Citrullination of matrisomal proteins in health and diseases

Abstract

Proteins once translated are subjected to post-translational modifications (PTMs) that can critically modify their characteristics. Citrullination is a unique type of PTM that is catalysed by peptidylarginine deiminase (PAD) enzymes, which regulate a multitude of physiological functions such as apoptosis, gene expression and immune response by altering the structure and function of cellular proteins. However, emerging data have unravelled compelling evidence to support that PAD-mediated citrullination is not exclusive to cellular proteins; rather citrullination of extracellular matrix (ECM) proteins also plays a major contributing role in various physiological/pathological conditions. Here, we discuss putative mechanisms for citrullination-induced alterations in the function of ECM proteins. Further, we put emphasis on influential roles of ECM citrullination in various pathological scenarios to underscore the clinical potential of its manipulation in human diseases.

This article is part of the Theo Murphy meeting issue ‘The virtues and vices of protein citrullination’.

1. Introduction

The structural rigidity of tissues is maintained by the dynamic scaffolding functions of extracellular matrix (ECM). The ECM proteome (also known as matrisome) roughly comprises 300 structural and functional ECM proteins which work in concert to stabilize the non-cellular component and provide structural support to the embedded cells [1,2]. The formation, composition, and integrity of ECM are heavily influenced by environmental cues, such as injury and inflammation, and are further governed by the activity of ECM-modifying enzymes that can alter the protein composition. The final characteristics of ECM proteome determine the stiffness, tensile strength, rigidity and elasticity of tissues. These physical characteristics critically influence the behaviour, proliferation, adhesion and signalling of the embedded cells, which in turn, modulate the architecture and dynamics of ECM. This highly regulated, two-way interaction between ECM and the embedded cells is essential for maintaining cellular homeostasis. Thus, any alteration in the formation, dynamics, integrity, composition, assembly and modification of ECM will have significant impact on the overall health of the host. However, ECM proteome is highly dynamic in nature and is constantly subjected to remodelling processes through multiple mechanisms. Similar to cellular proteins, ECM proteins are also liable to post-translational modifications (PTMs). Several PTMs, such as phosphorylation and glycosylation, have been shown to shape the characteristics of the ECM proteome. Citrullination is a unique PTM, whose roles have begun to be unfolded. Here, we review the emerging data on the impacts of citrullination on the function of ECM proteins and, through which, on health and diseases.

2. Citrullination and extracellular matrix proteins

(a) . Peptidylarginine deiminases

Citrullination, is irreversible deimination of the guanidino group of peptidylarginine side chains to form a ureido group (figure 1), thereby neutralizing the positive charge of arginine. This process is catalysed by a family of calcium-dependent enzymes known as peptidylarginine deiminases (PADs). There are five PAD enzymes in mammals including PAD1–4 and PAD6, which have divergent tissue expression profiles. Recent studies have demonstrated that PADs can target and modulate the function of various intracellular and transmembrane proteins, thereby regulating many physiological and pathological processes. An extensive review of the function of PADs is out of the scope of this review. Interested readers are requested to refer other excellent reviews in this issue and other journals for details.

Figure 1.

A schematic diagram of protein citrullination.

(b) . Extracellular matrix proteins

The bulk of ECM is constituted of core fibrous proteins, glycoproteins and proteoglycans (PGs). The fibrous proteins, such as collagens, elastin (ELN) and fibronectin (FN), are highly crosslinked to form an ordered meshwork-like structure. Collagens are the most abundant matrix protein which provide a basic framework of the ECM. There are 28 different types of collagens identified in vertebrates including fibril-forming collagens (type I, II and III), network-forming collagens (type IV, VI, VIII and X), fibril-associated collagens with interruptions in their triple helices (including type VII, IX, XII, XIV, XVI, XIX, XX, XXI, XXII, XXIV), membrane-associated collagens with interruptions in their triple helices (including XIII, XVII, XXIII, XXV) and multiple triple helix domains and interruptions (including type XV and XVIII) [3]. Elastic fibers provide elasticity to the tissues undergoing repeated stretch, such as lungs, skin, aorta and large arteries. These fibers are highly resilient and stable with an estimated half-life of roughly 70–80 years and are composed of a core of ELN protein stabilized by microfibril bundles. FN and fibrinogen are two important matrix glycoproteins, with the former popularly known as ‘biological glue’ owing to its predominant role in cellular adhesion, attachment and migration, apart from the integrin signalling, and maintenance of cellular rigidity. Fibrinogen, on the other hand, primarily acts as platelet aggregator and a precursor for the formation of fibrin clots to maintain hemostasis. PGs are glycosylated proteins that provide hydration properties, mechanical resistance to compression and bind multiple growth factors such as fibroblast growth factor (FGF) and transforming growth factor-beta (TGF-β).

