The SCREEN-RA study is an ongoing Swiss cohort, which enrolls and follows first-degree relatives (FDRs) of RA patients (detailed description in (Gilbert et al., 2021)). Hence, the participants of SCREEN-RA do not have a RA diagnosis at inclusion, but have increased genetical risk of developing the disease. The majority of them are healthy, while a fraction of the participants has autoimmunity associated with RA (meaning the presence of anti-citrullinated protein autoantibodies (ACPA) and/or rheumatoid factor (RF), and/or anti-ra33 antibodies in the serum), or clinically suspect symptoms, based on EULAR definition. In this study, we included serum samples from the following subgroups of the SCREEN-RA cohort: first-degree relatives without any autoantibodies or symptoms associated with RA (named FDR 1, serving as control group), first-degree relatives with presence of ACPA and/or RF in the serum (named FDR 2), and individuals with symptoms and signs associated with possible RA with/without presence of ACPA and/or RF (named FDR 3).

The Swiss Clinical Quality Management (SCQM) RA cohort is another Swiss registry, founded in 1997 with the support of Swiss regulatory authorities. SCQM is thus a prospective, longitudinal cohort of Swiss RA patients. The participants are enrolled by their practitioner rheumatologist after being diagnosed. At inclusion, RA patients provide a serum sample which is divided in multiple aliquots and stored in the SCQM biobank.

The Rheuma-VOR cohort is an ongoing German cohort, which recruits untreated patients with new onset of different rheumatic disorders. The cohort is dedicated to arthritis and to inflammatory back pain. Therefore, general practitioners could refer patients with either inflammatory back pain fulfilling the Berlin criteria or with symptoms of arthritis (swollen joints, morning stiffness > 30 min, elevated C-reactive protein (CRP) and/or erythrocyte sedimentation rate (ESR)) to rheumatologists. The rheumatologist diagnosed RA or other rheumatic disorders when the patient had been examined and all the laboratory values including RF, ACPA, CRP, ESR, and antinuclear antibodies (ANA) were available. The diagnosis of RA was based on the judgment of the rheumatologist. Stool samples as well as serum samples were obtained at the time of their first medical examination and various clinical data as well as dietary habits are recorded. Within this study, samples were included from newly diagnosed rheumatoid arthritis (RA), psoriatic arthritis (PsA), axial spondyloarthritis (axSpA), or reactive arthritis (ReA) patients. Patients diagnosed with non-rheumatic diseases (NRD) including arthralgia, lumbago, arthrosis, angioedema, tendopathy, and nonspecific backpain served as a control group. Serum samples from healthy blood donors (BD) were included as an additional control group for measuring serum biomarkers. Individuals were recruited via the German Hospital conducting the Rheuma-VOR recruitment. Additionally, only individuals who have not taken any medication in 7 days preceding the donation were permitted to donate blood. Due to obligatory anonymization of these blood donors, age and gender were the only collected parameters.

Serum samples from IBD patients were obtained from the University Medical Center Hamburg-Eppendorf in Germany. Individuals diagnosed with ulcerative colitis or Crohn’s disease were recruited within the scope of a human study approved by the local ethical committee. Ulcerative colitis patients with a Mayo Score ≥3 were considered as active disease patients, whereas the Bradshaw index was used to classify active Crohn’s disease (Bradshaw index ≥5) and Crohn’s disease patients in remission (Bradshaw index 0-4).

P. copri strains were isolated from fecal samples of P. copri positive donors previously identified by 16S rRNA gene sequencing. Fresh fecal samples were stored in sealed glass vials prefilled with 50% glycerol/BHI (Brain Heart Infusion) medium in -80°C. For isolation, the fecal sample was thawed and streaked out on BHI blood agar plates supplemented with vancomycin by performing dilution ranging from 10⁰ to 10^-3. The plates were incubated at 37°C inside an anaerobic chamber for 48-72h. Single colonies were picked into BHI medium supplemented with fetal bovine serum (FBS) and Vitamin K3 (BHI-S) and were further incubated at 37°C until grown. The resulting cultures were screened by PCR using P. copri specific primers (P_copri_69F: CATCGAAAGCTTGCTTTTGATGGGC, P_copri_853R: CTTGGCCGCTGACCTGTTCAGA). After species confirmation by Sanger sequencing, the obtained cultures were aliquoted into sealed glass vials mixed with 50% glycerol/BHI medium and were immediately frozen in –80°C.

