Analysis of Complement Gene Expression, Clinical Associations, and Biodistribution of Complement Proteins in the Synovium of Early Rheumatoid Arthritis Patients Reveals Unique Pathophysiologic Features

Rheumatoid arthritis (RA) is a complex, heterogeneous, and chronic immune-mediated joint disease marked by synovial inflammation as well as bone and cartilage destruction (1). RA-related disabilities in patients in the United States have been projected to increase during the next 25 y by 40% (2). Although substantial progress has been made in the development of new treatments for patients with RA, currently there are no therapeutics based on the complement system (CS) that are approved for use in RA. Thus, although the pathway is known to be centrally involved in the pathogenesis of experimental models of RA (3), the roles of specific complement factors in both early RA (<1 y of clinically apparent arthritis) and chronic disease are unknown.

The generation of anti-citrullinated protein Abs (ACPAs) (4, 5) in the early developmental (asymptomatic) preclinical phase of RA has a strong predictive value for progressing to clinically apparent disease (6). Despite these findings, major questions remain as to why the presence of ACPAs and their associated effector mechanisms do not lead to synovitis for many years, and likewise leaves their precise role in the transition from asymptomatic to symptomatic disease unknown. Similarly, rheumatoid factor (RF) and autoantibodies against the Fc portion of IgG are also present long before synovitis development (7). Nonetheless, once synovitis develops, both ACPAs and RF are associated with more severe disease and joint damage (8, 9).

In addition to being one of the major pathways of innate immunity that can be engaged by autoantibodies, the CS can generate several effector responses relevant to RA such as leukocyte chemotaxis, phagocytosis, inflammation, cell damage, microorganism opsonization, and activation of the adaptive immune responses (10, 11). To accomplish these roles, the CS is initiated and amplified by three interrelated pathways: the classical pathway (CP), lectin pathway (LP), and alternative pathway (AP). The CP is activated by binding of IgG or IgM with C1q, followed by C1s protease-driven C4 and C2 cleavage; notably, C1q can also directly activate the CP through binding with apoptotic cells, necrotic cells, and the pentraxin C-reactive protein (12, 13). In humans the LP also utilizes proteases to cleave C4 and C2, while target recognition is driven by mannose-binding lectin 2 (MBL2), ficolins (FCN1, FCN2, and FCN3), and/or collectins (CL-10, CL-11, and CL-12) (14, 15). The AP is constantly active at low levels and has four major complement components: complement component 3 (C3), complement factor B (CFB), complement factor D (CFD), and complement factor properdin (CFP) (3).

All three CS pathways converge to cleave C3, through multiprotein complexes designated C3 convertases, into the chemotactic peptide C3a and target-bound C3b. This step is followed by cleavage of C3b by factor I and cofactors into iC3b and then C3dg plus C3c (complement component c fragment of C3). C3c is released into the local environment, but iC3b and C3dg remain bound to the target surface and can exert signaling functions through engagement of specific receptors on inflammatory cells (16). The cleavage of C3 into C3b leads to the formation of a C5 convertase that cleaves C5 into C5a and C5b. C5a, another chemotactic factor, activates C5aR-positive cells and is found in higher levels in the synovial fluid of RA patients (17, 18). Anti-C5a or oral anti-C5aR1 inhibitor ameliorates arthritis development in mouse and rat models of RA (19, 20). C5aR1 deficiency protects mice from collagen Ab-induced arthritis (21), and we have shown that combined C5-C5aR1 blockade also protects from arthritis in this model (22). C5b further nucleates the C5b-9 complex, also known as the membrane attack complex (MAC), which is responsible for cell lysis and direct nucleated cell activation (23).

To prevent damage to the host, the CS is finely regulated by fluid phase and cell surface–bound molecules that control complement activation at these sites and noncellular surfaces such as cartilage and matrix (11, 24–28). Decay-accelerating factor (DAF, CD55) dissociates C3 and C5 convertases, while membrane cofactor protein (MCP, CD46) acts with factor I to cleave and both inactivate as well as generate receptor ligands for both C3 and C4. Among the most important fluid phase regulators are human complement factor H (CFH) and factor H-like 1 (FHL-1), both of which are encoded by the single CFH gene through alternative splicing. In addition, factor H–related proteins, including CFH-related protein (CFHR)1, CFHR2, CFHR3, CFHR4, and CFHR5, are encoded by their individual CFHR genes (24, 29–31). CFH and CFHR genes are primarily transcribed in the liver, and proteins are found in the serum/plasma; however, many cells such as retinal pigment epithelial cells, PBLs, myoblasts, rhabdomyosarcoma cells, fibroblasts, umbilical vein endothelial cells, glomerular mesangial cells, neurons, and glial cells also express these proteins (32). We have shown that endogenous CFH makes a significant contribution to inhibition of the AP in a mouse model of RA through binding to sites of immune complex formation and complement activation (33).

