Gingipains from Porphyromonas gingivalis – Complex domain structures confer diverse functions

Abstract

Gingipains, a group of arginine or lysine specific cysteine proteinases (also known as RgpA, RgpB and Kgp), have been recognized as major virulence factors in Porphyromonas gingivalis. This bacterium is one of a handful of pathogens that cause chronic periodontitis. Gingipains are involved in adherence to and colonization of epithelial cells, haemagglutination and haemolysis of erythrocytes, disruption and manipulation of the inflammatory response, and the degradation of host proteins and tissues. RgpA and Kgp are multi-domain proteins composed of catalytic domains and haemagglutinin/adhesin (HA) regions. The structure of the HA regions have previously been defined by a gingipain domain structure hypothesis which is a set of putative domain boundaries derived from the sequences of fragments of these proteins extracted from the cell surface. However, multiple sequence alignments and hidden Markov models predict an alternative domain architecture for the HA regions of gingipains. In this alternate model, two or three repeats of the so-called “cleaved adhesin” domains (and one other undefined domain in some strains) are the modules which constitute the substructure of the HA regions. Recombinant forms of these putative cleaved adhesin domains are indeed stable folded protein modules and recently determined crystal structures support the hypothesis of a modular organisation of the HA region. Based on the observed K2 and K3 structures as well as multiple sequence alignments, it is proposed that all the cleaved adhesin domains in gingipains will share the same β-sandwich jelly roll fold. The new domain model of the structure for gingipains and the haemagglutinin (HagA) proteins of P. gingivalis will guide future functional studies of these virulence factors.

Introduction

Periodontal disease

Periodontal diseases are caused by bacterial infections in the periodontium which is composed of gingiva, periodontal ligament, cementum and alveolar bone, all of which support and anchor the teeth [1]. Among these diseases, gingivitis and chronic periodontitis are the two most common forms of periodontal diseases. Gingivitis is non-destructive to the tooth tissues and most people can recover from it through effective oral hygiene which removes plaque. However, gingivitis can further progress to destructive periodontitis in some patients perhaps in part due to their unique host responses to the pathogenic organisms in the plaque and other factors [2]. Chronic periodontitis is the most common form of periodontitis [3]. It has been defined as “an infectious disease resulting in inflammation within the supporting tissue of the teeth, progressive de-attachment and bone loss” [4]. The inflammation of the gums has been considered to be a result of a direct host immune response to the increasing amount of bacterial pathogens in the dental plaque formed on the gingival margin [2, 5].

Recently, public health concerns have been raised as more and more evidence accumulates to demonstrate that periodontal diseases are also associated with cardiovascular diseases [6], diabetes mellitus [7–8], rheumatoid arthritis [9–10] and Alzheimer disease [11–13].

The generally accepted specific plaque hypothesis considers that only a selected few species among several hundred oral microorganisms are actively pathogenic and they are responsible for the inflammatory events and the destructive processes [14–16]. These studies have shown that only 10–20 bacterial pathogens in the dental plaque, the so-called periodontopathic microbiota, are responsible for destructive periodontitis, including Porphyromonas gingivalis, Bacteroids forsythus, Treponema denticola, Prevotella intermedia, Prevotella nigrescens, Fusobacterium nucleatum, Peptostreptococcus micros, Campylobacter rectus, Treponema pectinovorum, Treponema vincentii, Eikenella corrodens and Actinobacillus actinomycetemcomitans [15, 17–19]. Among these pathogens, Porphyromonas gingivalis has been considered to be a major contributor to the development of chronic periodontitis [20–22] and has been widely studied.

An unusual oral pathogen, Porphyromonas gingivalis

P. gingivalis is commonly detected in chronically inflamed periodontal lesions and the proportion of this bacterium in the anaerobically cultivable flora in subgingiva can be as high as 50% [23]. P. gingivalis is a non-motile Gram-negative anaerobic bacillus [18]. In contrast to other members of the genus Porphyromonas, many of which are able to grow on complex carbohydrates, P. gingivalis is asaccharolytic and reliant upon nitrogenous substrates such as proteins or peptides as nutrients and for metabolic energy [24–26]. P. gingivalis is able to produce a large amount of proteinases to degrade proteins from host or other microorganisms in order to meet its special nutritional requirements [5, 18]. Importantly, this bacterium needs exogenous haem for growth due to the lack of a haem biosynthesis pathway [27–28]. P. gingivalis is mostly found in bleeding chronic periodontal lesions, where haemoglobin from ruptured erythrocytes provides a very convenient and abundant haem source. When growing on blood agar plates P. gingivalis colonies are initially white to creamy in colour, but turn dark red to black after 6–10 days [26]. The black pigmentation has been verified as an accumulation of iron (III) protoporphyrin IX in the form of the µ-oxo dimer [Fe(III)PPIX]2O on the bacterial cell surface [29]. Among the common laboratory strains and clinical isolates of P. gingivalis, strains W83, W12 and W50 are found to be more virulent than strains 381, HG66 and ATCC33277 [30].

The major virulence factors of P. gingivalis include fimbriae, capsule, outer membrane vesicles, lipopolysaccharide (LPS), toxic metabolites and proteinases [18, 24–25] P. gingivalis expresses a group of endopeptidases called gingipains on the outer membranes which are responsible for at least 85% of the proteolytic activity and 99% of the “trypsin-like” activity produced by the bacterium [25]. The primary aim of the expression of these proteinases by this pathogen is to digest proteins for nutrition, but gingipains are also found to be involved in the destruction of the host periodontal matrix and alveolar bone, host cell adhesion and invasion, and in dysregulation of the host immune response [5].

