An interplay of structure and intrinsic disorder in the functionality of peptidylarginine deiminases, a family of key autoimmunity-related enzymes

Calcium as a key regulator of PAD activity

Peptidylarginine deiminase enzymes are a family of calcium-dependent enzymes that catalyze the conversion of arginine residue of substrate into citrulline amino acid, which is accompanied by loss of positive-charged arginine (basic) and producing non-charged citrulline (neutral) [6, 29]. The PAD-driving citrullination requires supraphysiological levels of calcium since these enzymes are inactive under normal physiological conditions, where the calcium concentration is low. The normal intracellular level of calcium is different from the extracellular level. The physiological extracellular concentration is relatively low (1.1–1.3 mM) in plasma, and (0.49–0.98 mM) in synovial fluid [31, 32], whereas the physiological intracellular cytosolic calcium level is very low (10⁻⁵–10⁻³ mM) [29]. A study was conducted to evaluate the effect of different bivalent ions on the activation of PAD enzymes, which revealed that calcium is required for the catalytic activity of PADs [33]. This study showed that the calcium concentration required for enzyme to reach the half-maximal activation (K_a) is (0.04–0.06 mM). In other studies, the threshold level of calcium to activate PADs was reported to be around 10⁻² mM, which is still noticeably higher than the cellular physiological calcium level, and is, therefore, insufficient to induce enzymatic activity [34]. Therefore, the citrullination of substrate proteins and PAD enzyme activation need large amounts of calcium, which can be provided through an influx of the extracellular calcium, or by releasing it from intracellular calcium stores [5]. Increased calcium levels can occur when calcium homeostasis is dysregulated, such as in cell death and epidermal differentiation [35, 36]. Interestingly, different experiments suggest that other co-factors, such as phosphatidylserine, can reduce the calcium requirement for PAD activation. Phosphatidylserine is a phospholipid component of the cell membrane, specifically in the fatty layer of myelin sheath surrounding the neural cells. This phospholipid is released during myelin disorder or demyelination, and was found to minimize the calcium requirement for PAD activity and increase the reaction velocity. Experimental data showed that phosphatidylserine can reduce the calcium level required for the half-maximal activity from (0.04–0.06) to (0.015–0.02 mM) [33].

Periodontitis

Periodontitis is a bacterial infection of the mucosal gum that is characterized by the inflammation and destruction of gingival and subgingival tissues. Several microorganisms are known to cause this disease, with the main pathogenic species of this infection being Gram-negative bacteria [71]. According to the disease progression and levels of destruction, periodontitis can be classified as a chronic or aggressive infection [72]. The biofilm of subgingival tissues in periodontitis lesions showed the presence of a variety of bacteria species, with the major bacteria causing the most aggressive destruction being Aggregatibacter actinomycetemcomitans (A. actinomycetemcomitans), Porphyromonas gingivalis (P. gingivalis), and Prevotella intermedia (P. intermedia) [73]. Several studies linked periodontitis to RA inflammation and autoimmunity. These studies showed that aggressive periodontitis, through its causative bacteria, can induce profound citrullination of different proteins. This process produced modified proteins that have been detected in the synovial fluid of RA joints providing a source of autoantigens for autoimmunity in RA patients [74]. Previous studies have identified P. gingivalis as an oral pathogen that can cause periodontitis and induce the citrullination of mucosal protein such as α-enolase (434 residue-long protein with the PPIDR^®_PONDRVSL2 of 16.6%), keratin, and vimentin. This pathogen expresses bacterial PAD enzymes (PPAD) that are calcium-independent, unlike human PADs, which require calcium for activation [75, 76].

Recently, several studies presented significant evidence linking the aggressive periodontitis caused by A. actinomycetemcomitans with the activation of human PADs, particularly PAD4 and PAD2 [77]. The abnormal activation of PADs influenced by periodontitis leads to hypercitrullination of histone proteins in neutrophils, which is known to trigger autoimmunity in RA. Neutrophils are hematopoietic cells residing the periodontal mucosa and involved in RA inflammation as a main source of autoantigen and citrullinated protein [78, 79]. Recent findings identified the mucosal surface of different tissues of the body as an initiation site for RA, such as periodontium [80]. In the infection of periodontal tissues by A. actinomycetemcomitans, this pathogen releases its virulence factor, known as bacterial pore-forming leukotoxin A (LtxA) [81]. Exposure of neutrophils to LtxA affects their membrane integrity, causing membrane destabilization and allowing for an increased influx of high amounts of extracellular calcium into leukocytes, resulting in rapid and abnormal activation of PADs [82]. The dysregulated activation of PADs induces hypercitrullination of histones. These citrullinated proteins are detected as autoantigens in the synovium of RA that cause inflammation and autoimmunity [81]. Furthermore, the neutrophils with highly activated PADs (PAD2 and PAD4) infiltrate into synovial joint and release high amounts of these activated PADs into the extracellular environment, where they stimulate the citrullination of extracellular proteins such as fibrin and antithrombin, which, in turn, can trigger an immune response and cause inflammation at synovial joints [2, 60].