Matrisome is highly diverse among different tissues and the characteristic composition of matrix proteins present in a particular tissue determines its topography, architecture, stiffness, tensile strength, rigidity, elasticity and other properties. For instance, highly elastic tissues such as lungs and aorta have abundant elastin content while the stiff tissue such as skin are enriched with fibrous protein, collagen. Matrisomal proteins interact with the embedded cells in a bidirectional manner through involvement of cell surface receptors primarily including integrins, discoidin domain receptors, syndecans and laminins. Integrins constitute the largest family of cell surface receptors which are structurally heterodimers of α and β subunits, with 24 distinct conformations depending on the arrangement of their α and β subunits. The interaction of ECM proteins with cell surface receptors induces a conformational change in the receptor causing modification in its cytoplasmic tail which leads to downstream effector functions and activation of signalling pathways. The cells, in turn, can influence the matrisome either by directly modulating the secretion of extracellular proteins, or indirectly, through different regulatory mechanisms. This continuous two-way interaction makes the matrisome highly dynamic in nature and keeps on altering its composition depending on the physiological or pathological status. For a detailed description of different ECM components, interested readers are requested to refer a recent excellent review on the topic [4].

3. Citrullination of extracellular matrix proteins

Although citrullination was detailed in the late 1960s, it has remained challenging to identify citrullinated proteins and map citrullinated residues. Several commercial antibodies and chemical probes, such as phenylglyoxal, have been widely used to detect citrullinated proteins. The results must be interpreted with caution because those reagents can cross-react to peptidylhomocitrulline [5], which is generated through a non-enzymatic modification of peptidyllysine. Many of the results were not confirmed on samples deficient in PADs or in the presence of PAD inhibitors in vivo or ex vivo. A better approach would be to directly analyse tissues/cells with mass spectrometry (MS) either with or without prior enrichment of citrullinated proteins; however, it can still be incredibly challenging to distinguish citrullination from deamidation of asparagine or glutamine, which can result in false interpretation. Several MS studies of various tissues, particularly synovial tissue/fluid from patients with rheumatoid arthritis (RA) have been published over the past decade [6–12]. The results have so far indicated the following: (i) brain and lungs contain the highest number of citrullinated proteins among 30 human tissues but the numbers of citrullinated proteins correlate poorly with the expression of PAD enzymes; (ii) many important ECM proteins, such as collagens, FN, PGs, matrix metalloproteinases (MMPs), serine protease inhibitors (SERPINs), fibrillin-1 (FBN1), and fibulin-4 (FBLN4, a.k.a. EFEMP1), are substrates of PADs. Depending on the source of samples and method of detection, ECM make up 1% (healthy human peripheral mast cells) to 50% (synovial fluid from RA patients) of citrullinome; and (iii) there is an expansion of citrullinome in synovial fluid from patients with RA compared to that from healthy controls. In addition, several citrullinated ECM proteins, such as collagens, FN, fibrinogen/fibrin, and vimentin can serve as autoantigens in RA. Nevertheless, the citrullinated residues are not mapped in most of the studies and the functional impacts of citrullination on the majority of the proteins remain unknown. Furthermore, ECM proteins often undergo conformational changes in response to alterations in biophysical properties of microenvironment. Such conformational changes may seclude canonical citrullination sites or expose cryptic sites.

