Peptidylarginine deiminases in citrullination, gene regulation, health and pathogenesis

1.1 The early discovery of PADs as protein citrullination enzymes

Citrulline residues were detected in early 1960s from polypeptide hydrolysates of hair inner root sheath cells and medullary cells [1]. Since there is no citrulline tRNA in vivo, citrulline can only be produced by enzymatic modifications after protein synthesis. Subsequently, the peptidylarginine deiminases (PADs, also called PADIs) were identified, which convert protein arginine residues to citrulline in a calcium dependent manner ([2], reviewed in [3]). PADs are involved nomal functions in the immune and reproduction systems as well as in the skin (see recent review in these particular areas in [4–6]). In each of the mammalian vertebrate genomes, five highly conserved PADs exist, including PAD1, 2, 3, 4 and 6. (Fig. 1A). In the last decade or so, more research on the gene features, tissue-specific distribution and preferred substrates of PADs has been performed. Human PAD4 can bind five calcium ions (Fig. 1B). Several flexible parts of PAD4 form stable secondary structures after the binding of calcium and substate, indicating that calcium stablizes the conformation of PAD4 and may facilitate the formation of the active site cleft (Fig. 1C and 1D). The calcium binding sites in PAD4 are conserved in PAD1, −2, and −3 except for PAD6. Thus, calcium is an important regulator of the active PAD enzymes [7].

1.2 The five PADs

2.1. PAD4 in gene regulation

In eukaryotic cells, nuclear DNA is organized with two of each histones H3, H4, H2B, and H2A to form a nucleosome core particle, which is further organized with the linker DNA and histone H1 to form the 10 nm chromatin fiber and folded to form the higher order chromatin structures. The nuclear structure and 3D organization is under continuous remodeling to adapt to the physiological and environmental changes that the cells are exposed to. Because of this structural organization, histone modifications, including methylation, acetylation, phosphorylation and citrullination, work as a signaling network to provide the on the off signal for gene expression and/or a landing platform for effector protein binding [42–44].

Histone modifying enzymes with opposite activities counteract each other’s effect, such as histone acetyltransferases (HATs) and histone deacetylases (HDACs), kinases and phosphatases [45, 46]. Until the Lys demethylases, such as LSD1 and JmjC domain-containing dioxygenase, as well as the Arg deiminase PAD4 were identified [47, 48], histone methylation on Arg and Lys residues was considered as static rather than dynamic because of the low turnover rate of the methyl groups [49]. PAD4 antagonizes CARM1 (also called PRMT4) and PRMT1 mediated histone H3 and H4 Arg methylation through a reaction dubbed as demethylimination in reflecting the removal of the methyl-imine group from monomethyl-arginine residues [33]. The activity of PAD4 on asymmetrical dimethyl-arginine residues is very low. We have observed that PAD4 prefers methyl-Arg in histone proteins over methyl-Arg in short peptides as substrates, suggesting the substrate has an allosteric effect on PAD4 [33]. CARM1 and PRMT1 function as transcription coactivators by catalyzing histone Arg monomethylation and asymmetrical dimethylation [42, 50–53]. By antagonizing Arg methylation, PAD4 functions as a transcription corepressor. In the case of ER target genes in the breast cancer MCF-7 cells, PAD4 regulates histone Arg methylation via its citrullination activity on the gene promoters [33, 54]. Interestingly, the modification of histone Arg residues on the ER (estrogen receptor) target promoters fluctuates over time after estradiol treatment, whereby the increase in histone citrullination correlates with the decrease in histone Arg methylation, indicating that opposite enzymes are alternatively working on the ER target gene promoters [55].

In addition to the ER target genes, our group has found that PAD4 interacts with the tumor suppressor and transcription factor p53 and functions as a corepressor to regulate the expression of multiple p53 target genes [56, 57]. Before DNA damage, a high level of histone citrullination and PAD4 was detected on the promoter of p53 target genes, such as p21/CIP1/WAF1, GADD45 and PUMA [56, 57]. After DNA damage, PAD4 association and histone citrullination decreases on these gene promoters with a concomitant increase in histone Arg methylation, suggesting that citrullination and arginine methylation counteract each other’s function to regulate gene expression.