Therefore, this is the mechanism by which the periodontal microbe A. actinomycetemcomitans is implicated in the activation of PADs and autoimmunity triggering in RA patients [74]. In addition, serological investigations of RA patients with a previous history of periodontitis infection by A. actinomycetemcomitans showed the presence of anti-LtxA antibodies in association with positivity for rheumatoid factor (RF) and anticitrullinated peptide antibodies (ACPA) [83]. In another context, the hypercitrullination of neutrophils histones is very important in NET formation and the NETosis process. NETosis is a physiological defense mechanism of neutrophils to kill bacteria through the formation of an extracellular chromatin network known as NET. Nuclear chromatin decondensation is mediated by the hypercitrullination of histones under activity of PADs, mainly PAD4 [25, 68, 84].

Redox conditions

Reduction–oxidation (redox) balance involves an interplay of oxidizing and reducing molecules that play an important role in many biological processes, including enzyme regulation [85]. Many researchers have suggested that the catalytic activity of PADs is regulated by the redox condition. The oxidizing species (oxidants) increase the oxidative status of tissue and inhibit the catalytic activity of PADs, whereas the reducing agents (reductants) decrease the oxidative state and enhance the catalytic activity of these enzymes. Therefore, redox balance is very important for the regulation of the activity of PAD enzymes.

Bicarbonate ${\text{HCO}}_3^{-}$

Bicarbonate is a major component of the body buffering system, and has a critical role in maintaining the physiological acid–base balance of the blood, where its serum concentration is about 17–29 mM [129, 130]. Bicarbonate is an alkaline substance that increases the pH of body systems, and is used for treatment and correction of acidosis conditions [131]. The impact of bicarbonate on the enzymatic activity of PADs and the efficiency of citrullination was assessed by studying the effect of bicarbonate on the histone and fibrinogen citrullination that was catalyzed by the recombinant PADs (PAD2/PAD4). In neutrophils, where the citrullination of histone H3 is important for the chromatin decondensation and NET formation, bicarbonate was found to increase the activity of PAD independently of pH [132]. It was pointed out that very different conditions are used for the analysis of PAD activity and for the assessing inhibitors of this activity [132]. In fact, citrullination assays can be conducted either in simple buffers (e.g., Tris–HCl pH 7–8 or HEPES) or in tissue culture media (e.g., Hank’s balanced salt solution (HBSS), RPMI medium, or Dulbecco’s modified eagle medium (DMEM)) containing very different bicarbonate concentrations. Since PAD2 and PAD4 possess the optimal activity for histone H3 citrullination at acidic pH (6.4–7.2) and since the presence of bicarbonate gives rise to the alkaline condition, the effect of bicarbonate on citrullination catalyzed by these PADs was evaluated in vitro using DPBS/HEPES (adjusted to pH 7.2) with 0.9 mM calcium [132]. Under these conditions, the addition of low amounts of bicarbonate enhances citrullination with both recombinant PAD4 and PAD2 [132]. These studies also showed that the PAD4 enzymatic activity was found to be more markedly affected by bicarbonate than that of PAD2. These findings proved that bicarbonate increases the activity of PADs independently of pH [132].

Oligomeric state of different human PADs

All five PAD human enzymes are highly conserved and exhibit high sequence identity [6]. As is shown in the Supplementary Materials, the average identity of their amino acid sequences is 50.8 ± 5.3%, ranging from 44.56% between PAD2 and PAD6 to 58.02% between PAD1 and PAD3. Interestingly, the high identity levels are not evenly distributed within the sequences of these proteins, with their C-terminal regions being much more conserved than the N-terminal regions of human PAD isozymes (see Supplementary Materials).