Complement activation is essential for disease progression in active and passive transfer mouse models of RA, and activated complement fragments have been found in the synovium of RA patients (34–36). Thus, CS pathways and its proteins can in principle play an important role in the pathogenesis of RA (3), but the exact role locally in the synovium of RA patients is unknown. Recently, we have reported that subjects with classified RA and ongoing joint inflammation demonstrate evidence through analysis of blood-based CS biomarkers of systemic complement activation, but RA-related autoantibody-positive subjects at risk for future RA do not (37). These results suggested that the CS becomes involved in joint disease once synovitis has begun and may contribute to that process locally. CS gene expression and its relationships with DAS28-ESR (disease activity score in 28 joints for erythrocyte sedimentation rate) and the presence of complement proteins locally in synovial biopsies from early RA patients have not been previously examined. However, prior studies have shown using in situ hybridization, immunohistochemistry, and Western blot analysis the local presence of C3, FB, C5b-9, C3aR, and C5aR in the synovium of patients with longstanding chronic RA (38). It also remains uncertain whether complement proteins present in the synovial fluid and synovium in RA are primarily serum derived or synthesized locally by the synovial mononuclear phagocytes or by cells of the synovium (39–42).

In the Pathobiology of Early Arthritis Cohort (PEAC) study, a comparison of directly matched peripheral blood and synovium from patients with enthesitis-related arthritis (ERA) was reported, providing comprehensive RNA sequencing analysis of potential molecular pathways that could drive early RA progression (43). This study developed a remarkable data exploration Web site tool to dissect gene signatures in the synovium and blood in a manner that is integrated with clinical and histologic phenotyping (43). The authors used ultrasound-guided synovial biopsies from early RA patients and identified transcriptional subgroups in the synovium linked to three synovial pathotypes, including fibroblastic rich (also known as the pauci-immune pathotype), macrophage rich (also known as the diffuse myeloid pathotype) and T cell, B cell, myeloid, and plasma cell rich (also known as the lympho-myeloid pathotype) (43).

In this early RA synovium study, variations in complement gene and protein expression were not specifically reported. Based on a combination of prior human RA and murine model studies, we hypothesized that excessive CS activation or dysregulated complement gene and protein expression in the synovium of early RA patients might contribute to the disease pathogenicity during the transition of disease from no synovitis in preclinical RA to synovitis in RA.

In the current study, we explored CS gene and protein expression locally in the synovium from early RA patients and also evaluated the presence of regulatory proteins and receptors. Our results are consistent with an important role for the CS in the synovium in early RA and, importantly, that regional or cell-specific synovial tissue dysregulation associated with an imbalance of activation versus regulatory proteins likely plays an important role in this phase of disease.

To evaluate the expression of CS and related genes (Tables I, II, and III), we evaluated data and performed analyses from the PEAC study Web site (https://peac.hpc.qmul.ac.uk/) (43). In the PEAC study, histological, clinical parameter, and ultrasound analyses for blood and synovium were performed in patients with early RA at baseline before treatment initiation and repeated after 6 mo of treatment. We correlated baseline expression levels with DAS28-ESR in the paired synovial and blood compartments from early RA patients (n = 90), integrated with deep phenotypic profiling (43). Briefly, in this study a total of 46 (51%) synovial biopsies were classified as lympho-myeloid (B cells, T cells, plasmacytoid dendritic cells, NK cells), 21 (23%) as diffuse-myeloid (macrophages, basophils, eosinophils, neutrophils), 17 (19%) as pauci-immune fibroid (synoviocytes, no immune cells), and 6 (7%) as unclassified by histological analysis. The correlations between DAS28-ESR and individual complement factor gene expression from early RA synovial biopsies and blood, regardless of the pathotypes, were analyzed. Due to the functional connection of complement proteins with FcγRs (44, 45), we also examined the correlations between DAS28-ESR and individual FcγR gene expression in the synovium and blood (Table IV).