(a) . Spatial citrullination of extracellular matrix proteins

The identification of citrullinated ECM proteins also raises a particularly important question. Does their citrullination take place intracellularly or extracellularly? PADs do not contain signal peptides or transmembrane regions and are considered intracellular proteins. However, it has been demonstrated that neutrophils undergoing necrosis or NETosis, but not apoptosis, can release PAD2 and PAD4 into extracellular space [13]. In addition, active PAD4 on the surface and PAD2 in culture supernatant of intact neutrophils have been detected [14]. Furthermore, release of PAD2 through extracellular vesicles has been observed in tumour cells [15]. Given the high calcium concentration in extracellular space, it is logical to assume that the citrullination of ECM proteins takes place extracellularly. However, we recently demonstrated that the level of intracellular phenylglyoxal-tagged fibulin-5 (FBLN5), presumably containing citrullinated FBLN5, was reduced by a pan-PAD inhibitor, BB-Cl-amidine (BB-Cl), in a dose-dependent manner while the level of total intracellular FBLN5 remained unchanged [16]. This observation strongly suggests that the citrullination of FBLN5 does occur intracellularly but does not exclude the possibility that FBLN5 can undergo additional citrullination after it is secreted.

(b) . Mechanisms by which citrullination modulates the functions of extracellular matrix proteins

Published data has clearly demonstrated that PADs can modulate many biological processes; however, their mechanism of action is still not fully understood. Here, we will discuss and postulate how citrullination can alter the functional landscape of ECM proteins.

(i) . Interfering with interactions between extracellular matrix proteins and integrins (figure 2)

Figure 2.

A schematic diagram depicting that citrullination interferes with the interaction between ECM proteins and integrins.

Many ECM proteins interact with integrins on the surface of embedded cells through arginine-containing motifs, such as RGD (Arg-Gly-Asp) in FN and fibrinogen and GFOGER in collagens. Such interaction mediates the two-way signalling between ECM and embedded cells, and consequentially regulates the cell-generated tension, adhesion, and motility. Citrullination of the arginine residue within those motifs in ECM proteins have been confirmed and can alter the function of the motifs [17,18]. Sipila et al., demonstrated that PAD-mediated citrullination of collagen II decreased the integrin-mediated cell adhesion of synovial fibroblasts and mesenchymal stem cells by 50% and 40%, respectively. The decrease in adhesion was observed owing to decreased interaction of integrin α₁₀β₁ and α₁₁β₁ to the GFOGER sequence motif [18]. However, out of the four β1 integrin collagen receptors, binding of only two receptors (α₁₀β₁ and α₁₁β₁) was affected by citrullinated collagen II. The plausible explanation for this selective binding seems to be the differences in the structural organization. In the case of non-citrullinated collagen, the α₁₀ and α₁₁ I-domains can interact with the charged arginine which is flanked by two negative residues of α₁₀ and α₁₁ I-domains of the GFOGER sequence. On the other hand, there is no charge neutralization for integrin binding to citrullinated collagen, resulting in a weaker interaction [19]. Later, it was shown that integrin-binding motif, GAOGER of collagen III can also be citrullinated leading to reduced cellular adhesion [8].

Several studies have demonstrated that citrullination of FN attenuates its interaction with integrins [8,20–22], thereby influencing the behaviour of embedded cells, such as fibroblasts. Fibroblasts adhered less well on citrullinated FN (cit-FN)-coated glass and were smaller in size and yet more stiff. However, conflicting data exist as to how cit-FN affects other behaviours of fibroblasts and the signalling events downstream of integrins. For example, Shelef et al. demonstrated impaired migration of synovial fibroblasts on cit-FN and attenuated phosphorylation of focal adhesion kinase (FAK) and paxillin [21], whereas Stefanelli et al. reported superior migration and wound healing ability of fibroblasts upon interacting with cit-FN and heightened activation of both the FAK-Src and kindlin-integrin linked kinase-glycogen synthase kinase signalling pathways [22]. Stefanelli et al. further showed that cit-FN favours the use of α₅β₁ integrin over α_vβ₃ at focal adhesions and expedites their turnover. How to reconcile the contradictory observations is still unclear. Intriguingly, FN has 24 enzyme-specific citrullination sites with 14 out of them residing in the region known for physiological functions. However, citrullination of the RGD motif has never been detected in endogenous FN or even after in vitro citrullination, so far [8,22]. Instead, the influence of cit-FN on the behaviour of fibroblasts was proposed to be caused by the citrullination of the N-terminal NGR motif or the 9-10FN-III, including R1479, R1476 and R1452, which is remarkably close to the RGD motif.

(ii) . Alteration in proteolysis

The functions of many ECM proteins are subjective to regulation by proteolysis. In some cases, proteolysis is required for activation of proteins, such as collagens and TGF-β1, while in other case, such as FN, proteolytic modification can inhibit its activity [23]. The activity of the ECM proteases is further subjected to regulation by various protease inhibitors. Hence the balance between proteases and their inhibitors critically influences the ECM properties. This balance can be influenced by citrullination through at least two mechanisms.

