Article Text
PDF
Abstract
Objectives An increased prevalence of periodontitis and perturbation of the oral microbiome has been identified in patients with rheumatoid arthritis (RA). The periodontal pathogen Porphyromonas gingivalis may cause local citrullination of proteins, potentially triggering anti-citrullinated protein antibody production. However, it is not known if oral dysbiosis precedes the onset of clinical arthritis. This study comprehensively characterised the oral microbiome in anti-cyclic citrullinated peptide (anti-CCP) positive at-risk individuals without clinical synovitis (CCP+at risk).
Methods Subgingival plaque was collected from periodontally healthy and diseased sites in 48 CCP+at risk, 26 early RA and 32 asymptomatic healthy control (HC) individuals. DNA libraries were sequenced on the Illumina HiSeq 3000 platform. Taxonomic profile and functional capability of the subgingival microbiome were compared between groups.
Results At periodontally healthy sites, CCP+at risk individuals had significantly lower microbial richness compared with HC and early RA groups (p=0.004 and 0.021). Microbial community alterations were found at phylum, genus and species levels. A large proportion of the community differed significantly in membership (523 species; 35.6%) and structure (575 species; 39.1%) comparing CCP+at risk and HC groups. Certain core species, including P. gingivalis, had higher relative abundance in the CCP+at risk group. Seventeen clusters of orthologous gene functional units were significantly over-represented in the CCP+at risk group compared with HC (adjusted p value <0.05).
Conclusion Anti-CCP positive at-risk individuals have dysbiotic subgingival microbiomes and increased abundance of P. gingivalis compared with controls. This supports the hypothesis that the oral microbiome and specifically P. gingivalis are important in RA initiation.
- rheumatoid arthritis
- anti-CCP
- early rheumatoid arthritis
Statistics from Altmetric.com
Key messages
What is already known about this subject?
Patients with rheumatoid arthritis have increased periodontal disease and a perturbed oral microbiome. The periodontal pathogen Porphyromonas gingivalis is able to citrullinate proteins via its peptidylarginine deiminase enzyme and can generate citrullinated antigens that may drive the autoimmune response in RA.
Periodontitis and P. gingivalis were increased before joint inflammation in individuals at risk of RA, supporting the concept of periodontal inflammation and P. gingivalis as important risk factors in RA initiation.
What does this study add?
This is the first study to demonstrate dysbiosis, including an increase of P. gingivalis, in the periodontally healthy microbiome (and altered diseased subgingival microbiomes) of individuals at risk of developing RA compared with healthy controls.
How might this impact on clinical practice or future developments?
Our results indicate that dysbiosis in the subgingival microbiome precedes the onset of joint inflammation in at-risk individuals. This dysbiosis, together with the increase of P. gingivalis, may play an important role in the initiation of RA.
Taken together with our previous findings, periodontal disease and the observed oral dysbiosis could be targets for future preventive interventions in individuals at risk of RA. Investigation of the overall metabolic capability of the subgingival microbiome may provide novel insights into the pathogenesis of RA.
Introduction
Individuals at-risk of rheumatoid arthritis (RA) often have anti-citrullinated protein antibodies (ACPA) well before the development of joint inflammation.1 2 Where the initiation of RA autoimmunity occurs is a critical question with significant implications for future preventative strategies. Recent data have implicated mucosal sites and the local microbiome and there has been considerable focus on the role of the oral mucosa and periodontium.3 4
There is an increased prevalence of periodontitis in patients with both early and established RA.5 The subgingival microbiota in periodontitis, in particular the periodontal pathogens Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans, may play a critical role in RA pathogenesis; P. gingivalis by contributing to ACPA production through citrullination of proteins via its peptidylarginine deiminase enzyme (PAD) and A. actinomycetemcomitans by inducing leukotoxic hypercitrullination.6–8 We recently reported an increased prevalence of periodontal inflammation and P. gingivalis in anti-cyclic citrullinated peptide (anti-CCP) positive at-risk individuals without arthritis (CCP+at risk), supporting the concept that periodontal inflammation and P. gingivalis precede joint inflammation, as important risk factors in RA initiation.9 A. actinomycetemcomitans did not emerge as similarly significantly associated with at-risk individuals; A. actinomycetemcomitans is particularly important in severe generalised periodontitis,10 which we did not see in our cohort.
