Rheumatoid arthritis (RA) is a chronic autoimmune disease that is characterized by painful inflammation of synovial joints. RA joint symptoms are characterized by symmetric inflammation of joints and change in synovial extracellular matrix, usually beginning in the metacarpal or metatarsal joints. The local findings include a pannus formation, i.e., invasive synovitis, formation of rheumatoid nodules and even pseudotumors of the synovial membrane and influx of inflammatory cells. The pathological process leading to RA is not fully understood. Antibodies binding to citrullinated proteins (ACPA) are a rheumatoid arthritis-specific finding that can precede clinical disease onset by several years [1], which suggests that these antibodies and their antigens could play a role in the disease pathogenesis. In addition to the joint tissues, citrullination in the lungs in response to smoking has been suggested as an inducing event (reviewed in [2]). Five different isoenzymes of citrullinating peptidylarginine deiminases (PADs) are found in humans. Of these, PAD2 and PAD4 have been shown to be present in RA synovial tissue [3]. PAD3 expression has been connected to tissue damage in chick embryonic neural stem cells [4]. However, so far the exact molecular mechanism in pathogenesis and the anatomical origin of these citrullinated and homocitrullinated proteins is not known. To elucidate these processes we selected PADs 2, 3 and 4 and myeloperoxidase as target antigens for the immunological stainings in this study.

The ACPA antigens identified so far are proteins occurring naturally in the human body. In RA, citrullinated fibrin, vimentin and histones are the focus of interest and are considered possible autoantigens for citrulline-binding antibodies [5]. Fibrin deposits are common in inflamed synovium, histone citrullination is present in extracellular trap structures created by neutrophils [6, 7] and vimentin is citrullinated during apoptosis. It has also been shown that in cell culture conditions citrullination of fibronectin alters synovial fibroblast behavior and may affect how these cells adhere to and invade the joint [8]. Since ACPA may develop years before disease onset, the processes leading to antibody development are an obvious subject of investigation. Recently, also antibodies binding to carbamylated [9] and specifically to homocitrullinated sequences have been found in RA sera [10], but also processes leading to formation of these antibodies are unknown. It has been suggested that ACPA and carbamylated protein-binding antibodies are distinct systems [11] although overlapping of the presence of both of these antibodies is frequent in RA [9, 10].

Citrulline and homocitrulline in proteins are products of different posttranslational modifications (arginine citrullination and lysine carbamylation respectively) but have similar effects on protein structure and may create epitopes that induce autoantibody responses. The formation of citrulline and homocitrulline contributes to protein immunogenity [12, 13]. Antibodies binding to citrulline- and homocitrulline-containing antigens contain partly overlapping binding sites in RA sera [9, 10, 14] and in animal models [13] suggesting some relationship between these two modifications in RA.

Lately, the role of neutrophils in RA has raised interest as neutrophil extracellular trap (NET) formation has been suggested as a source of citrullinated antigens in synovial fluid and pannus [7]. Myeloperoxidase, an enzyme that is present in the granules of neutrophils, has a function in host defense. Myeloperoxidase uses negatively charged ions (for example iodine, chloride and cyanate) [15, 16] to distort, for example, bacterial proteins to inhibit their function. The addition of cyanate to lysine side chain creates homocitrulline, which makes NETs interesting also in respect of origin of homocitrulline-binding antibodies.

We have already shown that citrulline and homocitrulline could be found simultaneously in the synovial tissues of a single ACPA-positive RA patient [17]. In this sequential study we show that the synovial tissues of seropositive RA patients are characterized by necrotic areas that contain high levels of citrulline and homocitrulline and PAD and myeloperoxidase enzymes. We also show the presence of myeloperoxidase in the necrotic areas. Necrotic areas are more frequent in the metatarsal than in the knee synovial tissues in seropositive RA patient tissues and absent in seronegative RA and in osteoarthritis (OA) patient synovial tissues.

Since ACPA and carbamylated protein antibodies are a sign of disease severity and a predictor of joint damage in RA, we wanted to study the expression patterns citrulline- and homocitrulline-containing proteins in tissue from RA and control patients. We had the opportunity of collecting and comparing RA synovial tissues from metatarsal and knee joints and found interesting differences between these two anatomical locations both in histology and citrulline and homocitrulline formation. Expression of citrullinating peptidyl arginine deiminases 2, 3 and 4 and homocitrullination-facilitating neutrophil enzyme myeloperoxidase were studied, because these enzymes contribute to formation of these altered proteins.