Abolishing cleavage sites (figure 3a)

Figure 3.

Schematic diagrams depicting that citrullination modulates the remodelling of ECM composition by abolishing cleavage sites of arginine-directed proteases in ECM proteins (a) and regulating the activity of proteases in cis (the left model in (b)) or in trans (the right model in (b)).

Many proteases target arginine residues. PADs-mediated citrullination, by converting arginine to citrulline, has been shown to abolish the cleavage sites of the arginine-targeting proteases [24], and can potentially alter the proteolysis of ECM proteins. For example, FBLN5 can be cleaved by several proteases, including MMPs and porcine pancreatic protease [25–27]. The cleaved FBLN5 is unable to interact with lysyl oxidase like-1 (LOXL1) or participate in the formation of elastic fibre, i.e. elastogenesis [28], which will be discussed in detail later. We have found that lung FBLN5 is citrullinated at young mice (six weeks old) but not in adult mice (three month old) and that FBLN5 secreted by PAD2-deficient or BB-Cl-treated fibroblasts undergo spontaneous cleavage in culture [16]. Accordingly, PAD2 knock out (PAD2KO) or BB-Cl-treated fibroblasts had a profound defect in forming elastic fibre. Thus, PAD2-mediated citrullination promotes elastogenesis through protecting FBLN5 from proteolysis. Although the protease responsible for FBLN5 cleavage in the absence of PAD2 is still unclear, one of the known cleavage sites of FBLN5 is between R77 and G78, a potential target site of PADs. It remains to be determined whether R77 is citrullinated and whether the cleavage of FBLN5 in the absence of PADs does occur between R77 and G78. Given the similar structure and overlapping function among fibulins, it is highly possible that the cleavage and activity of other fibulins are also regulated by citrullination.

Modulating the activity of proteases

Citrullination can modulate protease activity in cis (the left model of figure 3b). Several MMPs are substrates of PADs [29,30]. For example, MMP1, MMP13 and MMP9 can be citrullinated by PAD2 or PAD4 in vitro and citrullinated MMP9 has been detected in synovial fluid from RA. Although the specific impact of citrullination on the function of MMPs remains to be elucidated, cit-proMMP9 is more efficiently cleaved and converted to MMP9 by MMP3. In addition, cit-MMPs have a higher affinity for gelatin [30]. Citrullination can also influence protease activity in trans (the right model of figure 3b). For example, FN inhibits the activity of ADAMTS4, a protease that degrades PGs in articular cartilage. Citrullination of FN markedly attenuates this inhibitory function [31], thereby enhancing the activity of ADAMTS4. In addition, citrullination can inhibit the activity of several SERPINs, which can act on MMPs and vice versa. Thus, citrullination can shape the characteristics and function of ECM proteins through the complicate proteases/protease inhibitors network.

(iii) . Modulating the activity of transforming growth factor-beta (figure 4)

Figure 4.

A schematic diagram showing that citrullination inhibits not only the release of active TGF-β through interfering with the interaction between LAP and integrins but also the binding of active TGF-β to TGF-β-RII.

The TGF-β family, including TGF-β1, TGF-β2, and TGF-β3, is the most important pleiotropic cytokine family involved in ECM biology. They are produced as preproproteins containing a N-terminal prodomain (latency-associated peptide, LAP) and the C-terminal growth factor. LAP is cleaved from the growth factor in endoplasmic reticulum but remains noncovalently tethered to the growth factor to form the small latent complex (SLC), which further form disulfide bonds with latent transforming growth factor β binding proteins (LTBPs), resulting in the large latent complex (LLC). LLC is then secreted into ECM but is biologically inactive. The release of active TGF-β in ECM can be achieved by proteolysis of LAP or structural deformation of LLC elicited by the interaction between the RGD domain of LAP and integrins [32].