The characteristics of early RA patients (n = 23) from the AMP study are shown in Table V. Synovial biopsies from osteoarthritis (OA) patients and other human tissue samples were used as negative controls. Synovial tissue sections from these subjects were used for multiplexed immunohistochemistry (MIHC) and imaging analysis. The disease duration of early RA patients from which we obtained synovial biopsies was <12 mo (22 subjects), and one subject was ≤14 mo. Individuals were evaluated at the time of biopsy for ACPA positivity and other blood markers. A written informed consent was obtained from all early RA patients according to the approved Institutional Review Board protocol.

Briefly, synovial biopsies (<1–2 mm) (n = 23) from early RA male (n = 9) and female patients (n = 14) were obtained under local anesthesia from the knee, ankle, wrist, or metacarpophalageal joints (43, 46). Synovial biopsies were fixed in 10% neutral buffered formalin and embedded in paraffin wax, followed by sectioning and staining with H&E. Formalin-fixed, paraformaldehyde-embedded synovial sections were also processed for seven-color complement-related proteins as well as for immune cell MIHC staining and imaging analysis.

Prior to MIHC analysis, one section from each early RA synovial biopsy was used to examine the quality and the presence of synovial lining. Formalin cross-links epitopes, so to unmask epitopes and to detect various complement proteins in the early RA biopsies, we used for MIHC the following Abs as well as Ag retrieval procedures. For C3c, cell conditioner 1 (CC1) (Ventana Medical Systems), a mild Tris-based buffer, was used for an Ag retrieval using a Ventana stainer. The dilution of primary C3c Ab (LSBio) used was 1:800 and was incubated for 16 min at 37°C. Detection of C3c was done using Ventana iVIEW diaminobenzidine (DAB), secondary replaced with rabbit ImmPRESS polymer (Vector Laboratories), and enzyme replaced with rabbit ImmPRESS polymer at 50% strength (Vector Laboratories). For CFH, 10 mM sodium citrate solution (pH 6.0) was used for 10 min at 110°C. The dilution of primary CFH Ab (Abcam) used was 1:200 and was incubated for 32 min at 37°C. Detection of CFH was done with Ventana iVIEW DAB, substituting IgG with full-strength goat ImmPRESS (Vector Laboratories), and substituting streptavidin-HRP with half-strength goat ImmPRESS (diluted in PBS), all on the Ventana Benchmark XT. For CFHR4, 10 mM sodium citrate solution (pH 6.0) was used for 10 min at 110°C. The dilution of primary CFHR4 Ab (4E2) (a gift from Dr. Paul Morgan, U.K.) used was 1:400 and was incubated for 32 min at 37°C. Detection of CFHR4 was done with Ventana UltraView DAB, all on the Ventana Benchmark XT. For MBL2, BORG (Biocare Medical), a Tris-based solution, was used for Ag retrieval for 10 min at 110°C in a Decloaking Chamber (Biocare Medical) with a 10-min cool down. The dilution of primary MBL2 Ab (Abcam) used was 1:100 and was incubated for 32 min at 37°C. Detection of MBL2 was done using Ventana UltraView DAB universal polymer for mouse and rabbit Abs. For FCN3, 10 mM sodium citrate solution was used as mentioned above. The dilution of the primary FCN3 Ab (MyBioSource) (rabbit polyclonal) used was 1:500 and was incubated for 32 min at 37°C. Detection of FCN3 was done using Ventana UltraView DAB. For CFB, Ag retrieval was done using 10 mM sodium citrate (pH 6.0) + 0.1% Tween 20 with incubation for 10 min at 110°C. The dilution of primary CFB Ab (Invitrogen) (rabbit polyclonal) used was 1:1300 incubated for 32 min at 37°C. The detection of CFB was done using Ventana UltraView DAB. For C5b-9, cell conditioner 1 (CC1) (Ventana Medical Systems), a mild Ag retrieval procedure, was used on the Ventana stainer. The dilution of the primary C5b-9 (MAC) (Abcam) used was 1:500 for 16 min at 37°C. The detection of C5b-9 was done using Ventana iVIEW DAB, secondary replaced with rabbit ImmPRESS polymer (Vector Laboratories), and enzyme replaced with rabbit ImmPRESS polymer at 50% strength (Vector Laboratories). All IHC incubations were performed on the Ventana Benchmark XT autostainer and counterstaining was done with Harris hematoxylin with a 2-min incubation. Various human tissue samples (surgical discard) such as synovium, kidney, placenta, colon, lungs, liver, and spleen were used as negative and positive controls to examine the specificity of the complement Abs. All sections were examined, scanned, and imaged.