The structural analysis of PADs provides novel data that help in explaining the functional mechanisms of these hydrolase enzymes [133]. Currently, crystal structures are known for human PAD1 (PDB ID: 5HP5) [134], PAD2 (PDB ID: 4N28) [135], PAD3 (PDB ID: 6CE1), and PAD4 (PDB ID: 3APN) [4] (see Fig. 2). Although in their crystal structures PAD2, PAD3, and PAD4 appear as homodimeric proteins with an elongated rubber boot structure, in which the monomers contact each other to form a head-to-tail dimer, the analysis of the PAD1 structure revealed that this protein exists as a monomer (see Figs. 2, 3). This conclusion was further validated by the small angle X-ray scattering analysis (SAXS) of the solution structures of human PAD1 and PAD3, which supported the monomeric and dimeric states of these proteins, respectively [134]. The crystal structure of monomeric PAD1 differs from the crystal structures of PAD2, PAD3, and PAD4 by the presence of an elongated N-terminal tail that appears to prevent the formation of the homodimer [134]. Furthermore, instead of the arginine residue at position 8, which is conserved in PAD2, PAD3, and PAD4 and which was shown to be one of the key residues stabilizing the homodimeric structure of PAD4, since its substitution to glutamate (or glutamine) caused dissociation (or destabilization) of a homodimer [136], PAD1 contains a glutamine residue [134]. Since the N-terminal region of PAD6 has extra eight residues in comparison with other human PADs, and since its residue at position 16 that corresponds to position 8 of other PADs is not an arginine (see Supplementary Materials), it is likely that this PAD also exists as a monomer.

It was also pointed out that the monomeric molecule of PAD1 (and especially its N terminus) was noticeably more flexible than PAD2, PAD3, and PAD4, and that this structural flexibility can explain the poor X-ray diffraction of PAD1 crystals [134]. Since PAD enzymes that undergo dimerization are characterized by the formation of head-to-tail homodimer module, which involve the contact between two monomer molecules [136], it is likely that the creation of such homodimers causes structural stabilization of the N-terminal regions of PAD2, PAD3, and PAD4. The homodimerization of PAD isozymes produces a unique dimer interface stabilized by hydrophobic interactions and salt bridges between the adjacent monomers [86, 136]. In human PAD4 (see Fig. 4), among the residues Arg8, Tyr237, Asp273, Glu281, Tyr435, Arg544, and Asp547 located at the dimer interface surface, ionic interaction between Arg8 and Asp547 and the hydrophobic interactions driven by Tyr435 are crucial for dimerization of this protein [136]. In fact, it was pointed out that the PAD4 dimer is stabilized by the ion-pairing of Arg8 of subunit A with Asp547 of subunit B, as well as the ion-pairing of Asp273 of subunit A with Arg544 of subunit B, and formation of a hydrogen-bonding network between the Tyr237 and Glu281 of subunit A with Tyr435 of subunit B [136]. The homodimerization of PAD isozymes is important for the increase in their calcium binding capability and achieving a full enzyme activity. In fact, the disarrangement of the dimeric form of human PAD4 was shown to decrease the catalytic activity of this enzyme up to 75% [136]. On the other hand, the monomeric structure of the PAD1 isotype, characterized by increased flexibility, can explain the greater variation of its substrate proteins and broader distribution of PAD1 in different tissues [134]. It seems that homodimerization of PADs forming homodimers is independent of calcium binding. At least for human PAD2, it was shown that a stable head-to-tail dimer is formed by both apo- and holo-forms of this protein [135]. On the other hand, the analysis of human PAD4 revealed that homodimerization of this protein is required for its regulatory mechanism since the enzymatic activity of this protein and the cooperative binding of calcium ions were both affected by the disruption of the dimer interface [86, 136], suggesting the presence of some allosteric communication between the dimerization interface and the calcium-binding regions/residues.

Structural and functional modules of PADs

Despite the aforementioned differences in the quaternary organization, all the known structures of PAD isozymes are rather similar at the protomer level. In fact, these proteins are characterized by the elongated structure containing three domains: two N-terminal immunoglobulin (IgG)-like domains IgG1 and IgG2, and a C-terminal catalytic domain (Figs. 3, 4). The N-terminal domain in PAD1 spans amino acid residues 1–300 (IgG1 (residues 1–115) and IgG2 (residues 116–300)), being composed of 19 β-strands and 4 α-helices that contain three positions designated for the binding of calcium ions (Ca1, Ca2 and Ca3) [4]. It is worth noting that PAD4 is the only isotype of PADs that possesses a unique nuclear localization signal (NLS) located within its N-terminal half, close to the IgG1 domain (Fig. 4) [4, 69]. The presence of NLS is a specific feature of PAD4 and is accountable for its expression in the nucleus, where it facilitates the citrullination of nuclear proteins [69]. In fact, all PAD isotypes have been found to reside in the cell cytoplasm except PAD4 and PAD2, which can exist in both cytoplasm and nucleus [6, 69]. The N-terminal regions of PADs are characterized by the highest sequence (see above) and structure (see below) variability. On other hand, the C-terminal (catalytic) domains are highly conserved and are similar in all PAD isozymes (1–4) except PAD6, which lacks the catalytic activity [4]. This domain consists of five βββαβ modules, named as α/β propeller motif [137]. In all human PADs, the C-terminal domain is directly linked to the IgG2 domain on the same monomer. In the dimeric PADs (PAD2, PAD3, and PAD4), this catalytic domain forms a contact between IgG1 and the domain of the other monomer during the dimerization process to form a head-to-tail homodimer. Functionally, this domain contains two sites for calcium ions binding (Ca1 and Ca2), which is important for the formation of active site cleft and subsequent deamination of the substrate arginine [4].

