A subset of antibodies targeting citrullinated proteins confers protection from rheumatoid arthritis

Abstract

Although elevated levels of anti-citrullinated protein antibodies (ACPAs) are a hallmark of rheumatoid arthritis (RA), the in vivo functions of these antibodies remain unclear. Here, we have expressed monoclonal ACPAs derived from patients with RA, and analyzed their functions in mice, as well as their specificities. None of the ACPAs showed arthritogenicity nor induced pain-associated behavior in mice. However, one of the antibodies, clone E4, protected mice from antibody-induced arthritis. E4 showed a binding pattern restricted to skin, macrophages and dendritic cells in lymphoid tissue, and cartilage derived from mouse and human arthritic joints. Proteomic analysis confirmed that E4 strongly binds to macrophages and certain RA synovial fluid proteins such as α-enolase. The protective effect of E4 was epitope-specific and dependent on the interaction between E4-citrullinated α-enolase immune complexes with FCGR2B on macrophages, resulting in increased IL-10 secretion and reduced osteoclastogenesis. These findings suggest that a subset of ACPAs have therapeutic potential in RA.

Fig. 1. Reactivity of monoclonal antibodies to modified/non-modified cyclic peptides.

A library of 108 cyclic 17-mer COL2 peptides covering the whole sequence of mature human COL2 was synthesized (54 citrullinated and 54 corresponding non-modified arginine COL2 peptides) along with other indicated peptides, analysis of antibody specificities was carried out using established bead-based multiplex immunoassay (Luminex) and Log10-transformed MFI values at 1.0 μg/ml antibody concentration are presented as heatmap.

Fig. 2. Effects of mAbs in arthritis mouse models.

A Common ACPAs do not induce arthritis. A dose of 4 mg per mouse for each antibody was intraperitoneally injected on day 0 (n = 3 for LPS and 4 for others), followed by a boost with LPS on day 3. Arthritis scores are analyzed using Mann–Whitney test (two-tailed) and presented as mean ± SD. B E4 does not induce pain-like behavior. 2 mg of E4 and ACC1, and 4 mg of ACC4 or saline were injected on day 0 (n = 6 for ACC1, E4 and saline, and 12 for ACC4), mechanical allodynia was assessed by paw withdrawal response. Data were assessed using paired t test (two-tailed) and presented as mean ± SEM. C, D E4 protects Cia9i mice against collagen-antibody-induced arthritis (CAIA). Arthritogenic COL2 antibody M2139 (3 mg) was injected either (C) alone (n = 5 for M2139 + E4m and 6 for others) or (D) in combination with another COL2 antibody ACC1 (3 mg) (n = 6) through i.v or i.p on day 0, followed by a boost with LPS on day 3. E4/E4m/L2 (3 mg) were injected with or without COL2 antibodies on day 0. Data were analyzed using Mann–Whitney test (two-tailed) and presented as mean ± SD. E Representative H&E and tartrate-resistant acid phosphatase (TRAP) staining on joint tissue sections from the CAIA experiment. Scale bars represent 100 µm. F Histological assessment for each group (n = 5 for group iii and 6 for others). Data are analyzed using one-way ANOVA and presented as mean ± SEM. G Quantification of osteoclast numbers in cartilage sections from the TRAP staining (n = 5 for group i and 6 for others). Data are assessed using one-way ANOVA and presented as mean ± SEM. H E4 protects mice against glucose-6-phosphate isomerase (G6PI) induced arthritis (GPIA). DBA/1 mice were subcutaneously immunized with hGPI_325-339 peptide (10 μg/mouse) on day 0 together with E4 (2 mg, n = 9) or PBS (n = 15). Data are analyzed using Mann–Whitney test (two-tailed) and presented as mean ± SEM.

Fig. 3. Reactivity of E4 to mouse/human joint tissues.

