Skip to main content

Protein citrullination as a link between periodontal diseases and rheumatoid arthritis (RA) and target for development of novel drugs to treat RA

Final Report Summary - GUMS AND JOINTS (Protein citrullination as a link between periodontal diseases and rheumatoid arthritis (RA) and target for development of novel drugs to treat RA.)

Executive Summary:
"Gums & Joints" was a collaborative research project constituting of 10 international groups and 2 SME working together to investigate the possible causative link between periodontal disease (PD) and rheumatoid arthritis (RA).

P. gingivalis infection is strongly associated to the development of PD, a disease affecting 30% of the world population.Interestingly P. gingivalis is the only bacterium expressing the enzyme peptidylarginine deiminase (PAD) which converts arginine to citrulline, a process referred to as citrullination. Antibodies torwards such citrullinated proteins, i.e ACPAs/aCCP, are of central importance in RA, a chronic autoimmune disease which affect 0.5-1% of the population. The presence of ACPAs or aCCP autoantibodies is not only highly specific for RA but the presence is also related to a more severe and destructive disease progression.

During the lifetime of the project interdisciplinary cooperation with medical and dental clinicians, epidemiologists, bio-scientists, and industrial scientists has resulted in more than 70 published peer-review articles. By employing state-of-the-art methodologies and well-established cohorts the consortium has managed to discover key findings that may help to unveil the etiology of rheumatoid arthritis.

In short our findings include that self reported periodontitis can not be trusted, but through linking the EIRA cohort to the Swedish dental health registry it was discovered that age and smoking constitutes risk factors for periodontal disease in RA patients and controls. Further, presence of anti-RgpB antibodies is not suited as a marker for PD, although the presence of these antibodies is associated with ACPA positive and ACPA negative RA.
We solved the structure for PPAD after succeeding with purification of recombinant protein in sufficient quantities needed for crystallization. This further aided in the development of specific inhibitors. In addition, several inhibitors for human PADs have been characterized.

The potential for using detection of PPAD as screening tool for P. gingivalis infection in RA patients were investigated and it was found that the activity was higher in gingival crevicular washes from periodontitis patients as compared periodontally healthy RA patients or controls.

Project Context and Objectives:
Periodontitis is the most prevalent infectious inflammatory disease of mankind. It is estimated that up to 30% of the adult population suffers from periodontitis, and in approximately 8% a severe form of the disease ultimately leads to tooth loss. If morbidity caused by tooth loss due to advanced periodontitis is not enough, mounting evidence suggests a causative link between periodontal disease and rheumatoid arthritis.
It is now generally accepted that a relatively small consortium of bacteria (P. gingivalis, T. denticola and T. forsythia), described as the “red complex”, is strongly associated with the pathological changes in the periodontium.

Proliferation of “red complex” species on the subgingival tooth surface initiates a chronic inflammation of the periodontium. It is now understood that periodontal tissue damage is a consequence of a futile attempt by the host’s innate immune response to eradicate microbial invaders.
Rheumatoid arthritis (RA), affecting 0.5-1% worldwide, is a severe systemic autoimmune disease, where patients suffer chronic joint inflammation, causing pain and disability, comorbidities and increased mortality. The specificity and regulation of the underlying autoimmune reactions in RA have for long been elusive, preventing a deeper understanding of disease processes and the development of a cure. However, the discovery of antibodies reactive with citrullinated proteins has given us an opportunity, and a biological tool, to improve this situation. These antibodies, present in approximately 70% of patients, are (i) disease-specific; (ii) associated with a more severe disease course, suggesting a pathogenic involvement; and (iii) are present before clinical onset, which may indicate that the initial loss of immune tolerance to citrullinated proteins occur as a consequence of an inflammatory event outside the joint.
Interestingly, clinical studies of RA and periodontitis have provided strong evidence for a significant association between the two diseases. Patients with long-standing, active RA have a substantially increased frequency of periodontitis as compared with healthy subjects. Vice versa, patients with periodontal disease have higher prevalence of RA than patients without periodontitis. The expression of peptidyl-arginine deiminase (PAD) by P. gingivalis (unique for this bacterium) (PPAD), which is involved in citrullination, may constitute a mechanistic link explaining the clinical correlation between RA and periodontitis. Protein citrullination by P. gingivalis may expose new protein epitopes.
It has been shown that PPAD can efficiently deiminate C-terminal Arg residues of bradykinin to inactivate the activity of this kinin. Several other biologically active peptides have C-terminal arginine residues important for their activity. Therefore, citrullination of these peptides by PPAD may directly affect the inflammatory response. Further, citrullination of C-terminal arginines in peptides generated through the action of Arg-specific gingipain should facilitate P. gingivalis adherence to host connective tissues and to other dental plaque bacteria since these interactions are compromised by peptides bearing C-terminal arginine. Finally, generated by PPAD, citrullinated proteins and peptides can trigger autoimmune reaction to endogenously modified proteins and initiate RA as described above.

Our aim in this project is to examine the role of P. gingivalis in the pathogenesis of RA and periodontitis taking a multipronged approach:
• A multidisciplinary examination (genetics, epidemiology, molecular immunology and animal models) of susceptibility factors and immune responses in RA and periodontitis to identify etiologic pathways in both diseases
• Detailed analysis of the structure and function of P. gingivalis PAD and design of specific inhibitors as novel treatments
• Analysis, coordination and dissemination of data

