At a simplistic level, the success of immunity relies on distinguishing ‘self’ from ‘non-self’, originating primarily by purging the autoreactive repertoire both in central lymphoid organs, the thymus and bone marrow, as well as by peripheral tolerance mechanisms. Classical mechanisms of immune tolerance rely on the processing and presentation of self-peptides by antigen presenting cells (APCs), resulting in deletion or anergy of the autoimmune repertoire. However, it is clear that many self-proteins, or post-translational modifications (PTMs) to self-proteins, are not expressed in primary lymphoid organs. For example, PTMs specific to peripheral tissues may not be expressed in the thymus and thus these PTMs never tolerize or delete the emerging thymocytes. While we know that proteins are fundamentally assembled from 20 amino acid structures, the addition of PTMs pushes that group above 140 structurally unique amino acids (1). These observations have altered the concepts and the breadth of immune tolerance to self-proteins, particularly in the perspective of autoimmune syndromes. Many factors influence the rate at which PTMs arise in self-proteins, including the amino acid sequence, flanking amino acid motifs, and variables such as the tissue microenvironments. PTMs also alter specific peptides that arise by antigen processing and, subsequently, the specificity of ongoing B- and T-lymphocyte immunity (2–4).

Some PTMs arise by enzymatic processes, as with citrullination reviewed herein, or with N-linked glycosylation or phosphorylation, critical to biological functions of many host proteins. In contrast, some modifications arise spontaneously, under physiologic pH and temperature, such as the formation of isoaspartyl modifications [reviewed by (5)]. Also, some modifications can be processed either enzymatically or spontaneously, such as deamidation (6). As illustrated herein, PTMs may trigger aberrant autoimmunity as well as alter the biological functions of self-proteins in selected tissues, including transcriptional and translational events. The text that follows will review the conditions that elicit citrulline PTMs, including their role in biological and immunological processes, with specific focus on their implications in type 1 diabetes (T1D).

Citrullination, also known as deimination, is a PTM in which an arginine residue is converted into a citrulline ( Figure 1 ). This modification, which was first described in 1939 (7), leads to a loss of one positive charge and a reduction in mass of 0.984 Da per modified arginine. This type of modification can alter intra- and inter-molecular interactions of the protein (8), having an impact on its structure, function and its interaction with other proteins (9). Citrullination is catalyzed by peptidylarginine deiminase (PAD) enzymes in a Ca²⁺-dependent manner (10), which is thought to be an irreversible process (11).

A role for citrullinated self-proteins has been associated with several autoimmune diseases, such as T1D, rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), multiple sclerosis (MS), psoriasis, Sjögren’s syndrome (SS), antiphospholipid syndrome (ALS) and inflammatory bowel disease (IBD) ( Table 1 ). Among these different autoimmune diseases, citrullination as an autoimmune biomarker in RA has been most extensively described. RA is a chronic autoimmune disease characterized by inflammation of the synovial joints. Proteomic analysis of the cellular and soluble components of RA synovium identified the full RA citrullinome, with more than 100 citrullinated proteins, amongst which vimentin, enolase, fibrinogen and fibronectin (46). Some of these were shown to induce autoantibodies and/or autoreactive T-cell responses in RA (11, 22, 29, 30). The presence of anti-citrullinated protein antibodies (ACPAs) in the serum of RA patients is one of the most specific diagnostic marker for the disease (47). ACPAs can be detected years before clinical symptoms appear (48). Apart from being a prognostic biomarker for disease development, some ACPAs have been described as useful in predicting the severity of joint destruction during the first five years after RA onset (49). ACPAs can also be detected in a small percentage of patients with SS (3 to 9.9%), and the presence of such autoantibodies indicates a risk of developing RA (50). As listed in Table 1 , emerging citrullinated target proteins are proposed as autoimmune disease specific biomarkers. Interestingly, the serological level of citrullinated vimentin in patients with ulcerative colitis (UC) was decreased compared to patients with another important IBD, Crohn’s disease (CD), and non-IBD control subjects. Therefore, citrullinated vimentin has also been suggested as differentiating marker between CD and UC to improve the diagnostic accuracy in IBD (44).