As was already pointed out, compared to the C-terminal domains, which are similar and highly conserved in all PAD isotypes, the structure of N-terminal domains of these proteins displays significant differences [134]. Comparison of the crystal structure of human PAD1 (PDB ID: 5HP5) with the crystal structures of the Ca²⁺-bound forms of human PAD2 (PDB ID: 4N2B) and PAD4 (PDB ID: 1WD9) revealed that the RMSDs of their Cα atoms were 3.4 Å and 2.1 Å, respectively [134]. It was also pointed out that the structures of the C-terminal domains of these three PADs were better superimposed than the overall structures, showing RMSDs of 1.4 Å and 1.3 Å for PAD1–PAD2 and PAD1–PAD4 structural alignments [134]. This observation indicated that the C-terminal domains of the PAD1, PAD2, and PAD4 isozymes were structurally better conserved than their IgG1 and IgG2 domains [134]. In line with this conclusion, Fig. 5 represents the results of the multiple structural alignment of four human PAD isozymes, PAD1 (PDB ID: 5HP5), PAD2 (4N28), PAD3 (PDB ID: 6CE1), and PAD4 (PDB ID: 3APN) conducted by the MultiProt algorithm [138]. Here, the multiple structural alignment over the 421 C-terminal residues of all four structures gave an RMSD of 1.27 Å, indicating remarkable structural similarity of the catalytic and IgG2 domains of human PADs. Figure 5 also shows that the IgG2 domains of human PADs possessed the largest structural variability. It was suggested that N-terminal domains may have important roles in structural stabilization, substrate specificity, and the formation of protein–protein interactions [4].

Calcium dependency of PAD function and peculiarities of calcium binding

Peptidylarginine deiminase enzymes are cysteine hydrolases that catalyze the conversion of arginine into citrulline in the process known as citrullination or deamination. The catalytic function of PADs is triggered by a calcium ion, indicating that these proteins are active only under certain conditions characterized by the presence of high calcium levels. The calcium binding to PAD enzyme results in the formation of the active site cleft at C-terminal domain and induces conformational changes in the N-terminal subdomains [133]. Each PAD isotype has its specific binding sites for calcium, which vary in the number from PAD to another. Furthermore, PADs vary in the concentration of calcium required for their half maximal activity, K_0.5, which, at pH 7.6, ranges from 140 ± 90 μM in PAD1 [11] to 160 ± 20 μM in PAD2 [135], to 550 ± 80 μM in PAD3 [11], and to 280 ± 50 μM in PAD4 [139]. Although intracellular concentrations of calcium typically increase from ∼200 nM in resting cells to ∼1 µM in activated cells [140], these calcium levels are still noticeably lower than the K_0.5 values for human PADs, suggesting the existence of alternative mechanisms of PAD activation.

The earliest experiments focused on the calcium dependency of PADs and calcium binding sites suggested that at least three calcium ions can bind to rabbit PAD2 [33]. Later, advanced studies identified the number and locations of calcium-binding sites for all PADs, except PAD6 isotype that does not contain the catalytic site dependent on calcium binding [141]. Calcium binding induces the formation of an active site pocket (substrate binding site) in the (C-terminal) catalytic domain of PAD1, PAD2, PAD3, and PAD4. This catalytic domain contains the binding sites for Ca1 and Ca2. Amino acid residues for Ca1 and Ca2 binding are highly conserved and, therefore, the active site is similar in all activated PADs and appears to involve four residues: Cys645 (Cys647 in PAD2), Asp350, Asp473, and His471 [11, 142, 143].