Project Results:
The consortia have used epidemiological approaches to study genetic and environmental risk factors in distinct subsets of RA, to efficiently address the possibility that microbes associated with chronic periodontitis (i.e. Porphyromonas gingivalis) may be driving immune reactions (i.e. the production of anti-citrullinated protein antibodies – ACPA) of critical importance in a large proportion of RA patients. The hypothesis underlying this study is based on the fact that the oral pathogen P. gingivalis expresses its own unique peptidyl arginine deminase (PAD) enzyme, capable of citrullinating its own as well as human proteins, potentially producing citrullinated neo-epitopes in the inflamed gingival tissue, which may trigger B cells to produce low affinity ACPA. In genetically predisposed individuals (e.g. carriers of HLA-DRB1 shared epitope (SE) alleles), pathogenic T cells may subsequently be activated and help B cells to produce high affinity ACPA, which through epitope-spreading could cross-react with citrullinated epitopes exposed in the joints. This in turn may trigger a chronic inflammation and destruction, through the formation of immune complexes and the activation of synoviocytes and osteoclasts.
In order to address these scientific questions we have focused on the large and well-characterised Swedish population-based case-control cohort EIRA (the Epidemiological Investigation of RA), where data on genes and environmental exposures was already available. We have performed extensive serological analyses and a large linking analysis of patient registries, and then used epidemiology, biostatistics and bioinformatics, to convert the genetic, environmental and serological data into meaningful patterns, which are used to formulate new models of aetiology and pathogenicity in the development and progression of RA.
1) The identification and characterisation of the PD-RA population; achieved through linking of EIRA with the Swedish Dental Health Registry (DHR)
2) The identification and characterisation of the anti-P. gingivalis antibody positive RA population; achieved after screening EIRA for the presence of anti-P. gingivalis IgG, using the new P. gingivalis-specific ELISA developed within the Gums & Joints consortia

In order to get a comprehensive picture of the ACPA response in early RA, we investigated the presence of antibodies targeting the second generation of cyclic citrullinated peptide(s) (CCP2) as well as citrullinated peptides from α-enolase, vimentin, fibrinogen and collagen type II in EIRA RA cases (n=1985) by ELISA. To further increase our understanding of the aetiology and pathogenesis of RA, we studied well-known genetic and environmental risk factors in different subsets of RA, defined by the different ACPA profiles. Association of genetic and environmental risk factors with different subsets of RA was calculated by means of logistic regression analysis, in a case-control setting, using 2252 matched EIRA controls. The majority of ACPA positive patients were confined to the CCP2 positive subset, although 18% of CCP2 negative patients were positive for at least one ACPA fine-specificity. In total, 17 RA subsets could be identified based on their different ACPA fine-specificity profiles (figure 2). Large differences in association with genetic and environmental determinants were observed between subsets. Strongest association of HLA-DRB1 SE, PTPN22 and smoking was identified for the RA subset defined by the presence of antibodies targeting citrullinated α-enolase peptide 1 (CEP-1) and citrullinated vimentin. From this study we could conclude that HLA-DRB1 SE genes, the PTPN22 risk allele and cigarette smoking associated with the presence of specific ACPA-reactivities, rather than total levels of anti-CCP2 antibodies, i.e. these risk factors seem to influence the specificity rather than the magnitude of the ACPA response.

To further characterize ACPA positive and ACPA negative RA, we investigated the antibody response to carbamylated proteins (anti-CarP) in EIRA (n=1985) and in a Dutch early arthritis cohort (n=846), in relation to CCP2 status, HLA-DRB1 SE and smoking. Anti-CarP antibodies were detected in approximately 40% of RA patients, mainly in CCP2 positive disease, but also – to a lesser extent – in CCP negative disease. Importantly, no associations between anti-CarP antibodies and SE, other HLA-DRB1 alleles, PTPN22 or smoking could be identified, when analyses were adjusted for anti-CCP2 antibody status.
Linking of patient registries to identify PD in EIRA RA cases and controls
It was found that self-reported periodontitis in the EIRA questionnaire can not be trusted to identify PD among RA cases and controls, therefore we proceeded with another approach, namely to link the EIRA registry with the Swedish Dental Health Registry (DHR), where dental diagnostic- and treatment codes on a majority (>80%) of the adult Swedish population have been registered since 2008. This linkage of registers was made possible due to the unique 10-digit national registration number issued to all Swedish residents, and because all treating dentists in Sweden report diagnostic- and treatment codes for numerous dental and orofacial conditions for their patients according to a system for reimbursement. In total, 2740 EIRA RA cases and 3942 EIRA controls were identified in DHR, which corresponds to 86% of all EIRA subjects.

Dental records confirmed 90% of the periodontitis diagnoses in DHR among RA cases, and 88% among controls. Approximately 70% of all EIRA subjects, identified in DHR, had at least one of the diagnostic codes related to periodontal disease. The most frequent diagnostic codes observed were for gingivitis (33% of RA and 35% of controls) and periodontitis (33% of RA and 32% of controls). No differences in prevalence between RA cases and controls were found for any of these codes or for any of the other diagnostic codes (i.e. increased risk for periodontitis, peri-implantitis or increased risk for peri-implantitis). When stratifying the analyses for presence or absence of ACPA among RA cases, no significant differences were observed. Also the treatment codes for periodontitis or increased risk for periodontitis did not differ between RA and controls, or between ACPA positive and ACPA negative RA. We could, on the other hand, confirm both age and smoking as risk factors for periodontal disease also in the EIRA study population, for both RA cases and controls.
Since EIRA was initiated in 1996, and DHR only in 2008, most EIRA RA cases identified in DHR would have had RA for a number of years when the linking of registers was performed. Hence, we have not been able to investigate whether periodontitis is a risk factor for RA. Instead, we have investigated whether RA is a risk factor for developing periodontitis, and while we can verify that smoking and ageing are risk factor for the development of periodontitis, we found no evidence of an increased risk for developing periodontitis in patients with established RA, compared to controls, and no effect of ACPA. We aim to perform a new linking analysis of EIRA with DHR in a few years, when more periodontal data would be registered in DHR on EIRA subjects, from time points before their enrolment in EIRA, i.e. before the RA diagnosis. Hence, such a study would focus on newly diagnosed RA, rather than established RA, and may give new insight into a possible relationship between RA and periodontitis, not identified here.

