutoff values of each AHA were calculated by Youden's index, we obtained sensitivity/specificity, positive predictive value (PPV+)/negative predictive value (PPV−), positive likelihood ratio (LR+), and diagnostic odds ratio (DOR) as shown in condition 1 of Table 2. The LR+ values were 1.79 (95% CI 1.23-2.62) of AHA1 to 3.44 (95% CI 2.69-4.11) of AHA3, leading to only a small shift in pre-and post-test probability. The DOR of AHA6 and AHA7 were relatively high, yielding 10.24 (95% CI 5.40-19.42) and 11.35 (95% CI 5.65-22.79), respectively. Next, the same statistical indices were calculated at high specificity conditions (96-98%) for RA as shown in condition 2 of Table 2. Increments of the LR+ values, except for those in AHA2 and AHA4, resulted in moderate shifts in preand post-test probability, although the values of sensitivity were as low as 2.7% for AHA4 and 24.3% for AHA3. The DOR values also changed (e.g., from 7.02 to 12.7 of AHA3 and 11.4 to 4.57 of AHA7).

Expansion and diversification of the AHA response in RA
To assess the expansion of the AHA response in RA, we studied how many positive AHAs were obtained within each RA patient or HC (Fig. 4). The AHA exceeding each optimal cutoff value was accepted as positive, and the cumulative number of positive AHA response in RA patients and HC was counted. The IgG and IgA AHAs were classified into five grades (0-4) according to the Fig. 1 Comparison of AHA levels between HC and RA patients. AHA1, AHA2, AHA3, and AHA4 represent IgG AHA against TCZ IgG1 F(ab') 2pepsin , TCZ IgG1 F(ab') 2MMP-3 , NTZ IgG4 F(ab') 2pepsin , and NTZ IgG4 F(ab') 2MMP-3 , respectively. AHA5, AHA6, AHA7, and AHA8 represent IgA AHAs against TCZ IgG1 F(ab') 2pepsin , TCZ IgG1 F(ab') 2MMP-3 , NTZ IgG4 F(ab') 2pepsin , and NTZ IgG4 F(ab') 2MMP-3 , respectively. Differences in levels between HC and RA were analyzed with the Mann-Whitney U test cumulative number. The grade of IgG AHAs in RA showed a tendency to be higher than that in HC (Fig. 4a). Furthermore, the highest grade (cumulative number of recognized epitopes, 4) of IgA AHAs was over 50% in RA, indicating an extensive recognition profile (Fig. 4b).
These results suggested that the AHA repertoire in RA is expanded and diversified.

Potential of AHAs as independent diagnostic markers for RA
We performed a logistic regression analysis to further assess the potential ability of AHAs as diagnostic markers for RA. Univariate analysis using 106 RA patients and 81 HC showed that AHAs, except AHA2, could be selected for multivariate analysis as shown in analysis 1 of Table 3. A multivariate logistic regression analysis revealed that AHA3 (IgG AHA against NTZ IgG4 F(ab') 2pepsin ) and AHA5 (IgA AHA against TCZ IgG1 F(ab') 2pepsin ) were selected as independent variables contributing to RA diagnosis. Additionally, logistic regression analysis using AHAs of 18 patients with DNRA revealed that only AHA3 was selected as an independent variable (odds ratio 1.18, 95% CI 1.06-1.32; P = 0.003) as shown in analysis 2 of Table 3.
To strengthen the potential of AHAs as independent markers for RA, we performed additional experiments using sera from 61 patients with rheumatic diseases excluding RA (non-RA) and sera from 59 patients with RA. The Mann-Whitney U test displayed that AHA3 level in RA was significantly higher than that in non-RA (P = 2.69 × 10 −4 , Additional file 1: Figure S3). Finally, logistic regression analyses revealed that only AHA3 was selected as an independent marker for RA diagnosis (Table 4).