The gingipain proteins

Gingipains are cysteine proteinases that belong to the peptidase family C25 [31]. There are three types of gingipains in P. gingivalis: lysine-specific gingipain (Kgp), arginine-specific gingipain A ( RgpA), and arginine-specific gingipain B (RgpB) [32]. Gingipains are principally located on the outer membranes and outer membrane vesicles of all P. gingivalis strains except for HG66, which produces and secretes soluble forms of gingipains into the extracellular milieu [5, 25, 33–34]. Genomic studies show that gingipains are encoded by individual gene loci (kgpA, rgpA and rgpB), found in the genomes of all P. gingivalis strains [35–39]. The proteins encoded by kgp or rgpA genes consist of a signal peptide, an N-terminal pro-fragment, a Lys-specific or Arg-specific catalytic domain and a large C-terminal haemagglutinin/adhesin (HA) region (Fig. 1) [35, 40]. In contrast, the protein encoded by the rgpB gene consists of a signal peptide, an N-terminal pro-fragment, an Arg-specific catalytic domain and a small C-terminal fragment [41]. The C-terminal fragment of the RgpB protein shows a significant deletion in the HA region when compared with the C-terminal region in the RgpA protein [41]. In addition to the catalytic domains the differences between these proteins are found in the HA regions. It is thought that the adhesion properties of these varying regions confer the biological specificities of these proteins.

Fig. 1.

Schematic diagram of the gingipain domain structure hypothesis. The colour matches indicate that the sequences are highly similar. This figure was adapted from [38, 42, 44–45]

The gingipain domain structure hypothesis

Sequence analyses of purified gingipains

Studies of the possible domain organization of the structures of gingipains have been conducted by analysing the proteins purified from culture media and outer membranes using SDS-PAGE, peptide mass fingerprinting and N-terminal sequencing [42–44]. The proposed gingipain domain structure hypothesis which is supported by genetic analysis of cloned kgp and rgp from strain W50 as well as from the translated peptide sequences [38, 43] is presented in Figure 1. These peptides were also designated as HA1, HA2, HA3 and HA4 by DeCarlo et al. in their studies (Fig. 1) [45]. Further details on the putative domain boundaries have been provided by subsequent analysis of the outer membrane proteins from P. gingivalis W50 using the combined techniques of two dimensional gel electrophoresis, N-terminal sequencing and peptide mass fingerprinting [44]. It was proposed that the Kgp44 region might consist of three parts: Kgp14, Kgp13 and Kgp20 (Fig. 1) [44]. Moreover, it was observed that gingipains are present as noncovalently associated complexes because purified gingipains only fragmented into to smaller masses when they were boiled [42]. The presence of various detergents alone was insufficient to mediate this separation [46].

Sequence homologies of gingipains from different P. gingivalis strains

The corresponding translated amino acid sequences of RgpA or RgpB proteins found in different strains are more than 98% identical [47–48]. The catalytic domains of RgpA (RgpAcat) and RgpB (RgpBcat) proteins typically share ~90% sequence identity, but the pro-peptides are only ~75% identical in these two proteins (Fig. 2).

Fig. 2.

Schematic diagram summarizing the gingipain domain structure hypothesis for gingipains and HagA from the HG66, W50, 381, W83 and W83v strains of P. gingivalis. The proteins are divided into domains of sequence similarity bounded by proteolytic processing sites and coloured and labelled accordingly. Arrows indicate the proteolytic processing sites. The sequences in the regions between two vertical or slant dash lines are compared and the amino acid sequence identity in percentage is indicated. Similar sequences are indicated by the same colour. A part of the data was adapted from [25, 53]

In comparison with other gingipains, the catalytic domain of Kgp (Kgpcat) is only 20–30% identical to those of RgpAcat and RgpBcat. However, in the C-terminal regions of both RgpAcat and Kgpcat, there is a highly conserved sequence containing 31 residues. This conserved sequence contains the adhesin binding motif 1 (ABM1), which is also in part observed in the Rgp44/Kgp39 regions and the Rgp17/Kgp44 regions [38, 49].

Despite the differences in the catalytic domains, RgpA from strain HG66 and Kgp from strain HG66 have almost identical sequences in the HA regions with the exception of the N-terminal 146 residues in the Kgp39 and Rgp44 fragments, respectively. The Kgp proteins from strains W12, W50 and W83 are 99% identical in sequence, but differ from the Kgps in strains HG66 and 381 in the Kgp44 regions [5] (Fig. 2). The catalytic domain of Kgp W83v lacks the pro-peptide and part of the N-terminal sequence, while the remaining sequence is identical to that of KgpW83 except for a 154-residue fragment in the Kgp39 domain (Fig. 2). Strains HG66, 381, W83 and W83v are considered to be distinct lysine gingipain (kgp) biovars due to their different kgp gene structures [50].

The Kgp15 and Rgp15 regions in all of the gingipains are virtually identical (sharing 97% sequence identity) with only a few residue substitutions including G1186/D, S1204/T, S1208/A and N1288/D as numbered when found in Kgp W83. The last 72 residues in the C-terminus of RgpB are found to be partially conserved in the C-termini of RgpA and Kgp. This conserved sequence has been suggested to function as an anchor that attaches the gingipains to the cell outer membrane. The soluble form of RgpB does not contain this peptide and it is probably removed by proteolysis [31, 34]. This 72-residue peptide has also been suggested to play critical roles in maintaining the correct folding of the catalytic domain of RgpB and the posttranslational glycosylation of this protein [51].

The Kgp15 and Rgp15 regions in all of the gingipains are virtually identical (sharing 97% sequence identity) with only a few residue substitutions including G1186/D, S1204/T, S1208/A and N1288/D as numbered when found in Kgp W83. The last 72 residues in the C-terminus of RgpB are found to be partially conserved in the C-termini of RgpA and Kgp. This conserved sequence has been suggested to function as an anchor that attaches the gingipains to the cell outer membrane. The soluble form of RgpB does not contain this peptide and it is probably removed by proteolysis [31, 34]. This 72-residue peptide has also been suggested to play critical roles in maintaining the correct folding of the catalytic domain of RgpB and the posttranslational glycosylation of this protein [51].