Anti-gingipain antibodies in RA, in relation to autoantibodies, smoking & genetics
The anti-RgpB IgG response was measured in sera from 1974 EIRA RA cases and 377 EIRA controls. Anti-RgpB antibody levels were significantly elevated in RA (p<0.0001) compared to controls (figure 4 B), and using the cut-off value from the non-PD controls, 23% of RA patients but only 9.4% of the non-RA controls, were considered anti-RgpB antibody positive. Logistic regression analysis showed a significant association between elevated anti-RgpB IgG levels and RA (OR=3.0; 95% CI: 2.0-4.4) which was even stronger than the well-known association between smoking and RA (OR=1.37; 95% CI: 1.0-1.6).
Dividing RA further, based on presence/absence of ACPA, demonstrated significantly higher anti-RgpB IgG levels in ACPA positive, compared to ACPA negative RA (figure 4 C). The association of anti-RgpB antibodies was also significantly stronger (p=0.017) with ACPA positive RA (OR=3.2; 95% CI: 2.2-4.8) than with ACPA negative RA (OR=2.4; 95% CI: 1.5-3.6). On the contrary, none of the investigated ACPA fine-specificities (ACPA targeting citrullinated epitopes on α-enolase, vimentin, fibrinogen or collagen type II), nor rheumatoid factor or anti-CarP antibodies showed any specific association with anti-RgpB IgG, beyond the effect of the overall ACPA-response, measured as anti-CCP2 antibodies (data not shown).
When investigating the anti-RgpB IgG response in relation to smoking, it was clear that the association of anti-RgpB IgG with RA was independent of smoking, we could even observe an inverse association in the control population, in line with previous reports demonstrating lower anti-P. gingivalis antibody levels in smokers compared to non-smokers (Vlachojannis C et al, J Clin Periodontol 2010). Interestingly, there was an additive effect and an interaction between anti-RgpB IgG and smoking in ACPA positive RA, but not in ACPA negative RA.
There was no association between anti-RgpB IgG and HLA-DRB1 SE. Instead, both anti-RgpB antibodies and SE were (independently) associated with both ACPA positive and ACPA negative RA. In the ACPA positive subset, there was also an additive effect and an interaction between anti-RgpB IgG and SE (table 3). No interaction could be observed between anti-RgpB IgG exposure and the PTPN22 polymorphism that has been linked to RA and other autoimmune conditions (data not shown). The data from this study supports the hypothesis of Porphyromonas gingivalis as an aetiological agent in the development of ACPA positive RA, where smoking and HLA-DRB1 SE alleles already constitute well-established risk factors.

Anti-citrullinated P. PAD antibodies in RA
In addition to the RgpB ELISA assay, we have also developed another P. gingivalis-specific ELISA within the consortium, based on citrullinated P. gingivalis peptidyl arginine deiminase (P. PAD) peptides (CPPs). This antigen was selected since it was shown that RA patients had a heightened antibody response to auto-citrullinated P. PAD, but not to uncitrullinated P. PAD. The specificity of the anti-citrullinated P. PAD antibody response was determined using 13 synthetic citrullinated peptides spanning the P. PAD molecule. The antibody responses to five of these peptides were further investigated in a pilot cohort of 65 PD patients, 59 periodontally healthy controls, 100 ACPA positive EIRA RA cases, 100 ACPA negative EIRA RA cases and 100 EIRA controls. The antibody responses to four of these peptides were significantly increased in ACPA positive RA, compared to all other subsets. This was particularly evident for CPP3 (figure 5 A). The antibody response to one of the peptides did not differ between groups, while the antibody response to CPP5 was significantly elevated in both ACPA positive RA, and (to a lesser extent) in the PD subset (figure 5 B).
Subsequently, CPP3 and CCP5, as well as the arginine-containing versions of these P. PAD peptides (RPP3 and RPP5), were added to a multiplex assay based on the ISAC technology (figure 6), in collaboration with Thermo Fisher Scientific, former Phadia. To this date, nearly 3000 EIRA RA cases and 400 EIRA controls have been screened on the ISAC chip, and the data are currently being analysed. In order to learn more about the potential role of Porphyromonas gingivalis in the aetiology of RA, the antibody response to CPP3/RPP3 and CPP5/RPP5 will be analysed in relation to all known RA autoantibodies, anti-RgpB antibodies, smoking habits and HLA-DRB1 SE alleles.

Genetic, environmental and serological characterization of the RA-PD subset
As described in detail in the linking analysis, the RA-PD subset differed from the periodontally healthy RA subset in two aspects only:

1. RA patients with PD were significantly older than periodontally healthy RA patients
2. RA patients with PD were more likely to be smokers when compared to periodontally healthy RA patients

These two differences were also identified among EIRA controls with or without PD, and are in line with current knowledge concerning risk factors for chronic periodontitis Notably, PD was not more prevalent in RA compared to controls. Based on this finding we conclude that RA is not a risk factor for the development of PD. Finally, presence of ACPA did not associate with periodontal disease in the RA population. Hence, we could not identify a link between ACPA positive RA and periodontal disease.
Importantly, we investigated established RA in this study. In other words, patients would have had RA for a number of years prior to the development of PD. Therefore, we speculate that the anti-rheumatic treatment could have had a dampening effect on the future development of PD, potentially masking an association between RA and periodontal disease. Potentially, this would be of particular significance among ACPA positive RA patients, who suffer from a more severe and destructive form of the disease, when compared to ACPA negative RA patients, hence may be more likely to receive aggressive treatment. The role of anti-rheumatic treatment on the development of PD will be examined in future projects, linking EIRA, DHR and ARTIS (anti-rheumatic treatment in Sweden) registries. We will also perform a new linking analysis, with focus on treatment naïve early RA cases instead of established RA, to avoid the potential masking effect of anti-rheumatic treatments.
Since we could not use DHR to diagnose PD in EIRA RA cases and controls before enrolment in the EIRA study, and we failed to validate the EIRA questionnaire when answers were compared to dental records, and since we could not use anti-P. gingivalis antibodies as a surrogate marker for periodontitis either, we cannot say anything about PD as a risk factor for future RA. However, as described in detail in the P. gingivalis ELISA study, we could clearly show that RA patients, in particular ACPA positive RA patients, had a heightened antibody response to two P. gingivalis-specific antigens: arginine gingipain B and citrullinated P. PAD. Therefor we have also been able to characterize the “anti-P. gingivalis antibody positive RA subset”, in terms of genetics, autoimmunity and smoking.
This analysis revealed that the frequency of smokers was equal among anti-RgpB IgG positive RA patients and anti-RgpB IgG negative RA patients. Likewise, the prevalence of HLA-DRB1 SE positive individuals did not differ between anti-RgpB IgG positive and negative RA patients. However, ACPA positivity was more common in anti-RgpB IgG positive RA, compared to anti-RgpB IgG negative RA. Moreover, in ACPA positive RA, there were statistical interactions - that can be interpreted as biological interactions (Rothman, Am J Epidemiol, 1980) - between elevated anti-RgpB IgG levels and smoking and between elevated anti-RgpB IgG levels and HLA-DRB1 SE, data which may suggest a role for Porphyromonas gingivalis in the aetiopathology of ACPA positive RA.
The RA subset characterized by an antibody response to CPP3/RPP3 and/or CPP5/RPP5 remains to be defined, and subsequently studied in relation to the anti-RgpB antibody response, ACPA, smoking, and genetic risk factors.

