P. gingivalis strains ATCC 33277 and W83 (kindly provided by ACTA, Amsterdam, The Netherlands) were cultured on blood agar plates under anaerobic conditions at 37°C. At day 7, colonies were transferred to anaerobe basal broth (CM0957, ThermoFisher, Waltham, USA) and cultured for 24h until they reached mid-log growth. Gram staining confirmed purity of the cultures. The bacterial cultures (OD₆₉₀ 0.1) were heat-inactivated at 60°C for 60min.

Buffy coats were obtained from healthy donors after written consent (Sanquin Blood Bank, Nijmegen, the Netherlands). Peripheral blood mononuclear cells (PBMCs) were isolated using density centrifugation over Ficoll-Paque PLUS (GE Healthcare Biosciences, Chicago, USA). Cells were resuspended in Dutch modified RPMI culture medium (Invitrogen, CA, USA) supplemented with 2mmol/L glutamine, 10mg/mL gentamicin and 1mmol/L pyruvate.

A validated trained immunity experimental design was used (14, 15). After 1h incubation with 5x10⁵ PBMCs per well, cells were washed thrice to remove non-adherent cells in flat-bottom 96-well plates (Corning, NY, USA). Adherent monocytes were stimulated with 1x10⁴/mL P. gingivalis W83, 1x10⁴/mL P. gingivalis ATCC, or RPMI only. After 24h incubation, the cells were washed with warm PBS and incubated in culture medium with 10% pooled human serum. At day 6, cells were restimulated with RPMI, 10ng/mL Escherichia coli lipopolysaccharide (LPS; serotype 055:B5 Sigma-Aldrich, St. Louis, USA), and 10μg/mL Pam3CysK4 (P3C; EMC Microcollections, Tübingen, Germany). At day 7, supernatants were collected and stored at −20°C.

In separate experiments, P. gingivalis stimulation was combined with freshly oxidized LDL [10µg/mL, prepared as described previously (16)] for 24h in adherent monocytes. In brief, LDL was isolated from healthy volunteers under sterile conditions and stored at -80 with saccharose and EDTA to prevent oxidation. Upon thawing, LDL was dialyzed with sterile PBS for 7h and oxidated overnight with filter sterilized copper sulfate. Contamination by endotoxins has been previously ruled out by the original protocol (13, 16). After 24h stimulation, the stimuli were washed away and the similar protocol as noted above was followed.

Cytotoxicity was determined after 24h of incubation with P. gingivalis using a lactate dehydrogenase cytotoxicity assay (LDH; CytoTox-96-Non-Radioactive-Cytotoxicity-Assay, Promega, USA) on fresh supernatants.

Foam cell formation was evaluated after 4h starvation and 24h oxLDL (25µg/mL) exposure at day 6 of the trained immunity experiment. On day 7, Oil Red O staining was performed on adherent macrophages. Intracellular ApoB concentration was measured using enzyme-linked immunosorbent assay (ELISA) as described in (17).

Cytokine production in supernatants was determined using DuoSet ELISA for TNFα and IL-6 according to the manufacturers’ instructions (R&D Systems, Minneapolis, USA).

Participants (40-80 years) were recruited among patients of the Dentistry department of Radboud University Medical Center, the Netherlands from 03-2018 till 02-2019. At the department of Dentistry standard dental care is provided and the patients represent the average population in the area, i.e. they are not referred because of dental problems. Exclusion criteria were previous CVD, auto-immune or auto-inflammatory diseases (including diabetes mellitus), chronic immunomodulatory drug use, chronic kidney disease (MDRD<45ml/min) or liver disease (ALAT>135U/l). Additionally, participants with an infection (>38,5°C or antibiotic treatment), hospital admission, or vaccination within 1-month prior study entry were excluded. We attempted to match control participants based on age, sex, body mass index (BMI) and smoking. The study protocol was approved by the Institutional Review Board Arnhem/Nijmegen, the Netherlands (NL61840.091.17) and conform STROBE Guidelines. All participants gave written informed consent.

The Dutch Periodontal Screening Index (DPSI) and probing pocket depth (PPD) were recorded by an experienced periodontist [AP]. Subjects with DPSI scores 0-2 were classified as having no PD (category A), those with DPSI score 3- as having mild PD (category B), and subjects with DPSI scores 3+ and 4 as having severe PD (category C) (18). The presence of alveolar bone loss on paired bitewings, defined as >2 mm distance between the cement-enamel junction and alveolar bone crest (19) was used to confirm the clinical diagnosis. Subjects with DPSI category C and advanced alveolar bone loss were assigned to the patient group, i.e. participants with severe PD. Subjects with DPSI category A and B without radiographic alveolar bone loss were assigned to the control group, i.e. participants with no-to-mild PD, which is representative for the normal population in this age category.