The P. copri type strain (DSM 18205), Porphyromonas gingivalis (DSM 20709), Prevotella intermedia (DSM 20706) and Prevotella melaninogenica (DSM 7089) were obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures GmbH. All DSMZ strains as well as P. copri in-house strains were cultured and maintained in BHI-S medium inside an anaerobic chamber.

All strains were maintained inside an anaerobic chamber using BHI-S medium. Bacterial cultures were grown freshly from an overnight inoculation until exponentially phase and until reaching an OD600 ranging from 0.4 to 0.8. An aliquot of the culture was plated in serial dilutions (10^-4, 10^-5, 10^-6) on BHI blood agar plates and served later on as quantification control. The rest of the culture was taken out of the chamber, was pelleted by centrifugation and was washed once with PBS. To inactivate the bacterial cells, the pellet was further resuspended in 1% Formalin and kept in the fridge. After 24h, Formalin was removed by centrifugation, the pellet was washed twice with PBS and the resulting solution was normalized to 4x10⁶ cells/ml by quantifying the colony forming units (CFU) on the previously prepared BHI blood agar plates. Bacterial solutions were kept at -20°C for long-term storage. The identity of all cultures was additionally checked by Sanger sequencing. Because Prevotella intermedia failed to be pelleted, we exceptionally used a fresh culture instead of a Formalin fixed culture for subsequent anti-bacterial ELISA assays.

Corning ELISA plates (Microplates, Cat: 3690) were coated with 25µl of the previously described bacterial solutions per well (resulting in a final concentration of 10⁵ bacterial cells/well) and were incubated overnight at 4°C. The next day, plates were washed twice with PBS solution supplemented with 0.05% Tween 20 (PBST) and were blocked with PBS + 5% milk powder for 1h at RT on a shaker. After washing the plates twice with PBST, 25µl of patient serum samples were added as duplicates to the plates (dilution 1:100 in PBS + 5% milk) and were incubated for 2h at RT on a shaker. Following washing four times with PBST, 25µl diluted HRP-conjugated anti-human IgG or IgA were added to the plate and were incubated for 1.5h at RT on a shaker (anti-human IgG used for oral pathobionts: Goat anti-human IgG antibody, GenScript, Cat. No.: A00166, dilution 1:5,000; anti-human IgG used for P. copri strains: Goat anti-human IgG, Invitrogen, Ref: SA5-10283, dilution 1:10,000; anti-human IgA: Goat anti-human IgA, BioLegend, Cat: 411002, dilution 1:1,000). After washing 3 times with PBST and twice with PBS, TMB mix was added to the wells (25µl/well) and was incubated for 10 min (for IgG) or 15 min (for IgA). After adding 25µl of H2SO4, the plates were measured at the ELISA reader at 450 nm against 570 nm as a reference. All samples were measured in duplicates and additional four control serum samples were included on each plate for inter-plate standardization.

Total serum IgG was measured using the IgG human uncoated ELISA Kit from Invitrogen, Cat: 88-50550, following the manufacturer’s instructions using a serum dilution of 1:100,000. Serum calprotectin was quantified using the RayBio Human S100-A8/A9 complex ELISA Kit from RayBiotech, Cat: ELH-S100A8-9, following the manufacturer’s instructions. Serum zonulin was measured using the IDK Zonulin ELISA Kit from Immundiagnostik AG, Cat: K5601, following the manufacturer’s instructions.

The relative abundances of P. copri in the fecal samples from the SCREEN-RA cohort have been obtained from previously generated data by our group (Alpizar-Rodriguez et al., 2019). Using identical workflows and analyzing pipelines as described by Alpizar-Rodriguez et al., amplification and sequencing of the 16S rRNA gene as well as the analysis of P. copri relative abundance were performed from the fecal samples of Rheuma-VOR and SCQM cohort. Sequencing data of all samples used in this study was deposited to a public repository with the Accession no./Bioproject: PRJEB55184.