The C-terminal HA sequences have also been found in the haemagglutinin, HagA (Fig. 2). Han et al. [52] have reported that the hagA gene of P. gingivalis strain 381 encodes four large contiguous direct peptide repeats: Y416-D868, Y869-D1324, Y1325-D1780 and Y1781-D2236. These four peptide repeats share 94–100% sequence identity and are 95% identical to the fragment Y534-D992 found in Kgp W83v. In addition, the region of Y2237-K2628 in the C-terminus of HagA is 97% identical to the region of Y1341-K1732 in the C-terminus of Kgp W83 (Fig. 2).

The structure of the catalytic domain of RgpB

The structure of only one catalytic domain in the gingipain protein family, the crystal structure of RgpB, has been reported. The RgpB protein was purified from the culture medium of P. gingivalis strain HG66 in a soluble form and the crystal structure represents an archetypical gingipain catalytic domain consisting of 435 residues [31].

The overall structure of RgpB was described as a “crooked one-root tooth” with a spherical “crown” as the N-terminal catalytic domain (351 residues) and the “root” being the C-terminal Ig-like domain (84 residues) (Fig. 3). The catalytic domain is composed of two sub-domains A and B with similar α/β open sheet topologies formed by a central β-sheet and two α-helices flanked on either sides. The structure of RgpB in complex with D-Phe-Phe-Arg-chloromethylketone (FFRCMK) inhibitor (PDB entry 1CVR) has also been reported [31]. The position of the FFRCMK inhibitor defines an active site formed by H211 and C244 in sub-domain B. The Cys-His catalytic dyad acts to bind and cleave the Arg-Xaa substrate. The residue Arg in FFR-CMK forms a covalent bond with C244 Sγ via its methylene group [31]. The side chain carboxylate group of residue D163 in RgpB forms a salt bridge with the guanidyl group of the Arg of the FFRCMK inhibitor. Several hydrogen bonds connect the inhibitor to residues G212, Q282 and W284 which are also close to the active site. A Zn2+ ion is observed close to the active site interacting with H211 and E152 but only in the RgpB-FFRCMK complex structure. Although E152 in RgpB is replaced by an Asp in Kgp, G212 and the catalytic residues H211 and C244 are conserved in the Kgp catalytic domain sequence [31]. This suggests that similar active sites might be shared by these two enzymes despite the fact that they share only about 22% overall sequence identity [5]. The C-terminal domain of RgpB has a β-barrel topology formed by seven β-strands which is similar to the members of the immunoglobulin superfamily (IgSF). IgSF domains are known to be involved in the recognition, binding and adhesion processes of cells [31].

Fig. 3.

Representation of the structure of RgpB in complex with D-Phe-Phe-Arg-chloromethylketone (FFRCMK) inhibitor (PDB entry 1CVR). The RgpB and FFRCMK [31] are represented as ribbon and sticks, respectively. The sub-domains A and B of the catalytic domain are coloured in blue and deep-teal, respectively, the C-terminal Ig-like domain is in red colour, and the FFRCMK is coloured orange. Six Ca²⁺ and two Zn²⁺ ions are shown as spheres and coloured gold and green, respectively. The side chains of residues E152, H211 and C244 are shown as red sticks

The closest structural homologues of RgpB are the caspases (Asp-specific cysteine proteases), in particular caspase-1 and caspase-3, with their catalytic domains being partially superimposable. Residues H211 and C244 found in the active site of RgpB are structurally conserved in the active sites of these caspases, indicating that these enzymes might be evolutionally related [31]. The catalytic domains of gingipains are chemically highly specific and cleave polypeptides at either lysine or arginine residues while the biologically relevant targets are selected in situ by the adhesion/binding properties of the associated HA regions.

Investigating the structure of the HA region of gingipains

Until recently attempts using recombinant proteins to understand the structure and function of the HA regions have focussed upon the gingipain domain structure hypothesis. For example, the recombinant Kgp15/Rgp15 protein construct has been studied extensively. Proposals have been made regarding direct haemoglobin and haem/haemin interactions being mediated by this polypeptide [29, 54–56]. However, structural studies of this protein construct have been unsuccessful, possibly because it tends to aggregate when it is concentrated above 1 mg/ml in solution (D. Langley, personal communication). Also, recombinant Rgp17and Rgp27 of RgpA from P. gingivalis ATC33277 have been purified in very low yields (30–40 µg purified protein per litre culture) and the recombinant product appears not to be suitable for structural studies [57].

Despite these studies of the binding activities of synthetic protein constructs, most being designed as fragments based upon the gingipain domain structure hypothesis, there is presently no clear understanding of the contribution made by the HA regions to gingipain function. The complexity of the HA structure and the consequent practical challenges of its protein biochemistry is a barrier that has until recently impeded this research effort.

The biological functions of gingipains

Gingipains in adhesion and haemagglutination

P. gingivalis is able to agglutinate erythrocytes and this is considered to be evidence that P. gingivalis adheres to host tissue [58–61]. Fimbriae and LPS have both been implicated as factors in the adherence of P. gingivalis to epithelia, erythrocytes and other bacteria [62–65]. These factors are complemented by the outer membrane expressed gingipains and HagA proteins which also play important roles in cell adhesion. It has been observed that both intact P. gingivalis and purified gingipains from cell membranes or growth culture are able to agglutinate erythrocytes and bind to haemoglobin [46]. The observation that two triple mutants, rgpA– rgpB– kgp– and rgpA– kgp– hagA–, lack haemagglutination and haemoglobin binding ability indicates the importance of the HA regions in mediating these activities [66].