Periodontitis is a complex disease, mediated by consortia of co-operating bacteria and the host responses to them. While P. gingivalis is a keystone pathogen that increases the risk of periodontitis, it depends on the activities of other members of the subgingival microbiome to establish within the community and express full virulence. Thus, to fully understand the role of periodontitis in RA pathogenesis, it is important to study the entire bacterial community. Although certain taxa and compositional and functional alterations were identified in RA-associated oral microbiomes,11–13 it is difficult to clarify the cause and effect of these findings once clinical arthritis has developed. Furthermore, RA treatment is also likely to influence the oral microbiome.12
We therefore sought to comprehensively characterise the oral microbiome in CCP+at risk individuals without clinical arthritis; we aimed to report differences in the metagenomes, characterised by a shotgun metagenomic approach, sampled from periodontally healthy and diseased subgingival sites of CCP+at risk individuals, patients with early RA and healthy controls (HCs).
Materials and methods
HCs, CCP+at risk individuals with musculoskeletal symptoms but no clinical synovitis and patients with anti-CCP positive early RA (within the first 3 months of disease-modifying anti-rheumatic drug (DMARD) therapy) were recruited. The three groups were balanced for age, sex and smoking status (online supplemental table S1).9 Periodontal assessments and subgingival plaque sampling were performed by three experienced dentists.9 According to the latest Classification of Periodontal Diseases and Conditions, periodontally healthy sites were defined as sites with ≤3 mm probing depth and no bleeding on probing.14 Diseased sites were those with ≥4 mm probing depth and ≥2 mm clinical attachment loss.15 Subgingival plaque samples from a maximum of three healthy and three diseased sites were analysed for each participant using shotgun metagenomics sequencing (Illumina HiSeq 3000). Microbial diversity and community composition were compared between three groups. Periodontitis is a dysbiotic disease, with significant differences comparing microbiomes from healthy and diseased subgingival sites. The term dysbiosis is also used here to describe microbiomes from healthy sites that are distinct in composition from those of healthy sites from the HC group. Further details are given in the online supplemental material.
Supplemental material
Results
Microbial diversity
Within periodontally healthy sites, the CCP+at risk group showed a significantly lower Abundance Coverage Estimator value compared with the HC group (p=0.004) and the early RA group (p=0.021), indicating decreased estimated microbial richness of the subgingival microbiome (figure 1).
Comparison of α-diversity in healthy control (HC), CCP+at risk and early RA groups using samples from periodontally healthy sites and diseased sites. Abundance Coverage Estimator (ACE) index was significantly decreased in the CCP+at risk group compared with the HC group in periodontally healthy sites (Kruskal-Wallis test). CCP, cycliccitrullinated peptide; RA, rheumatoid arthritis.
Bacterial community composition
Overall, 28 bacterial phyla, 593 genera and 1472 species were identified. Significantly altered community composition was found in the CCP+at risk group at different taxonomic levels. In periodontally healthy sites, phylum Synergistetes was found with significantly higher relative abundance in the CCP+at risk group compared with other groups (online supplemental figure S1a).