PAD1 has been identified to contain four calcium-binding sites (Ca1, Ca2, Ca3 and Ca4) [134], whereas PAD2 isotypes possess six binding sites for calcium ions (Ca1–Ca6). Physical and chemical studies of PAD2 revealed that the binding site for Ca6 is a unique feature of this protein, and this site possesses the highest affinity for calcium among the other sites in PAD2. The amino acid residues (Asp122 and Asp123) accountable for the binding of Ca6 were not detected in PAD1 and PAD4, and appear to exist in PAD2 only. Furthermore, the Ca3, Ca4, and Ca5 sites were found to control the binding of the calcium ion at the Ca2-binding site, which is critical for the PAD enzymatic activity, since calcium binding to Ca2 stimulates the formation of the active site and promotes the interactions needed for the catalytic activity of the enzyme. Therefore, it is speculated that the binding of calcium to the binding sites of PAD2 occurs in a specific order, first binding at the Ca6 site followed by saturation of Ca1, then Ca3, Ca4, and Ca5 sites, and finally binding to the Ca2 site [134, 135]. Furthermore, crystallographic analysis conducted in the presence of different calcium concentrations revealed the presence of several structurally different calcium-bound configurations of human PAD2, and suggested the existence of a “calcium switch” that controls PAD2 activity [135]. This “calcium switch” is controlled by the binding of calcium to the eight acidic residues that comprise the Ca3–5 sites leading to the suppression of the electrostatic repulsions. Importantly, in its apo-form, this PAD2 region comprising the Ca3–5 sites is disordered (constitutes a region of missing electron density), but undergoes calcium-induced disorder-to-order transition. This promotes calcium binding at the Ca2 site by modulating the conformation of a loop (residues 369–389), as well as the subsequent movement of R347 out of the substrate-binding pocket, and C647 into the active site to generate the catalytically competent state [135].

Structural analysis of PAD4 and PAD3 isozymes identified the presence of five calcium binding sites (Ca1, Ca2, Ca3, Ca4, and Ca5). Considering the PAD4 isotype, Ca3, Ca4, and Ca5 sites are located within the IgG2 subdomain of the N-terminal domain, whereas Ca1 and Ca2 exist in the bottom of the C-terminal domain. The significant role of Ca3, Ca4, and Ca5 is the positioning of Ca2 in the proper site, while Ca2 has a critical role in the positioning and alignment of residues required for interaction of the substrates and inhibitors with the active cleft. The functional importance of Ca1 is in placing the catalytic residue Asp350 in the proper position [4, 144].

Specific structural features of the active site cleft

The formation of the active site involves the movement and rearrangement of the four catalytic residues into the catalytic site after calcium binding. This process is mediated by conformational changes around the substrate-binding site (active site), converting it from an open biconcave to a closed and very rigid cleft site [4, 136]. Four residues, Cys645 (Cys647 in PAD2), Asp350, Asp473, and His471, are the main catalytic residues in the active site that are required for the catalytic mechanism of the enzyme to convert arginine substrate into citrulline (Fig. 6) [4, 136].

Two residues, Asp350 and Asp473, are important for the orientation and positioning of the guanidine group of the substrate (arginine) into the active site. Each of these aspartate residues form hydrogen bonds with the nitrogen atoms at the side chain of the guanidine substrate. Cys645 and His473 residues exist on the opposite positions in the active site. Cys645 initiates the catalytic mechanism by the nucleophilic attack of thiolate (–S^–) on the nitrogen of guanidine substrate. This nucleophilic attack results in the formation of the tetrahedral intermediate. His471 plays an important role in stabilizing the tetrahedral intermediate through proton donation to release the ammonia. In addition, His471 acts by the activation of the water molecule involved in the hydrolysis and formation of the second tetrahedral intermediate that collapses for the production of citrulline [11, 142, 143].

Brief history of the development of PAD inhibitors

Brahn and colleagues conducted the earliest investigation on the inactivation of citrullination without exact knowledge of the molecular mechanisms of PAD activity [166]. These researchers have investigated the effect of Taxol (Paclitaxel), a microtubules stabilizer, in collagen-induced arthritis (CIA). Taxol is an anti-cancer drug used as a chemotherapeutic agent to stop tumor cell division by preventing the microtubules de-polymerization and increasing the microtubules’ stability [167]. The results of Brahn et al. showed that Taxol, in addition to its role in preventing cell mitosis, can suppress and reduce the severity of CIA in an animal model of arthritis. The usefulness of this drug for CIA intervention was demonstrated by the significant decrease in the levels of characteristic antibodies [166]. However, no evidence was given to show the correlation between Taxol treatment and inhibition of PAD activity.