Based on the data generated we conclude the following:
1. RA is (most likely) not a risk factor for the development of periodontitis
2. Porphyromonas gingivalis remains a credible candidate for triggering and/or driving autoimmunity and autoimmune disease in a subset of RA

It is well established today that the ACPA response antedate RA symptoms and diagnosis by several years, and that the ACPA response increase (in terms of antibody levels, affinity, isotype usage, glycosylation pattern and number of recognised epitopes) six to twelve month prior to diagnosis. However, we still do not know where in this disease process Porphyromonas gingivalis may act.
All we know today is that the anti-P. gingivalis antibody response is present in early RA, within one month of diagnosis (which is when RA patients are recruited to the EIRA study, and donate blood for genetic analysis as well as serological analysis).

Post-translationally modified (PTM) proteins in the development of RA
For investigating post-translationally modified proteins our efforts have been mostly directed towards detection and assignment of sites of citrullination for in-vitro citrullinated LL-37, EGF, C5a and in-vitro carbamylated LL-37, EGF and bradykinin. Mass spectrometry based assays have been developed for detection of the in-vitro modified proteins. These have been used to determine the degree and specificity of modification both in terms of citrullination and carbamylation. For identification of citrullination we have used a large extent of manual spectral inspection in several projects. These projects aimed at examining the citrullination patterns of PPAD on suspected target proteins, such as LL-37, EGF, C5a-anaphylatoxin, Bradykinin, Staphopain A+B etc.
PPAD induced citrullination of the C5a C-terminal was detected by tandem mass spectrometry. Two different variants of the C-terminal were generated by either Trypsin or Clostripain treatment. Citrullination of the C-terminal was detected by a combination of a 1 Da mass shift, MS obtained sequence information (Peptide Score) and a clear shift in retention time compared to the untreated sample. Only the citrullinated version of the C-terminal was detected in the PPAD treated sample indicating a complete citrullination at the C-terminal. The same C5a specific assay was used to show that P.gingivalis OMV extracts where able to generate and citrullinate C5a.
We have characterized the complex pattern of in-vitro carbamylation of LL-37 and studied the peptide after different degrees of modification. By establishing a targeted mass spectrometry assay we were able to differentiate LL-37 with zero to seven carbamylations. The carbamylation of LL-37 results in a complex mixture of different LL-37 species. At least 23 different forms of LL-37 arise after incubation with 50 mM KOCN. The many different species of LL-37 arise through degree and position of the carbamylations. From these data, it is clear that the N-terminal and all Lys residues within LL-37 can be carbamylated, with the N-terminal showing the highest tendency to become modified.
Cathelicidin LL-37 plays an essential role in innate immunity by killing invading microorganisms and regulating the inflammatory response. These activities depend on the cationic character of the peptide, which is conferred by five arginine and lysine residues. Since LL-37 is released in large amounts in response to inflammatory stimuli, it is highly plausible that it is a good substrate for PADs, which are secreted into the inflammatory milieu. Therefore, the activity of PAD4 expressed in E. coli system was verified using synthetic LL-37. Depending on the incubation time and enzyme concentration, deimination of the Arg residues within LL-37 generated different levels of citrullination as shown by amino acid sequence analysis of the peptide incubated with PAD2, PAD4 and rabbit PAD. Since the level of citrullination may variably affect the ability of the peptide to neutralise LPS we verified whether the degree of citrullination cause abrogation of the anti-inflammatory activity of LL-37. We found that, when citrullinated, LL-37 was at least 40 times less efficient at neutralizing the proinflammatory activity of LPS due to a marked decrease in its affinity for endotoxin. Also, the ability of citrullinated LL-37 to quench macrophage responses to lipoteichoic acid and poly(I:C) signaling via TLR2 and TLR3, respectively, was significantly reduced. Furthermore, in stark contrast to native LL-37, the modified peptide completely lost the ability to prevent morbidity and mortality in a mouse model of d-galactosamine-sensitized endotoxin shock (Figure 3). Importantly, serum from septic mice showed increased PAD activity, which strongly correlated with the level of citrullination, indicating that PAD-driven protein modification occurs in vivo. Therefore, we hypothesized that PAD-mediated citrullination of the arginine residues within LL-37 will abrogate its immunomodulatory functions. As reported previously using Mass-spectrometry approach we were analysing synovial fluids to map the presence citrullination and carbamylation on peptides and proteins. This work is still ongoing.

Production of recombinant PADs
Despite using an array of different expression plasmids and strains of Escherichia coli all attempts to produce PAD2 in bacteria failed to deliver in a reproducible manner any significant amounts of the enzyme. The low yield of PAD2 expression was also a problem in the Pichia pastoris expression system. In this case only a minute amount of the enzyme were obtained and heroic efforts to optimize the expression were met with failure. To add insult to injury the purified protein had low specific activity. Therefore, in the last attempt we cloned the PAD2 coding sequence into expression vectors used in the eukaryotic Leishmania tarantolae expression system. First constructs yielding expression of PAD2 with C- and N-terminal hexahistidine tag will be tested. This work is in progress and no results are yet available.
We have attempted to express PAD4 in P. pastoris, Leishmania tarantolae and human cell-line 297T to improve the yield of expression and obtained homogenous preparations of the enzyme of high specific activity. Unfortunately, in vain trials to express PAD4 in the yeast system turned unsatisfactory.
A trial using L. tarantolae did not generate enough of the recombinant protein despite correct genetic constructs, as verified by double-stranded sequencing. In the case of expression in human cells we have failed to obtain a genetic construct of PAD4 coding sequence with fused sequence encoding the Fc tag necessary for recombinant protein purification. Due to these failures we have carried out of our research using rPAD4 produced in initially worked-out E. coli expression system. Purified GST-PAD4 was purified to homogeneity and had high specific activity. The results obtained using the tagged enzyme were verified in a set of experiments in which purified in minute amounts rPAD4 was applied.
In addition, major progress was achieved with human peptidylarginine deiminase 4 (PAD4), which was successfully overexpressed in E. coli strain BL21 (DE3) cells with the C-terminal hexahistidine tag. The recombinant enzyme was purified by affinity chromatography on Ni-Sepharose, followed by ion-exchange chromatography and a mass-exclusion chromatography.

