Ten animals were used in this study. We immunized four H6 animals (carrying the shared epitope) with citrullinated peptides (n = 2) or native arginine peptides (n = 2) and two non-H6 animals were immunized with citrullinated peptides. Four control animals were used: two without systemic immunization and two with intradermal immunization with recombinant human myelin oligodendrocyte glycoprotein (rhMOG). A summary of the design is in Table S1 in Supplementary Material.

Adult captive-bred 3- to 5-year-old female cynomolgus macaques (M. fascicularis) were used. The study was approved by the regional animal care and ethics committee (Comité Régional d’Ethique sur l’Expérimentation Animale Île de France Sud, Fontenay-Aux-Roses, France; decision #A13_026). The CEA Institute was approved as compliant with ETS123 recommendations for animal breeding (European Union Directive 2010/63/EU, September 22, 2010) and with Standards for Human Care and Use of Laboratory Animals (Animal Welfare Assurance, OLAW no. #A5826-01). The study was also approved by the French department of education and research (MENESR; study no. 02769.01) as defined in French law “décret 2013-118 from 2013 Feb 1st.” At the end of each study, sedated animals were euthanized by intravenous injection of a lethal dose of pentobarbital.

Cynomolgus macaques (M. fascicularis) from Mauritius have a well-characterized genetic background and limited MHC diversity because of their isolation. Microsatellite analysis revealed six common haplotypes (H1–6) of MHC class II (8). The H6 genotype in the DRB1 region shares the same amino acid sequence as the 70–74 sequence of the human shared epitope DRB1 04-04 (QRRAA) (Table S2 in Supplementary Material). H6 animals were identified by a DNA microsatellite technique as described in Ref. (9). Presence of the shared epitope was confirmed after whole-blood RNA extraction with two specifically designed RT-PCR primers and probes followed by sequencing (Table S3 in Supplementary Material). Thus, the amino acid sequence on the macaque Mafa class II DRB alleles in positions 11 and 70–74 could be precisely identified.

Several proteins identified as targets of ACPAs include fibrinogen, vimentin, aggrecan, and enolase. Certain peptides from these proteins can induce a citrulline-specific T-cell response in patients with RA who carry the shared epitope (5, 10–12). We chose four peptides that are presented by DRB1 04-04 (5). These peptides originating from vimentin 59-71 (Vim59), vimentin 66-78 (Vim66) fibrinogen alpha chain 79-91 (Fg79), and aggrecan 89-103 (Agg89) (Table S4 in Supplementary Material) were synthesized in native or citrullinated form, the arginyl substituted by citrullyl residues (Genscript, Piscataway, NJ, USA).

Monkeys were immunized intradermally, each peptide injected individually, at 500 μg/peptide. Peptides were adjuvated with CpG and Montanide^® (kindly provided by Seppic, Puteaux, France) and injected in the back of the animals every 2 weeks for a total of four injections.

Peripheral blood mononuclear cells (PBMCs) were collected, and the T-cell response was assessed by comparing treatment with the pool of four citrullinated peptides used for immunization of animals to that of native peptides. T-cell response against citrullinated peptides and against arginine peptides was assessed using a macaque anti-interferon-gamma (IFNγ) Enzyme Linked Immunospot (ELISPOT) kit (Mabtech, Nacka Strand, Sweden). In total, 200 000 cells were exposed to a pool of 2 µg/mL of each peptide for 18 h. Positive controls were phorbol myristate acetate and ionomycin. Negative controls (cells with medium only) were subtracted from the sample for each animal. Positive spots were defined as 40 wide with 20 intensity using an AID EliSpot reader (Autoimmun Diagnostika, Strassberg, Germany). B-cell response against citrullinated peptides and arginine peptides was assessed using an in-house-designed ELISA (13). Briefly, peptides were passively coated overnight. After washing, blocking was performed for 1 h at 4°C with 2% bovine serum albumin (BSA). After washing, serum diluted in 2% BSA was incubated for 1 h at 4°C. Goat anti-monkey IgG (BioRad, Watford, UK) was then added to wells and revealed using 3,3′,5,5′;-tetramethylbenzidine (Lifetech, Villebon Sur Yvette, France). Results are displayed as ratio of optical densities to the preimmunization state for each animal. Ratios >2 were considered positive. Cytokine analysis was performed on plasma collected in heparin lithium tubes using a non-human primate 23-cytokine multiplex assay (Millipore, Guyancourt, France, Table S5 in Supplementary Material).

