Introduction
Citrulline-containing proteins are generated through posttranslational modification of arginine residues in a reaction catalyzed by the Ca
2+-dependent peptidyl arginine deiminases (PADs). There are five PAD family members, but only PAD2 and PAD4 expression are closely linked with inflammation in RA synovial tissue [
1,
2]. While PAD2 is broadly expressed across tissue types, including by immune cells, PAD4 exhibits an expression pattern restricted to immune cell types, in particular macrophages and granulocytes [
1,
3].
Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by systemic inflammation, chronic synovitis, joint destruction and bone loss, affecting approximately 2% of the world population [
4]. Plasma and synovial biopsy specimens from patients with RA contain high levels of citrullinated proteins [
5,
6], and anti-citrullinated peptide antibodies (ACPAs) exhibit high specificity and sensitivity as diagnostic markers of the disease [
7]. Anti-citrulline peptide antibodies can appear before disease onset and correlate with the most erosive form of RA [
8].
PAD4 shows elevated expression in RA [
1,
9], and RA patients generate high affinity anti-PAD4 autoantibodies which correlate with more severe disease [
10‐
12]. Further, variants of PAD4 are linked to RA in several Japanese and Korean cohorts, although this association has not held up in most North American and European study groups, despite the prevalence of ACPA in all ethnic groups [
13]. Thus, the development of autoantibodies to citrullinated epitopes and PAD4 and elevated PAD4 expression in RA, suggests that aberrant PAD activity may contribute to disease pathogenesis.
The offspring of an intercross between the KRN TCR transgenic mouse specific for a bovine RNase (42-56) in the context of I-A
k and the I-A
g7-expressing non-obese diabetic (NOD) background, spontaneously develop a progressive, inflammatory joint disease with features similar to human RA (K/BxN mice) [
14]. The autoantigen in this model is glucose-6-phosphate isomerase (GPI), a ubiquitous cytoplasmic enzyme [
15]. Treatment with the sera of K/BxN mice or purified anti-GPI autoantibodies is sufficient to transfer disease to healthy animals, even in animals devoid of B and T cells [
14,
15]. Because autoantibodies are passively transferred, this model focuses on immune recruitment and joint destruction (effector phase), rather than the breaking of immune tolerance (priming phase). Innate immune signals are critical for this model because mice deficient in the alternative complement pathway, the C5a receptor, the CXCR2 chemokine receptor, interleukin-1 receptor (IL-1R), and myeloid differentiation primary response protein (MyD88) are resistant to disease [
14,
16,
17]. Further, arthritis was not sustained in toll-like receptor 4 (TLR4) mutant mice [
16]. Passively transferred arthritis also requires the presence of mast cells and neutrophils [
18‐
21].
Neutrophils are among the first immune cell types to accumulate during an inflammatory response [
22]. In response to inflammatory stimuli, neutrophils decondense their chromatin and actively expel their DNA-producing neutrophil extracellular traps (NETs) that are decorated with granular and nuclear proteins, including citrullinated histones [
23,
24]. Incubation of neutrophils with phorbol 12-myristate 13-acetate (PMA), hydrogen peroxide, lipopolysaccharide (LPS), bacteria, and yeast induces NET formation [
24‐
27]. Our lab and others have shown that PAD4 is essential for the production of NETs and NET-associated histone citrullination [
24,
25,
27]. PAD4-mediated histone citrullination is thought to play a mechanical role in NET formation, where the conversion of positively charged arginine residues into the neutral citrulline amino acid by PAD4 promotes chromatin decondensation [
24]. PAD4-mediated NET formation is critical for controlling at least a subset of bacterial infections because PAD4-deficient mice are more susceptible to infectious disease in a necrotizing fasciitis model [
27].
NET formation, although critical for the full activation of the innate immune response [
27], has also been implicated in inflammatory disease pathogenesis, including the autoimmune disorder lupus [
28], cystic fibrosis ([
29‐
31], sepsis [
32], and thrombosis [
33]. Interestingly, it has been suggested that NETs offer a possible mechanism by which PAD4 may be liberated from the cell to generate citrullinated antigens and exacerbate inflammation [
9,
34]. Recently, Dwievedi
et al. described hypercitrullination in neutrophils from arthritic patients, as well as the specific reactivity of arthritic serum to activated neutrophils and citrullinated histones [
34]. It is unclear whether PAD4-induced NET formation plays a role in the RA inflammatory process.
Recently, using a genome-wide screen of mice, Johnsen
et al. identified the region containing
Padi4 as being putatively involved in the development of K/BxN arthritis and demonstrated that increases in the transcripts of both PAD2 and PAD4 correlates with increased severity of disease [
35]. Because the K/BxN model is neutrophil dependent, PAD4 activity is required for NET formation and PAD4 overexpression correlates with rheumatoid arthritis in patients, we sought to determine the extent to which PAD4 would contribute to the effector phase of arthritis using the K/BxN-serum transfer model [
20,
25,
27,
36].
Discussion
In this study, we find that the PAD4 enzyme is active within the joint tissue of arthritic mice and leads to NET formation. However, the disease course and histological features of the arthritic joints following K/BxN serum injection were comparable between WT and PAD4-deficient mice, demonstrating that PAD4 is not required for the effector phase of arthritis. Our data are consistent with the findings by Willis
et al., showing that the PAD inhibitor Cl-amidine provides therapeutic benefit in the collagen-induced arthritis model but has no benefit when arthritic disease is induced by the administration of anti-collagen antibodies. Collagen induced arthritis and injection of anti-collagen antibodies represent the priming and effector phases of disease, respectively. We have now shown that PAD4 is active but not required for disease, in another model of the effector phase, the K/BxN serum transfer model, which is consistent with the finding that PAD inhibitors have no effect on disease when induced by administration by anti-collagen antibodies [
40]. Interestingly, Cl-amidine preferentially inhibits PADs 1 and 4 over PAD3, and its effects have not been reported for PADs 2 or 6 [
41].