The phylogenetic analysis was conducted using whole genome sequencing data from the P. copri strains used in this study and genome data from previously published P. copri strains (Tett et al., 2019) as a reference. Phylogenetic analysis was conducted using bacterial marker genes predicted and aligned by GTDBTk version 1.7.0. The tree was built using the neighbor-joining approach with 1000 boot steps in MEGA-X.

Statistical analysis was performed using GraphPad Prism (GraphPad Software, Inc.). If not further indicated, data are presented as mean ± SD (standard deviation) and statistical differences were analyzed using Kruskal-Wallis test. P values ≤ 0.05 were considered as significant: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To study calprotectin levels in preclinical phases of RA, 165 serum samples from the Swiss SCREEN-RA and SCQM cohorts were characterized. Specifically, samples from the two cohorts were divided into four groups: i) first degree relatives of RA patients without any ACPA or RF positivity and disease symptoms as controls (FDR 1), ii) first-degree relatives with RA associated auto-immunity defined by serum ACPA and/or RF positivity (FDR 2), iii) first-degree relatives having RA specific symptoms with/without ACPA and/or RF autoantibodies (FDR 3), and iii) chronic RA patients (cRA) (see Table 1 for details). To study the new onset phase, serum samples were obtained from 178 individuals recruited via the complementary Rheuma-VOR study being newly diagnosed with a rheumatic disorder, including rheumatoid arthritis (RA), psoriatic arthritis (PsA), axial spondyloarthritis (axSpA), or reactive arthritis (ReA). Moreover, within this cohort also patients with unspecific non-rheumatic diseases (NRD), such as arthralgia, lumbago, and back pain were recruited as the diagnosis occurred after recruitment. Finally, serum samples from healthy blood donors (BD) recruited via the same German hospital as the Rheuma-VOR patients were additionally included. Detailed information about the individuals’ clinical metadata are listed in Table 1 .

Calprotectin levels were determined in all samples using a commercially available ELISA kit. While the FDR 1 group served as control for the other groups from the SCREEN-RA/SCQM cohorts, we included the NRD serum samples as well as the BD sera as controls for the Rheuma-VOR patients. In line with previously published data, calprotectin levels were elevated in cRA patients ( Figure 1A ) and in new-onset RA patients ( Figure 1B ). However, we did neither detect a significant increase in calprotectin levels in first-degree individuals at risk for developing RA, nor in the different SpA subforms, when comparing to the respective control groups. This indicates calprotectin to be rather specifically increased in patients with new onset or chronic RA, but not in earlier disease phases. Moreover, in contrast to previous studies we were not able to detect elevated calprotectin levels in newly diagnosed SpA. To further compare calprotectin levels with the commonly used inflammatory marker C-reactive protein (CRP), we obtained serum CRP values of patients from the Rheuma-VOR cohort. Serum CRP levels were significantly increased in RA and PsA patients compared to the NRD control group ( Figure 1C ) and positively correlated with serum calprotectin concentration ( Figure 1D ). As RF and ACPA autoantibodies have been associated to disease activity and inflammation in RA patients (Aletaha et al., 2015; de Brito Rocha et al., 2019), we divided FDR individuals and cRA patients from the SCREEN-RA/SCQM cohort as well as new-onset RA patient from the Rheuma-VOR cohort based on their RF and ACPA seropositivity and investigated potential differences in serum calprotectin and CRP levels. ACPA and RF presence both resulted in higher calprotectin concentrations in individuals from the SCREEN-RA/SCQM cohort, although differences between RF positive and negative individuals did not reach significance ( Figure 1E ). In contrast, calprotectin levels in new onset RA patients from the Rheuma-VOR cohort did not differ based on RF and ACPA presence/absence. Similarly, serum CRP levels did not differentiate RF or ACPA seropositive and –negative RA patients ( Figure 1F ). These results however need to be interpreted with caution, as the division of RA patients from the Rheuma-VOR cohort based on autoantibody presence resulted in a low sample size in the RF and ACPA negative group. Overall, this data suggests seropositivity in ACPA to be associated to systemic inflammation measured by calprotectin.