Extensive studies have been carried out to identify the colonization and haemagglutination motifs within the gingipain amino acid sequences. Monoclonal antibody MAb 61BG1.3 raised against formalinized P. gingivalis W83 cells, was shown to be able to inhibit host re-colonization in immunized periodontal patients for 6-9 months [67–69]. Using serially truncated recombinant proteins and western blots, the “colonization epitope” recognized by MAb 61BG1.3 was mapped to residues G907-T931 of RgpA [69–70]. The peptide G907-T931 was recognised as a haemagglutination motif by using truncated recombinant proteins and another monoclonal antibody MAb 1A1, raised against formalinized P. gingivalis DCR 2015, inhibit the haemagglutination mediated by P. gingivalis W50 culture supernatant [68]. Later, a small peptide “PVQNLT”, which presents within the G907-T931 sequence was characterized as a haemagglutinin-associated short motif by using monoclonal antibody MAb-Pg-vc raised against outer membrane vesicles of P. gingivalis strain 381 [71]. This short motif is present as “PVQNLT” or “PVKNLK” at the C-terminus end of RgpAcat and Kgpcat domains, in the Rgp44 and Kgp39 regions and in the Rgp17 and Kgp44 regions. Another synthetic peptide containing residues G1083-T1102 of RgpA strongly inhibited the haemagglutination mediated by intact W50 cells with a minimum inhibitory concentration of <1.5 µM [70]. Moreover, synthetic peptides spanning residues P785-K840, A934-N973, G974-F1042 and R1093-V1112 of RgpA are all recognized by antibodies within human sera taken from patients with periodontitis [70].

Recently, Sakai et al. found that a recombinant full length Rgp44720-1138 protein (referred to as rHGP44A) did not display haemagglutination activity [59] even though it contained a previously identified haemagglutinating determinant (residues G1083-T1102) [70]. Surprisingly, a second truncated form, Rgp44720-1081 (referred to as rHGP44B), did, however, display haemagglutination activity. A synthetic peptide corresponding to Rgp residues G1083-T1102 bound to Rgp44720-1081 in a dose-dependent manner and subsequently inhibited the haemagglutination mediated by Rgp44720-1081 [59]. It was suggested that the peptide G1083-T1102 functioned as a haemagglutination suppressor rather than a haemagglutination determinant [59]. Additionally, an anti-Rgp44720-1138 antibody was found to block the haemagglutination mediated by P. gingivalis with a range of concentrations of 2.5–10 µg/well [59]. In contrast, anti-HbR, an antibody raised against the Rgp15/Kgp15 region, did not block P. gingivalis-mediated haemagglutination, suggesting that Rgp44 region played more important role in haemagglutination than the Rgp15/Kgp15 region [59].

In addition to ABM1 which was proposed to mediate the formation of RgpA-Kgp complexes [38], O’Brien-Simpson et al. have characterised another two adhesin binding motifs, ABM2 (S886-A914 in Kgp) and ABM3 (V939-F972 in RgpA), using competitive ELISA and antibodies raised against synthetic peptides [49]. The anti-ABM2 antibody inhibited the binding of RgpA-Kgp complexes to fibrinogen, fibronectin and collagen type V but did not inhibit the binding to haemoglobin. In contrast, the anti-ABM3 antibody only inhibited the binding of gingipain complexes to haemoglobin but not to the other proteins [49]. The epitopes in ABM2 and ABM3 recognized by these antibodies were further identified to be the short motifs “EGLTAT-TFEEDGVAA” and “GTPNPNPNPNPNPNPGT”, respectively [49]. The binding of gingipains to fibrinogen, fibronectin and laminin were also reported by Pike et al. [72], and the apparent disassociation constants (KD) of Kgp and RgpA binding to fibrinogen were estimated to be 4 nM and 8.5 nM, respectively [72]. The sequence “EGLTAT-TFEEDGVAA” presents three times in both the Kgp and RgpA sequences [49]. The “FEED” motif has been reported to be an essential sequence for proteins binding to fibronectin [73]. These colonization and haemagglutination determinants, and ABM1, ABM2 and ABM3 peptides are also found in HagA sequences [49, 53].

Gingipains in haemolysis

The haemagglutination mediated by P. gingivalis is a primary step for this pathogen in its adherence to erythrocytes. P. gingivalis is known to lyse erythrocytes and the key haemolysin has been identified as being associated with the outer membrane gingipains [74]. Both Rgp and Kgp have been reported to possess haemolytic activity [61, 75], but the receptors on the erythrocytes are still not known. The band 3 protein from the membranes of erythrocytes has been considered to be one possible target because the purified band 3 protein can bind to P. gingivalis cells and inhibits the haemagglutination mediated by a protein purified from the culture supernatant of strain W83 [76]. Sakai et al. reported that a recombinant protein Rgp44720-1081 was able to bind to the erythrocyte membrane protein glycophorin A with a KD of 30 µM [59].

Gingipains in haem acquisition

P. gingivalis has an obligate requirement for exogenous iron and protoporphyrin IX (PPIX) for its growth due to gene defects. Genes hemA, hemL, hemB and hemC encoding the enzymes that are involved in haem biosynthesis pathway cannot be identified in the genome of P. gingivalis [27]. While P. gingivalis can utilize haem from other sources such as serum albumin, transferrin, lactoferrin, haemopexin, catalase, lactoperoxidase and cytochrome [77–78], haemoglobin is the preferred haem and iron source [79].

P. gingivalis is known to be able to accumulate haem on the cell surface and the black pigmentation is a typical phenotype for this pathogen when grown on blood agar plates [80–81]. The kgp– mutants lack the black pigmentation, indicating that Kgp is involved in haem capture [66, 75]. Moreover, Lewis et al. [75] have confirmed that the kgp+ rgpA– mutant strain has higher hydrolysis efficiency of haemoglobin than the kgp– rgpA+ mutant strain. Secondly, purified Kgp from W83 vesicles have a haemaglobinase activity which is comparable to haemoglobin hydrolysis by intact outer membrane vesicles which contain both active Kgp and Rgp [75]. Therefore, Kgp has been designated as the major active haemoglobinase in P. gingivalis [75].