Among the top 20 most predominant genera in periodontally healthy sites (figure 2A), Bifidobacterium and Porphyromonas were present with significantly increased relative abundance in the CCP+at risk group (p=0.027, 0.033). In pairwise comparison, 523 species (35.6% of the community) differed significantly in membership and 575 species (39.1%) differed significantly in structure, comparing the CCP+at risk and HC groups. Less difference was found in the community membership (62 species, 4.2%) and structure (42 species, 2.9%) comparing the early RA and HC groups (figure 3A). Certain significant differences were also found between groups in periodontally diseased sites, for example, the abundance of phylum Chlorobi was increased in the HC group compared with other groups (online supplemental figure S1b) (corrected p<0.05). The genus Porphyromonas was significantly higher in the CCP+at risk group compared with other groups (p=0.015), and Capnocytophaga, Cardiobacterium, Neisseria and Streptococcus were significantly more abundant in the early RA group (p=0.009, 0.003, 0.024, 0.003) (figure 2B). At species level, only 1.4% and 5.7% of the microbial community differed significantly in membership and structure between the CCP+at risk and HC groups (figure 3B).
Taxonomic profiles for the 20 most abundant genera in subgingival plaque from periodontally healthy and diseased sites in healthy control (HC), CCP+at risk and early RA groups. Relative abundance of the 20 most abundant genera within (A) periodontally healthy sites and (B) diseased sites was plotted for each group. The permutation test (one-sided signassoc function, indicspecies R-package) was used to find the genera with significantly different relative abundances between groups. * corrected p<0.05 (Sidak’s correction). CCP,cyclic citrullinated peptide; RA, rheumatoid arthritis.
Phylogenetic tree representing normalised mean relative abundance of species (stacked bar chart) in the subgingival microbiome of (A) periodontally healthy and (B) periodontally diseased sites (phylogenetic tree constructed using the webserver iTOL.embl.de). CCP,cyclic citrullinated peptide; HC, healthy control; RA, rheumatoid arthritis.
Core microbiome
The core microbiome, of which the species were present in at least 80% of the samples in each group, was used to compare stable associations between groups. Within periodontally healthy sites (figure 4A), 81 species were identified in the core microbiome of all study participants. The core microbiome from the CCP+at risk group was much less diverse than that of the HC or early RA group. There was no core species exclusively belonging to the CCP+at risk group, unlike the HC and early RA groups, which had 35 and 79 exclusive core species, respectively. In the periodontally diseased sites (figure 4B), 42 species were found in the core microbiome of all groups. Importantly, 6, 2 and 190 species were identified as uniquely belonging to the HC, CCP+at risk and early RA core microbiomes, respectively (online supplemental tables S2-S3). Certain species were significantly more abundant in each group compared with the other groups within periodontally healthy or diseased sites (online supplemental table S4). In particular, within both periodontally healthy and diseased sites, Arthrobacter chlorophenolicus and P. gingivalis were significantly more abundant in CCP+at risk individuals.
Overlap analysis of the group-specific and shared core species.Core species in each group of periodontally healthy and diseased site samples were identified, respectively (>80% prevalence). Number of group-specific and shared core species were visualised for (A) healthy sites and (B) diseased sites. CCP,cyclic citrullinated peptide; HC, healthy control; RA, rheumatoid arthritis.
Bacterial co-occurrence networks in subgingival microbiomes
In periodontally healthy sites, Spearman’s correlation analysis identified 347, 83 and 1024 edges as strong (q<−0.7 or >0.7) and significant (corrected p<0.01) pairwise correlations between nodes (species) in each the HC, CCP+at risk and early RA groups, respectively (online supplemental figure S2). In periodontally diseased sites, there were 49, 139 and 365 edges identified in HC, CCP+at risk and early RA groups, respectively (online supplemental figure S3). The edge/node ratio (density) of the network represents the number of co-occurrence instances in a microbial community; in the early RA group, this was higher than that of other groups in both periodontally healthy and diseased sites, reflecting a dysbiosis of the subgingival microbiome in early RA patients (online supplemental table S5).