With the aim to evaluate zonulin levels, a marker for intestinal permeability, in new-onset RA patients and to compare them to patients with other newly-diagnosed rheumatic diseases, zonulin levels were quantified in 163 serum samples from the Rheuma-VOR cohort using a commercial ELISA kit. The samples from the SCREEN-RA/SCQM cohorts were not suitable for this analysis due to their extended storage (>3 years) at -80°C, which interferes with the quantification of zonulin. Notably, we detected significantly increased zonulin concentrations in the serum of new-onset RA patients compared to blood donors ( Figure 1G ). Interestingly, the zonulin levels of PsA, axSpA, ReA, and NRD patients were slightly increased compared to the BD controls, but did not differ significantly. In order to compare zonulin levels in these cohorts to patients with a disease affecting the gastrointestinal mucosa, serum samples from IBD patients divided into active ulcerative colitis (UC A, n=14), active Crohn’s disease (CD A, n=13) and Crohn’s disease in remission (CD R, n=12) were further included as controls (metadata in Table 1 ). While zonulin secretion in the rheumatic groups were similar to active IBD patients, CD patients in remission overall displayed the highest levels. Hence, we could corroborate previous findings identifying increased serum zonulin levels already in new-onset RA patients, but determined that they are not elevated in other rheumatic diseases. That strengthens zonulin as another potential specific biomarker for RA and suggests that RA and SpA differ in the presence of biomarkers reflecting acute mucosal inflammation during the new-onset phase before the initiation of treatment. Investigation of serum zonulin levels in RF and ACPA seropositive and seronegative patients revealed significantly increased zonulin levels in RF positive patients and a minor increase in RA patients with ACPA autoantibodies ( Figure 1H ). This indicates intestinal permeability to correlate with RA autoantibody seropositivity.

In order to extend the characterization of biomarkers to differentiate rheumatic diseases in our cohort, an ELISA assay was established to determine the serum IgG reactivity against previously described oral and gut-derived bacteria. To this end, the IgG antibody responses against the four bacterial strains Prevotella copri, Porphyromonas gingivalis, Prevotella intermedia, and Prevotella melaninogenica were measured in serum samples from the SCREEN-RA/SCQM studies as well as a subset of serum samples from the Rheuma-VOR cohort ( Table 2 ). In contrast to previous studies, we did neither detect elevated IgG responses in cRA or NORA patients against the three oral pathobionts P. gingivalis ( Figure 2A ), P. melaninogenica ( Figure 2B ), and P. intermedia ( Figure 2C ), nor against the P. copri type strain (further named PDSM) ( Figure 2D ), when being compared to other patients or respective control groups. Moreover, none of the other preclinical or new onset patient groups showed significant differences in their IgG response compared to controls.

Since previous studies investigating the antibody reactivity of P. copri solely focused on the type strain isolated from a healthy Japanese donor (Hayashi et al., 2007), we expanded the P. copri strain panel to compare the immunostimulatory potential of additional distinct strains belonging to the P. copri complex. Specifically, we tested the serum antibody binding capacity of four additional P. copri strains from our internal laboratory collection, which were previously isolated from fecal samples of healthy individuals (HDD04, HDE04 and HDD05) and from a patient with new onset of rheumatoid arthritis (RPC01). While the strains P. copri PDSM, HDD04 and RPC01 clustered based on the previous defined classifications (Tett et al., 2019) into clade A, strain HDE04 and HDD05 belonged to clade C and D, respectively. Whole genome phylogeny of the five strains within the P. copri complex are depicted in the supplementary material ( Supplementary Figure 1 ). Similar to our results with strain PDSM, the IgG reactivity against RPC01, HDD04, HDE04 and HDD05 did not differ between distinct rheumatic groups, neither between patient groups from the SCREEN-RA/SCQM nor from the Rheuma-VOR cohort ( Figure 3A ). However, using Spearman’s correlation analysis, a highly significant (P < 0.0001) moderate to strong positive correlation (r = 0.42 – 0.64) was detected between the five P. copri strains ( Figure 3B ). Remarkably, by comparing the overall strain-specific IgG antibody reactivity in all serum samples, we found highly significant differences between all the P. copri isolates, with P. copri strain RPC01 clearly reaching the highest signals ( Figure 3C ). We obtained similar results measuring IgA reactivity profiles against all five P. copri strains ( Supplementary Figure 2 ). While IgA responses did not significantly differ between different disease types from the SCREEN-RA/SCQM or the Rheuma-VOR cohort ( Supplementary Figures 2A, B ), we also identified the highest IgA signal in both cohorts, independently of disease status, against the P. copri strain RPC01 ( Supplementary Figures 2C, D ). Together, this demonstrates a high variability between P. copri strains in their ability to trigger immune responses in their host, which appears not to be driven by subspecies clustering. Moreover, the observed increased immune recognition of P. copri RPC01 compared to other strains suggests a higher immunogenic potential of the RA patient strain compared to strains isolated from healthy individuals.