The haemoglobin binding abilities of RgpA and Kgp have been examined using both purified monomeric and heterodimeric proteins [44]. Both forms of RgpA and Kgp bind strongly to haemoglobin with dissociation constants (KD values) of ~2–4 nM [23]. This binding affinity can be destroyed by boiling, indicating that the haemoglobin binding activity of gingipains requires folded proteins [45]. Both purified RgpA and Kgp have been observed to bind to haemin [45, 82]. In contrast, RgpB, which lacks the HA region, has little or no binding activity to either haemin or haemoglobin, further confirming that the HA regions play a critical role for gingipains to bind with haemoglobin and haem [82].

The haemoglobin-binding receptor in gingipains has been identified as Kgp15/Rgp15 region [54, 83], which is also referred to as the HA2 domain [45]. The Kgp15/Rgp15 region consists of 135 residues with a theoretical molecular weight of 15 kD, but the recombinant protein (referred to as rKgp15 or rRgp15) runs anomalously as a single band at 19 kD on the SDS-PAGE under reducing conditions [54–55]. rKgp15 has been found to bind to haemoglobin in a pH-dependent manner with a KD of 50 nM at pH 5.5 [54], and the rRgp15 binds to haemoglobin and haemin with KD values of 4 nM and 16 nM, respectively [45].

Studies by Smalley et al. revealed that rKgp15 is able to convert monomeric haem (Fe(III) PPIX.OH) to µ-oxo bishaem ([Fe(III)PPIX]2O), the main component in the black pigment deposited on the cell surface by P. gingivalis [29]. The accumulation of µ-oxo bishaem on the cell surface is proposed to be a defence mechanism used by P. gingivalis against the ingress of oxygen and reactive oxygen species such as H2O2, and a mechanism for iron and haem storage [84].

Olczak et al. have examined the binding of gingipains with haemoglobin, non-metal porphyrins and metal porphyrins [82]. They found that Kgp and RgpA bound more tightly to haemoglobin than to the porphyrins, implicating the involvement of different mechanisms in regulating the binding capacity of gingipains to haemoglobin and porphyrin. It was suggested that the gingipain–haemoglobin binding might occur via protein–protein interactions [82]. In contrast, DeCarlo et al. have observed that the binding of rKgp15, RgpA and Kgp to haemoglobin and haemin was inhibited by porphyrin PPIX [45]. Further studies using modified porphyrins suggested that rKgp15 and porphyrin recognition is mediated via the vinyl side chain [55].

Accumulated evidence supports a critical role for Kgp in binding of and in the degradation of haemoglobin, haem capture and the accumulation of the black pigmentation [54, 75, 82]. However, recent studies by Smalley et al. in dicated a specific role for RgpA in the black pigmentation of P. gingivalis, as that the pigmentation process requires both Rgp and Kgp activities [85]. They have proposed that Kgp and Rgp co-operate to degrade haemoglobin, with where RgpA mediating the conversion of oxyhaemoglobin to methaemoglobin and Kgp rapidly degrading methaemoglobin to generate µ-oxo bisheam [86].

Several mechanisms have been proposed to explain haem transportation in P. gingivalis. Olczak et al. have demonstrated that gingipains are able to bind to haem receptor proteins such as HmuR [82]. They have proposed that haemoglobin is degraded by Kgp, upon which the released haem is then initially captured by gingipains and delivered to the outer membrane receptor HmuR [82]. Recently, Gao et al. have identified another haemophore-like haem-binding protein, HusA, which can bind to dimeric haem with a KD of ~1 nM [87]. These authors have proposed that the haem transportation involves the following steps: 1) Kgp degradation of host methaemoglobin and capture of the released monomeric haem; 2) Kgp converts the monomeric haem to µ-oxo dimeric haem; 3) the dimeric haem is transferred to the haem receptor HusA and subsequently transferred to transmembrane protein HusB; 4) the dimeric haem is transported to the periplasm by HusB [87]. Although these proposals provide some biochemical and molecular insight into haem transportation, the detailed mechanisms need yet to be further elucidated.

Gingipains in the degradation of host proteins and in house keeping

Gingipains, the “trypsin-like” cysteine proteinases, are capable of cleaving or degrading a range of host proteins including plasma proteins, extracellular matrix proteins, cytokines and host cell surface proteins [25]. Fifteen P. gingivalis strains grow well in KGB minimal medium containing bovine serum albumin as the sole carbon and nitrogen source, and W50 and ATC33277 also grow well in a medium with bovine γ-immunoglobulin as the sole carbon and nitrogen source [88]. However, strains with rgpA or kgp gene deficiencies and an rgpA– kgp– RgpB– triple mutant do not grow in media supplied with human serum as the sole carbon and energy source, indicating the important roles of RgpA and Kgp in degradation of human serum proteins such as albumin, transferrin and IgG [89]. The binding and degradation of haemoglobin by gingipains is similarly linked to acquisition of peptide nutrients and the essential haem.

Aside from the dual roles of nutrient release and haem capture, gingipains are also responsible for the housekeeping work of gingipain maturation. Almost all of the proteolytic processing sites in gingipains and HagA are Arg-Xaa bonds except for one Lys-Xaa site that is found at the C-terminus of Kgp15/Rgp15. Perhaps this indicates the greater importance of RgpA and RgpB than Kgp in maturation and self-processing [25]. Gingipains also process profimbrilin to fimbrilin, the building block subunit used in the assembly of the fimbria [90].

Gingipains and the dysregulation of the host immune response

As an intruder to the human body, P. gingivalis is naturally a target of the host immune system including the initial non-specific innate immune response and the specific adaptive immune response [5]. The innate immune system provides the first-line of defence against infections by activating the complement system and employing antibacterial peptides, neutrophils and tissue-resident macrophages. The adaptive immune system, which is activated by the innate immune response subsequently defends the host against bacterial invasion by using antibodies, activated macrophages and cytotoxic T cells [34]. To survive a vigorous attack from the host immune system, P. gingivalis manipulates both the layers of host defense and gingipains are key players in the process.