PPAD and other proteins of interest
During the duration of the project we produced, crystallized and solved the structure of the large number of proteins. These included proteins relevant to rheumatoid arthritis, bacterial infection and periodontitis. In some cases, in which human proteins or those from odontopathogens could not be obtained, orthologues from other species were successfully assayed. This is common practice in structural studies, as the structures of close orthologs are very similar and allow drawing conclusions as to mechanism and function. These were:
• Human mitochondrial transcription factor A (TFAM)
• Drosophila melanogaster carboxypeptidase D splicing variant 1B (CPD-1B)
• Human thrombin-activatable fibrinolysis inhibitor (TAFI)
• Rhipicephalus bursa tick-derived carboxypeptidase inhibitor (TCI)
• Tannerella forsythia karilysin catalytic domain (Kly18)
• Bacteroides fragilis (pro)fragilysin-3 (Bft-3)
• Staphylococcus aureus HmrA (HMRA)
• Bacillus thermoproteolyticus thermolysin (TL)
• Galleria mellonella metalloproteinase inhibitor (IMPI)
• Human alpha-2-macroglobulin (A2M)

Throughout last year, we expanded the list of proteins directly or indirectly relevant to rheumatoid arthritis and bacterial infection and periodontitis. We made substantial progress and reported the production of large amounts of purified material and crystals suitable for X-ray diffraction of 11 proteins and complexes. These were:

• Methanococcus jannaschii projannalysin-1
• Pyrococcus abyssi proabylysin
• Staphylococcus aureus MecR2
• Porphyromonas gingivalis gingipain R in complex with its pro-domain
• Tannerella forsythia karilysin in complex with a peptidic inhibitor
• Geobacillus lithuanicus U32 peptidase C-terminal domain
• Streptomyces caespitosus sermetstatin
• S. caespitosus sermetstatin in complex with Bacillus subtilis subtilisin
• S. caespitosus sermetstatin in complex with S. caespitosus snapalysin
• Human meprin beta
• Human meprin beta zymogen

In addition, significant progress was made with the structure determination of P. gingivalis peptidylarginine deiminase (PPAD), which we produced as purified material in large amounts. We encounter several obstacles before reaching our goal of 100mg of protein suitable for crystallization. This part of the projects was divided in two items: (i) identification and optimisation of the most appropriate conditions for protein crystallisation and (ii) collection and examination of the formed crystals in synchrotron radiation facilities to obtain appropriate data for determination of a structural model.
At this point, three different protein preparations had been delivered and, depending on their purity, they were further purified by ion exchange chromatography or size exclusion chromatography.
An intitial screen of PDB database for already determined structures of homologs showed that there wasn't any other model to be used for solving PPAD structure by molecular replacement. Therefore, a SAD experiment (Single Anomalous Diffraction) was performed in order to determine the phases necessary for solving the structure. Around 25 crystals were derivatised with heavy atoms (e.g. mercury acetate, mercury chloride, tetrakis(acetoxymercury) methane and phenyl mercury acetate) and tested as before for diffraction. In this case, crystals derivatised with phenyl mercury acetate gave the best diffraction and anomalous signal.
Finall, we made the most relevant achievement of the project with the structure determination of PPAD. In addition, we also obtained the structures of PPAD complexes with substrates, inhibitors, and products, which helped us to determine the exact working mechanism of the protein and the design of specific inhibitors.

In addition to this main project, we also made substantial progress and report hereby the production of large amounts of purified material and crystals suitable for X-ray diffraction of several proteins and complexes, as well as the structures of 2 of them. These are:
• Porphyromonas gingivalis gingipain K in complex with a covalent inhibitor
• Methanococcus jannaschii selecase
• Several α2-macroglobulin forms

Altogether, the funding provided by the present project has enabled us to push the boundaries of our knowledge at the molecular level of key players in rheumatoid arthritis and bacterial infection and periodontitis. This information is expected to contribute to the design of novel therapeutic agents in the future, thus contributing to the relief of the European population affected by these health issues.

PAD expression in primary neutrophils.
Neutrophils were isolated from subjects with chronic periodontitis were recruited from patients referred to the periodontal department of Birmingham’s Dental Hospital.
In our analyses, PAD4 was expressed at increased levels in Fn stimulated cells at both 4 hrs and 24hrs, potentially reflecting its role in NET formation in response to bacterial stimulation. Interestingly expression levels were however decreased in patient samples (stimulated and unstimulated) compared with controls and when treated with IFN, which is found at higher levels in patient plasma.

This decrease in expression was also seen for PAD2 in patient samples compared with controls and IFNa treated cells. These decreases may be a reflection of PADs role in inflammation and its attenuation. PADs are reported to be both antibacterial and anti-inflammatory and PADs -2 and -4 have been shown to citrullinate IL8 which results in decreases in neutrophil recruitment from tissues and thus dampens tissue inflammation (Proost et al. 2008). Interestingly PAD2 activity has also been shown to reduce NF-κB activity in RAW 264.7 macrophages (Lee et al, 2010) and subsequently may also modulate inflammation via this mechanism.

CSE treatment of cells, that has previously been shown to ‘stress’ neutrophils and induce ROS production, resulted in decreased PAD2 and 4 expressions. This may be due to changes caused within the cells by the CSE or reflect the pro-inflammatory state of the cells.

