All participants provided written informed consent prior to study enrollment according to the Declaration of Helsinki (General Assembly October 2008), and this study was conducted with the approval of the Medical Ethical Committee of the University Medical Center Groningen (METc UMCG 2011/010).

RA disease activity was measured with the Disease Activity Score 28 joint count (DAS28) 34. Other parameters were disease duration of RA, smoking status (current, former, or never), BMI, and RA medication, including steroids and anti-tumor necrosis factor biologic agents. Assessment of the socioeconomic status was made according to highest self-reported level of received education.

Periodontal condition was examined by using DPSI. The DPSI score was taken by a periodontist blinded for the diagnosis of RA. In the general dental practice, the DPSI score was taken by the dentist at the first visit of the patient.

Peripheral blood and subgingival samples were obtained from the RA and non-RA controls at the day of clinical examination. Subgingival samples were taken by using sterile paper points 35. Microbiological sampling included selection of the deepest bleeding periodontal pocket in each quadrant of the dentition on the basis of pocket probing depth measurements. If there were no bleeding periodontal pockets, mesial sites of the first molars or, in absence of a first molar, the mesial site from the adjacent anterior tooth in the dental arch was selected. Sample sites were isolated with cotton rolls, and supragingival plaque was carefully removed with curettes and cotton pallets. Subsequently, two paper points were inserted to the depth of the pocket and left in place for 10 seconds. All paper points per subject were pooled in reduced transport fluid 36 and processed for microbiological examination immediately after sampling. GCF samples were collected in the same way. The deepest non-bleeding site per quadrant was used to collect GCF to avoid blood contamination of the sample. GCF samples were discarded for further analysis if they were visibly contaminated with blood. Paper points for the GCF sample were pooled per subject in phosphate-buffered saline (PBS) with 1% bovine serum albumin. The paper points were removed after centrifuging at 28,000 g for 10 minutes, and the supernatant was stored at -20°C until use.

Serum was investigated for C-reactive protein (CRP) concentration by enzyme-linked immunosorbent assay (ELISA) (DuoSet; R&D Systems, Abingdon, UK). IgM-RF (in international units per milliliter) was measured by using an in-house validated ELISA (cutoff point for positivity: 25 U/mL) 37. Total IgG anti-cyclic citrullinated protein (anti-CCP) antibodies (ACPAs) (in units per milliliter) were determined by using the Phadia analyzer (Phadia Laboratory Systems, Phadia AB, Uppsala, Sweden) with an upper detection limit of 340 U/mL (cutoff point for positivity: 10 U/mL). Antibody levels to five synthetic native peptides representing the best-established auto-antigens in RA 38 (that is, enolase-1 fibrinogen-1, fibrinogen-2, fibrinogen-3, and vimentin-1 and their citrullinated forms) were measured by using the same ELISA as described by van de Stadt and colleagues 39.

IgA, IgG, and IgM antibodies to P. gingivalis were determined by using an in-house developed ELISA, in which a pooled lysate of four randomly selected clinical isolates of P. gingivalis from patients with periodontitis was used as an antigen. These monocultures were suspended in ice-cold PBS with protease inhibitors (Complete Mini EDTA-free Protease Inhibitor Cocktail Tablets; Roche Diagnostics, Basel, Switzerland). After centrifuging and discarding of the supernatant, pellets were resuspended in an ice-cold lysis buffer containing non-denaturing detergent (Noninet P-40; Sigma-Aldrich, St. Louis, MO, USA) and sonicated for 15 minutes (Bioruptor Standard sonication device; Diagenode s.a., Liège, Belgium). Protein concentration was determined by using the BCA Protein Assay Kit (Pierce Protein Biology Products, Thermo Fisher Scientific, Rockford, IL, USA). Microtiter plates (Costar 96-Well; Corning, Amsterdam, The Netherlands) were coated overnight with 1 μg/mL of antigen. Standard curves were made from protein standard (N protein-standard SL; Siemens Healthcare Diagnostics, Den Haag, The Netherlands) diluted twofold (highest standard curve values were for 300, 25, and 250 ng/mL for IgA, IgG, and IgM, respectively). For the standard curves, the following capture antibodies were used: monoclonal anti-human IgA (1:4,000, clone GA-112; Sigma-Aldrich, Zwijndrecht, The Netherlands), monoclonal anti-human IgM (1:10,000, clone MB-11, μ-chain-specific; Sigma-Aldrich), and goat anti-human IgG (1:5,000, F(ab')₂ fragment-specific; Jackson ImmunoResearch Europe Ltd., Newmarket, UK), respectively. Serum samples were incubated in fourfold serial dilutions (1:100, 1:400, 1:1,600, and 1:6,400). Detection of anti-P. gingivalis antibodies was carried out with horseradish peroxidase-labeled goat anti-human IgA, mouse anti-human IgG (Fc fragment-specific, clone JDC-10), and mouse anti-human IgM (μ-chain-specific, clone SA-DA4; all from SouthernBiotech, Birmingham, AL, USA) followed by color reaction with tetramythylbenzidin and hydrogen peroxide. Absorbance was read at 450 nm in an EMax microplate reader and corrected for background binding, and concentration of antibodies was calculated by SOFTmax PRO software (Molecular Devices, Sunnyvale, CA, USA) according to the IgA, IgG, or IgM standard curves included on each ELISA plate and expressed in nanograms per milliliter. As an internal control, two serum samples with a repeatedly high and a low titer were tested at each plate. The variation coefficients were 22% for IgA, 20% for IgG, and 25% for IgM anti-P. gingivalis. Interference of IgM-RF was investigated by spiking samples with sera with known high titers of RF. Adding RF had no measurable effect on anti-P. gingivalis titers.