Citrullination of the RGD domain of LAP has been shown to weaken its interaction with α_vβ₆ integrin, thereby attenuating the release of active TGF-β1 [17]. This mechanism very likely also regulates the release of active TGF-β3 given the sequence and structural similarities of the RGD motif between these two TGF-β members. In addition, TGF-β1 uses two arginine residues, namely R25 and R94, to interact with TGF-β receptor 2 (TGF-β-RII), and citrullinated TGF-β1 binds poorly to TGF-β-RII in vitro. Whether the poor binding is owing to the citrullination of R25 and/or R94 remains to be determined. Of note, these two arginine residues do not participate in the TGF-β1/TGF-β-RI interaction and are replaced by two lysine residues in TGF-β2. Thus, citrullination very likely preferentially attenuates the activity of TGF-β1 and TGF-β3 on TGF-β-RII. However, the functions of TGF-β cytokines are context- and condition-dependent, making it particularly challenging to predict the eventual impacts of citrullination on the function of TGF-β. Furthermore, it was observed that other ECM-bound growth factors such as FGF and platelet-derived growth factor had at least one arginine residue at their receptor interaction sites, suggesting a wider reach of citrullination [17].

(iv) . Intrinsically altering the property/function of extracellular matrix proteins

Citrullination has been shown to intrinsically alter the property/function of nuclear, cytoplasmic and membrane proteins through modulating their DNA binding, protein-protein interaction, and intracellular localization independently of proteolysis. Thus, it is logical to assume that similar effects can be seen in ECM proteins. Surprisingly, such effects of PADs have so far been demonstrated only in a few ECM proteins. For example, citrullination of fibrinogen/fibrin has been well documented and shown to alter the property of blood clots [33,34]. Interestingly, fibrinogen/fibrin is also an endogenous ligand of toll-like receptor 4 (TLR4), and cit-fibrinogen/fibrin appears to be more potent in activating TLR4 than its native counterpart [35,36] (figure 5). Activation of TLR4 has also been reported with citrullinated vimentin [37]. However, the mechanisms mediating the preferential effect on the TLR4 pathway is still unknown. In addition, as discussed later, cit-fibrinogen/fibrin can also regulate the differentiation and function of osteoclasts, macrophages, and bone marrow mesenchymal stem cells independently of TLR4.

Figure 5.

A schematic diagram showing citrullination of fibrinogen/fibrin alters TLR4 signaling.

4. Citrullination of extracellular matrix proteins in human diseases

Given the regulatory roles of citrullination in ECM proteins, it has become clear that aberrant citrullination can also contribute to human diseases through modulating the characteristics of ECM. Here we will discuss the potential roles of ECM citrullination in several human diseases.

(a) . Aberrant homeostasis of elastic fibre (cutis laxa and secondary emphysema)

Citrullination can critically regulate elastogenesis. The building block of elastic fibre, i.e. tropoelastin (TE) is transported to the plasma membrane where it is bound by FBLN4 and FBLN5, which bring in lysyl oxidase (LOX) and LOXL1 proteins to facilitate initial crosslinking of TE and the formation of TE microaggregates. The TE microaggregates are then recruited by FBLN5 and LTBP4 to microfibril consisting of FN and FBN. Once the TE microaggregates are deposited on the microfibril complex, they interact with other proteins including LTBP-1 and LTBP-2, and undergo the final crosslinking, which is also mediated by LOX and LOXL1, eventually forming the mature elastic fibre. Impaired elastogenesis caused by loss-of-function mutations in the elastogenic genes, such as Eln, Fbln5, Ltbp4, and Fbln4, results in a wide spectrum of developmental defects in mice, including severe emphysema, loose skin, tortuous aorta, and genital prolapse [38,39], and cutis laxa in humans. Cutis laxa can also be acquired owing to destruction of the existing elastic fibre, which can be seen in patients with multiple myeloma, infection, or mastocytosis. In addition, destruction of elastic fibre is a hallmark of secondary emphysema seen in chronic obstructive pulmonary disease (COPD) commonly caused by cigarette smoking.