Indeed, Johnsen
et al. identified the
Padi locus, which contains the PAD4 gene along with the genes for PADs 1, 2, 3, and 6, as being linked with disease in the K/BxN serum transfer model [
35]. PAD2 and PAD4 are the most likely candidates to regulate the effector stage of arthritis as they are both expressed in immune cell types, whereas the expression of PADs 1, 3, and 6 are restricted to the epidermis, hair follicle, and oocyte, respectively. Indeed, increased splenic expression of both PAD2 and PAD4 correlated with disease severity in the K/BxN model [
35]. Further, within the
PAD region, a SNP found within the
Padi2 locus showed the most significant association with disease [
35]. We speculate that the loss of PAD2 and PAD4 together may produce a more apparent phenotype in the K/BxN model [
40]. PAD2 KO mice have been reported [
42], however the PAD2 and PAD4 loci are approximately 1 centimorgan apart, and therefore, the recombination frequency between the targeted PAD2 and PAD4 alleles would be quite small, making the generation of
Padi2/Padi4 DKO mice unlikely. While our data demonstrate that PAD4 is not required for the development of the K/BxN serum-transfer model, it is possible that there might be redundant activity of other PADs or they may independently contribute to the pathogenesis of antibody-mediated arthritis.
The blood of RA patients contains autoantibodies directed against a number of self-antigens. Many autoantibodies in RA are directed against citrullinated proteins. In fact, the presence of anti-citrulline antibodies is a better predictor of RA than rheumatoid factor [
43]. Variants of PAD4 are linked to RA in several Japanese and Korean cohorts, and the mRNA of a disease-associated allele is more stable than a non-disease associated allele [
36,
44]. It has been proposed that PAD4 is linked to RA because PAD4 citrullination of peptides leads to a breakdown in tolerance to self-antigens [
43,
45]. In support of this, treatment of mice with the PAD inhibitor Cl-amidine in the collagen-induced arthritis (CIA) model reduces the levels of citrulline found in the serum and synovial tissue, diminishes the formation of autoantibodies, and ameliorates disease [
40]. Thus, it will be interesting to determine whether the effects of Cl-amidine are attributable to PAD4 activity by examining the susceptibility of PAD4 KO mice to arthritic disease using the CIA model.
Incubation of human neutrophils with lipopolysaccharides (LPS), TNFα, N-formyl-methionine- leucine-phenylalanine (fMLP), or lipoteichoic acid and murine neutrophils with LPS or bacteria, has been shown to induce histone deimination and NET formation, marking PAD4 activity [
24‐
27]. Further, we detected deiminated histone H4 in lung leukocytes isolated from influenza-infected mice [
25]. In this report, we find that PAD4 activity is readily detected within the affected arthritic joint. In WT mice receiving K/BxN serum, the presence of deiminated histones corresponded primarily to the infiltrating cells of the joint sublining, which is consistent with the expression pattern of PAD4 found in patients with RA [
46,
47]. The stimulus that induces PAD activity during autoimmune-mediated inflammation is undefined. However, the LPS receptor TLR4, has been linked to animal models of arthritis, perhaps because of the activation of TLR4 by endogenous ligands, such as Tenascin-C [
16,
48,
49].
NETs possess potent microbicidal capabilities but have also been implicated in chronic inflammatory diseases [
23,
50]. NET formation is linked to cystic fibrosis [
29‐
31]. Similarly, the formation of NETs contributes to endothelial and tissue injury during sepsis [
32]. In autoimmune small-vessel vasculitis, anti-neutrophil cytoplasm antibodies (ANCA) trigger the formation of NETs, promoting necrotic inflammation of the blood vessels [
51]. Systemic lupus erythematous (SLE) is a systemic autoimmune disease characterized by the formation of pathogenic immune complexes. When activated by autoantibodies, neutrophils isolated from patients with SLE produce NETs, exposing immunostimulatory proteins and potential autoantigens and leading to the induction of Type I interferons by plasmacytoid dendritic cells [
28,
52,
53]. Collectively, these results support the notion that NET production can contribute to disease pathogenesis in inflammatory conditions. While hypercitrullination of neutrophil histones has been reported in patients with RA [
34], it is unclear whether NETs have a role in RA inflammation. Our results suggest that PAD4 activity and subsequent NET formation is present in the K/BxN serum transfer model, but is not required for this model of effector phase of disease.
Our data demonstrate that PAD4 is not necessary for the antibody-dependent, effector stage of arthritis. It is also possible that compensation by other PAD family members, PAD2 in particular, may mask the function of PAD4 in arthritis, although we note that PAD2 expression is not upregulated in PAD4-deficient neutrophils [
25]. Finally, since the
Padi locus is linked to disease severity in the K/BxN serum transfer model, it may be necessary to eliminate several PAD family members, either by targeting multiple locations within the PAD locus or by combining treatment with specific PAD inhibitors with targeted PAD alleles. Further studies will be necessary to dissect the role of PAD4 in the priming phase of arthritis.
Competing interests
No financial support or benefits were received from commercial sources for this study. The authors declare no financial interests or conflicts of interest related to this work.
Authors' contributions
KM conceived the experiments. AR and KM designed the experiments, analyzed data and prepared the manuscript, with input from SA, SH and MC. AR, SA and SH prepared reagents and executed experiments. MC scored the histological samples. The manuscript was read and approved by all authors.