In paired samples of serum and GCF of patients with RA, the presence of IgG-ACPAs was assessed by using the anti-CCP2 ELISA (Euro Diagnostica, Nijmegen, The Netherlands). Because the HLA-DRB1 shared epitope (SE) is a known genetic risk factor for RA and a possible genetic risk factor for periodontitis 40, HLA-DRB1-SE-containing alleles were genotyped from genomic DNA in the patients with RA. The presence of an RA SE was analyzed by a polymerase chain reaction-based sequence-specific oligonucleotide probe hybridization (SSOP) approach by using a commercial kit (Hologic Gen-Probe Incorporated, San Diego, CA, USA) and Luminex xMAP technology (Luminex Corporation, Austin, TX, USA) in accordance with the instructions of the manufacturer.

Microbiological samples were processed by using standard anaerobic culture techniques 4142. Paper point samples were vortexed for 2 minutes, and appropriate 10-fold serial dilutions (100 μL) in PBS were plated on blood agar plates (Oxoid no. 2; Oxoid Limited, Basingstoke, UK), which were supplemented with horse blood (5% vol/vol), hemin (5 mg/L), and menadione (1 mg/L) and incubated in 80% N₂, 10% H₂, and 10% CO₂ at 37°C for up to 14 days. Besides the presence and proportions of P. gingivalis, those of other established periodontal pathogens, including Prevotella intermedia, Fusobacterium nucleatum, Parvimonas micra, Tannerella forsythia, and Campylobacter rectus, were assessed 43. Identification was based on microscopic morphology, Gram reaction, and the production of a set of metabolic enzymes (API/ID 32; BioMérieux, La Balme Les Grottes, France). Additional tests for identification included detection of a trypsin-like activity based on the degradation of benzoyl-DL-arginine-2-naphthylamide (Sigma-Aldrich) 44. Finally, the total number of colony-forming units per sample was determined.

Data were analyzed by using GraphPad Prism 4 and Instat 3 (GraphPad Software, Inc., San Diego, CA, USA). For group comparisons among two groups, the unpaired two-tailed t test for variables with normal distribution and the two-tailed Mann-Whitney test for skewed variables were used. For group comparisons among three groups, Kruskal-Wallis one-way analysis of variance was performed. The Fisher exact test or the chi-squared test for independence was used to analyze contingency tables. Significance level α was 0.05.

Two hundred thirty-nine consecutive patients with RA were invited to participate in this study. Of the 167 patients who provided written informed consent, 66 patients met at least one of the exclusion criteria. Of the 101 included patients, six patients were excluded from analysis because of incomplete data. The final cohort consisted of 95 patients with RA. In addition, 203 consecutive patients were invited to join the control group. Of this group, 108 dentate persons without RA provided written informed consent. Twenty-eight of them had to be expelled on the basis of the exclusion criteria. The remaining 80 patients served as non-RA controls for blood and subgingival samples, matched for age, gender, BMI, and periodontal and smoking status (n = 44), or as healthy controls (no periodontitis and no cultivable P. gingivalis, n = 36). Characteristics of the predominantly (99%) Caucasian patients with RA, non-RA controls, and the healthy controls are summarized in Table 1. The general control population for estimating the prevalence of periodontitis consisted of 420 individuals matched for age and gender with the patients with RA (age of 56 ± 5.5 years, and 68% were female). The socioeconomic status of this group as assessed by the highest self-reported level of received education was comparable between the control population and the patients with RA (low: 27% versus 36%, middle: 37% versus 30%, and high: 36% versus 34%).

Forty-three percent of the patients with RA had moderate periodontitis, and 27% had severe periodontitis. These numbers are significantly higher than, respectively, the 18% and 12% found in the control population (P < 0.001). Compared with the control population, the relative risk for having RA and severe periodontitis was 3.7 (95% confidence interval of 2.4 to 5.9) (Table 2).