To address this issue, an important study was conducted in 1998 by Moscarello and colleagues to investigate the inhibitory effects of Taxol on the activity of PAD enzyme in brain tissues [168]. This PAD was later identified as PAD2 responsible for the citrullination of neural proteins, such as myelin basic protein (MBP) causing demyelinating disease, i.e., multiple sclerosis (MS) [15]. Treatment of PAD enzyme with Taxol in bovine brain resulted in reduced demyelination. Furthermore, the observed remyelination was correlated with the decreased PAD activity. Unfortunately, demyelination was eventually observed, indicating that the inhibition reaction was reversible, and the potency of this drug was very weak [168, 169]. Therefore, further investigations were required to find effective and irreversible PAD inhibitors with high potency. The development of such new irreversible PAD inhibitors was inspired by the arginine-like substrate, N-benzoyl-l-arginine amide (BAA), that has been modified for higher efficiency and compatibility with the PAD active site (Fig. 10) [142, 170]. These efforts employed BAA as a scaffold for the production of bioavailable, irreversible, and highly potent PAD inhibitors [171, 172]. N-benzoyl-l-arginine amide is a small synthetic molecule mimicking arginine residue, which was originally used in experiments as a substrate for PAD enzymes. Later, this compound was used as a mechanistic probe for the PAD activation and a basis for the generation of PAD inhibitors [86, 142]. All mammalian PAD isozymes have a high structural and sequence similarity [6]. Human PADs possess similar active sites that contain four highly conserved residues (Cys645, His471, Asp350, and Asp473) required for the PAD catalytic activity and conversion of arginine into citrulline [4], with an exception of PAD2 that has Cys647 instead of Cys645 [173]. This structural similarity of PADs and the identity and rigidity of their active sites define the structure-based approach broadly utilized in a search for the PAD inhibitors.

F- and Cl-amidines are the first potent and irreversible PAD inhibitors

The reversibility of inhibition and the lack of potency of the earliest inhibitors defined the efforts to generate new irreversible and bioavailable PAD inhibitors with high potency and selectivity. In arginine amino acid, the guanidine group basic side chain contains three amino groups that make positive charge and represent the target site for activated (calcium bound) PADs [174]. N-benzoyl-l-arginine amide is the mechanistic substrate probe that is structurally analogous to arginine and contains a similar guanidine group. The substitution of one amino group with halides (F⁻ or Cl⁻) was the main step in the formation of (halo)acetamidine-containing compounds, 2-fluro-acetamidine (F-amidine) and 2-chloro-acetamidine (Cl-amidine) inhibitors (Fig. 11). The (halo)acetamidine warhead of these inhibitors is similar in size to the guanidine of substrate [170]. The F- and Cl-amidine keep the charge (positivity) at neutral pH and preserve most of the hydrogen binding capability (three H-bonds) between guanidine and the catalytic sites (Asp350 and Asp473) of activated PADs [171]. The (halo)acetamidine warhead of F- and Cl-amidine reacts and modifies the cysteine residue of the active site of PAD4 and other isotypes to form a stable thiol adduct that leads to the inhibition of PAD catalytic activity [175]. Studies showed that both inhibitors are bioavailable and react with PAD4 and other PADs only in the activated calcium-bound state, and have similar potency for the irreversible inactivation of these enzymes by the modification and alkylation of the Cys645, one of the four catalytic residues of PAD active site [170, 171, 175]. Structurally, both inhibitors are identical except for the substitution of fluorine with chlorine, and they have a side chain of three methylene units [171, 176].

The efficiency of these inhibitors is attributed to the correct positioning of their warhead for the nucleophile attack on the PAD–cysteine thiolate. This proper position of the warhead is defined by the side chain length. Although most of the high potency PAD inhibitors have a side chain length between two and four methylene units, the optimal potency was observed in compounds containing side chain of three units, such as F- and Cl-amidines, whereas the increase or decrease in the side chain length reduced the potency of PAD inhibitors [171]. Interestingly, Cl-amidine has a higher potency than F-amidine by fourfolds. This was related to the withdrawing nature and identity of the leaving groups, i.e., fluorine and chlorine [170, 171]. Concerning selectivity and specificity, experimental studies suggested that the Cl-amidine is a pan PAD inhibitor that can react with most of the PAD isotypes, whereas F- amidine is a more selective inhibitor that acts on PAD1 and PAD4 only [11]. Cl-amidine is a highly effective and safe drug in reducing the severity and treatment of several cancers and autoimmune diseases, such as RA (where Cl-amidine reduces the severity of inflammation by 55% [86, 177]), ulcerative colitis (UC), and nerve damage. Because of all these observations, Cl-amidine continues to serve as the most widely used inhibitor and acts as the basis for the production of new inhibitors [175]. In addition, Cl-amidine showed a high potential in treating many cancers, possessing a cytotoxic effect against tumor cells and inhibiting cell differentiation, e.g. in breast cancer, where it inhibits the PAD2 enzyme [178].