As neutrophils, in particular upon release of DNA in the form of neutrophil extracellular traps (NETS) are discussed as a source of citrullinated proteins, we have initially investigated levels of free extracellular DNA in the synovial fluid of RA patients.
Quantification of untreated, freshly obtained extracellular DNA in SF of patients with a range of arthritides confirmed and extended previous findings that free extracellular DNA can be found in RA SF. Significantly higher DNA levels were found in SF of patients with RA compared to patients with osteoarthritis (OA) (p<0.001) and psoriatic arthritis (PsA) (p<0.01). Validation experiments in which the DNA signal of untreated, fresh SF samples was measured before and after centrifugation of cells showed that about 84% of the signal derived from cell-free SF samples suggesting that necrotic cells with permeable membranes contributed only a marginal proportion of the DNA signal.

Levels of free DNA in SF samples showed a strong correlation not only with concentrations of total SF cells but also, more importantly, with neutrophil cell counts in the synovial fluid of RA patients, whereas PsA SF samples showed a trend of low neutrophil concentrations. In contrast, no correlation with macrophage cell counts could be observed - cells which were also reported to release extracellular traps (ETs) containing DNA.
In frozen sections of synovial tissue, neutrophil aggregates were observed on the surface of the synovial lining, facing the joint cavity. DNA staining revealed extracellular DNA associated within these aggregates. Immunostaining of smear preparations of RA SF showed a network of extracellular DNA strands similar to that reported for NETs in a range of studies. DNA co-localized with neutrophil elastase, a major constituent of NETs. These results are consistent with the concept that extracellular DNA and NETs are associated with the localization and number of neutrophils present in the SF of patients with inflammatory arthritis.

In vitro PAD release during NETosis
Since NETosis involves release of cellular components, we investigated whether neutrophils undergoing NETosis would release PAD into the extracellular space. The release of unbound PAD into the supernatant and its attachment to NETs were thus both assessed. For this purpose, an in vitro assay of NET isolation and detection was developed based on a previously published method. Both PAD2 and PAD4 were detected in the supernatant and in the DNase-I treated NET fraction, suggesting that PAD2 and PAD4 are both released freely diffusible as well as bound to NETs. A small number of mainly granular and nuclear proteins could be confirmed to be present in NETs compared to the wide range of proteins released into the supernatant during NETosis. In addition, both PAD2 and PAD4 could be detected in both the supernatant and in the DNase-I treated NET fraction from 7 patients: while PAD2 was more abundant in the supernatant, PAD4 was more abundant in the NET fraction. To test whether the activation of PAD isoforms during NETosis would result in the generation of citrullinated proteins, proteins from culture supernatants were chemically modified and detected using an anti modified citrulline antibody (AMC), a large number of citrullinated proteins are released from PMA-stimulated neutrophils into the supernatant after 4 hours compared to only small numbers of citrullinated proteins present in the DNase-I treated NET fraction The distribution of PAD4 in stimulated neutrophils was visualized by immunofluorescence staining. PAD4 localization in resting peripheral blood neutrophils was found to be restricted to the cell nucleus. Upon stimulation a proportion of the cells had changed their nuclear morphology. Nuclei were rounded and decondensed indicating a stage of NETosis directly preceding DNA release. At this stage PAD4 was not detectable in the nuclei.
These results suggests that neutrophils going into NETosis in vitro can release enzymatically activate PAD which are able to citrullinate a large number of neutrophil derived proteins de novo during cell activation.

PAD in the synovial fluid of RA patients are enzymatically active and correlate with DNA and neutrophil concentrations
PAD2 and PAD4 proteins were detected in the cell-free SF of patients with RA in addition to the presence of neutrophil elastase. Whereas PAD4 seemed to vary in its expression pattern between patients, the amount of PAD2 was found to be more consistent. Furthermore, PAD enzymatic activity was significantly higher in the SF of RA patients than in that of OA patients at supraphysiological calcium concentrations (p<0.05). Interestingly, this difference could also be observed with total, cell-free and non-diluted SF samples, however at about 250 fold lower PAD activity in the RA SF samples when accounting for the dilution. Furthermore, PAD activity strongly correlated not only with the level of extracellular DNA in the SF (r = 0.8; p < 0.001) but also with neutrophil cell counts (r = 0.8; p = 0.002) in untreated, fresh SF samples.

PPAD abolishes anaphylatoxin C5a activity
PPAD induced citrullination of a critical C-terminal arginine of the anaphylatoxin C5a disabled the protein function. Treatment of C5a with PPAD in vitro resulted in decreased chemotaxis of human neutrophils and diminished calcium signaling in monocytic cell line U937 transfected with the C5a receptor (C5aR) and loaded with a fluorescent intracellular calcium probe: Fura-2 AM. Moreover, a low degree of citrullination of internal arginine residues by PPAD was also detected using mass spectrometry. Further, after treatment of C5 with outer membrane vesicles naturally shed by P. gingivalis, we observed generation of C5a totally citrullinated at the C-terminal Arg-74 residue (Arg74Cit). In stark contrast, only native C5a was detected after treatment with PPAD-null outer membrane vesicles. Our study suggests reduced antibacterial and proinflammatory capacity of citrullinated C5a, achieved via lower level of chemotactic potential of the modified molecule, and weaker cell activation. In the context of previous studies, which showed crosstalk between C5aR and Toll-like receptors, as well as enhanced arthritis development in mice infected with PPAD-expressing P. gingivalis.
The role of PPAD in proinflammatory reaction of gingival fibroblasts
To investigate the contribution of PPAD to the interaction of P. gingivalis with primary human gingival fibroblasts (PHGF) and P. gingivalis-induced synthesis of prostaglandin E2 (PGE2), PHGF were infected with wild-type P. gingivalis ATCC 33277, an isogenic PPAD-knockout strain (∆ppad) or a mutated strain (C351A) expressing an inactive enzyme in which the catalytic cysteine has been mutated to alanine (PPADC351A). Cells were infected in medium containing the mutants alone or in medium supplemented with purified, active PPAD. PHGF infection was assessed by colony-forming assay, microscopic analysis and flow cytometry. Expression of cyclo-oxygenase 2 (COX-2) and microsomal PGE synthase-1 (mPGES-1), key factors in the prostaglandin synthesis pathway, was examined by quantitative reverse transcription polymerase chain reaction (qRT-PCR), while PGE2 synthesis was evaluated by enzyme immunoassay. PHGF were infected more efficiently by wild-type P. gingivalis than by the ∆ppad strain, which correlated with strong induction of COX-2 and mPGES-1 expression by wild-type P. gingivalis, but not by the PPAD activity-null mutant strains (Δppad and C351A). The impaired ability of the Δppad strain to adhere to and/or invade PHGF and both Δppad and C351A to stimulate the PGE2 -synthesis pathway was fully restored by the addition of purified PPAD. The latter effect was strongly inhibited by aspirin. Collectively, our results implicate PPAD activity, but not PPAD itself, as an important factor for gingival fibroblast infection and activation of PGE2 synthesis, the latter of which may strongly contribute to bone resorption and eventual tooth loss.