Cationic antimicrobial peptides including human β-defensins, cathelicidin LL-37 peptide and neutrophil-derived α-defensins form the first line of innate immune defence in the oral cavity. Pathogen invasion induces the expression of these antimicrobial peptides [91]. On the other hand, gingipains have been found to degrade these induced antimicrobial peptides, enabling the survival of this pathogen in the periodontal lesions [34].

The innate immune responses summon leukocytes, mainly neutrophils and tissue-resident macrophages, to eliminate the intruding pathogens. Gingipains have been found to mediate a dysfunction in the phagocytes [5] by cleaving the formyl-methionyl-leucyl-phenylanine (FMLP) receptors, resulting in the inactivation of neutrophils and an inability to recognize the invading pathogens [92]. The observation that Rgps are able to strongly suppress the chemiluminescence responses of neutrophils to serum-activated zymosan [93] implies a role for gingipains in attenuating the bactericidal activities of neutrophils. Additionally, Kgp has been reported to be involved in the degradation of a C5a receptor (C5aR) on the surface of phagocytes, resulting in a suppression of neutrophil migration to infected sites [94–95].

The complement system plays a significant role in the innate immune response via three pathways: the classical, the lectin and alternative pathways. This cascade system includes about 20 glycoproteins that work in a sequence to form the membrane attack complex (MAC), which is a complex composed of C5b, C6, C7, C8 and C9 factors [96]. Upon activation, the MAC aggregates to form transmembrane channels on the bacterial cell surface, leading to the death of the bacteria caused by ruptured membranes [96]. Gingipains have been found to be able to degrade C3, C4 and C5, preventing the activation of C5b and the subsequent formation of MAC [34].

Bleeding on probing is an important clinical sign in the diagnosis of periodontitis [3, 97]. RgpA has been found to stimulate the activation of thrombin in the periodontal lesion by directly activating factor X, factor IX, prothrombin and protein C in the coagulation cascade [34]. The uncontrolled activation of thrombin stimulates the production of prostaglandin, platelet-activating factor and interleukin-1, which are major factors in tissue inflammation and destruction [5]. Additionally, Kgp digests fibrinogen and fibrin efficiently, which are major components of plasma and responsible for blood coagulation [34, 53]. Hence, by working in unison the gingipains contribute to bleeding by deregulating the host coagulation system.

Rgps are also able to activate prekallikrein to generate kallikrein, a protease which is responsible for the formation of bradykinin from high molecular weight kininogen [5, 34]. Additionally, Rgps and Kgp together are able to release kallidin directly from kininogen [34]. Bradykinin and kallidin, also known as kinins, enhance the vascular permeability and the generation of gingival crevicular fluid [5, 34]. They also activate prostaglandin synthesis in periodontal-ligament cells and osteoblasts, which leads to alveolar bone resorption [34].

P. gingivalis is also known to evade the host immune system by deregulating the cytokine network [98–99], which comprises signalling proteins and their receptors. Gingipains have been shown to degrade and inactive many of the components of the cytokine network including the interlukins (IL), IL-1β, IL-4, IL-6, IL-8 and IL-12, the IL-6 receptor, interferon-α, tumour necrosis factor-α(TNF-α), CD4, CD8, CD14 and CD54 [34]. This results in the disruption of critical communications within the immune system. By deregulating the host cytokine network, the bacterium is able to evade destruction by immune cells.

Gingipains in tissue matrix destruction

It is well known that gingipains are able to degrade a number of extracellular matrix proteins in vitro, including laminin, fibronectin, collagen types III, IV and V, and possibly collagen type I which is the major component of collagen fibers in gingival connecting tissue [5, 100]. However, it has been pointed out that the degradation of matrix proteins by gingipain proteolytic activity alone is insufficient to cause the destruction of the gingival connective tissue. For this to occur, the endogenous matrix-degrading metalloproteinases also make a significant contribution [5, 34]. Matrix metalloproteinases (MMPs) are a group of zinc- and calcium-dependent endopeptidases produced as zymogens (pro-MMps) by many cells including dendritic cells, periodontal ligament cells, gingival fibroblasts and the oral mucosa [34]. The pro-MMPs are activated by the removal of the pro-fragment, usually through proteolysis regulated by the host. The activities of MMPs are inhibited by tissue inhibitor metalloproteinases (TIMPs). It has been reported that Rgp is able to up regulate the expression of MMP-1 from human gingival fibroblasts [101] and to activate MMPs directly [102–103]. Such actions could result in an imbalance between MMPs and TIMPs, which would contribute to the destruction of the matrix tissue.

Neutrophils, which are attracted by host inflammatory responses, infiltrate the periodontal tissue and release elastase, cathepsin G, protease 3, MMP-8 and MMP-9 [34]. Such enzymes are normally repressed by inhibitors pro duced by the host. However, once again it has been reported that gingipains are able to proteolytically inactivate these inhibitors [104–107], leading to destructive proteolysis of the periodontal tissue.

The modular structure of the HA regions of gingipains

The Protein families (Pfam) database

Sequence alignments between RgpA, Kgp and HagA have shown that some peptides are highly conserved in their HA regions. The sequence similarities have been thoroughly analysed in the Pfam database and an alternative domain architecture has been predicted for gingipains. The Pfam database (http://pfam.sanger.ac.uk/) is a large and comprehensive collection of protein families constructed using multiple sequence alignments and profile hidden Markov models [108]. Pfam-A contains manually curated families of a high quality. The construction of each Pfam-A family requires an annotation, a seed alignment, a profile hidden Markov model and a multiple-sequence alignment. The seed alignment is made by aligning a set of sequence representatives for which there is reliable evidence that they are family members and that the sequences are not very similar. The seed alignment is then used to build up a profile hidden Markov model. This profile hidden Markov model is then used to search homologues in any sequence database. Subsequently, a full alignment is made by aligning the sequence homologues with the profile hidden Markov model. A domain family is then assigned for all of the aligned sequences. The domain families with a similar fold or predicted fold are included in a superfamily or clan. For each protein sequence entry, a schematic domain organisation is made up based on the predicted domains.