RA patients with severe periodontitis had significantly higher DAS28 scores (P < 0.001) than RA patients with no or moderate periodontitis (Figure 1). CRP levels in RA patients with severe periodontitis were higher than in RA patients without periodontitis (marginally significant, P = 0.05). In none of the DPSI categories was there a difference in DAS28 scores between smokers and non-smokers or former smokers. Also, no difference was found in RA disease duration, although patients with severe periodontitis were significantly older (P < 0.01). In the control population, there were no age differences between the DPSI categories.

No differences were found in the presence or absence of the HLA-DRB1-SE in RA patients with no, moderate, or severe periodontitis (n = 78). In RA patients with no, moderate, or severe periodontitis, 58%, 66%, and 62%, respectively, were positive for HLA-DRB1-SE, and 29%, 21%, and 15%, respectively, had two alleles.

Between RA patients with no, moderate, or severe periodontitis, no differences were seen in IgM-RF and ACPA levels or in reactivity against the citrullinated peptides enolase-1, fibrinogen-1, -2, and -3, and vimentin-1 (Figure 2). Between RA patients culture-positive or -negative for P. gingivalis, only reactivity against citrullinated fibrinogen-2 was different; reactivity was higher in P. gingivalis-positive patients (P < 0.01). A small number of RA patients with moderate (n = 2) or severe (n = 2) periodontitis and culture-positive for P. gingivalis had a higher reactivity against citrullinated α-enolase compared with RA patients without periodontitis and without subgingival P. gingivalis. This P. gingivalis culture-positive subgroup also had high IgM-RF (mean 539 ± 796 U/mL), ACPAs (mean 340 ± 0 U/mL), and anti-P. gingivalis titers (IgM: mean 93 ± 135, IgG: 190 ± 357, IgA: 61 ± 81 mg/L).

Overall, patients with RA showed higher IgM anti-P. gingivalis titers compared with non-RA controls (P < 0.05). There were no differences in anti-P. gingivalis titers between RA patients and non-RA controls with no or moderate periodontitis; however, RA patients with severe periodontitis showed both higher IgG and IgM anti-P. gingivalis titers compared with non-RA controls with severe periodontitis (P < 0.05) (Figure 3). RA patients with moderate periodontitis have a less pronounced anti-P. gingivalis response compared with RA patients with severe periodontitis but a higher one than non-RA controls with severe periodontitis for IgG (borderline significant, P = 0.06) and IgM (P < 0.05). Both RA patients and non-RA controls culture-positive for P. gingivalis showed higher anti-P. gingivalis titers compared with their culture-negative counterparts.

In patients with RA, serum levels of IgM-RF and ACPAs showed a strong correlation (ρ = 0.51, P < 0.0001), as did IgG anti-P. gingivalis with IgM anti-P. gingivalis (ρ = 0.41, P < 0.0001) and IgA anti-P. gingivalis (ρ = 0.66, P < 0.0001). Medication used for RA was not of influence on P. gingivalis titers. In non-RA controls, IgG anti-P. gingivalis only correlated with IgA anti-P. gingivalis (ρ = 0.65, P < 0.0001).

In patients with RA, there was a weak but significant correlation between IgM anti-P. gingivalis and IgM-RF (ρ = 0.33, P < 0.01), ACPAs (ρ = 0.24, P < 0.05), and the citrullinated peptides fibrinogen-1 (ρ = 0.27, P < 0.05) and fibrinogen-3 (ρ = 0.22, P < 0.05). Likewise, IgG anti-P. gingivalis was correlated with IgM-RF (ρ = 0.26, P < 0.05). In RA patients with severe periodontitis, there were no correlations between IgG, IgM, and IgA anti-P. gingivalis titers and IgM-RF and ACPAs. In patients with RA, ACPA levels in serum were comparable to ACPA levels in paired GCF samples which were not visibly contaminated with blood (n = 45, ρ = 0.89, P < 0.0001) (Figure 4).

The subgingival prevalence of P. gingivalis was not different between patients with RA (16%) and non-RA controls (20%). In both groups, P. gingivalis was infrequently cultured in the absence of periodontitis (6% to 12%). None of the other identified periodontal pathogens differed in prevalence between RA patients and non-RA controls (P. intermedia: 66% versus 73%, T. forsythia: 84% versus 70%, P. micra: 88% versus 89%, F. nucleatum: 100% versus 95%, and C. rectus: 33% versus 41%). However, the prevalence of anaerobic Gram-negative rods other than P. gingivalis and P. intermedia was higher in patients with RA (27% versus 8%, P < 0.05), as was the total bacterial load (in colony-forming units per milliliter) (P < 0.05), notwithstanding the comparable mean probing pocket depth between RA patients and matched non-RA controls (4.2 ± 1.0 mm).