To gain deeper insights into the differences between groups, the hubs in each network were identified by ranking the top 20 nodes with the maximal clique centrality (MCC) algorithm. In the periodontally healthy sites (figure 5A), the cluster of Neisseria spp. by which the network of HC group was dominated was not found in the hubs of other groups. Species including Filifactor alocis, Campylobacter rectus, Porphyromonas endodontalis and Treponema vincentii formed the network hubs for both HC and CCP+at risk groups, while the early RA group showed entirely different network hubs. Within the periodontally diseased sites (figure 5B), Actinomyces viscosus and Actinomyces urogenitalis were identified in the network hubs of all groups, indicating an implication in the development of periodontal disease irrespective of RA status. Intriguingly, the periodontal pathogen A. actinomycetemcomitans, which may also initiate protein citrullination in RA, was one of the hubs of the early RA group.
Identification in plaque from periodontally healthy and diseased sites of hubs in the networks of healthy control (HC), CCP+at risk and early RA groups. The top 20 nodes (species) ranked by maximal clique centrality were displayed in circular layout for each group from (A) periodontally healthy and (B) diseased site samples. Nodes are coloured based on rank; dark colour denotes high ranks. Green dashed line: HC; orange: CCP+at risk; blue: early RA. CCP,cyclic citrullinated peptide; RA, rheumatoid arthritis.
Functional capabilities of subgingival plaque microbiomes
Abundances of 3034 clusters of orthologous genes (COGs) functional units were normalised and compared between groups. Within periodontally healthy sites, 17 functional units were significantly over-represented in the CCP+at risk group compared with the HC group and 5 functional units were significantly over-represented in the early RA group compared with the HC group (online supplemental table S6) (corrected p<0.05). In periodontally diseased sites, significant differences were found comparing the early RA group with the HC and CCP+at risk groups (online supplemental table S7). The functional unit of ‘PAD and related enzymes’ were detected in 65.6%, 68.8% and 69.2% of samples in the HC, CCP+at risk and early RA groups from periodontally healthy sites and in 55.6%, 69.2% and 56.3% of each group from diseased sites. No significant difference was found in the normalised counts between groups either in periodontally healthy or in diseased sites (figure 6).
Normalised count of peptidylarginine deiminase enzyme (PAD) and related enzymes in healthy control (HC), CCP+at risk and early RA groups using samples from periodontally healthy sites and diseased sites. Abundance of PAD and related enzymes was normalised by sequencing depth and compared between groups using the Wald test in DESeq2 R package. No significant difference was found between groups either in (A) periodontally healthy or in (B) diseased sites (corrected p>0.05). CCP,cyclic citrullinated peptide; RA, rheumatoid arthritis.
Discussion
Although intensively studied, the mechanisms of disease initiation and development of autoimmunity in RA are still unclear.16 ACPA are highly specific for RA and can be detected years before joint inflammation, suggesting a preclinical phase of RA, which could be a window of opportunity for disease prevention.17 We previously showed that periodontitis and P. gingivalis were increased before clinical or subclinical joint inflammation in individuals at risk of RA.9 Other studies have identified increased periodontitis in the first-degree relatives of patients with RA.18 19 Compared with HCs, the alterations in the subgingival microbial community of patients with RA has been reported in different studies,11–13 suggesting a potential role of oral microbial dysbiosis in RA development. However, it is unknown if subgingival microbial dysbiosis precedes the onset of RA. The present study, to our knowledge, is the first comprehensive characterisation of the subgingival microbiome from both periodontally healthy and diseased sites in at-risk individuals. To preclude the effect of established periodontitis on the subgingival microbiome, analysis was performed on the samples from shallow gingival sulci (3 mm depth or less) with no bleeding on probing. This study comprised a relatively small sample size but participant groups were well balanced for age, sex and smoking status. Other variables currently being investigated for possible associations with periodontal disease (eg, body mass index, race, alcohol, education level) may also influence the subgingival microbiome. Larger sample size will be needed to more completely define the role of the subgingival microbiome in the development and progression of RA.