Peptidylarginine deiminases are a group of 5 enzymes encoded by genes localized on chromosome 1p36.1 in human and chromosome 4E1 in mice, located in a cluster of 350 kb and 240 kb, respectively (51). They were first described in 1977 as the enzymes mediating the conversion of arginine into the non-classical amino acid citrulline in proteins, from studies performed in mammalian hair follicles (52). The enzyme responsible for this reaction was for the first time partially purified by Fujisaki and Sugawara in 1981 and named peptidylarginine deiminase (PAD) (53). Since then, five different PAD isozymes have been described in mammals, designated as PAD1 to PAD4 and PAD6, which display 50-70% sequence identity (54). PADs replace the primary ketamine group of arginine (=NH) by a ketone group (=O) and yield ammonia as a side-product ( Figure 1 ), leading to the loss of one positive charge of the target protein. This can lead to changes in the function and fate of the citrullinated target protein.

Although the PAD enzymes are widely expressed throughout the body, and have been implicated in a variety of physiological processes, each isozyme has specific tissue distribution, functions, and substrates under physiological conditions (8) ( Figure 2 and Table 2 ). As such, PAD1 is predominantly expressed in skin epidermis, uterus and hair follicles, targeting keratin and filaggrin (86). PAD2 is the ubiquitous member of the family, being expressed in multiple tissues such as brain, skeletal muscle, spleen, uterus, secretory glands and leukocytes (60, 65, 71). Importantly, PAD2 and PAD4 are the only PAD isozymes expressed in immune cells. Amongst other targets, PAD2 citrullinates myelin basic protein (MBP) in brain, vimentin in skeletal muscle and macrophages, actin in neutrophils and histones in various cell types (71). PAD3 is found in hair follicles and epidermis, and citrullinates filaggrin, trichohyalin, apoptosis-inducing factor (AIF) and vimentin (86). PAD4 is expressed in leukocytes, mainly granulocytes (like neutrophils and eosinophils), and monocytes and macrophages (63, 73). It is the only PAD isotype with a nuclear localization signal sequence, which is located at its N-terminus. PAD4 targets several nuclear proteins, such as histones, nucleophosmin and nuclear lamin C (71). With its high expression in neutrophils, PAD4 plays an important role in the generation of neutrophil extracellular traps (NETs) and thereby in the first line of defense against bacterial pathogenic invaders (see in more detail below). PAD6 is mainly expressed in eggs, embryo and ovary. It is the only PAD isozyme for which until today no protein substrates have been identified (86) and, in vitro, no catalytic activity can be measured (89). Also, no known association with autoimmunity or other diseases has been reported for PAD6.

Although all 5 PAD enzymes target arginine residues in proteins, they do have different substrate specificities. The basis for this difference is not fully understood, but cannot solely be explained by their difference in tissue distribution or subcellular location (54). Additional influencing factors include the enzyme kinetics, the conformation of the secondary structure of the target protein and the flanking amino acid composition surrounding the arginine residues, as revealed from in vitro studies with recombinant PAD enzymes. In regard to the latter, it was shown for instance that a glutamic acid accompanying the arginine residue decreases the chance of being citrullinated, whereas a flanking aspartic acid residue increases the citrullination level (91). In this regard, PAD4 was found to have a higher substrate specificity as compared to PAD2 (92, 93).