The role of histone acetylation in gene activation has been long established. A landmark 2004 paper with rich in vitro biochemical analyses showed that protein Arg methyltransferases and histone acetyltransferase p300/CBP cooperatively activate the p53-mediated transcription [42]. Reversely, we found that PAD4 and HDAC2 associate with p53 in a dynamic fashion [56]. Before DNA damage, both PAD4 and HDAC2 are associated with p53 target gene promoters, while they dissociate from the gene promoters after DNA damage allowing the activation of the p53 target genes, such as p21, GADD45 and PUMA [56]. Taken together, we propose that histone Arg modification in concert with histone Lys acetylation forms a molecular switch on the p53 target gene promoters for gene regulation (Fig. 2).

Although the corepressor function of PAD4 has been well established, it may also play a coactivator role in a promoter context dependent manner. A genome wide ChIP-chip study of PAD4 promoter association in MCF-7 cells found that PAD4 is enriched on the promoter regions of actively transcribed genes [58]. Motif analyses found that many of the PAD4 bound genes contain potential binding sites for Elk-1, a member of the ETS oncogene family [58]. It was proposed that PAD4 interacts with and citrullinates Elk-1 thereby facilitating Elk-1 phosphorylation to activate transcription [58]. Additionally, PAD4 can target histone H3 Arg8 for citrullination and subsequently affects the binding of HP1 (heterochromatin protein 1) to the nearby H3 Lys9 methylation site [59, 60]. The dissociation of HP1 from its binding cognate sites after citrullination activates transcription in multiple sclerosis patients [60]. In addition, the dissociation of HP1 likely regulates chromatin decondensation during the formation of neutrophil extracellular traps (NETs) [59].

The hypothesis of the “histone code” predicts that histone modifications may function in a combinatorial manner to regulate chromatin biology [61]. Two adjacent modification sites were proposed to form a binary code to antagonize or synergize the function of each other [62]. The effect of histone H3 Ser10 phosphorylation on the function of H3 Lys9 methylation and HP1 in cell cycle and gene regulation was well studied [63, 64]. The new finding of the effect of H3 Arg8 citrullination on HP1 binding to H3 Lys9 methylation greatly enriched the binary code concept and highlights the dense cell signaling information that a cluster of histone residues carries (Fig. 3).

2.2. PAD2 in gene regulation

That PAD2 can citrullinate histone H4 Arg3 in vitro raises a role of PAD2-mediated histone citrullination in transcriptional regulation [21]. The expression of PAD2 is regulated by estrogen in vertebrate uterus and pituitary gland [65–67]. Recent data from the Coonrod’s group support that PAD2 responses to cellular signals to regulate transcription via histone citrullination. Only during the diestrus phase of the reproduction cycle, PAD2 in mammary gland epithelial cells was found to citrullinate the histone H3 N-terminus [19]. Moreover, PAD2 is detected in human breast luminal epithelial cells, and associates with target gene promoters in the ERα positive breast cancer MCF-7 cells to regulate histone H3 Arg26 citrullination and transcription [20]. Estradiol stimulates the recruitment of PAD2 to the estrogen-response element of estrogen receptor alpha (ERα) target gene promoters. Since PAD2 does not have a nuclear localization signal, its association with ERα may facilitate its translocation from the cytosolic to the nuclear compartment [20]. As such, PAD2 likely mediates chromatin decondensation and activation of target gene transcription during cellular response to estrogen stimulation [23]. These studies indicate that PAD2 functions as an epigenetic regulator of gene activity and plays a potential role in breast cancer progression.

3.1. NETs, a highly decondensed form of chromatin

Peripheral blood neutrophils serve as the first line of defense against microbial infection [68–70]. Although neutrophils are well known for their antimicrobial function via phagocytosis, it was reported in 2004 that a distinct mechanism of killing bacteria in the extracellular space using a meshwork of highly decondensed chromatin in association with antimicrobial proteins (e.g., myeloperoxidase and neutrophil elastase) [68]. Upon encountering bacteria, a certain percentage of neutrophils release their chromatin into the extracellular space, forming chromatin-based structures called neutrophil extracellular traps (NETs) (reviewed by [71]). NET formation can be induced by IL-8, PMA, LPS and diverse microbes, such as bacteria and fungi, as well as protozoan parasites [68, 72]. As a mechanism of immune defense, NETs could mediate the microbial death and/or limits pathogen spread in host [68, 73, 74]. Although the role of NETs in innate immunity and human disease has gained much attention, the mechanism regulating the NET formation is less clear. The activation of the intracellular signaling pathways, such as the MAPK pathway and the ROS (reactive oxygen species) signaling pathways, has been implicated in the formation of NETs [75–77]. Since NET formation directly leads to cell death (dubbed netosis) and the release of nuclear and cytoplasmic components into the extracellular space has harmful consequences to the surrounding normal tissues, this process needs deliberate regulation.