Development of new specific PAD inhibitors

Because of the variability of PAD isozymes and their related disorders, there is a steady need in the specific inhibitors with high potency, bioavailability, and increased selectivity. In this endeavor, F- and Cl-amidines serve as the benchmark for generating new PAD inhibitors based on the utilization of the structure–activity relationships (SAR). The structural components of these first generation of PAD inhibitors involved three parts: warhead, side chain, and backbone (Fig. 11) [144]. The studies on the substrate specificity and the SAR identification for the different PAD isotypes (PAD1-4) represent the foundation for the efforts to develop the second generation of inhibitors with improved potency, bioavailability, and selectivity [11].

The success of the administration of F- and Cl-amidines inspired researchers to develop derivatives from these inhibitors, such as N-α-2carboxyl-benzoyl-N5-2-fluoro-1-iminoethyl-l-ornithine amide (o-F-amidine) and N-α-2-carboxyl-benzoyl-N5-2-chloro-1-iminoethyl-l-ornithine amide (o-Cl-amidine). Comparing to the parent inhibitors (F- and Cl-amidine), these new PAD inhibitors have a higher potency (up to 65-fold) for all PAD types (PAD1-4) and also show enhanced selectivity (up to 25-fold) [176]. These improved inhibitors have been produced by the addition of carboxylic ortho acid to the benzyl group at the backbone (Fig. 12) [176]. Detailed analysis revealed that the incorporation of this group increased the potency of o-Cl-amidine threefold and fourfold for PAD1 and PAD2, respectively. In contrast, the potency of o-F-amidine inhibitor was increased 65-, 20-, 39- and 11-fold for PAD 1, 2, 3 and 4, respectively. In context of selectivity, o-Cl-amidine possessed a decreased selectivity for PADs 2 and 3 and increased selectivity for PADs 1 and 4. On the other hand, the selectivity of o-F-amidine was markedly increased for all PADs, mainly for PADs 1, 2, and 3. These data suggested that o-F-amidine can be considered as a primary PAD1-selective inhibitor, whereas o-Cl-amidine is a PAD1- and PAD4-selective inhibitor [176]. In line with the improved potency and selectivity of these inhibitors, several studies reported the great success in using o-F-amidine and o-Cl-amidine in the treatment of certain cancers and inflammatory autoimmune diseases. In fact, due to the selectivity of o-Cl-amidine for PAD4, it represents a potential therapy for inflammatory RA and many cancers, such as breast, lung, and colon cancers. Interestingly, the o-Cl-amidine inhibitor inactivates PAD4 and reduces its effect as a suppressor of the p53 gene, thereby upregulating the expression of p53, a protein that plays a crucial role in preventing tumorigenesis [3, 179].

Advanced studies were conducted to find more potent and selective PAD inhibitors. Here, the screening of a library of 264 haloacetamidine containing tripeptides resulted in the identification of threonine–aspartate F-amidine (TFDA) (see Fig. 12c), as the most potent, selective, and irreversible PAD4 inhibitor. Contrary to previous generations of PAD inhibitors, TFDA interacts with the side chains of Q346, R374, and R639 of PAD4. As a result, this drug possesses a great selectivity for PAD4, which is 52-fold higher than that for PAD2, and 15-fold and 65-fold more than that for PADs 1 and 3, respectively [144, 165]. The specificity of TFDA interaction with PAD4 is determined by the fact that this drug binds to R374 and R639. In fact, R639 is unique to PAD4, since in PAD1 and PAD3, the corresponding site is taken by leucine, whereas in PAD2 and PAD6, it is occupied by phenylalanine and glutamate, respectively. Similarly, although in PAD1 and PAD4, R374 is conserved, the corresponding position in PAD2 and PAD3 is occupied by a glycine and by an alanine in PAD6 [165].