A Immunofluorescence (IF) staining of E4/L2/E4m on naïve or CAIA DBA/1 mouse joint tissues taken on day 15 of CAIA. Antibodies were stained with goat-anti-mouse IgG antibody conjugated with CF488A and DNA was stained with Hoechst. Images were captured by confocal microscopy at ×10 magnification and the scale bars represent 200 μm. B IF staining of E4/L2/E4/ACC4 on naïve or arthritic joint tissue from FCGR2B KO mice taken on day 1 of CAIA. The staining was performed as above. Images were captured by confocal microscopy at ×10 magnification and the scale bars represent 200 μm. C IF staining of E4/L2/E4m on citrullinated extracellular matrix (ECM) from ATDC5 chondrocytes. Cells were cultured for 14 days with insulin-transferrin-selenium (ITS) following a treatment with recombinant human PAD4 (10 μg/ml) or PBS overnight before staining. E4/L2/E4m was stained as above, the citrullinated CII was labeled using biotinylated ACC4 antibody and stained with streptavidin-PE, and DNA was stained with Hoechst33342. Images were captured by confocal microscopy at 20× magnification. The scale bars represent 100 μm; D antibody-bound areas (n = 3 independent samples) were analyzed using ImageJ software. Data were assessed using one-way ANOVA and presented as mean ± SD. E IF staining of E4/L2/E4m on human cartilage explants from RA patients (RA01 & RA02) and a non-RA donor (HC01). Antibodies were stained as above. Images were captured by confocal microscopy at 10× magnification. The scale bars represent 200 μm.

Fig. 4. Reactivity of E4 to non-joint tissues/cells.

A E4 binding to skin tissues from naïve or CAIA arthritic joints. Immunofluorescence (IF) staining was performed using the indicated mIgG2b antibodies on skin tissue from naïve or arthritic joints from DBA/1 mice with CAIA (day 15), the antibodies were detected using a goat anti-mouse IgG secondary antibody conjugated with CF488A and DNA was stained with Hoechst. Images were captured by confocal microscopy at 10× magnification. Scale bars represent 200 µm. B E4 binding to naïve and inflamed lung tissue. IHC/H&E staining was performed using the indicated antibodies on either healthy or mannan-induced inflamed lung tissue taken from B10.Q mice. Images were captured by light microscopy at ×20 magnification and scale bars represent 100 µm. C E4 binding to human and (D) murine thymus tissue from B10.Q mice. Immunofluorescence (IF) staining of the thymus from two individuals or mice was performed using the indicated biotin-labeled antibodies labeled. For in vivo binding, biotinylated antibodies were pre-injected. Antibodies and cell nuclei were visualized by Streptavidin Alexa Fluor 555 (red) and Hoechst33342 (blue), respectively. Images were captured by confocal microscopy at ×10, ×20 or ×40 magnification. Scale bars represent 50 µm. E–G The binding of E4 to splenocytes in vivo. Splenocytes from naïve or CAIA arthritic mice injected with the indicated biotinylated antibodies (n = 3 or 4) were analyzed by flow cytometry. E, F Represent the general gating strategy. Macrophages are in the F4/80⁺Ly6G⁻ population gated on streptavidin-APC⁺ cells, dendritic cells in the CD11c⁺CD11b⁻ population gated on F4/80⁻Ly6G⁻ cells, monocytes in the Ly6C⁺streptavidin-APC⁺ population, and neutrophils in the F4/80^-Ly6G⁺ population gated on streptavidin-APC⁺ cells. Data are assessed by Mann–Whitney test (two-tailed) and presented in (G) as mean ± SD.

Fig. 5. Interaction between E4 and macrophages.

A Removal of Fc N-glycan abolishes the protective effect of E4 on CAIA. Arthritogenic antibody cocktail (3 mg M2139 + 3 mg ACC1) together with 3 mg of either E4, E4m, or E4-EndoS (Fc N-glycan cleaved) were injected to B10.Cia9i mice (n = 4 for M2139 + ACC1 and 5 for others) on day 0, followed by intraperitoneal injection of LPS on day 3. Arthritis scores are assessed using Mann–Whitney test (two-tailed) and presented as mean ± SD. B E4 does not protect FCGR2B KO mice from CAIA. Indicated antibodies were injected to FCGR2B KO mice (n = 6) on day 0 and LPS on day 5. Data were analyzed by Mann–Whitney test (two-tailed) and presented as mean ± SD. C, D E4 and L2 bind to LPS-stimulated macrophages. Mouse bone marrow-derived macrophages (BMDMs) were differentiated with M-CSF (20 ng/ml) for 6 days followed by overnight stimulation with LPS (100 ng/ml) or PBS (n = 3 biologically independent mice). Biotinylated antibodies (20 μg/ml) were added on day 7 for 1 h before flow cytometry. CD11b⁺F4/80⁺CD38⁺ macrophages bound by antibodies were gated by streptavidin⁺ population. Data are analyzed using one-way ANOVA and presented as mean ± SD. E E4 does not bind to FCGR2B KO macrophages. BMDMs from wildtype and FCGR2B knockout mice were similarly differentiated and stimulated with LPS. Macrophages were labeled with F4/80-specific antibody, and the binding of E4/E4m/L2 to macrophages was visualized using goat anti-mouse IgG-AF488. Confocal microscopy images are presented as single channels and merges, magnification ×20. Scale bars represent 100 µm. F, G E4 attenuates osteoclastogenesis. BMDMs were differentiated for 7 days, followed by 7 days of culture with 50 ng/ml RANKL and 5 μg/ml indicated antibodies before IF staining (n = 5 mice for RANKL group and 3 for others). Osteoclasts were visualized by phalloidin (green) and Hoechst 33342 (blue), confocal microscopy images were presented as merges, scale bars represent 100 µm, and the number of osteoclasts per field was counted. Data are analyzed using one-way ANOVA and presented as mean ± SEM.