The most important regulator of PAD activity is calcium, with micro to millimolar calcium concentrations needed for their full activation (53, 54, 94). However, under physiological conditions, with intracellular cytosolic calcium concentrations usually around 100-fold lower, a basal PAD activity can also be measured, in line with the occurrence of low levels of citrullination and their role during normal physiological processes. This raises questions as to how exactly calcium is mediating the activity of PADs. One thought is that calcium may alter the confirmation of PADs. Under low calcium conditions, PADs may be present in a conformation that selects only high efficiency substrates, thus allowing citrullinations to happen, but limiting aberrant citrullination events (11). In the presence of high calcium, all calcium binding sites of the PAD enzyme are occupied, promoting extensive rearrangement and dimerization of the enzyme, leading to its full activation, as shown for PAD4 (95, 96). Alternatively or additionally, other factors are known to regulate PAD enzymes (9), which could act as cofactors modulating their calcium sensitivity and specificity (11). Considering the importance of calcium in PAD activation, pathways that alter intra- or extracellular calcium levels, such as endoplasmic reticulum (ER) and inflammatory stress, are important in many of the autoimmune diseases in which citrullination is implicated (see in more detail below) and underscore the important link between citrullination and stress pathways. Indeed, changes in calcium fluxes may lead to activation of PAD activity in stressed or dying cells. Moreover, the externalization and activation of intracellular PAD enzymes into the extracellular space may occur in the surroundings of dying cells. This may explain the citrullination of extracellular proteins, as observed for instance during NET formation (NETosis) (see in detail below) or the citrullination of beta-cell proteins as observed in T1D (15). Similar observations in RA show that PADs are highly activated by Ca²⁺ ion deposition in the inflamed joints, particularly during apoptotic cell death (97, 98), leading to elevated levels of citrullinated proteins and peptides.

The first antibody-based methodology for citrullinated protein detection was described in 1992 by Senshu et al. (99). The antibody does not recognize citrulline residues directly but instead binds a chemically modified form, diacetyl monoxime and antipyrine derivatized-citrulline, now available in commercialized kit form. The approach utilizes protein samples separated by SDS-PAGE, transferred to a microporous membrane. The membrane bound citrulline-containing proteins are modified in the presence of 2,3-butanedione monoxime and antipyrine in acid condition. In a similar concept known as the “Senshu” method, Moelants et al. developed a sandwich ELISA format to detect citrullinated proteins utilizing antibody recognizing 2,3-butanedione-modified citrulline (100). Another citrulline-specific labeling chemical probe, rhodamine-tagged phenylglyoxal derivative (Rh-PG), was developed for detection of citrullinated proteins in both purified protein sample and complex mixtures including serum (101, 102).

There are now commercially available anti-peptidyl citrulline antibodies for detection of citrullinated proteins including mouse monoclonal IgM antibody, clone F95, from Millipore (MABN328), and rabbit polyclonal antibody from Upstate (07-377) or Abcam (ab10092 and ab6464). Clone F95 antibody is extensively used in tissue staining, immunoblot and ELISA (25, 103, 104). The specificity and sensitivity of the above commercial anti-peptidyl citrulline antibodies for differentiating arginine citrullination and lysine carbamylation were recently discussed (105). Of note, carbamylation is a non-enzymatic PTM converting lysine to homocitrulline, which shares similar structural features with citrulline. For specific citrullinated target proteins, a variety of commercially available or in-house produced antibodies are available, such as those specific to citrullinated-histones, hypoxia-inducible factor 1-α, vimentin, fibrinogen, MBP and GRP78.

Upon citrullination, 0.984 Da mass increase from the parent peptide can be identified by mass spectrometry. As this is exactly the same increase in mass as for deamidated peptides, caution is needed when analyzing liquid chromatography-tandem mass spectrometry (LC-MS/MS) data, not to misinterpret a citrullinated peptide from a deamidated peptide on a closeby N- or Q residue. On top of that, C¹³ isotopes, with an increase in mass of 1 Da compared to C¹³ parent ions, can also be wrongfully identified as deamidated or citrullinated residues, even with the use of very accurate mass spectrometry instrumentation and strict settings. A recent study by Callebaut et al. (6) outlines critical parameters to detect deamidated residues by LC-MS/MS, through minimizing artificial in vitro occurring deamidations and manual inspection of spectra. This same method can also be applied for critical evaluation of citrullinated peptides from LC-MS/MS analysis. Another issue to be considered is that the loss of one positive charge of the arginine residue due to citrullination will affect protease cleavage efficiency for proteases which cleave after arginine, such as trypsin. This will result in mis-cleavages by trypsin, in case the arginine is citrullinated (106). Of particular interest is the dual search delta score method developed by Qian laboratory which integrates several critical parameters for identifying citrullinated and deamidated peptides in an automated way, thereby decreasing false discovery rates (FDR) (107). A different method used for facilitating LC-MS/MS-based detection of citrullinated proteins is chemical derivatization of citrulline residues by 2,3-butanedione alone or combined with antipyrine, resulting in a mass increase of 50 or 238 Da, respectively (108, 109). Chemical derivatization of citrulline residues can also be used for enrichment of citrullinated proteins, through the use of biotin-conjugated phenylglyoxal (BPG), prior to LC-MS/MS analysis (101, 102). This method, however, requires large amounts of starting material. In addition, a BPG-based ELISA platform recently developed will validate the mass spectrometry proteomic data for citrullination detection (110).