After confirming the integrity and architecture of the synovial biopsies, simultaneous staining for the seven CS proteins as above was performed. To detect and quantitate levels of seven complement proteins and inhibitors using MIHC, early RA synovial sections after dewaxing (Leica Biosystems) were heat treated in epitope retrieval solution 2 (ER2, EDTA buffer with pH 8.9–9.1) or epitope retrieval solution 1 (ER1, citrate buffer with pH 5.9–6.1) depending on the Ab for 20 min at 93°C (Leica Biosystems), blocked in Ab diluent (Akoya Biosciences, Marlborough, MA), and incubated for 30 min with the primary Ab, 10 min with HRP-conjugated secondary polymer (anti-rabbit and anti-mouse, Akoya Biosciences), and 10 min with HRP-reactive OPAL fluorescent reagents (Akoya Biosciences). Slides were washed between staining steps with BOND Wash (Leica Biosystems) and stripped between each round of staining with heat treatment in Ag retrieval buffer. After the final heat treatment in Ag retrieval buffer, the slides were stained with spectral DAPI (Akoya Biosciences) and coverslipped with ProLong Diamond mounting media (Thermo Fisher Scientific). To detect cell nuclei in the synovium, DAPI was used.

After MIHC staining, whole-slide scans of the synovial biopsies were imaged on the Vectra Polaris automated quantitative pathology imaging system (Akoya Biosciences) using a ×20 objective with 0.5-μm resolution. In parallel, tissue sections from human tonsil, human kidney, and human synovium from OA patients after knee replacement were used as a positive control for all complement protein analysis. To compare directly complement protein expression, independent panels were created and then merged, consolidated, and analyzed in R studio using the phenoptrReports plug-in (Akoya Biosciences). The seven color images for complement proteins were analyzed with InForm software (v2.4.8, Akoya Biosciences) to unmix adjacent fluorochromes and subtract autofluorescence, followed by quantification using a specific algorithm for complement proteins.

The systemic levels of CFHR4 protein levels using sera from our previous study (37), not related to PEAC or AMP, from healthy controls (n = 32), at-risk RA subjects (n = 18), and RA patients (n = 16) were measured using ELISA according to the procedure described (47). Briefly, for a CFHR4 ELISA we used monoclonal anti-FHR4 Ab 4E9 (5 μg/ml) diluted in 0.1 M carbonate buffer (pH 9.6) to coat wells. Sera from all three groups were diluted (1:40) and incubated for 1.5 h at 37°C. After washing, CFHR4 protein was detected using secondary HRP-conjugated anti–CFHR-4 mAb (clone 17) (1 µg/ml). Absorbance was measured at 492 nm. CFH levels were also measured by ELISA previously in the sera from these three groups as a part of separate study and were used for comparison with CFHR4 (37).

Transcript abundances of synovial and whole-blood biopsies from early treatment-naive RA patients were loaded into R using methods previously described (43). Count data were normalized with batch, sex, and pathotype as covariates using DESeq2 (48). Gene expression data then underwent a regularized log transformation. Using the volcano3D package, the differential gene expression shown in Supplemental Fig. 1 was mapped to pathotype vectors and plotted on three axes for lympho-myeloid (L), diffuse-myeloid (M), and pauci-immune fibroid (F) using polar coordinates in the horizontal plane (49). To assess the correlation between complement and FcγR gene expression and DAS28-ESR from PEAC data, adjusted p (p-adj) values from Spearman correlation in the synovium and blood were used, regardless of the specific pathotypes. To compare the ratios of CFH to CFHR4 proteins among control, at-risk, and RA subjects, a one-way ANOVA test was used, and data were plotted using GraphPad Prism 5.