PPAD inactivates LL-37 bactericidal activity
Arginine residues contribute to cationic character of LL-37 and their enzymatic deimination to the neutral citrulline residue strongly affects bactericidal activity of the peptide. We found that preincubation of LL-37 with PPAD hindered ability of LL-37 to kill E. coli in concentration- and time-dependent manner. Our results were supported by SEM imaging observation showing no effect of PPAD-pretreated LL-37 on bacterial morphology. Collectively there results clearly show that PPAD can efficiently inactivate the LL-37 antibacterial activity. Described result states a part of manuscript entitled A novel pathway to alter function of antibacterial peptides: Citrullination abates bactericidal activity of LL-37 by Sroka et al.
Our studies towards this goal have focused on elucidation of a suggested causative link between periodontal infection and rheumatoid arthritis via bacteria-dependent induction of a pathogenic autoimmune response to citrullinated epitopes. To this end we have studied in more details the role of P. gingivalis in the pathogenesis of RA in mice of different genetic background (C57BL/6, BALB/c, and DBA/1) and inoculation site using chamber, oral and intratracheal infection models. Joints inflammation was induced only in DBA/1 (see table I). The inflammation was characterized by influx of neutrophils into affected joints.
Partner 8 (GLAS) obtained PAD4 deficient mice (kind gift of Kerri Mowen, The Scripps Research Institute, La Jolla, CA), to allow study of endogenous PAD enzymes. PADs are vital for normal physiological processes, with roles related to skin physiology, immune function and gene regulation. However, they have also been associated with several chronic immune-mediated diseases such as rheumatoid arthritis. The PAD4 enzyme is mainly expressed in cells of the hematopoietic lineage and may play an important role in host defence against bacterial infections due its role in neutrophil extracellular traps (NETs) formation. We confirmed a deficiency in NET production in PAD4 deficient neutrophils. We demonstrated that animals lacking PAD4 respond normally following infection with P. gingivalis in terms of anti-P. gingivalis IgG antibody production, and B and T cell activation, and appear to exhibit similar levels of alveolar bone loss to normal control mice (Figure 9), although the differences in bone loss between infected and control groups are less as the mice are on a C57Bl/6 background which are less susceptible to oral infection induced periodontitis than BALB/c animals.

To assess the contribution of PAD4 activity to rheumatoid arthritis initiation, and how periodontitis might influence arthritis in the absence of PAD4, control C57BL/6 or PAD4 deficient mice were orally infected with P. gingivalis W83 or sham infected with carrier alone, and subsequently experimental arthritis initiated. The previously used T cell transfer/breach of tolerance model of arthritis was chosen to investigate the early stages of arthritis, with a T cell induced breach of immune tolerance. Moreover, this model is more reliable than CIA in C57Bl/6 mice – which is the background of the PAD4 deficient animals). Following initiation of arthritis alone (not in combination with periodontitis), there was a slight, statistically insignificant trend suggesting an mild exacerbation of experimental arthritis in PAD4 knockout mice compared with wild type mice, in terms of footpad swelling and clinical score. However, the immunohistological analysis of the affected paws, the alveolar bone loss analysis and the serum anti-P. gingivalis and anti-OVA IgG antibody titres were similar in knock out and wild type animals. Concurrent oral infection with P. gingivalis had no effect on any measures of arthritis in either C57Bl/6 or PAD4 deficient mice, and although the infected animals demonstrated anti-P. gingivalis serum antibody responses there were otherwise no differences in the antibody responses.
Experimental arthritis (EA) was induced by adoptive transfer of Th1 ovalbumin specific T cells followed by immunisation with ovalbumin. Experimental arthritis was then induced by subcutaneous injection of heat aggregated ovalbumin or PBS into the footpad. The experiments were carried out in either Wild-type C57BL/6 mice or PAD4 deficient. In another experiment, the induction of EA was preceded by infection with P. gingivalis W83 (PD) two weeks prior to arthritis induction.
These findings suggest that there is limited contribution of PAD4 to the progression of periodontitis in isolation, and limited contribution of endogenous PAD4 to the initiation of arthritis, at least in these animal models. To a large extent these findings echo those throughout these studies, that there are a variety of sources of citrullinated antigens in inflammation and there are complex mechanimsm mediating the breach of immune tolerance to these antigens.

Synthesis of inhibitory compounds.
The crystal the crystal structure of PPAD, which had not been available to the consortium previous to the initiation of this grant, was solved. The active site of PPAD was determined; it is a targetable pocket accessible from the surface. Based on the PPAD crystal structure, we performed in silico screens using proprietary software and the in house data base containing more than 6 million commercially available compounds. As done for hPAD, also a pharmacophore model was built on ligand/substrate information from literature and another set of in silico screens was carried out using this information. Hit lists were analysed and structures predicted to inhibit PPAD activity were clustered into chemical classes. A selection of the 1000 best in silico compounds was evaluated by a medicinal chemist and binding poses in the active site were analysed. From this pool of compounds altogether 220 were delivered for testing in the in vitro assay.
Obtained compounds inhibited hPAD4 more potently than the reference compound and thus represent another chemical class for further medchem optimisation. hPAD inhibitors were synthesised from several chemical classes (guanidine mimetics and pyrrologuanidines, various sulfonamides, benzylamides, methylene-thiazolidines, peptidomimetics, and non-electrophilic compounds) in a hit-to-lead program for identifying compounds suitable for in vivo examination in a proof-of-concept animal model, serving then as starting point for a subsequent lead optimisation program. Several compounds were upscaled for closer investigation of their physico-chemical, toxicological and pharmacokinetic properties and one compound was finally produced in the multi-gram scale for an in vivo efficacy model.
Partner 4 also provided compounds for testing on the Protein Arginine Deiminase from Porphyromonas gingivalis, whereas partner 11 worked on compounds active against the carbonic anhydrase of Porphyromonas gingivalis, a new potential target enzyme from this bacterium.
Based on the solved crystal structure of PPAD, high-throughput in silico screenings for PAD inhibitors were performed. 220 in silico hits were acquired and tested for PPAD inhibition in the in vitro assay established. Three compounds showed significant reduction of PPAD activity and will serve as starting points for a subsequent hit-to-lead medchem program. Compounds selected by structure based in silico selection from an in-house database were evaluated on their inhibitory efficacy against carbonic anhydrase, a new target from P. gingivalis.
Lead compounds selected within the scope of the previous tasks underwent further testing in studies directly evaluating their toxicological, physico-chemical, and pharmacokinetic properties. After in vitro ADME and toxicological characterisation (e.g. cellular toxicity, microsomal stability) suitable substances were selected for PK studies in rodents. The best two compounds were identified as leads within their chemical classes and were applied in an appropriate formulation orally and intraperitoneally, respectively, to female Swiss mice.