Simply put, autoimmunity is initiated when cellular and soluble components of the immune system interact to trigger the recognition and robust response to self-proteins leading to tissue pathology. Additionally, a large number of heritable genetic risk traits have been defined by genome wide association studies (GWAS) in many autoimmune syndromes. However, autoimmune syndromes and T1D, in particular, are not entirely explained by a defined collection of heritable genetic traits. Indeed, poorly defined environmental influences and epigenetic factors, which may or may not be inherited, also influence the early onset and progression of T1D (111). Of importance, the amplification of PTMs in T1D autoimmunity is clearly linked to oxidative tissue environments. Reactive oxygen species (ROS), including superoxide anion (O₂ ^.-), hydrogen radicals (OH.), and hydrogen peroxide (H₂O₂) are a product of a dynamic balance of endogenous anti-oxidant cellular compounds that control their tissue concentrations and biological effects. These anti-oxidants include superoxide dismutase (SOD), glutathione peroxidase, catalase, peroxiredoxins, as well as other small molecule anti-oxidants, including vitamins E and C. There are potentially a number of sources of specific ROS in tissue autoimmunity, including the infiltration of activated phagocytic cells (neutrophils, macrophages and dendritic cells) which have been demonstrated to be important in the progression and tissue pathology of T1D as well as many other autoimmune syndromes.

Oxidative stress can amplify the modification of certain proteins, or protein motifs, or, alternatively, alter metabolic pathways. As described in detail below, there are secondary effects of ROS on apoptosis, NETosis, and cellular metabolic pathways, affecting the progression of autoimmune responses and tissue pathology. ROS affected proteins may be changed in solubility, reduce the ability of proteins to be cleared, or increase immunogenicity. As detailed below, oxidation and subsequent citrullination can also provoke changes at both the level of DNA transcription and translation.

Citrullination of histone H3 by PAD4 in granulocytes, leading to formation of NETosis, is a process which is promoted by intracellular ROS (85, 112). Also, exogenous H₂O₂ was shown to induce citrullination and NETosis in mouse neutrophils (85). Depending on the stimulus used, the level and time course of ROS production was shown to be important for subsequent H3 citrullination and NET formation (113). However, seemingly contradictory to this notion, in vitro studies showed a direct inhibitory effect of PAD2 and PAD4 activity in the presence of H₂O₂ concentrations above 40 µM. In addition, PMA-stimulated leukocytes could effectively citrullinate recombinant fibrinogen, although this process was markedly enhanced when ROS formation was inhibited by the NADPH oxidase inhibitor diphenyleneiodonium (DPI) (114). These findings suggest that although ROS is important intracellularly for mediating citrullination of histones, supraphysiological levels of ROS may inhibit citrullination extracellularly. Discrepancies in different studies on the exact role of ROS in PAD activation and citrullination may need further examination to resolve the impact of intra- vs. extracellular ROS concentrations.

Another important factor mediating PAD activity is the local redox balance, with a reducing environment needed for efficient PAD activation. As such, in vitro studies, making use of the non-physiological reducing agent dithiothreitol (DTT), have shown that PADs can be activated under reducing conditions. Also more physiological reducing agents, like thioredoxin (115) and reduced glutathione (GSH) (116) can activate PAD enzymes. Kinetic characterization of PAD4 using thioredoxin as reducing agent produced results equivalent to those obtained with DTT (115).