The potential role of aberrant citrullination in emphysematous lung diseases is further supported by increased lung compliance observed in young PAD2KO mice [16], which also spontaneously develop emphysematous changes when they are six months or older. One explanation for this phenotype is impaired elastogenesis owing to enhanced susceptibility of FBLN5 to proteolysis in the absence of PAD2. The impairment is probably subtle but leads to inferior quality of elastic fibre, resulting in increased lung compliance in young mice and emphysematous changes only in adult mice. PAD2 may also promote elastogenesis through FBLN5 cleavage-insensitive mechanisms. FBLN5 physically interacts with several other elastogenic proteins, including ELN, FN, LTBP4, FBN1/2, and LOXL1, mainly through the fibulin module at its C-terminus [28,40]. In addition, FBLN5 very likely operates in dimeric or multimeric forms [25]. Intriguingly, the fibulin module contains at least three confirmed citrullination sites (R326, R351, and R408). It is possible that citrullination of FBLN5 at its fibulin module is critical for its proper interaction with the other elastogenic proteins or its multimerization. Furthermore, PAD2 may promote elastogenesis through FBLN5-independent mechanisms by citrullinating other elastogenic proteins, such as FBLN4, which contains two arginine residues equivalent to the R351 and R408 of FBLN5, or LTPB4, which is present along with FBLN5 in the protein fraction enriched with citrullinated proteins obtained from the lung of young mice [16].

Interestingly, upregulation of FBLN5, LTBP4 and ELN has been observed in COPD lungs [41], probably representing the body's attempts of regenerating elastic fibre; however, no regeneration of functional elastic fibre has ever been observed. As physiological citrullination of FBLN5 in lung occurs only at a young age, it is possible that the failure to regenerate functional elastic fibre in COPD lungs is owing to hypocitrullination of FBLN5. Paradoxically, cigarette smoking has been shown to increase the expression of PAD2 and the degree of citrullination in bronchial lavage [42]; and the total level of citrullinated lung proteins was reportedly higher in COPD patients compared to non-smokers [43]. These observations suggest that if citrullination is essential for regenerating functional elastic fibre, it must be directed to the right cells, right proteins and right arginine residues.

(b) . Rheumatoid arthritis

RA is an autoimmune disease characterized by joint inflammation in a symmetrical pattern. The joint inflammation is mediated by a set of inflammatory cytokines, including tumour necrosis factor alpha (TNF-α), IL(interleukin)-6, and IL-1, which can be produced by immune cells and fibroblast-like synoviocytes (FLS). Approximately 60% of patients with RA have anti-citrullinated protein antibodies (ACPA) in their serum, which can predate the onset of clinical symptoms by decades and are associated with poor prognosis. Several major genetic and environmental risk factors of RA are associated with local or systemic hypercitrullination [44–47]; citrullination but no other types of PTM in synovial fluid correlated with the activity of inflammation in RA [8]. In addition, deficiency of PAD2 or PAD4 attenuates joint inflammation in animal models of inflammatory arthritis [48–51]. The beneficial effects of PAD deficiency have been attributed to the functional impact of PADs intrinsic to immune cells or on the activity/expression of inflammatory cytokines. Nevertheless, PADs can also contribute to the pathogenesis of RA through acting on ECM proteins. For example, the aforementioned effects of citrullination on the proteases/protease inhibitors network in ECM can potentially facilitate joint damage in RA. In addition, many of the citrullinated antigens recognized by ACPAs are ECM proteins, such as collagens and FN. A recent study further demonstrates that citrullination of vimentin and fibrinogen results in the generation of neoantigens composed of a native sequence of these two proteins. Such native neoantigens are able to activate T cells from ACPA + RA patients [52]. Emerging data have further suggested active pathogenic roles for some of the citrullinated ECM proteins in RA.

Extracellular aggregates of cit-FN were detected in RA but not osteoarthritis (OA) synovial tissue [53], and cit-FN was detected in the serum of RA patients but not healthy controls or lupus patients [20]. As described above, the status of FN citrullination critically influences the behaviour of fibroblasts. Several studies have strongly suggested that FN has a protective role in RA. It can induce the apoptosis of FLS and HL-60 [20,53], a human neutrophil-like cell line. This effect is associated with increased expression of caspase-3 but reduced expression of survivin and cyclin-B1 [53]. In addition, FN binds and inhibits the activity of ADAMTS4 [31], a metalloprotease that can degrade aggrecan and contribute to cartilage destruction seen in RA. These protective effects of FN are attenuated by PAD2 and/or PAD4-mediated citrullination. Furthermore, cit-FN compared to native FN induces the expression of TNF-α, IL-1 and IL-17 in FLS [53]. In agreement with these observations, the citrullination level of R234 in FN, which is within the integrin-interacting isoDGR motif, strongly correlates with the number of leukocytes in synovial fluid [8].