In CCP+at risk individuals, significant alterations were found in the composition of the periodontally healthy subgingival microbiome at different levels, which distinguished this group from matched controls and patients with early RA. In agreement with the present study, compositional change of salivary microbiota and decreased microbial diversity were found in individuals at high risk for RA in a recent study.20
Most previous studies utilised 16S rRNA gene sequencing to analyse the oral microbiome of RA patients.11 13 20 However, a major limitation of this method is that only a single region of the bacterial genome can be sequenced and it is difficult to distinguish the species when their 16S rRNA gene sequences display high similarities.21 The present study utilised shotgun metagenomics, which has several advantages including more confident identification of bacterial species, increased detection of diversity and prediction of genes.22
P. gingivalis may contribute to RA aetiology via the citrullination of local antigens by its PAD.7 23 While some previous studies have examined the association between P. gingivalis, and established RA, few have looked at P. gingivalis in individuals at risk of RA. Studies determining levels of antibodies against P. gingivalis, or its virulence determinants, in HC, at-risk or established RA groups have been equivocal, possibly due to methodological and sampling differences.7 24–28 A recent study demonstrated decreased levels of P. gingivalis in the saliva of high-risk individuals compared with HCs using 16S rRNA gene sequencing.20 Analysis of the microbiome of saliva and supragingival dental plaque using shotgun sequencing revealed P. gingivalis to be enriched in HCs rather than patients with RA.12 In another study, periodontitis, but not the subgingival presence of P. gingivalis, was more prevalent in patients who later progressed to classifiable RA.29 de Smit et al 30 concluded that, while there was evidence that periodontitis may precede symptomatic RA, there was insufficient evidence to confirm a role specifically for P. gingivalis in disease progression. Thus, while the link between periodontitis and RA is established, the specific roles of P. gingivalis or its PAD have been less clear. Our data indicate that anti-CCP positive at-risk individuals have increased abundance of P. gingivalis compared with HCs.
A lower abundance of P. gingivalis as well as alterations in microbial composition and functional capability were found in the early RA group, which may be related to the inflammatory burden of RA. Lopez-Oliva et al 13 proposed RA may act as a condition shaping the subgingival microbiome, particularly promoting the growth of certain organisms. Moreover, these patients were receiving DMARDs, although for less than 3 months. It is likely that RA therapy, particularly drugs with additional antibacterial properties,31 32 can influence the subgingival microbiome. RA regimes with immunomodulatory effects may influence both the development of the subgingival microbiome and the progression of periodontitis.33 34 A recent shotgun sequencing study identified alterations in the oral microbiome in patients with RA, which were partially restored by DMARD treatment.12
The presence and abundance of PAD and related enzymes (the COG functional unit representing a family of orthologous protein-coding genes) were similar between groups. This is interesting given the differences that were observed between the groups in P. gingivalis abundance. Although P. gingivalis was once considered unique among prokaryotes in producing a PAD, PAD homologues were recently found in other Porphyromonas species.35 Thus, the PAD in the subgingival microbiomes may arise from a range of species, not all of which may express PAD at the levels and with similar activity to the P. gingivalis PAD. A recent study also reported variations in the active site of PAD detected in clinical isolates of P. gingivalis, one of which was associated with increased in vitro activity.36 Our data cannot reveal differences in the expression or activity of PADs, or P. gingivalis PAD specifically. Detailed comparison of the active P. gingivalis PAD site and potential enzyme activity in different groups related to RA status would be an important area for future work.