Clearly, ER stress is a key factor in the amplification of PTMs in T1D. The insulin producing beta-cell, i.e. the immune targeted cell in T1D, is particularly sensitive to ER stress, because of its highly developed ER needed to cope with the high demands for protein translation and folding (117–119), in response to acute changes in blood glucose levels. As such, beta-cells can increase the translation of preproinsulin up to 25-fold (120), reaching almost 1 million molecules of preproinsulin per minute (121), when blood glucose levels are high. In order to restore the equilibrium between the cellular demand for protein synthesis and the ER folding capacity, cells under ER stress activate the unfolded protein response (UPR) (122). The UPR is activated by three transmembrane protein sensors, activating transcription factor 6 (ATF6), protein kinase RNA-like ER kinase (PERK) and inositol-requiring kinase 1 (IRE1). Under physiological conditions, these three sensors are kept in an inactive state through binding to the chaperone glucose-regulated protein 78 (GRP78; also known as binding immunoglobulin protein (BiP) or heat shock 70kDa protein 5 (HSPA5)) (123). When misfolded proteins accumulate in the ER lumen, GRP78 releases from these sensors inducing their activation and downstream signaling (123) leading to ER stress attenuation. When this UPR fails to restore ER homeostasis, usually in conditions of intense or chronic stress, this adaptive UPR will change to a terminal UPR (124), activating pro-apoptotic signaling pathways that lead to cell death (122).

Next to ER stress induced by high demands of protein translation, typical for secretory cells such as beta-cells, environmental factors associated with T1D can trigger additional ER stress in beta-cells (125), including coxsackie viral infection, dysglycemia, inflammation, ROS and exposure to chemicals such as streptozotocin and alloxan (125, 126). Apart from inducing beta-cell dysfunction and death, ER stress can also induce PTMs in beta-cells, a process that has been described for thapsigargin induced ER stress, showing increased activation of tissue transglutaminase (TGM2) and PAD, both Ca²⁺ dependent enzymes (127, 128). Also, inflammatory cytokines, which act at least in part through activation of ER stress pathways, were shown to induce citrullination of beta-cell proteins (15, 16, 118, 129) ( Figure 3 ). Interestingly, thapsigargin induced ER stress of human beta-cells leads to increased immunogenicity, as measured by IFNγ response of T-cell clones specific for deamidated peptides (127, 128).

Several investigators have identified significant differences between the T- and B-cell responses that develop against PTM self-proteins, or cryptic epitopes (5, 13, 128, 168–170). In general, T-cell responses to PTM determinants tend to be specific for the modified peptide only and more rarely cross-react with the unmodified form of the peptide. This concept was originally demonstrated in mice immunized with the isoaspartyl PTM form (isoAsp) of snRNP D lupus autoantigen protein, showing T cells only proliferate in response to the isoAsp PTM, but are unresponsive to the native (Asp) peptide (168). Alternatively, B-cell and autoantibody responses are often more promiscuous in their binding to both the PTM modified and native self-protein. This phenomenon may be due to the features of antibodies to bind flanking amino acid sequences in both modified and native protein forms. As one example, human SLE and lupus-prone MRL/lpr mice exhibit autoantibodies that bind both isoAsp and Asp forms of histone H2B (171). Moreover, it was demonstrated that autoimmune responses originating from the PTM self-protein diversify in an intra- and extra-molecular manner to other self-protein determinants. Therefore, breaking immune tolerance to a PTM self-protein promotes ‘epitope spreading’, a mechanism where autoimmunity diversifies to epitopes beyond the initial site(s) initiating the response (170). Similarly, there is both intra- and inter-molecular B- and T-cell epitope spreading in T1D, SLE, and MS (168, 172). Classical studies have described epitope spreading of autoantibody responses with the progression of autoimmune disease (173). As illustrated earlier, apoptotic and necrotic cells are rich sources of PTM-altered self-proteins in various microenvironments, notably conditions of oxidative stress (174) or altered pH (175).