Sequence alignments between RgpA, Kgp and HagA have shown that some peptides are highly conserved in their haemagglutin/adhesin regions. The sequence similarities have been thoroughly analysed and presented in the Pfam database and a modular architecture has been predicted for the HA regions of gingipains.

The cleaved adhesin domain family

There are presently a total of 102 entries in the cleaved adhesin family (PF07675). These cleaved adhesin domains are found in 47 protein entries from P. gingivalis, Psychroflexus torques, Herpetosiphon aurantiacus, Candidatus Cloacamonas acidaminovorans and Flavobacteria bacterium BAL38, with 20 unique domain organizations. Within the 30 protein sequences from P. gingivalis, 15 are full length gingipains (RgpA and Kgp) and three are HagA proteins from strain W83, 381 and ATC33277 containing eight or ten cleaved adhesin domains. The alternative domain structure model is presented for RgpA, Kgp and HagA in Fig. 4 and the set of Pfam models chosen are representative of all of the adhesin modules found in disease related strains of P. gingivalis. In order to refer to the cleaved adhesin domain model, we have named the modules K1, K2 and K3 in Kgp, R1 and R2 in RgpA and A1 to A10 in HagA. The domain structures of full length RgpA or Kgp predicted by Pfam comprise two regions: a catalytic domain region and a cleaved adhesin domain region. The catalytic domain regions of RgpA and Kgp have a similar construction, composed of one signal peptide, one C25 propeptide, one C25 peptidase domain and one peptidase C terminal Ig-like domain. The cleaved adhesin domain regions of RgpA and a Kgp HG66 contain two cleaved adhesin domains and two DUF2436 domains of unknown function which flank the two cleaved adhesins, while three cleaved adhesin domains and one DUF2436 domain are found in all other Kgps (Fig. 4).

Fig. 4.

Domain structures of RgpA, Kgp and HagA. RgpA is from strain W50, Kgps are from strains HG66, 381, W83 and W83v. HagA is from strain 381. The domains are presented as colored boxes and residues at the N-terminus and C-terminus for each domain are labeled at the top of each box. The cleaved adhesin domains are assigned to be R1 and R2 in RgpA, K1, K2, K3, K1* and K3* in Kgps and A1–A10 in HagA. The green, yellow and blue bars represent the flanking domain locations of three adhesion binding motifs ABM1, ABM2 and ABM3. The regions labelled as DUF2436 are possibly domains of unknown function

In a comparison of the Pfam predicted domain model with the gingipain domain structure hypothesis (Fig. 2) it is evident that the domain organizations in the catalytic domain regions of the two models are basically the same, however the domain boundaries of the HA regions differ substantially. K2 and R2 domains, containing 172 residues in each, have the same N-terminal boundaries as those defined for Kgp15 and Rgp15. K1 and R1 domains are located in the C-terminal regions of Kgp39 or Rgp44, respectively, while the K3 domain and the C-terminal DUF2436 domain are located at the centre of the Rgp17–Rgp27 and Kgp44 regions and share an identical N-terminus with Kgp13.

Comparison of the cleaved adhesin modules found in gingipains

To eliminate degenerate sequences in the cleaved adhesin domain family from consideration, nine sequences from P. gingivalis were chosen as seeds to represent the sequence variety. The multiple amino acid sequence alignment including strain specific variations for all of the 110 sequences can be viewed at http://pfam.sanger.ac.uk.

Sequence alignments of the sequences of 31 cleaved adhesin domains from RgpA, Kgp and HagA from strains HG66, 381, W83, W83v and W50 have shown that they form five tight homology groups (labelled in five different colours and defined as groups 1–5 and shown in Fig. 5). K2 is conserved in all Kgps and K1 and R1 are identical in all strains. The sequences of K2, R2, A3, A5, A7 and A10 are 97–100% identical. Kgp W83v effectively has a K3–K2–K3 combination as the K1* and K3 domains of this protein are 94% identical. K3, A8 of HagA1 and A10 of HagA2 from W83 strains are almost identical, and K1*, A2, A4, A6 and A8 of HagA2 are similar (90% identical) to K3. The K3* domain is the most different in the set and sits alone being at best ~40% identical to modules in other groups. The sequence identities between members of the five groups are in a range of 30–74%. These close relationships indicate that this set of groups of modules is a result of gene duplication. The high level of conservation of amino acid sequence in these homologues may also indicate functional conservation in addition to the predicted conservation of structure.

Fig. 5.

Pairwise sequence comparisons of 31 cleaved adhesin domains from RgpA, Kgp and HagA from a complete family multiple sequence alignment. The conservation of amino sequences in RgpA, Kgp and the HagA proteins is shown by colour groups. Each colour group represents regions of amino acid sequence in P. gingivalis proteins which in pairwise alignments are more than 96% identical. The number on the line linking domains of different colour groups is the percentage of sequence identity between each pair. The five groups of adhesins are defined here as follows: Group 1 = K1 and R1; Group 2 = A2, A4, A6, A8, A10, K1* and K3; Group 3 = A1; Group 4 = A3, A5, A7, A9, K2 and R2; and Group 5 = K3*. The multiple sequence alignments are provided as an Appendix in supplementary material

The galactose-binding domain-like superfamily

The hidden Markov model predicts that the cleaved adhesin domain family belongs to the galactose-binding domain-like superfamily (GBD) (PFam Clan CL0202). Observed structures of the family members in the GBD superfamily indicate that they share a β-sandwich jelly roll topology despite sequence similarities between many of them being insignificant. Twenty-five protein families are found in this clan representing a diversity of biological functions. For example, the Ephrin receptor ligand binding domain family are involved in protein–protein interactions and bind to ephrins. The MAM domain family proteins are involved in cell–cell interactions, while carbohydrate binding module families CBM4–9, CBM6, CBM11 and CBM17–28 (as number by the CAZY database) are sub-domains of glycoside hydrolases, many of which are involved in plant cell wall degradation. It is not known whether the cleaved adhesin domain family is functionally related to any of the other families or if it is only structurally similar.