Since calcium binding causes a dramatic increase in PAD activity (e.g., the holo-form of PAD4 shows 10,000-fold higher activity than the apo-form of this protein [11]) resulting from substantial calcium-induced structural reordering, the representatives of the early generations of PAD inhibitors were exclusively designed to act on holo-PADs. A different approach was utilized by Lewis et al. [180], who used GSK’s DNA-encoded small-molecule libraries [181] to find PAD4 inhibitors in the absence or presence of added calcium [180]. This search generated a promising lead, GSK121, the optimization of which generated GSK199 and GSK484 molecules (see Fig. 13a, b), which showed, in the absence of calcium, the IC₅₀ values of 200 nM and 50 nM, respectively [180]. Furthermore, contrary to the irreversible binding of the halo-acetamidine-based inhibitors exclusively to the holo-PADs, the drugs of new generation, GSK199 and GSK484, bind to PAD4 reversibly and inhibited PAD4 both in the absence or the presence of 2 mM calcium (although GSK199 and GSK484 potencies to holo-PAD4 were notably lower (1 µM and 250 nM, respectively) in comparison with those to apo-form of the protein) [180]. To analyze the effect of GSK199 binding on the PAD4 structure, apo-PAD4 crystals were pre-soaked in a solution containing 100 mM imidazole, pH 7.8, 2 mM TCEP, 50 mM CaCl₂ and then transferred to the same solution with an additional 25 mM ligand. An attempt to soak compounds in apo-PAD4 crystals in the absence of CaCl₂ failed, causing crystal degradation [180]. Structural analysis of the resulting crystals revealed that GSK199 partially overlapped with a substrate and only a part of the five calcium sites of this protein was occupied. Furthermore, the formation of the PAD4–GSK199 complex was accompanied by a noticeable change in PAD4 structure. In fact, although in the absence of GSK199, the residues 633–645 of the apo-PAD4 were disordered, GSK199 binding resulted in the ordering of this region leading to the formation of β-hairpin structure. This new β-hairpin structure served as a lid allowing the hydrophobic residues Phe634 and Val643 to pack over the central part of the inhibitor [180]. Therefore, these new inhibitors bind to a novel conformation of the PAD4 active site, where part of the site is re-ordered to form a β-hairpin acting as a lid closing the bound drug [180]. However, similar to all other known PAD inhibitors, these representatives of GSK’s DNA-encoded small-molecule library bind directly to the enzyme active site.

A recent study by Lewallen et al. [182] represents the first attempt to find a PAD inhibitor capable of preferential binding the apo-enzyme. Here, a fluorescence polarization-activity-based protein profiling-based high throughput screening assay (FluoPol-ABPP HTS assay) [183] was used to screen the 1280-compound LOPAC library (Sigma-Aldrich Library Of Pharmacologically Active Compounds) to find potential inhibitors that bind apo-PAD2, holo-PAD2, or both [182]. This screen revealed the presence of several PAD2 inhibitors that can bind the apoenzyme, with the ruthenium red (see Fig. 13c) serving as calcium-sensitive inhibitor that interacts specifically with apo-PAD2, but loses its inhibitory potency in the presence of high calcium concentrations [182].

Inactivation mechanisms of PAD inhibitors

Most of the PAD inhibitors react and inhibit PADs after these enzymes have been activated by calcium binding, i.e., they target the active sites of the holo-PADs. Assuming that the active sites of all the catalytically active PAD isotypes are likely to be fairly identical, one can utilize the PAD4 enzyme as an ideal model for understanding the mechanism by which the inhibitor compounds would inhibit the activated calcium-bound enzymes [4]. Biochemical and structural studies have identified the components of the active site of PAD4 and the catalytic mechanism of the citrullination process, where the enzyme attacks the guanidine group of arginine residue to convert it into citrulline. The active site pocket of the activated PAD contains four major amino acid residues required for its catalytic activity: Cys645, His471, Asp350, and Asp473. Cys645 acts as the key residue for initiating the catalytic mechanism, which is mediated by the nucleophilic attack of cysteine thiolate on guanidine group of the arginine [4, 11]. Based on these data, two significant mechanisms were proposed to illustrate how PAD inhibitors (F- and Cl-amidine) can inactivate PAD4 enzyme (Fig. 14). In the first mechanism, the one-step of the inactivation process involves the direct displacement of the active site residue (Cys654) of PAD with the halide (F^– or Cl^–) of the warhead of inhibitor compound through process known as $S_{N}2$ mechanism. This replacement inhibits the formation of the covalent tetrahedral intermediate, which is required for the hydrolytic activity of PAD to convert the guanidine group of arginine into the ureido group of citrulline. The second mechanism of inactivation involves many steps, which start with the nucleophilic attack of cysteine thiolate on the amidinum carbon of the inhibitor warhead, resulting in the formation of a tetrahedral intermediate mimicking the result formed during a catalytic (hydrolysis) process. Then the thiolate undergoes replacement with the halide to form a three-membered sulfonium ring, which finally collapses to produce a stable thioether ring. Therefore, the inhibitor can tightly grasp the Cys645 and inhibit enzyme hydrolysis function [144, 171].