Correlations between DAS28-ESR and complement gene expression in the synovium and blood using PEAC data from early RA individuals were performed (Table I). Notably, correlations with specific complement activation pathway gene expression were found only in the synovium but not in the blood. With regard to pathways involved, mRNA levels from three genes, namely C2 (p-adj = 0.023), FCN1 (p-adj = 0.018), and CFB (p-adj = 0.052), uniquely belonging to each of the three of the complement initiation and amplification pathways, were significantly and positively correlated with DAS28-ESR in the synovium, but none was correlated in the blood (Table II). Representative correlations between CFB mRNA expression and DAS28-ESR in both synovium and blood are shown (Fig. 1A, 1B). Importantly, complement C5 mRNA expression (p-adj = 0.019) (Table II) (Fig. 1C, 1D), which has been the target of several prior clinical trials, as well as the additional MAC components C6 and C7 all exhibited an inverse synovial expression correlation with DAS28-ESR. In addition, for Collectin12 (p-adj = 0.02), an LP activation pathway component, gene expression in the synovium was inversely correlated with DAS28-ESR (Table II). Notably, no correlation of C3 mRNA expression, the major component of the CS upon which all three activation pathways are focused, with DAS28-ESR was observed in the synovium or blood (Table II).

To further extend these analyses, correlations between DAS28-ESR and gene expression of complement receptors, proteases, and regulatory proteins in the synovium and blood were explored (Table III). Similar to the complement activation pathway gene expression findings, there were several positive correlations detected only in the synovium. These included C3AR1 (p-adj = 0.03), C5AR1 (p-adj = 0.013), and CR1 (CD35) (p-adj = 0.0044) in the synovium and not in the blood; in contrast, no correlations were detected with C5AR2 in either location (Table III) (Fig. 2). Of note, the LP protease MASP-1 (r = 0.44, p-adj = 0.0042) and the complement regulatory protein CFH (r = 0.39, p-adj = 0.012) mRNAs were highly negatively correlated with DAS28-ESR (Table III). Interestingly, MCP (CD46), a complement regulator that is expressed on the membrane of all circulating and tissue cells except RBCs, was also negatively correlated (p-adj = 0.038) with DAS28-ESR (Table III). These results suggest that decreased expression of both protein regulators CFH and MCP/CD46 are associated with more active disease, perhaps due to insufficient ability to control the multiple synovial complement activation pathway mechanisms that are present. No such correlations for another complement receptor, CR2, was found at either site (Table III). Overall, positive correlations in the synovium between receptors for the activation fragments C3a and C5a, as well as an inverse correlation with regulators CFH and MCP, suggest that an imbalance between these components may be causally associated with increased disease activity.

As different pathotypes are present in the RA synovium, we sought to determine whether there were relationships between individual complement gene mRNA levels and those features. Notably, regarding CFB levels, significant differences were found between the lymphoid pathotype and fibroid pathotype, but again only in synovium and not in blood (p-adj < 0.0065) (Supplement Fig. 1). No correlations were seen among lymphoid, myeloid, and fibroid pathotypes and C3 gene expression in the synovium as well as in the blood (Supplement Fig. 2A, 2B). Interestingly, there were also highly significant (p < 0.05) differences across pathotypes regarding CR1 and C3AR1 gene expression in the synovium but not in the blood where the lymphoid pathotype had the highest expression followed by myeloid then fibroid (Supplement Fig. 2C–F). In addition, in the synovium, C5AR1 gene expression was significantly (p < 0.05) higher in the lymphoid versus fibroid pathotype and also higher in the myeloid compared with the fibroid pathotype (Supplement Fig. 2G, 2H).

Potential correlations between DAS28-ESR and FcγR gene expression were also explored in the synovium and blood from early RA patients (Table IV) (Supplement Fig. 3). Notably, there were positive correlations between DAS28-ESR and FcγR gene expression in the synovium but also not in the blood. Specifically, we found in the synovium that there were positive correlations between DAS28-ESR and activating FcγR1A gene expression (p-adj = 0.0069), FcγR1B (p-adj = 0.0085), FcγR2A (p-adj = 0.027), and FcγR3A (p-adj = 0.037), although there was only a trend for FcγR3B (p-adj = 0.092) (Supplement Fig. 3A, 3C, 3E, 3L, 3K). In contrast, no correlation was seen among lymphoid, myeloid, and fibroid pathotypes in the synovium or in the blood between DAS28-ESR and FcγR2B gene expression, which is an inhibitory receptor (Table IV) (Supplement Fig. 3B, 3D, 3F, 3H, 3J, 3K). However, on further analysis of the synovium there was a differential expression of FcγR2B in lymphoid and myeloid pathotypes compared with the fibroid (Supplement Fig. 3G). There was no differential FcγR2B expression among all three pathotypes, that is, lymphoid, myeloid, and fibroid pathotypes in blood (Supplemental Fig. 3H).