Although the haemagglutinin/adhesin regions of gingipains have been related to the virulence of P. gingivalis in colonization, haemagglutination, haemolysis, and haem acquisition, the exact mechanisms by which gingipains interact with the host remain unclear [34]. Biochemical studies based upon exploiting the predicted modular HA model combined with further structural studies of these modules may be the approach that will lead us to a comprehensive understanding of the mechanisms of gingipain function.

The structures of cleaved adhesin modules

The Pfam predicted domain structures of the HA regions of gingipains have provided new information on the probable domain boundaries and of possible similarities between the cleaved adhesin domains. Based on these models, a number of recombinant protein constructs suitable for structural studies have been the subject of close investigation. Several of these recombinant protein constructs can be produced as soluble proteins in vitro (Nan Li personal communication). The crystal structure of the W83 Kgp K2 (Group 4) module has recently been reported and refined at high resolution [109] (Fig. 6A). As predicted the K2 module folds as a β-sandwich jelly roll globular domain. This module shares a significant conservation of sequence with modular sections found in each of the Kgp, RgpA and HagA proteins expressed by this and other strains (Fig. 5). More recently the crystal structure of the K3 (Group 2) module of W83 Kgp refined at a resolution of 1.6 Å has been deposited at the PDB (code: 3M1H) (Fig. 6B). The K3 fold is also a β-sandwich module and as reported for K2 it is also stabilised by two Ca2+ ions (Fig. 6). It is not uncommon that protein domains such as K2 (Group 4) and K3 (Group 2) with only 33% amino acid sequence identity are observed to be such close structural homologues. This fold is intimately related to other families of the galactose-bind ing domain-like superfamily. Given the sequence identity of 71% between K3 and K1 (Group 1) in strain W83, the fold of the K1 module in Kgp will be almost identical. It can therefore be concluded that HA region of Kgp W83 is composed of a region unknown structure (the so-called DUF2436 domain) followed by a three repeats in tandem of homologous cleaved adhesin modules. These four domains are flanked and linked by sequences that represent adhesion binding motifs (ABM1, ABM2 and ABM3) (Fig. 4).

Fig. 6.

Ribbon representations of the crystal structures of cleaved adhesin modules coloured from N to C termini with a colour ramp from blue to red. A. K2 from strain W83 (PDB code: 3KM5) and B. K3 from strain W83 (PDB code: 3M1H). C. A projection of the three dimensional fold topology of the K3 structure. The β-strands in the two β-sheets which make up the jelly roll fold are coloured green and yellow, respectively, with two small β-strands located in the loop region coloured blue. As indicated by a visual inspection of A and B the topology of K3 as shown is almost identical to that reported for K2 [109]

As three of the cleaved adhesin domain group structures are not known (Groups 1, 2, 4) it is highly likely, given the known sequence similarities and identities (Fig. 5), that all the members of the P. gingivalis cleaved adhesin Groups 1–5 (Fig. 5) are indeed homologues.

Unravelling the functions of cleaved adhesin modules

As it has only recently become apparent that the HA regions in gingipains are composed of tandem repeats of discrete adhesin modules, functional studies of these discrete gingipain components have only just commenced. Possible functions of the K2 module have been investigated. Purified Kgp and the recombinant K2 module induce the in vitro haemolysis of erythrocytes with similar efficiencies. The observed dose-dependent responses can be augmented by the blocking of anion transport [109]. Further, cysteine-activated arginine gingipain RgpB, which degrades glycophorin A, sensitizes erythrocytes to the haemolytic effect of K2. Cleaved K2, similar to that found in extracted Kgp preparations, lacks the haemolytic activity indicating that autolysis of Kgp may be a staged process which is artificially enhanced by extraction of the protein [109]. This hypothesis assumes a function for K2 in the role of Kgp that enables the porphyrin auxotroph to acquire essential haem from erythrocytes.

Any likely hypothesis that addresses the basis of the polyfunctional nature of the HA regions and ginigpains is to be found in the different and cooperative binding properties of individual adhesin modules. Specific sets of modules may confer particular binding activities on gingipains, thereby targeting the catalytic domains to specific host proteins. Further, it might be that some sets of these modules act synergistically to bind to those protein targets. How these multi-domain proteins may be organized in the context of the bacterial surface, perhaps even associated with other partner proteins in gingipain complexes, is yet to be determined.

Future investigations of the specific recognition of host protein/carbohydrate targets by the cleaved modular adhesins found in the HA regions and related HagA proteins will contribute to a greater understanding of complex gingipain functions and their role in P. gingivalis biology. For the development of clinical applications, more knowledge is required about the specific role of gingipains as virulence factors, and particularly of their interactions with the host immune system.

Appendix

Multiple sequence alignments of cleaved adhesin domains of Kgp, RgpA and HagA

31 sequences of cleaved adhesin domains of Kgp W83, Kgp HG66, Kgp 381, Kgp W83v, RgpA HG66, HagAW83 and HagA 381 are aligned using the program ClustalW (Larkin et al. [110]).

The first column shows the assigned domain names corresponding to Figures 4 and 5, sequences are in five tight homology groups which were enclosed within boxes of different colours. K1/R1 group (red frame), K1’/K3/A2/A4/A6/A8/A10 group (blue frame), A1 group (green frame), K2/R2/A3/A5/A7/A9 group (orange frame), K3* group (purple frame).