Defined by the New Periodontal classification of 2018 (20), patients had PD-stage II-IV, whilst controls had gingivitis or PD-stage I.

Blood pressure was measured according to AHA guidelines (21). Fasting total cholesterol (Tchol), high-density lipoprotein cholesterol (HDLc), and triglycerides (TG) were measured using standardized methods, and LDL cholesterol (LDLc) was calculated with the Friedewald formula.

PBMCs were isolated by Ficoll-Paque density gradient centrifugation (GE Healthcare) and resuspended in supplemented RPMI. Cell composition was evaluated by Sysmex analyzer (Sysmex). Per well, 5x10⁵ PBMCs were stimulated for 24h in triplicate in round-bottom 96-well plates (Corning) with the following stimuli to assess the innate immune response: RPMI, 10ng/mL LPS and 10μg/mL P3C. Simultaneously, to assess the adaptive immune response, PBMCs were stimulated in duplicate for 7days with RPMI, 1×10⁶/mL Candida albicans conidia (UC820 strain), or 1x10⁶/mL Staphylococcus aureus (ATCC29213 strain) both with 10% human pool serum. After the incubation periods of 24h and 7days, supernatants were stored after plate centrifugation at −80°C.

Cytokine and chemokine concentrations were determined in supernatants using DuoSet ELISA for TNFα, IL-6, IL-8, IL-10, IL-1β, IL-17 and IL-22 (R&D), and interferon gamma (IFN-γ) (Sanquin). Plasma hsCRP concentrations were obtained (R&D). Circulating IL-1β, IL-1Ra, IL-6 and IL-18 concentrations were sensitively measured using SimplePlex cartridge on the Ella platform (Protein Simple, San Jose, USA).

Circulating immune cells and monocyte subpopulations were determined by their expression markers using a CYTOflex flow cytometer (Beckman Coulter). 50 µL EDTA blood was stained after the lysis-no-wash strategy (BD Pharm Lyse lysing buffer, Becton Dickinson) by monoclonal antibodies CD45 Krome Orange ([KO], clone J33; Beckman Coulter), HLA-DR PE (clone immu-357; Beckman Coulter), CD14 PC7 (clone 61D3 Bioscience), CD16 FITC (clone CB16; eBioscience), CD3 APC-Alexa750 (clone UCTH1; Beckman Coulter), CD56 APC (clone N901; Beckman Coulter), CD192 Brilliant Violet421 ([BV421] clone 48607; Becton Dickinson), CD11b BV785 (clone ICRF44; Biolegend), CD41 PC5.5 (clone Hip8; Biolegend) and measured with CytoFLEX flow cytometer (Beckman Coulter).

Single cells were analyzed by manual gating using Kaluza v2.1 software (Beckman Coulter) and unsupervised computational methods in parallel using FlowJo v10.6.2 software (Beckton Dickinson). The gating strategy applied is shown in Appendix Figure 1 , gates were set with the fluorescence-minus-one method (22, 23). In short, monocytes were selected based on CD45+ HLA-DR+ and monocyte scatter properties, then CD3+ T-lymphocytes and CD56+ NK-cells were excluded, and monocyte subsets were identified in the CD14/CD16 plot as percentage of gated. Data was analyzed with Kaluza 2.1 software (Beckman Coulter). Characterization of monocytes subsets is according to current recommendations (22, 23).

For unsupervised computation analyses, data files were randomly down sampled to 10,000 events per file and subsequently concatenated to a single file containing all events. Controls and patients were labelled accordingly to allow separation after analysis. Unsupervised clustering was performed on the expression values of all markers using the tSNE plugin in FlowJo v10.6.2 software (Becton Dickinson), using 300 iterations and a perplexity of 20. Manual gating of cell populations (based on gating strategy) was used to identify cell populations and check the separation quality of the unsupervised clustering. The contribution of periodontitis to circulating cell populations in the total tSNE was inspected by separating the patient and control group in two figures ( Figure 3 ). Visual differences were confirmed by manual gating and statistical analysis.