Animals were injected in the right knee with the citrullinated peptides Vim59, Vim66, Fg79, and Agg89 with or without incomplete Freund’s adjuvant (IFA) (Sigma-Aldrich, Saint-Quentin-Fallavier, France). Control animals were injected with IFA alone. Other control animals previously immunized with rhMOG (1–125) were injected intra-articularly with rhMOG + IFA. See experimental procedure in Method S6 in Supplementary Material.

Clinical inflammation was scored from 0 to 4. Systemic inflammation was measured by plasma C-reactive protein (CRP) level. Blood counting was performed at each blood sampling using an HMX Hematology analyzer (Beckman Coulter, Villepinte, France).

Tissue samples were fixed in 4% formalin, embedded in paraffin wax or 2-hydroxyethyl(glycol)methacrylate, then transversally cut into 5-μm thick sections. Sections were stained by a routine hematoxylin–eosin–safranin staining method. All samples were evaluated by a skilled pathologist in a double-blind manner, and all lesions were systematically recorded. Serial additional sections of the knee joint were stained with safranin O/fast green for better observation of osteochondral structures (Histalim, Toulouse, France).

When applicable, data were compared by non-parametric tests, and continuous values were compared by Mann–Whitney tests with GraphPad Prism 7. P < 0.05 was considered statistically significant.

At 4 and 13 weeks’ postimmunization, T-cell response to the pool of citrullinated peptides was greater for animals immunized with the citrullinated peptides (4/4 animals) than arginine peptides (0/2 animals) (Figure 1A). Likewise, T-cell response to the pool of arginine peptides was greater for animals immunized with the arginine version of the peptides (2/2 animals) than the citrulline version (1/4 animals) (Figure 1A). Therefore, after intradermal immunization, the T-cell response was rather citrulline- or arginine-specific depending on the type of peptides used for immunization. Analysis of responses to individual peptides showed a high response to citrullinated Vim59 and native Agg89, whereas all animals showed low or no response to citrullinated or native Fg79-91 (data not shown). Of note, the magnitude of the T-cell response was similar for H6 and non-H6 carriers (Figure 1B). Thus, H6 and non-H6 animals immunized with citrullinated peptides were pooled in the following experiments.

Serum antibody response against the four immunizing peptides was monitored individually during and after the immunization phase using ELISA. As for the T-cell response, the B-cell response, animals immunized with the citrulline or native version of the peptides were compared. All animals displayed a strong IgG response as compared with the preimmunization state (Figure 2). This response appeared between 4 and 6 weeks after the first immunization. For Vim59, the antibody response was systematically stronger against the citrullinated peptide than its arginine version (Figure 2A). However, the response was not exclusively dependent on citrullination because animals immunized with arginine peptides also showed a response against the citrullinated form (Figures 2A–C). In general, and contrary to the T-cell response, the B-cell response showed a high degree of cross-reactivity between antibodies against the citrullinated and the native peptide form, whatever the peptide used for immunization. Cross-reactivity was confirmed using an ELISA competition assay—preincubation with arginine or citrulline peptides inhibited anti-citrullinated response in a similar fashion (Figure S7 in Supplementary Material). Of note, all animals had a low or no response to citrullinated or native Fg79. Similar to the T-cell response, the presence of the shared epitope did not affect the B-cell response against the citrullinated or native version of peptides overall (Figure S8 in Supplementary Material).

During this first systemic immunization phase, despite a significant T- and B-cell response, no clinical manifestations of arthritis were noted. Because in humans, ACPA positivity can be asymptomatic for years before the occurrence of arthritis, second hits may be required to induce arthritis in immunized humans or animals. As a second hit, we chose unilateral injection of citrullinated peptides in the right knee 30 weeks after initial immunization. Five of the six animals previously immunized with citrullinated (n = 3) or arginine (n = 2) peptides were available for this second step. IA injection of citrullinated peptides without IFA did not induce clinical signs or local immune response. Then, previously immunized animals were unilaterally injected in the right knee with IFA alone (n = 2) or citrullinated peptides + IFA (n = 3). IFA alone led to a 4-day unilateral knee swelling. Conversely, with IA injection of citrullinated peptides + IFA, two of three animals showed a 7-week unilateral swelling with joint effusion (Figures 3A,B). No other joint was clinically involved. This prolonged mono-arthritis resulted from both previous systemic immunization with citrullinated peptides and knee injection of citrullinated peptides + IFA. Indeed, the same knee injection of citrullinated peptides in two non-immunized animals led to a 4-day transient swelling similar to IFA injection alone.