Other periodontal pathogens may also contribute to protein citrullination via routes different from P. gingivalis. The leukotoxin-A (LtxA) produced by A. actinomycetemcomitans has been implicated in inducing leukotoxic hypercitrullination, and exposure to A. actinomycetemcomitans was associated with ACPA.6 This species was not dominant in the present study; considerable variations in isolation rates of A. actinomycetemcomitans have been reported in the literature, which may be the consequence of geographical differences in prevalence and methodological differences.37 P. intermedia was recently reported to be associated with antibody responses to a novel citrullinated peptide related to RA,38 but abundance of this organism did not emerge in our analyses as different in the groups sampled. It is clear that the microbiome of these patients was highly perturbed compared with both HCs and CCP+at risk individuals and the influence of DMARDs and duration of therapy requires further consideration. Intriguingly, there were some species that have not previously been reported as abundant in the subgingival plaque of patients with early RA, for example, Neisseria gonorrhoeae (online supplemental table S4). This pathogen of the urogenital tract can adapt to display asymptomatic survival in the human nasopharynx and oropharynx, providing a potential reservoir for their further spread.39 40 There is evidence of widespread horizontal gene transfer in the genus Neisseria 41 and of commensal species sharing many gene sequences with closely related pathogenic species,42 and this may have impacted on our findings regarding the relative abundance of individual Neisseria species. In vitro culture and more in-depth analysis are necessary to clarify the presence of N. gonorrhoea and its potential contribution to oral microbial dysbiosis.
Several species were identified as hubs of the co-occurrence networks; those in the CCP+at risk group may be indirectly involved in the pathogenesis of RA via the interplay with P. gingivalis and possibly by supporting communities that promote citrullination by multiple routes. Among these hub species, Streptococcus spp are considered the principle early colonisers in dental plaque, and their colonisation influences the composition of maturing plaque.43 Fusobacteriumnucleatum, which was demonstrated to accelerate collagen-induced arthritis in mice, functions in a bridging complex between early and late colonisers such as P. gingivalis.44 A strong synergy was also observed between Treponema denticola and P. gingivalis in biofilm formation.45 Therefore, it is logical to consider the overall capacity of the microbial community in future work.
In conclusion, this study has demonstrated dysbiosis in the subgingival microbiome alongside the specific increase of P. gingivalis in individuals at risk of RA. We propose that these may play an important role in the initiation of RA and that periodontitis and the observed oral dysbiosis may be attractive targets for future preventative interventions, such as periodontal therapy, in individuals at risk of RA.
Acknowledgments
Diane Corscadden, Katie Mbara and Shabnum Rashid for laboratory support; Ashna Chavda for nursing support; Jenny Boards, Ian Weatherill, Chris Brooks, Jiawen Dou and Philip Luxford for administrative support.
References
Footnotes
Handling editor Josef S Smolen
Contributors ZC: conceptualisation, methodology, validation, formal analysis, investigation, data curation, writing, visualisation; TD: conceptualisation, methodology, validation, formal analysis, data curation, writing, supervision; KM: conceptualisation, methodology, validation, data curation, writing; JM: conceptualisation, methodology, validation, writing, supervision; LH, VC, AS, AT: conceptualisation, methodology, investigation, writing; PE: conceptualisation, writing, supervision, administration, funding; DD: conceptualisation, methodology, validation, writing, supervision, administration, funding.
Funding This research was supported by the NIHR infrastructure at Leeds. ZC was supported by a scholarship from the China Scholarship Council. Support was also provided by the Leeds Dental Clinical and Translational Research Unit.
Competing interests ZC reports scholarship from the China Scholarship Council during the conduct of the study. Dr TD and JM reports grants from Colgate Palmolive, outside the submitted work. KM reports grants from the National Institute for Health Research (NIHR) Leeds Biomedical Research Unit and grants from Leeds Biomedical Research Centre during the conduct of the study. DD reports grants from NIHR, grants from Wellcome Trust, during the conduct of the study; grants from Colgate Palmolive, outside the submitted work. PE reports grants and personal fees from Pfizer, Merck Sharp & Dohme, AbbVie, Bristol-Myers Squibb, Roche, Samsung, Sandoz, and Eli Lilly and Company; and personal fees from Novartis and UCB outside the submitted work.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available upon reasonable request. The shotgun sequencing data are uploaded in MG-rast (http://www.mg-rast.org). The data are available upon reasonable request.