Discussion
There have been conflicting reports concerning the specificity of anti-F(ab') 2 antibodies in human sera [5,16,17]. Nowadays, studies using human monoclonal IgG1 F(ab') 2 and synthetic peptide analogs of IgG1 hinge region has revealed that most anti-IgG F(ab') 2 antibodies target lower hinge epitopes but not idiotopes nor other epitopes in the IgG F(ab') 2 fragment [9,10,18]. However, as 30 patients with RA in this study were receiving or had been receiving TCZ, it is not straightforward to clarify whether anti-TCZ IgG F(ab') 2 antibodies target idiotopes or other epitopes except lower hinge epitopes. We once tried to detect the anti-idiotype antibodies by ELISA using TCZ IgG F(ab') 2 fragment as coating antigen in RA patients who were receiving TCZ. Whereas many anti-TCZ IgG F(ab') 2 antibodies were detected, the antibodies were not inhibited by intact TCZ, but only by TCZ IgG F(ab') 2 fragment. From these experimental findings, it is likely that most of anti-IgG F(ab') 2 antibodies in this study do not react with idiotopes in the biologics.
We found a more extended hinge epitope recognition profile in RA compared with HC. This finding might indicate the possibility of an epitope-spreading phenomenon in which the immune response extends to involve new intramolecular or intermolecular epitopes [27][28][29], although our study was cross-sectional. This phenomenon has been revealed by the observation of ACPA in RA, which occurs before clinical disease onset [30,31]. Further studies using preclinical and longitudinal RA patients are needed to confirm whether the same phenomenon is observed.
Compared with ACPA, AHAs were not useful for RA diagnosis because their LR+ values were less than 5 at optimal cutoff values, with anti-CCP2 LR+ values of 71.6 and 12.1 in HC and non-RA patients, respectively [32]. Meanwhile, the LR+ of RFs was not high, reported to be 4.86 [33], and seemed to approximate to the LR+ of AHAs. At a cutoff value of over 95% specificity, however, two AHAs (AHA3 and AHA6) revealed LR+ close to 10, which can lead to an increased probability of RA diagnosis.
Three AHAs (AHA3, AHA5, AHA8) were selected by univariate analysis, and then AHA3 (IgG anti-IgG4 F(ab') 2pepsin ) and AHA 5 (IgA anti-IgG1 F(ab') 2pepsin ) were selected as independent variables for RA diagnosis by multivariate logistic regression analysis as shown in Table 3. Additionally, only AHA3 was selected as an independent variable to conduct the multivariate analysis on DNRA. Furthermore, when patients with non-RA were subject as a substitute for HC, it strengthened a  The numbers of patients with RA and non-RA were 59 and 61, respectively. All of non-RA patients including six patients with Sjogren's syndrome (3 primary and 6 secondary) did not show persistent destructive polyarthritis, namely RAlike symptom or sign possibility of the conclusion that AHA3 was selected again as shown in Table 4. The resulting increased AHA reactivity against IgG4 F(ab') 2pepsin in RA highlights that (1) IgG4 is produced in the context of prolonged antigenic stimulation and (2) AHAs must be generated in vivo by other physiological proteases, except pepsin, since pepsin needs activation in acidic stomach conditions and do not reach circulation. In this regard, activated matrix metalloproteinase-7 (MMP-7) seems to be a candidate protease for human IgG cleavage so that it cleaves IgG4 at the same lower hinge site between F234 and L235 (EU numbering) as pepsin does [10,13]. Interestingly, the increase in MMP-7 seems to mainly originate from non-articular but extra-articular lesions, such as the nodules and lung of RA patients [12,34,35]. Although we cannot clearly explain the role of AHAs in RA, it has been proposed that several biological functions of AHAs, such as B cell suppression due to crosslinking the B cell receptor and FcγRIIb, complement amplification via the capture of dimeric C3b due to immune-complex formation of antigen-binding IgG F(ab') 2 and AHA, and the functional restoration of cleaved IgGs without Fc [7]. A possible suggestion in RA is IgG4 ACPA, the leading IgG subclass following IgG1 [36]. IgG4 ACPA would enervate the IgG1-mediated ACPA-associated pathogenic progression through activating the complement system and triggering Fcγ receptors. In this situation, it has been proposed that specific AHAs bind to IgG4 F(ab') 2 with ACPA reactivity, form immune complexes, and potentially lead to the progression of inflammatory processes [10]. Further studies are needed to evaluate the pathogenic or protective participation of AHA in addition to RF and ACPA in joints and/or lungs of RA patients.

Conclusions
We have discovered an extended epitope recognition profile of AHAs in RA, suggesting maturation of AHAproducing immune cells. IgG AHA against IgG4 F(ab') 2 generated by pepsin as an alternate protease of MMP-7, but not MMP-3, could be used as a potential diagnostic marker for RA, including seronegative RA.