The staining of PAD2, PAD3 and PAD4 enzymes localized to the lining in OA and both the lining and sublining layers of synovial tissue in RA. PAD2 was found in the synovial lining and sublining layers and endothelium of blood vessels. PAD4 was found in the lining and sublining layers of intact RA synovial tissue and diffusely in the necrotic tissue, but not in OA tissue. The diffuse PAD4 staining in the necrotic tissue could be due to release of PAD4 from disrupted neutrophils discussed in detail below, or alternatively unspecific binding of the antibody. Our results are in concordance with previous findings, where PAD2 and PAD4 were shown to be present in RA synovial tissue [3] and plastic adherent macrophage-like cells isolated from peripheral blood monocyte fraction [20]. PAD3 expression in the synovial tissue has not been shown previously but PAD3 mRNA was found recently in synovial tissue, not in cultured synovial cells [21]. In our specimens, PAD3 staining was strongest in the macrophage-like synoviocytes in the lining layer of OA and lining and sublining layers in RA synovial tissue (Figs. 6 and 7) co-localizing with macrophage marker KP1 (Fig. 11). This could explain the earlier PAD3 mRNA finding [21], since inflammatory macrophages are not long-lived in cell culture. Peripheral blood monocytes and macrophage-like adherent cells isolated from peripheral blood monocyte fraction did not seem to express PAD3 in an earlier study [20]. This could indicate that in addition to differentiation to macrophages also inflammatory activation is needed for PAD3 expression. In our study, some PAD3 staining was also detected in synovial stromal cells with more fibroblast-like morphology (Figs. 7d and 9d). The exact role of PAD3 in RA context remains to be clarified, but interestingly, PAD3 involvement in calcium-induced cell death pathway in neural stem cells was shown recently [4]. This earlier study suggested a role for PAD3 as a regulator of cell death or survival in neural stem cells. A similar role in regulation of synovial growth could be an interesting phenomenon from the perspective of RA disease progression also. To verify this hypothesis, the role and effect of PAD3 expression and activation in inflamed tissue needs further studies on functional level.

Citrulline- and homocitrulline-containing proteins localized to the necrotic areas of rheumatoid nodule and synovial tissue, as shown here by F95 staining and earlier by staining with anti-modified citrulline antibodies [22]. It is notable, that these antibodies, do not make a distinction between these two structurally homological amino acids [19] and thus some of the earlier findings concerning citrulline may involve homocitrulline as well when immunological detection methods have been used. Here, we used HPLC to distinguish the two amino acids in hydrolyzed tissue samples. Some necrotic areas could be found in seropositive RA knee synovial tissues. Fibrinoid necrosis was absent in the seronegative RA and OA knee synovial tissues, where only low levels of citrulline and homocitrulline was found. The highest content of both were found in the seropositive RA metatarsal synovial tissues that were characterized by fibrinoid necrosis (Figs. 1 and 2). In our data, the immunostaining results were consistent with the HPLC analysis showing staining by F95 antibody as well as PAD2, PAD3, PAD4 and myeloperoxidase (Figs. 8 and 9). The reliability of immunostaining of necrotic tissue is always a challenge and any conclusions should be considered critically. Therefore we prepared homogenates of necrotic tissue showing the binding of chemically modified citrulline- and homocitrulline-binding antibodies (KS350) to high molecular weight citrullinated or homocitrullinated proteins in this tissue. Digestion of DNA from the sample released some of the citrulline- or homocitrulline-containing proteins. Finding myeloperoxidase by Western blotting (Fig. 4), and in the histological sections of the necrotic area (Figs. 8 and 9), suggests that indeed neutrophils are involved in the seropositive RA chronic inflammation.