Fibrinogen is converted to insoluble fibrin during the formation of blood clots. Emerging data, however, has suggested an unexpected role of fibrinogen in RA. Immunization of C57BL/6 mice with human fibrinogen, which was purified from human plasma and was reactive to anti-citrulline antibody, elicited ACPA response and led to joint inflammation and cartilage loss upon intra-articular challenge in an IL-17/IL-23 dependent manner [54]. Citrullinated fibrin is an autoantigen of ACPAs and the level of cit-fibrinogen is higher in RA synovial fluid compared to that of OA fluid [35,55]. Fibrinogen/fibrin is an endogenous TLR4 ligand and cit-fibrinogen/fibrin is more potent than its native counterpart in inducing the expression of IL-6 and IL-8 in RA FLS or bone marrow mesenchymal stem cells, which have immune-modulatory function, through the TLR4/nuclear factor-kappa B (NF-kB) pathway [35,36], and the production of IL-1β, TNF-α and monocyte chemoattractant protein-1 by macrophages in vitro [56]. The formation of immune complex between cit-fibrinogen/fibrin and anti-cit-fibrinogen/fibrin, which probably occurs in patients with RA, can activate macrophages through both TLR4 and FcγR, thereby further potentiating the production of TNF-α [57]. In addition, exogenous fibrinogen inhibits the differentiation of osteoclasts from CD14+ monocytes. This inhibitory effect of fibrinogen is mitigated upon citrullination [55]. Furthermore, exogenous cit-fibrinogen reduces the production of indoleamine 2,3-dixoygenase by bone marrow mesenchymal stem cells and render them unable to suppress the proliferation of peripheral blood mononuclear cell and the expression of IL-17 by Th cells [36]. Thus, citrullination of fibrinogen very likely enhances joint inflammation and erosion seen in RA.

(c) . Organ fibrosis

Previous studies have demonstrated that the characteristics of ECM, such as proteolysis, cross-linking and glycosylation, can influence the process of organ fibrosis [58]. As citrullination can interfere with the release of active TGF-β and/or its binding to TGF-β-RII, one may postulate that citrullination would inhibit organ fibrosis. Unexpectedly, emerging data actually has strongly suggested a pathogenic role of citrullination in organ fibrosis. The transcript levels of PAD2 and PAD4 as well as the level of citrullinated proteins were higher in human idiopathic pulmonary fibrosis (IPF) [37,59]. PAD4KO mice were more resistant to age-related organ fibrosis [60], pressure overload-induced cardiac fibrosis [60], and bleomycin-induced pulmonary fibrosis [61], whereas PAD2KO mice were more resistant to pulmonary fibrosis caused by inhalation of cadmium (Cd)/carbon black (CB) [37], and knock-down of PAD2 exerted antifibrotic effects in lung fibroblasts [62]. In addition, higher levels of PAD2 and citrullinated proteins, particularly citrullinated glial fibrillary acidic protein, were detected in the liver tissue in the animal model of liver fibrosis triggered by bile duct ligation [63]. These observations have been attributed to cell-autonomous effects of PADs in neutrophils (PAD4), fibroblasts (PAD2 and PAD4), and macrophages (PAD2). For example, the pro-fibrotic effect of PAD4 was reportedly owing to the formation of NETs. Links between citrullination of ECM proteins and organ fibrosis are still missing; however, several studies have suggested that citrullinated vimentin may be one of the missing links.