Fig. 6. Reactivity of E4 ACPA to macrophage and RA synovial fluid (SF) proteins.

A–C Reactivity of E4 to mouse macrophage proteins. Naïve bone marrow-derived macrophages (BMDMs) (n = 3) were stimulated with LPS (500 ng/ml) overnight, and proteins were extracted for immunoprecipitation (IP) with ACPA. Numbers of captured proteins identified by mass spectrometry from each antibody are presented as (A) Venn diagram; correlation between log2-fold-change of E4/E4m (x-axis) and E4/L2 (y-axis) in (B) scatter plot; and proteins of interest in (C) boxplots. D, E Reactivity of E4 to human SF proteins. The CEP1-positive RA SF samples were subjected to IP with E4, L2 or E4m (n = 3). Captured proteins were presented as log2-fold-change (x-axis) and −Log10 of p value (p < 0.05, y-axis) in (D) volcano plot; and proteins of interest in (E) boxplots. Horizonal lines in the boxplots represent the median, 25th and 75th percentiles and whiskers represent measurements to the 5th and 95th percentiles. F Reactivity of ACPA to citrullinated human α-enolase (citENO1). ENO1 was citrullinated by hPAD4 with 100 mM Ca²⁺ and coated on ELISA plates. Antibody binding was detected using anti-mouse IgG Fc-HRP antibody with ABTS assay, results were presented as curve plot. G Reactivity of ACPAs to FCGR2B. CitENO1 was incubated with indicated biotinylated antibodies to form immune complexes (ICs). ICs were added to ELISA plate coated with recombinant FCGR2B and detected by Streptavidin-HRP. For unmodified ENO1/PAD4 controls, either PAD4 without ENO1 or ENO1 without PAD4 were used. Results were presented as curve plot. H E4-citENO1 IC increases IL-10 secretion from mouse macrophages. Wildtype or FCGR2B KO macrophages (n = 5) were treated with similarly prepared ICs (10 μg/ml) or LPS (500 ng/ml) overnight. Supernatants were concentrated for ELISA using the Europium assay. I E4 human IgG1 in complex with citENO1 increases IL-10 secretion from human macrophages. The hIgG1 E4-citENO1 IC was prepared similarly, the hIgG1 M2139-COL2 IC was used as control. The human CD14⁺CD16^- monocyte^-derived macrophages were treated with IC (10 μg/ml) or LPS (250 ng/ml) overnight(n = 3) and supernatants were collected for ELISA developed by ABTS assay. ELISA data are analyzed using one-way ANOVA and presented as mean ± SD.

Fig. 7. Schematic proposal for the function of E4 ACPA in effector arthritis.

In collagen antibody-induced arthritis (CAIA), macrophages are activated by LPS-mediated TLR signaling and arthritogenic immune complexes, leading to an increased secretion of PAD4. PAD4-mediated citrullination generates antigens that could be recognized by ACPAs. E4 binds to certain group of citrullinated antigens (e.g., ENO1) and forms local immune complexes that preferably interact with the FCGR2B on activated macrophages, delivering regulatory signals to macrophages, promoting IL-10 secretion and reducing osteoclastogenesis. The graph was created and licensed with Biorender.com.