With regard to other pathotype relationships, the expression of all FcγRs, except FcγR3B, in the synovium the lymphoid pathotype had the highest expression followed by myeloid pathotype then fibroid pathotype (Supplement Fig. 3). The finding that FcγR gene expression correlations in the early RA synovium but not in the blood suggests that these receptors, which similar to complement factors are effectors of the autoantibody response, also play an important role in the local synovial pathogenesis of RA.

The synovial (n = 87) and blood (n = 67) gene expression of 48 complement genes from PEAC data were correlated with the change in disease activity, measured by DAS28-ESR, between baseline and 6 mo to calculate Spearman’s rank correlation coefficients and corresponding p values (Fig. 3). The significance (–log₁₀ p values) multiplied by direction (sign r) of correlations in blood versus synovium is shown. Genes significantly (p < 0.05) altered in synovium only are shown in blue, those significant in blood only are in red, and those that are nonsignificant are in gray. Surprisingly, with regard to positive correlations with initial disease severity, the correlation bubble plot showed that only CFHR4 gene expression was significantly positively correlated with the ΔDAS28-ESR in the synovium after 6 mo of treatment, but not in the blood (Fig. 3A). The striking trend in the CFHR4 gene expression data show that its higher level of gene expression at the baseline, only in the synovium, is associated with less disease improvement at 6 mo.

To determine whether this was a phenotype also found in blood protein factor levels or was associated with differences in ratios between CFH and CFHR4, we determined the absolute levels of these proteins by ELISA in the serum samples from healthy controls, ACPA-positive at-risk individuals, and RA patients. Although the CFHR4 mRNA expression was statistically associated with higher disease activity in the synovium, no difference was found in the absolute levels of CFHR4 in the sera from these groups (Fig. 3B), which is consistent with the similar blood CFHR4 mRNA quantitation. Furthermore, there was not a difference in the absolute levels of CFH or in the ratios of CFH/CFHR4 (Fig. 3B). Synovial samples were not available to perform a similar type of longitudinal analysis. As CFHR4 is a complement modulator, these results suggest that it may play a local synovial role either in the control of complement activation or in another inflammatory pathway related to therapeutic responsiveness.

To examine complement protein levels in the synovium and assure that expression was present of relevant factors for which there were positive correlations between DAS28-ESR and mRNA expression, we examined using immunohistochemistry and MIHC staining synovial biopsies from early RA patients from the AMP RA/SLE study (50) (Fig. 4). In addition to complement factors identified by positive correlations with DAS28-ESR, we examined C3c and C5b-9, as murine studies have shown that both C3 and C5 convertases are required for full disease expression (3, 21, 35, 51).

Histopathology and immunohistochemical studies using early RA biopsies were performed to optimize detection of complement proteins (Fig. 4). Using this approach, we confirmed the individual presence of seven complement proteins and complexes, that is, C3c, CFHR4, CFB, CFH, FCN3, MBL2, and C5b-9 (MAC), in the synovial biopsies (Fig. 4B–H). Representative histopathological and immunohistochemical images along with DAPI are shown (Fig. 4A, 4I). DAPI was used in MIHC. Overall, we found that C3c and CFH proteins were highly expressed in all biopsies (Fig. 4B, 4E). C3c was often observed to be more predominantly present in the synovial membrane of early RA biopsies (Fig. 4B). MBL2 was present as was FCN3 (Fig. 4F, 4G). Notably, C5b-9 was detected only at modest levels in synovial biopsies (Fig. 4H). CFB was also present, predominantly in the subsynovial lining area of the biopsy shown (Fig. 4D). In parallel, we have used OA synovial biopsies and other biopsies (liver, kidney, and lung) as controls (Supplement Fig. 4). These data confirm that individual complement proteins identified in the mRNA analyses are indeed expressed in the synovium of early RA patients and thus could be playing a role locally in the joints in the development of arthritis and tissue damage.

To further confirm the interrelationships among the expression of complement genes and their respective proteins, MIHC and simultaneous imaging analysis was performed for the same seven complement proteins on synovial biopsies from early RA patients (Fig. 5). After individual staining of C3c, CFHR4, CFB, CFH, FCN3, MBL2, and C5b-9 (MAC), a composite image was generated from each synovial biopsy (Fig. 6A) (Table V).