Vimentin is a cytosolic protein but can also be secreted and modulate the biology of ECM. Citrullinated vimentin is one of the most common antigens of ACPAs and the serum level of a citrullinated MMP-degraded vimentin peptide, as measured with ELISA, correlated with the degree of fibrosis in a rat model of liver fibrosis and in human with hepatitus B virus infection or non-alcoholic fatty liver disease [64,65]. A recent study further shows that the level of cit-vimentin was higher in lung tissue (measured with Western blotting) and plasma (measured with ELISA) of patients with IPF than in controls, and demonstrates a critical role of cit-vimentin in the pathogenesis of IPF induced by Cd/CB [37]. Cd/CB are common components of cigarette smoke and air pollutants, which are known risk factors of IPF. Uptake of Cd/CB by macrophages increased the levels of PAD2 and subsequently cit-vimentin, which was secreted and acted on lung fibroblasts through the TLR4/NF-kB pathway, promoting their differentiation into myofibroblasts, invasiveness, and the expression of collagen, α-smooth muscle actin, and inflammatory cytokines, such as TGF-β1, connective tissue growth factor, and IL-8. Accordingly, inhalation of Cd/CB caused lung fibrosis in a PAD2-dependent manner and intra-tracheal treatment with cit-vimentin, but not native vimentin also led to TLR4-dependent fibrotic remodelling of lung in mice. Nevertheless, cit-vimentin can potentially contribute to organ fibrosis through additional mechanisms. For instance, the secreted cit-vimentin can have yet-to-be-discovered effects on the characteristics of ECM. In addition, fibroblasts express both PAD2 and vimentin, and lung fibroblasts produce cit-vimentin in a PAD2-dependent fashion [62]. The intracellular cit-vimentin may modulate the integrin signalling induced by ECM proteins in fibroblasts and facilitates their differentiation into myofibroblasts. This scenario can explain the intrinsic profibrotic effect of PAD2 in lung fibroblasts.

(d) . Tumour metastasis

Aberrant expression of PADs has been observed in many types of tumour cells and can intrinsically modulate the behaviour of tumour cells [66–68]. In addition, the intrinsic effect of PADs on immune cells can critically influence the growth, invasion and metastasis of tumours. A recent study elegantly demonstrates promotion of liver metastasis of colorectal cancer (CRC) by citrullination of ECM proteins [15]. Enrichment of citrullinated proteins, such as collagen I, FN1, and FBN1, was observed in matrisome from liver metastasis of human CRC but not in the original sites. The enzyme responsible for the enrichment of citrullinated matrisome was determined to be PAD4, which was produced by metastatic CRC, but not neutrophils, and released into ECM through extracellular vesicles. PAD4 once released by CRC citrullinated ECM proteins, specifically collagen I. Citrullinated collagen I promoted the adhesion of CRC cells, inhibited their motility, and induced weaker signals downstream of integrins, such as phosphorylation of FAK, extracellular signal-regulated kinase, and c-Jun N-terminal kinase, and facilitated mesenchymal-to-epithelial transition. More strikingly, pharmacological, or genetic inhibition of PAD4 inhibited liver metastasis of CRC in animals but had little impact on the tumour growth in the original sites. These observations highlight the importance of citrullination of ECM proteins in metastasis of CRC. It remains to be determined whether the effect of citrullinated collagen I is mediated by the citrullination of the GFOGER motif and whether this mechanism also regulates the metastasis of other types of cancer. A recent study re-analysing published proteomic data also identified citrullinated ECM proteins, including FN1, fibrinogen and collagen VI, in 16 out of 24 cancer-related datasets, Surprisingly, no citrullinated ECM protein was detected in spheroid culture of cancer cells even in the presence of stromal fibroblasts. Instead, the degree of ECM citrullination correlated with the presence of inflammatory cells [69]. Thus, the exact mechanism responsible for the citrullination of matrisome in cancer is still unclear. One possible explanation for the seemingly contradictory results is that the presence of inflammatory cells is necessary for metastatic cancer cells to express PADs and citrullinated ECM proteins.

5. Conclusion

Published data have demonstrated that citrullination can regulate the property and function of ECM through various mechanisms and, by doing so, is critical for maintaining tissue homeostasis. Nevertheless, the full spectrum of extracellular citrullinome and the functional consequence of each citrullination event are still poorly characterized. Many questions remain unanswered. For example, how do PADs select their targets? What is the impact of each PAD enzyme on the characteristics of ECM? Is the citrullination of ECM proteins temporally and spatially regulated and, if so, how? The progression in this field has been hindered by several technical limitations, such as lack of simple and reliable methods for detecting citrullination events and introducing site-specific citrullination in live cells or in vivo. Furthermore, potential overlapping functions of PADs and poor correlation of citrullination with PAD expression pose additional challenges. Major technical breakthroughs overcoming these hurdles will markedly advance the field of functional extracellular citrullinome and enable us to tap into the therapeutic potential of manipulating the matrisomal citrullination in various human diseases.

Data accessibility

This article has no additional data.

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

M.A.S.: conceptualization, data curation, formal analysis, validation, visualization, writing—original draft, writing—review and editing; I.-C.H.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing—original draft, writing—review and editing.

Both authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

We received no funding for this study.