After adhering to a 24h low-carbohydrate diet and 6h of fasting, participants underwent 2’-deoxy-2’-[¹⁸F]fluoro-D-glucose positron-emission-tomography ([¹⁸F]FDG PET) with low-dose non-contrast enhanced CT on a dedicated Siemens Biograph 40mCT scanner (Siemens Healthineers, Erlangen, Germany). Analyses were performed on reconstruction settings according to European guidelines (24) using Inveon Research Workspace v4.2 software. See supplement for details.

Periodontal [¹⁸F]FDG-uptake was determined in the left and right periodontium of maxilla-to-mandible ( Figure 4A , Appendix Figure 2 ). Vascular [¹⁸F]FDG-uptake was determined in the aorta ascendens, aorta descendens, abdominal aorta, the left and right common carotid arteries, and the left and right iliac arteries ( Figure 4B , illustration). Hematopoietic [¹⁸F]FDG-uptake was assessed in the spleen, lumbar vertebrae L2/L3, and the left and right medullary femur cavity ( Figure 4C , illustration). The mean standardized uptake value (SUV_mean) was calculated after correction for weight and [¹⁸F]FDG-dose (MBq). The target-to-background ratio (TBR) was calculated as the ratio of vascular wall SUV_mean and arterial bloodpool SUV_mean. Hematopoietic tissue TBR was expressed as ratio of liver SUV_mean.

In vitro data are reported as mean ± SEM. P. gingivalis-trained cells were compared to RPMI-incubated control cells using Wilcoxon signed-rank test. The in vivo study is exploratory, hence no sample size calculation is performed. In vivo continuous variables are reported as mean ± SD or median[interquartile range], depending on the data distribution. Normal distribution of the data was checked with the Shapiro-Wilk test, when the p-value reached <0.05 this assumption was violated. Patients and controls were compared for continuous variables using independent samples T-test (if normally distributed) or Mann-Whitney U test (if not normally distributed). Categorical variables are presented as percentage(number) and were compared using Χ²-test. For all study outcomes, a one-way ANCOVA was conducted to control for age differences between patients and controls. Beforehand, outliers were removed with > ± 3SD of Z-scores and thereafter log(10)-transformed. Missing values were not imputed. A two-sided p-value <0.05 was considered statistically significant. All data were analyzed using SPSS v25.0 (Chicago, USA).

We included participants with severe PD (DPSI 3.9 ± 0.2, n=14, ‘patients’) and participants with no-to-mild PD (DPSI 2.9 ± 0.3, n=14, ‘controls’), which is a representative of the general population ( Table 1 ). Patients were older than controls (mean 63 ± 6 versus 54 ± 10 years, p<0.05), but there was no significant different in classical cardiovascular risk factors. Patients only had a higher previous tobacco exposure (15.8 ± 18.2 versus 4.5 ± 5.7 pack years, p<0.05), but current smoking behavior was similar. Age is known to impact immune cell function (26) and future cardiovascular risk, therefore we corrected all comparisons for age.

Higher circulating leukocytes (median 6.1 versus 5.6 10⁶/mL, p<0.01) and a trend for higher lymphocytes numbers (1.8 versus 1.7 10⁶/mL, p=0.06) were found in patients compared to controls after age correction. In addition, a tendency for higher concentrations of circulating pro-inflammatory IL-6 and anti-inflammatory IL-1Ra cytokines were observed, but after age correction the statistical significance was lost (IL-6: 2.0 versus 1.4 pg/mL, p=0.08; IL-1Ra: 247 versus 171 pg/mL, p=0.10) ( Table 2 ).

Using flow cytometry, circulating immune cell populations were further identified based on their expression markers ( Figure 3 ). Unsupervised t-Stochastic Neighbor Embedding (tSNE) analysis showed a large overlap between both groups with potentially interesting differences in cell populations ( Figure 3B ). The most distinct population was a smaller cluster of T-cells with high CD41 expression in patients compared to controls ( Figure 3C , arrows). Further analysis of this lymphocyte-platelet interacting cell population was not allowed as the flow cytometry panel used was designed for monocyte activation analysis and did not include further T-cell markers. Manual gating of these populations and statistical analysis also revealed high similarity between groups ( Table 2 ).

No significant difference in the cytokine production capacity of PBMCs between patients and controls was observed ( Table 3 ), for both 24h stimulation with LPS or P3C stimulation (i.e. innate immune response) and 7 day stimulation with whole pathogens C. albicans or S. aureus (i.e. adaptive immune response).