In both of the collagen-induced arthritis (CIA) and experimental autoimmune encephalomyelitis (EAE) mouse models, citrullination of joint and brain proteins creates neoantigens that become additional targets in epitope spreading of autoimmune responses (172). Citrullination of aggrecan, vimentin, fibrinogen, and type II collagen, known target proteins in RA, initiates epitope spreading by promoting T-cell responses to both citrullinated peptides and the corresponding control peptides (176). In addition, citrullination-induced conformational changes of HSP90 protein unmasks cryptic epitopes to bypass B-cell tolerance in RA (177). Epitope spreading may also apply for citrullinated epitopes in triggering T1D autoimmunity ( Figure 3 ). This is shown for example for citrullinated GRP78, where circulating CD4⁺ T cells and autoantibodies against both naïve and citrullinated GRP78 peptides are detectable in patients with T1D, suggesting that epitope spreading due to citrullination also occurs in T1D (15).

It is clear that the specificity of ongoing immunity relies on the efficient and accurate antigen processing pathways in epitope generation (178). The presence (or absence) of a PTM of an amino acid residue has been shown to alter the recognition and cleavage by proteases. Additionally, the presence of PTM within an intracellularly processed peptide alters the binding to specific MHC motifs. For example, it has been demonstrated that the absence of N-glycosylation of the neuronal glutamate receptor subunit 3 in Rasmussen’s encephalitis, a severe form of pediatric epilepsy, exposes a granzyme B cleavage site. This is just one example of how PTMs may create a novel autoantigen (neoepitope) (179). As another example, most proteases and peptidases do not recognize the peptide linkage connecting isoAsp residues to its neighboring amino acid (180). Finally, Moss and coworkers demonstrated that the deamidation of asparagine residues in tetanus toxin C fragment inhibits the processing by asparagine endopeptidase and results in decreased antigen processing (181). The role of PTMs in antigen processing was examined years ago in studies of model proteins in immunity (168, 182–184). An isoaspartylated form of cytochrome c protein is cleaved differently by cathepsin D compared to the (normal) aspartyl form of the same protein (2). Granzyme B cleavage of autoantigens may also generate new epitopes based on the presence or absence of PTMs in self-protein (185). Simply put, the presence of PTMs that affect proteolytic enzyme recognition generates a completely new repertoire of peptides during antigen presentation.

Specific subsets of APCs and even the compartments within the cells in which the antigen is processed may shape the type of PTM acquired and whether modified peptide is presented on MHC. Ireland and colleagues (186) demonstrated that B-cell autophagy was required for the generation and presentation of a citrullinated peptide (but not required for the non-modified peptide form). Mamula and others have demonstrated the unique APC functions of B cells in presenting antigen to T cells. In particular, it is clear that B cells both present antigens directly, as well as transfer antigens to other APC, such as macrophages and dendritic cells. Thus, different APCs may dictate the self PTM epitopes generated and eventually presented by the immune system (187–189).

It is not fully understood as to how a PTM will affect HLA binding. The ‘fit’ of the PTM peptide versus native peptide for an HLA molecule can vary based on elements including charge and structural changes imposed by PTMs. The HLA binding of various PTM T1D autoantigens has been studied and reviewed by James et al. (12, 13). For deamidated and citrullinated residues specifically, the introduction of a negative charge enhances the binding affinity at specific positions in HLA-DR or HLA-DQ molecules. As such, HLA-DRB1*04:01 and DRB1*15:01 prefer citrulline at key positions of their binding motifs, whereas deamidation results in preferential loading onto HLA-DQ molecules (13). This was also shown for citrullinated peptides of vimentin, a RA autoantigen, which has greater affinity for HLA-DRB1*0401 than the unmodified peptide (190).