First, by imaging the entire synovium section from each patient, relative complement protein expression was quantified randomly and blindly by selecting entire sections for comparison. C3 was present in many cells, and CFH was present and highly expressed at levels significantly (p = 0.044) higher than for CFB (Fig. 5). There were significant differences in the levels of CFH versus CFHR4 (p = 0.008) (Fig. 5). MBL2 was significantly (p = 0.006) expressed in the subsynovial areas compared with the FCN3, but the presence of both suggests that in early RA the LP might be a relevant mechanism to initiate activation of the CS (data not shown). Further in-depth quantitative overlapping analysis of the composite images revealed that a substantial number of the cells in the synovium were CFH⁺ followed by C3c⁺/CFB⁻/CFH⁻ and C3c⁺/CFB⁻/CFH⁺, respectively. This overlapping complement protein expression analysis indicated that synovial cells also expressed more than one complement protein in the early RA biopsies directly involved in regulating the AP. Representative composite images of the complement protein expression data in the early RA synovial biopsies are shown (Fig. 6A, 6B).

An additional feature that was present in subdomains of biopsies was the lack of consistent expression of C3c relative to CFH protein (Fig. 7). In some biopsies CFH was clearly present in the synovial membrane, in the complete absence of C3c (Fig. 7B). In contrast, in other regions of early RA synovium, CFH and C3c were completely separated, reflecting a potential imbalance of activation versus regulation in a localized manner (Fig. 7C). More examples of regional imbalance and compartmentalization of complement proteins are shown in composite images (Fig. 7D–I). Of note, there was more C3c deposition in the interspace among adipocytes (Fig. 7E), whereas around blood vessels there was more CFH deposition (Fig. 7D). A clear compartmentalization of CFH and CFB along with C3c colocalization was seen in the subsynovial lining area and deep subsynovial lining areas, respectively (Fig. 7G). We also quantified the presence of complement proteins in the synovial membrane and subsynovial lining areas, respectively (Fig. 8A). A regional relationship was found in these areas, as CFH was significantly (p < 0.05) higher in both of these areas compared with CFB (Fig. 8B, 8C). CFH was also significantly (p < 0.046) higher in the synovial membrane than C3c; however, it was not statistically significant in the subsynovial lining area (Fig. 8C). Increasing trends of FCN3, MBL2, and MAC were seen in the subsynovial lining areas compared with the synovial lining, but these were not significant (Fig. 8B, 8C). In sum, the general relationships between activation (i.e., C3c and CFB) and regulation (i.e., CFH) are maintained, promoting overall control of the CS. However, as show in (Fig. 7, there are many individual areas within these regions that demonstrated striking differences, suggesting that there are context-dependent activation sites. This very localized imbalance of activation and regulatory proteins suggests that dysregulation may play a particularly important role in the response of some regions within the synovial tissue to complement-activating stimuli and generation of effector molecules.

In this study, we have explored the expression of the complement activation pathway, receptor and regulatory genes, and their correlations with DAS28-ESR in synovium and blood. We have also examined, focusing on factors that display clinical associations with disease activity or major roles in murine models, synovial protein expression using eight-color MIHC and imaging analysis in tissue from early RA patients in the AMP RA/SLE study and identified additional features that are likely to be involved in dysregulation of the system.

We first found that there were strong correlations between DAS28-ESR, as a measure of clinical disease activity, with a subset of complement gene mRNA expression, notably with activation pathway factors belonging to all three pathways of the CS. Importantly, the correlations between complement gene mRNA expression and DAS28-ESR were also exclusively found locally in the synovium and not in peripheral blood, emphasizing the importance of local complement activation and effector functions. In addition, the inverse correlations between synovial CFH and MCP (CD46) gene expression and clinical activity suggest that inadequate levels of regulatory proteins are present to control robust complement-activating factors.

These findings are striking relative to those in extensive studies of murine models of human RA. For instance, in the collagen Ab-induced arthritis (CAIA) model, all three pathways contribute to the initiation and then propagation of joint damage following treatment with anti–type II collagen mAbs (3). Additionally, engagement of both C3AR1 and C5AR1 is essential to the full development of joint damage (21). With regard to complement regulators, binding of mouse CFH to the cartilage surface is a major means by which the CS is controlled in the CAIA model (33). Thus, these similar findings in patients with RA compared with CAIA suggest that the CS is pathogenically related to synovial inflammation and tissue damage.