3.2. Histone hypercitrullination in NET formation

Chromatin in NETs is extremely decondensed, with a diameter ranging between 10–30 nm [68]. This degree of chromatin decondensation indicates that the neutrophil chromatin can be unfolded extensively to form the polymers of nucleosomes, i.e., the 10 nm chromatin fibers. A proteomic study of the NET protein components has found that even linker histone H1 is diminished from NET chromatin [78]. However, the chromatin decondensation mechanisms underlying NET formation remain further explored.

Retrospectively, our finding that PAD4 and histone hypercitrullination are involved in the formation of neutrophil extracellular traps is a serendipitous story. In the early months of 2004, we found that prolonged treatment of HL-60 granulocytes with calcium ionophore produced sticky chromatin that glues many of cells in a cluster. The gluy chromatin was stained with the DNA dye and very strongly stained with the histone H4Cit3 antibody, suggesting that there is a link between the high-degree of histone citrullination and the formation of NETs. The link of PAD4 and NET formation was eventually published in 2009 [79]. Around the time we were preparing to publish our report, the Radic group published immunostaining images to show that NET chromatin can be stained by a histone citrullination antibody [80]. It is now clear that one the normal physiological function of PAD4 in peripheral blood neutrophils is to regulate the formation of NETs.

Histone hypercitrullination is an important intermediate step for the formation of NETs. In HL-60 granulocytic cells, the inhibition of PAD4 and histone hypercitrullination decreases chromatin decondensation and the formation of NET like structures [79]. Importantly, citrullination of a 12-mer nucleosome array decreased the chromatin folding by the linker histone H1 [79]. The role of PAD4 in NET formation is further supported by genetic studies in mice. The PAD4 knockout mice can survive to adulthood and are fertile with a normal number of the neutrophil counts [81]. In PAD4 null neutrophils, the basal level of histone citrullination is undetectable. After stimulation with LPS, H₂O₂, PMA and bacteria, histone citrullination is greatly increased in wild type neutrophils but remains undetectable in the null cells [81]. The lack of histone citrullination is correlated with a lack of NET formation in the null cells, indicating the PAD4 is required for NET formation under these experimental conditions [81]. Furthermore, elevated PAD4 activity is sufficient for the chromatin decondensation to form NET-like structures. First, treatment of detergent permeated cells with GST-PAD4 but not its catalytic inactive mutant GST-PAD4 C645S induced the release of chromatin from HL-60 cells to form decondensed chromatin structures [79]. Surprisingly, forced overexpression of PAD4 in osteosarcoma U2OS cells was also sufficient for the formation of NET-like chromatin structures [59]. These results indicate that PAD4 activity is required and sufficient for the formation of decondensed chromatin, a process that is essential for the formation of NETs.

How does citrullination mechanistically unfold chromatin? First, the degree of histone citrullination associated with NET formation is very high. Based on Edman degradation experiments to analyze histone citrullination, we estimate that over 50% of histone H3 Arg8 can be converted to citrulline [33]. Since citrullination eliminates the positive charge of histones, this modification may weaken the interaction of histones and DNA thereby decreasing chromatin fiber compaction to facilitate NET formation. Second, PAD4 mediated histone hypercitrullination could exclude HP1β from binding to chromatin, offering a new mechanism for PAD4 mediated NET formation [59]. Third, citrullinated nucleosome arrays are not folded properly by the linker histone H1. Moreover, many other proteins can be targeted by PAD4 for citrullination. These other PAD4 substrates could paly a modulatory role for the process of NET formation.