To study the peptide specificity, IA injection of rhMOG + IFA in two animals previously immunized with rhMOG induced only mild transient swelling for 4 days, similar to that with IFA alone (Figure 3A). Finally, IA immunization with arginine peptides + IFA in a systemically citrullinated primed animal did not yield prolonged swelling. Therefore, only specific and citrullinated antigens could induce prolonged arthritis in an animal immunized against this antigen.

Among the three animals with an IA boost with citrullinated peptides, the two showing prolonged mono-arthritis were H6-positive and the one showing only a short episode of inflammation was not.

The T-cell response was striking with IA boost with citrullinated peptides + IFA versus IFA alone (3/3 vs 0/2 animals) (Figure 4A). The magnitude of the response after IA boost was two to six times higher than after the intradermal prime, and three boosts with similar quantities of peptides and after late intradermal boost (Figure 4B). This response retained a degree of specificity to citrullinated peptides, with a higher response against these citrullinated peptides versus the pool of arginine peptides. Control animals receiving IA IFA alone showed no response. The B-cell response, which was high before the IA boost, was increased in one of the three animals receiving an IA boost with citrullinated peptides + IFA and in none of the others (Figure 4C).

The striking immune activation after the IA boost with citrullinated peptides + IFA but not IFA alone suggests that the prolonged articular swelling was due to an immune response against the citrullinated peptides, whereas the transient swelling was due to a non-specific inflammatory response to IFA.

This finding is further emphasized by the level of systemic inflammation measured by CRP level in serum at day 1 after the boost. CRP level was higher for animals that received an IA boost with citrullinated peptides + IFA than IFA alone (3/3 animals > 100 mg/mL vs 2/2 animals < 60 mg/mL). rhMOG + IFA or citrullinated peptides + IFA, without a previous prime; all displayed significantly lower levels of CRP (Figure 5A).

This increase in CRP level occurred at the same time (day 1) as a reduced lymphocyte count and an increased neutrophil cell count in blood (Figure 5B). Among a panel of 23 analyzed cytokines (Table S5 in Supplementary Material), a significant increase in ratio of circulating cytokines on day 1 to the pre-IA injection level was limited to proinflammatory and angiogenic cytokines [interleukin (IL)-6, IL-1B, IL1-RA, and VEGF] and granulocyte-stimulating chemokines (GM-CSF and MIP1A) in animals that received an IA boost with citrullinated peptides + IFA versus IFA alone (Figure 5C).

Articular swelling was confirmed by taping effusion liquid. During the first phase of aspecific swelling that appeared in most animals, low cellularity with a predominance of granulocytes occurred (Figure 5D left panel). At later times (7–14 days postinjection), animals with prolonged swelling upon IFA + peptide injection presented high cellularity (Figure 5D middle and right panel).

Histological analysis of the synovial tissue of the injected knee performed at euthanasia 15 weeks after articular challenge revealed different types of lesions in the injected knee without any clear differences between an IA boost with citrullinated peptides + IFA or IFA alone: unspecific adjuvant-related granulomas (Figure 6A), diffuse mononuclear-cell and granulocyte infiltration of the synovial layer (Figure 6B) with lymphocyte infiltration foci (Figure 6C), and synovial proliferation (Figures 6D,E) without any bone or cartilage lesions (Figure 6D). Finally, systematic sampling of metacarpophalangeal and metatarsophalangeal joints revealed synovial proliferation in two animals with citrullinated peptides + IFA injection (Figure 6F): the animal that died prematurely and thus did not receive the IA boost and an animal with prolonged knee swelling after IA boost with citrullinated peptides + IFA. These last histological lesions were present despite the absence of clinical swelling in these joints.