Among other autoimmune syndromes, specific PTMs of MBP result in either low, intermediate or a similar affinity for MHC compared with the corresponding wild-type peptide (191). Notably, acetylated MBP peptide (Ac 1-11) is required to incite pathogenic T cells in murine MS, though the unmodified peptide binds MHC with virtually identical kinetics. Similarly, isoaspartic acid residues in cytochrome c or snRNP D peptides (SLE autoantigens) bind MHC class II in a manner identical to the unmodified peptides (168). However, immune tolerance is maintained to the native peptides. The overall lesson is that PTM self-peptides alter processing and binding by MHC in distinct manner from unmodified (native) peptide.

Given the involvement of citrullination and PAD enzymes in several autoimmune diseases, it has been of great interest to generate compounds targeting PAD activity, i.e. PAD inhibitors. The development of such compounds was made possible with the discovery of the crystal structure of PAD enzymes (95, 192). Since then several pan-PAD inhibitors have been developed, with increasing potency, specificity and metabolic stability, of which the 4 most extensively studied irreversible inhibitors are shown in Figure 4 . More recent efforts also include the development of specific PAD inhibitors, targeting for instance PAD2 or PAD4, as well as non-covalent reversible inhibitors. For a detailed overview of all developed PAD inhibitors we refer to a recent review specifically focusing on these developments (197).

Considering that PAD dysregulation is associated with several autoimmune diseases (86), numerous studies have evaluated the efficacy, safety and mechanism of action, mainly making use of irreversible pan-PAD inhibitors in animal models of autoimmunity, with promising results. As such, amelioration or even reversal of disease in case of intervention therapies, and delayed initiation or complete protection in case of preventive therapies, was shown in mouse models of RA, MS, SLE, UC, inflammatory bowel disease, and recently also in T1D [reviewed by (198, 199)]. In general, mechanistic insights from these studies have taught us that targeting PAD making use of pan-PAD inhibitors effectively decreases protein citrullination levels in the inflamed target tissues, as measured mainly by LC-MS/MS, PAD activity assays or Western blotting using anti-citrulline Ab (200–203). Such decreased protein citrullination can evidently have an effect on the autoreactive responses against citrullinated autoantigens, both in terms of autoantibody and T-cell responses, as shown in some published studies (200, 201). Next to this, it has become clear that other mechanisms are involved in the observed protection, pointing towards more general effects on innate and adaptive immunity. Not surprisingly, based on the high levels of PAD4 in neutrophils and their important role in histone 3 citrullination and induction of NETosis, as outlined above, in vivo PAD inhibition was shown to reduce NET formation and associated NET-induced tissue damage (204, 205). Other ameliorating disease effects were associated with a decrease in circulating pro-inflammatory cytokine levels, such as IL-6, TNFα and IL-1β (206, 207), an increase in Treg populations (200) or a shift in T-lymphocyte populations from Th1/Th17 towards Th2 (201, 208). The latter is thought to be mediated through direct inhibition of citrullination of the transcription factors RORγT and GATA3 (see also above) (72).

Of interest in the field of T1D, a recent study from the Overbergh laboratory showed a complete protection against diabetes development in the NOD mouse, by daily subcutaneous injections of BB-Cl-amidine (1µg/g body weight) (200). BB-Cl-amidine is a pan-PAD inhibitor that, similar to its mother compound (Cl-amidine), irreversibly inactivates PAD enzymes through covalent modification of an important cysteine for the activity of the enzymes (209). Remarkably, diabetes protection was observed when starting treatment at 8 weeks of age, a time point at which insulitis is already ongoing, but hyperglycemia has not developed (200). This observation tempts us to conclude that citrullination may play a role in amplification of the disease rather than being an initial trigger in breaking immune tolerance. BB-Cl-amidine treatment was associated with decreased citrullination levels in the pancreas as demonstrated by LC-MS/MS and western blotting with anti-citrulline Ab, as well as decreased levels of circulating autoantibodies against citrullinated GRP78, a known citrullinated autoantigen in T1D (as outlined above, Table 4 ) (15, 16). These findings confirm the effective direct action of BB-Cl-amidine on inhibition of PAD activity. Furthermore, bone-marrow derived neutrophils isolated from BB-Cl-amidine treated mice showed less potential for spontaneous NET formation when compared to the control group. Disease protection was associated with preservation of pancreatic insulin levels, although only a marginal reduction in insulitis was observed, suggesting a less aggressive form of insulitis, in line with reduced CD4⁺ effector memory T cells and reduced IFNγ-producing T cells in the pancreas infiltrates. In the periphery, a shift from Th1 towards Th2 cytokine levels and increased frequency of regulatory T cells was observed.