4.1 Rheumatoid Arthritis

RA is a chronic autoimmune disease featured with inflammatory synovium and infiltration of activated macrophages. The anti-citrullinated protein antibodies are the most specific autoantibodies present in the RA sera [88–90]. Most of these autoantibodies can be detected in the early stage of the disease [91, 92], which makes them useful diagnostic markers of RA [93–95]. Those citrullinated proteins, including fibrin, fibrinogen and vimentin, were produced in synovial fluid of the inflammatory joints [96–101]. The presence of citrulline residues in these proteins sends the immune cells a false alarm and initiates immune responses to generate anticitrulline antibodies against these proteins [102, 103]. Many PAD2 and PAD4 expressing leucocytes infiltrate into the inflammatory synovial tissues in RA patients and release large amount of PAD2 and PAD4 in the synovial fluid, which in turn produce high levels of citrullinated proteins [104, 105]. Since the calcium concentration in the extracellular space is at the millimolar levels, PAD4 released from the dying neutrophils during the NET formation process can be super-activated to citrullinate joint proteins [16]. In addition to these pathology links, the involvement of PAD4 in rheumatoid arthritis is also supported by human genetic studies, in particular in the oriental ethnic groups [106–108].

4.2 Multiple Sclerosis and Alzheimer’s

MS is mainly a PAD2 related inflammatory disorder in the central nervous system associated with excess citrullination of myelin basic protein (MBP), resulting in the demyelination of the myelin sheath and affecting the nerve signal transduction [109, 110]. About 18% of MBP protein is citrullinated at 6 out of its 19 Arg residues in healthy individuals, while 45% or more of MBP is citrullinated in MS patients [111]. Hypercitrullination of MBP weakens its interactions with phospholipid and disrupts the formation of normal multilayer myelin sheath structures. In addition, hypercitrullinated MBP is more susceptible to protein degradation by the protease cathepsin D [112–114]. Elevated PAD2 in the myelin sheath is related to an increase in MBP citrullination prior to the onset of MS symptoms [109, 115]. PAD2 overexpression in transgenic mice increases the amount of citrullinated MBP and accelerates the development of demyelination [116–118]. Moreover, upregulation of PAD2 and the presence of the inflammatory signals at the affected axon regions might further increase PAD4 locally to exacerbate the inflammatory disease [115, 119]. It is surprising that knockout of PAD2 did not produce a significant phenotype in the nervous system given the importance of MBP citrullination in the axon electrical signal transmission [120]. Whether other PAD enzymes, such as PAD4 play a redundant role with PAD2 in the central nervous system remain to be solved.

In addition to multiple sclerosis, PAD2 and PAD4 are also expressed in neurons [121]. Protein citrullination in the central nervous system may cause protein denaturation and precipitation, which in turn lead to the Alzheimer’s disease [121–123]. These studies highlight the possibility of PAD2 as a drug target for the treatment of nervous system diseases.

4.3 Cancer

Epigenetic modification plays an important role in tumorigenesis [124, 125]. Two major epigenetic alterations in cancer cells are histone modifications and DNA methylation. In this session we will focus on the effect of histone modifications on aberrant tumor suppressor gene silencing during tumorigenesis [126, 127]. Tumor suppressor p53 is one key transcription factor in maintaining cellular homeostasis. In response to diverse upstream signals, such as starvation, DNA damage and various stress signals, p53 regulates its downstream genes to cope with stress and control the cell fate [128–131]. Many of the p53 target genes are tumor suppressor genes that regulate the cell cycle arrest, programed cell death, and autophagy.

Under normal conditions, PAD4 is mainly expressed in the peripheral blood leukocytes. In pathology studies using a large cohort of human patient samples, PAD4 is overexpressed in a majority of tumor tissues, including osteosarcoma, colon adenocarcinoma, esophagus adenocarcinoma, ovary adenocarcinoma, pancreas adenocarcinoma, and stomach adenocarcinoma [29, 30]. Given that PAD4 functions as a corepressor of p53 to regulate its downstream tumor suppressor genes, the overexpression of PAD4 in tumor tissues suggests that it might be involved in tumorigenesis. Interestingly, the expression of PAD4 is directly regulated by p53, suggesting that PAD4 may form a negative feedback loop to regulate p53 [132]. As we have discussed in the previous section, PAD4 catalyzed histone citrullination coupled with HDAC2 catalyzed deacetylation represses p53 target gene expression [56]. Since HDACs are targets for cancer drug development, we recently devoted a solid effort to test if PAD4 is a druggable target for cancer treatment (see below).