Bioinformatics processing.
A human protein interaction maps of RA and PD have been built and the comparative studies of RA and PD protein networks have been performed and the necessary specific tools for this specific data analysis and project data integration were developed.
The network around the key proteins of RA, PD and AS has been modelled to be able to overcome the short comes of topological measurements. In the current analysis, models have been built applying machine learning supervised methods. The model generated has been queried by applying a perturbation on PD key proteins, consisting on activation or inhibition of the protein depending on which state is causing /contributing to the development of PD, and measuring how the key RA proteins are affected. From the pool of possible solutions we have selected to extreme clusters: i) the mechanisms that lead to a major affectation of RA proteins and ii) the mechanisms that lead to a minor affectation of RA proteins. The two groups of mechanistic solutions were compared to identify proteins with differential activity, that is, proteins that are more or less affected (activated / up-regulated or inhibited / down-regulated) in RA+ vs. RA-, as these proteins have a good chance of being potential biomarkers of the transition from PD to RA. This analysis identified 99 protein candidates that were reduced to 53 after removing those proteins that also appear when doing the same exercise but with AS as response. Out of the 53 candidates, 23 were previously identified in the overlap of the PD and RA networks and 30 are new ones. Disagreement between the predictions and the literature discarded 17 more candidates. Finally, 36 biomarker candidates have been identified. Supporting bibliographic evidences exist for some of the candidates. Although, existing data sets of gene expression are not suitable for validation, differential expression for 15 of the candidates has been observed in human peripheral blood cells of 114 RA individuals.
In conclusion, 36 biomarker candidates linking PD to RA have been identified using systems biology approaches. We could find supportive bibliographic evidences for some of the candidates. Although, existing data sets of gene expression are not suitable for validation, differential expression for 15 of the candidates has been observed in human peripheral blood cells of 114 RA individuals.

Potential Impact:
Periodontitis is a prevalent disease with 30% of the world population suffering from various levels of the disease, where 8% ultimately will lose their teeth. The implications of the disease are in many cases underestimated as the loss of teeth is not only debilitating and creating a social stigma but it also results in a substantial cost for the affected individual. Dental care for adults is, in the majority of cases, not covered by the national health care systems in Europe. This may lead to individuals not receiving prophylactic dental care to the extent that would be needed, and in the case of periodontitis when specialist care is needed the costs to be paid by the patient accumulate rapidly.
The importance of regular dental care is gaining focus as periodontitis to a greater extent has been found to be linked to several systemic inflammatory diseases. The Gums & Joints consortium has during the lifetime of the project investigated the possible role of P. gingivalis (oral bacteria associated with periodontitis) on the development of rheumatoid arthritis.
It has previously been found that patients with long-standing, active RA have a substantially increased frequency of periodontitis as compared with healthy subjects. Vice versa, patients with periodontal disease have higher prevalence of RA than patients without periodontitis. The expression of the peptidyl-arginine deiminase (PAD) enzyme by P. gingivalis (unique for this bacterium) (PPAD), which is involved in citrullination, may constitute a mechanistic link explaining the clinical correlation between RA and periodontitis where protein citrullination by P. gingivalis may expose new protein epitopes. The state of the teeth and surrounding gums is no longer just associated to the oral cavity but it is clear that it has a larger systemic effect. If treatment, or even prevention, of periodontitis can reduce the risk of RA or in the case of diagnosed arthritides, reduce the severity, then enormous gain may be obtained. Treatment of RA is not only costly but also life-long and any improvements in the early stage of the disease may not only change the course of the disease but also result in a lower financial burden for the society (cost of treatment as well as absence from work life).

The consortium has shed light on molecular pathways where P. gingivalis may contribute to the breach of tolerance in autoimmune arthritis, and investigated the connections between the two diseases using large patient cohorts. Further, large efforts have been put into purification of recombinant PAD as well as development of PAD and PPAD inhibitors. In addition, P. gingivalis carbonic anhydrase has been identified as a new possible drug target.

One of the main goals of the Gums & Joints consortium was to raise the public awareness and through the publication of over 70 peer-reviewed papers and several interviews in daily newspapers/monthly magazines the results and message of the consortium has reached a broad audience. In addition to in-print dissemination efforts has been made to reach the young adult population through participation in popular scientific television program, use of Facebook page, as well as organisation of several workshops.

In detail, the consortium has reached the scientific community through 77 peer-reviewed articles, 44 poster presentations, and 86 oral presentations. Further, efforts has been made to specifically reach the civil society and with four articles published in Scandinavian journals and one published in the French magazine Science & Vie. The project was also covered on the popular TV-program Schrödinger’s cat, broadcasted on the Norwegian NRK channel.
In March 2014 the Gums & Joints consortium successfully arranged a conference “Arthritis, Infections and Autoimmunity: Infections as an etiological factor in chronic inflammatory diseases with close to 200 attendees, many of them being immunology and biochemistry students. Later in the same year the consortia was able to organize additionally 2-day open workshop that took place in Krakow on October 16th-18th 2014. Workshop was entitled “On the Edge: Periodontal and autoimmune diseases”. In total there was 11 invited speakers and 56 people attending.

Related documents