Taken together, the promising results obtained with pan-PAD inhibition in the NOD mouse, as well as in other autoimmune mouse models, indicate that disease protection is mediated by effective inhibition of citrullination in the target inflamed tissue, leading to decreased autoreactive responses. This further underscores the observation that citrullination is not a specific disease-related event, but rather an inflammation-dependent process occurring preferentially in autoimmune target tissues (210). Next to this, the extensive data in different preclinical autoimmune models indicate that pan-PAD inhibitors also have more general effects on immune cells, thereby dampening inflammation and reshaping the immune response towards tolerance. Whereas this may be an additional benefit for the use of PAD inhibitors, caution is also warranted. Since pan-PAD inhibitors act both intra- and extracellularly, and citrullination is important in many physiological processes, such as epigenetic and transcriptional regulation, as outlined above, the concern about possible adverse side effects cannot be neglected. Therefore, more studies are needed to evaluate more specifically potential adverse effects before translating to the human situation. Of interest in this regard is the development of isozyme-specific PAD inhibitors, which are hoped to still be protective in disease development, but allow physiological citrullination to occur. Great progress has been made in this path, with development of a high number of PAD specific inhibitors. For instance, a PAD1-selective inhibitor (211), D-Cl-amidine, and a specific PAD2 inhibitor (194), AFM30a, both based on the structure of Cl-amidine. PAD4-specific inhibitors include TDFA (212) and GSK199, with GSK199 being a reversible PAD4 inhibitor that targets the apo state of the enzyme, showing potent inhibition of PAD4 at low concentration of calcium (0.2mM) (143). GSK199 has been demonstrated to be effective in blocking murine arthritis (213), however, more studies are needed to evaluate the efficacy and safety of such specific inhibitors.

In summary, we have attempted to illustrate and summarize various biochemical, immunologic, and transcriptional aspects of citrulline modifications. How citrullination alters these unique biologic processes in spontaneous autoimmune syndromes are beyond the prediction of simple genetics or other stochastic factors. Citrullination, as well as many other PTMs and cellular pathways, are affected by inflammatory cytokines and reactive oxygen species that inhabit the tissue microenvironments in T1D and other autoimmune diseases. More indirect pathways are also affected by citrulline modified proteins, including downstream transcriptional events. The emerging technologies of detecting citrulline modifications as well as other PTMs in proteomics and tissue analyses will undoubtable change the landscape of autoimmunity in the coming months and years. These analyses, notably the presence of serum anti-citrulline autoantibodies, have already contributed to the clinical diagnoses and assessment of the progression of disease, and tissue pathology. With the identification of specific biomarkers and an understanding of their origins, the field will now have potential therapeutic pathways as targets to modify these autoimmune diseases, including the reducing tissue inflammation and use of inhibitors of citrullination prior to destruction of the pancreatic islets.

M-LY, FS, MM, and LO conceived the concept and co-wrote the manuscript. All authors contributed to the article and approved the submitted version.

Research in this area in the LO lab is supported by JDRF (1-SRA-2019-809-S-B) and by IMI2-JU under grant agreement No 115797 (INNODIA) and No 945268 (INNODIA HARVEST). This Joint Undertaking receives support from the Union’s Horizon 2020 research and innovation program and “EFPIA”, ‘JDRF” and “The Leona M. and Harry B. Helmsley Charitable Trust”. M-LY and MM were supported by the JDRF (1-SRA-2020-977-S-B and 1-SRA-2020-981-S-B), NIH: AI48120-13 and DK104205-01.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.