Myeloperoxidase is an enzyme found in granules of intact neutrophils. When an object is too big to be ingested, neutrophils use their chromatin and the enzymes from their granules as the last available defense mechanism of forming extracellular traps (NETs) [23] and these have been suggested as sources of citrullinated autoantigens [7]. Histone citrullination by PAD-enzyme releases the chromatin packing [24]. RA neutrophils are prone to form NETs. NET components have been found in RA blood [25] and synovial fluid and active PAD enzymes [26] in synovial fluid. Also increased levels of myeloperoxidase have been found in RA sera and synovial fluid [27, 28]. We suppose that in seropositive RA synovium neutrophils respond to the massive inflammatory reaction or presence of histiocytes in and around the necrotic area. Myeloperoxidase staining could be found in the middle of the necrotic area and also in the intact tissue adjacent to the necrotic area. In the necrotic tissue, these distinct defense mechanisms target and modify proteins at the site. Our present data suggests, that myeloperoxidase from neutrophil granules is present at active sites of inflammation in seropositive RA synovial tissues and it could be from extracellular traps or secreted. Detection of NETosis in tissue specimens is a challenge. However, these events can create a situation where citrulline and homocitrulline are present simultaneously in large amount in host proteins that are further presented to the inflammatory cells. This could be a part of the process for RA autoantibody induction, and at least partly explain the wide variety of antibody specificities that can be found in RA sera. Overall, our data supports the idea of repeated auto-vaccination as a propagator of RA [29].

In relation to RA, myeloperoxidase could also have an independent role in autoantibody development. The levels of thiocyanate are elevated in blood of smokers [30] and myeloperoxidase facilitates homocitrulline formation through a chemical reaction originating from thiocyanate [31]. Antibodies binding to carbamylated proteins have been found in RA sera and these were also suggested to be related to joint damage [9]. Here we found high levels of antibodies binding to citrulline- and homocitrulline-containing collagen telopeptides. As shown before [10], the presence of these antibodies was overlapping with ACPA positivity, but was not in clear relation with tissue citrulline and homocitrulline levels. Here it should be noted that not all of the sera analyzed were taken at the time of the operation and thus the antibody titers may not directly reflect the situation at the time of the operation. In general, the levels of these antibodies were highest in the seropositive RA knee and MTP subgroups that showed also the highest local tissue levels of citrulline and homocitrulline. Thus, the presence of high levels of citrulline- and homocitrulline-containing proteins in ACPA and carbamylated antibody-positive individuals could lead to persistence of inflammation leading to progression of tissue damage. The chemical protein modifications are still not enough for antibody induction as shown recently [32], but probably also genetic predisposition is needed.

In clinical perspective, our data suggests that there are distinct processes in RA and OA synovial tissue that have differences even between anatomical locations. In the necrotic tissue, the altered self-proteins are presented to the inflammatory cells, inducing antibody response that could cause secondary inflammation in other joints leading to chronic inflammation and tissue erosions. The removal of the necrotic tissue is the only way of extinguishing the local inflammatory process that the necrotic tissue upholds. In RA, the local and systemic control of inflammation is disturbed. These findings however raise new questions on pathophysiology behind the necrotizing events in RA synovium and the clinical possibilities on how to prevent these events. Moreover, since in RA the body is not able to remove all necrotic tissue that upholds the inflammation, surgery might indeed still be needed, until the mechanism behind the formation of necrotic lesions is fully understood and curable by other means. For surgeons, this means that synovial biopsies should be taken while performing synovectomy and if necrosis is seen this suggests that the patient could benefit from more extensive surgery. Removal of the necrotic mass and activated synovial tissue could be a key element in alleviating the local symptoms of the disease.

Some study limitation issues need to be taken into account while interpreting these results. The RA patient tissues were collected from patients with a long-lasting disease with severe tissue alterations in the joints and there is a selection bias, since all patients had to undergo surgery. The patients’ history of smoking and the medication in use are confounding factors and the tissue piece used for analyses only presents a small area of the synovial tissue. In addition, the method used here for citrulline and homocitrulline detection is semi-quantitative and based on a chemical modification reaction after hydrolysis of the tissue. Some citrulline or homocitrulline might be lost during hydrolysis and the chemical modification reaction is done individually to each sample, which may somewhat affect the result, so the absolute amounts of citrulline and homocitrulline could be somewhat greater than reported here, if a direct determination method was used. Despite these shortcomings, significant differences between groups could be detected and hence, the data is reliable albeit limited due to currently usable methodology.

In conclusion, this study adds to the current knowledge by showing neutrophil enzyme myeloperoxidase in seropositive RA synovial tissue with homocitrulline. Disperse staining patterns of myeloperoxidase suggests NETosis or myeloperoxidase release from neutrophils. NETosis involves also peptidyl arginine deiminases and is a process creating citrullinated proteins released in the extracellular space. These findings can further explain the RA progression and disease mechanism.