To further understand whether current P. copri colonization and abundance levels influence anti-P. copri IgG antibody levels, we analyzed the abundance of P. copri in the fecal samples of the respective patients from the Rheuma-VOR and SCQM study by 16S rRNA gene sequencing and included previously published sequencing data from the SCREEN-RA study (Alpizar-Rodriguez et al., 2019).

Hypothesizing that a higher abundance or general presence of P. copri in the intestine of patients is able to trigger the immune system and eventually provokes the production of P. copri specific IgG in the serum of patients, we expected the serum antibody responses to correlate with intestinal prevalence. However, the relative abundance of P. copri in feces within our sample collections did neither significantly differ between FDR 1, FDR 2, FDR 3 or cRA patients, nor between new-onset patients with NRD, RA, PsA or axSpA ( Figure 4A ). Correlation between serum IgG response against P. copri strain RPC01 (chosen as representative of the five P. copri strains) and the relative abundance of fecal P. copri did not detect a significant relation between both variables in the different patient groups from SCREEN-RA/SCQM or Rheuma-VOR (Spearman’s correlation p value not significant, Table 3 ). Moreover, presence (relative abundance > 1%) or absence of P. copri in the fecal samples did not have an effect on the serum IgG responses against RPC01 in the different patient groups ( Figures 4B, C ). Hence, the current colonization of P. copri in the intestine unexpectedly appeared to not influence its serological recognition by the immune system.

In order to evaluate whether anti-P. copri IgG responses merely reflected total IgG levels in individuals, total IgG levels were determined in the sera of study participants. The comparison of total serum IgG levels between patient groups revealed a slight difference between the mean IgG concentrations in the serum of FDR 2 and FDR 3 individuals, but other diseases and disease subgroups did not differ from each other ( Figure 4D ). Importantly, linking the RPC01-specific IgG responses with the serum IgG concentrations for all participants of the study did identify a weak positive correlation between these two variables (Spearman correlation r = 0.159, P = 0.011 (*), Table 3 ). However, when dividing samples based on diagnoses, we only observed a significant, moderate positive correlation between both variables in the cRA group (Spearman r = 0.382, P = 0.009 (**)), but not others. We furthermore analyzed whether antibody levels correlated between different species of bacteria, which could reflect a generalized immune-reactivity against bacteria in individuals. We observed highly significant, weak to moderate positive correlations between serum IgG responses against RPC01 and P. gingivalis ( Figure 4E ), P. intermedia ( Figure 4F ), and P. melaninogenica ( Figure 4G ) across all samples, which was also partially reflected in the individual patient groups ( Table 4 ). In summary, these results suggest that serum IgG levels marginally determine the reactivity against P. copri across the different patient groups and that serum immune reactivity correlates between distinct bacterial species colonizing different mucosal sites.

Lastly, we assessed the potential influence of autoantibody presence or absence on the antibacterial IgG responses in the serum of FDR and cRA patients from SCREEN-RA/SCQM and new onset RA patients from Rheuma-VOR. Seropositivity for RF did not affect antibody reactivity against the oral pathobionts P. gingivalis, P. melaninogenica and P. intermedia but strikingly increased the response against P. copri strain RPC01 in chronic RA patients ( Figure 5A ). This effect was specific for RF, as we did not detect any differences in the IgG responses against the four bacteria based on ACPA seropositivity ( Figure 5B ). Hence, this data suggests the presence of RF autoantibodies in RA patients to enhance serological detection of P. copri and possibly indicates towards a direct interaction between RF autoantibodies and P. copri cells.