There are contrasting findings to CAIA, however, in human RA, especially as related to the surprising inverse relationship between complement C5 and other MAC gene expression and clinical disease activity. These findings do provide potential insight into why anti-C5 and C5AR1 blockade in patients with RA were not substantially effective (52, 53) and suggest that perhaps earlier effector mechanisms derived from C3 or C4 fragments, or earlier components such as CFB, C1 and MASPs, preferentially promote damage in RA itself.

Additional insights into mechanisms of complement dysregulation are provided by two findings. The first is that increased CFHR-4 expression at baseline is associated with less clinical improvement, and the second is that when examining synovial tissue directly, there is striking evidence of localized differences in the presence of AP proteins such as CFB, CFH, and C3c. These findings suggest that, although looking at whole-tissue mRNA expression provides some insights, it is likely that cell- or context-specific features govern the activation in subdomains within the inflamed synovium. This could be due to localized ACPA or RF binding, inadequate binding of CFH to absent tissue substrate GAG ligands, competition with CFH by CFHRs (47, 54, 55), or induced downregulation of regulatory proteins such as MCP (CD46). Further single-cell studies of complement factors and ex vivo/in vivo models are necessary to further understand the mechanisms underlying these findings.

With regard to other factors, it has been shown using deep phenotyping and RNA sequencing that there are differential levels of gene expression in different synovial pathotypes in early RA without treatment, and this gene expression is correlated with DAS28-ESR, C-reactive protein, and ACPAs (43). In RA synovium compared with the normal synovium, inflammation-related genes are increased (56, 57), but not much is known about complement genes and protein expression levels, and whether they might be altered in the presence of different RA pathotypes (43). Further analyses assessed whether there were differences in CFB gene expression across the three synovial pathotypes (Supplemental Fig. 1B). We found that CFB expression in the synovium, in the fibroid pathotype, was significantly lower (p ≤ 0.05) than in the lymphoid when compared to fibroid pathotype, with a trend toward a decrease as compared with the myeloid pathotype (Supplemental Fig. 1B).

Interestingly, no correlation of C3 gene expression with DAS28-ESR in the early RA synovium was seen, even though it is the most abundant complement protein normally present in the serum (Table II). However, CFP and CFH gene expression levels were positively and negatively correlated, respectively, with DAS28-ESR only in the synovium and not in the blood (Table II). With this finding, CFP might be directly engaged in complement activation at a very early stage of the disease, as its presence is known to be required for synovial injury in murine models (58). To this end, neutrophils are the main extrahepatic source of CFP, where in the synovium it can activate the AP. CFP by itself has been proposed to initiate C3 convertase formation and AP activation on microbial and other surfaces (59). The depressed levels of CFP indicating consumption have been reported in the synovial fluid from 21 RA patients (60), which would be consistent with our observations on activation of the AP in the joints locally compared systemically. The disconnect sometimes we do see between gene and protein expression might be due to the fact that proteins for proper function after secretion are modified due to posttranslation. We also evaluated the possibility that cartilage oligomeric matrix protein (COMP), a glycoprotein expressed on the cartilage, which also binds with the CFP to activate the complement (61), could play an important role; however, the correlations between COMP gene expression and DAS28-ESR in early RA synovium were significantly negative (data not shown). In sum, it is not entirely surprising that CFP and CFH, which are opposing AP regulators, have opposite associations with DAS28-ESR in the synovium from early RA. Ongoing work to identify CFP protein in early RA should be informative in this regard.

One of the surprising findings of the current study was that in early RA patients, after 6 mo follow-up and treatment, in contrast to CFH, a significant correlation existed between ΔDAS28-ESR and CFHR4 mRNA expression in the synovium (Fig. 3A). In contrast, no such correlation was found between DAS28-ESR and CFHR4 gene expression before treatment (Table II). Overall, the correlation between ΔDAS28-ESR and CFHR4 expression in the synovium, after 6 mo, shows that less improvement over time relative to the starting DAS28-ESR is directly associated with initially higher CFHR4 levels. We do not know whether the high CFHR4 mRNA expression in the synovium was the direct result of treatment, due to evolving pathotypes, or was directly involved in the lesser impressive response to therapy. Upon further investigation, we could not find any difference in the absolute levels of CFHR4 and CFH in the serum from healthy controls, at-risk subjects, and RA patients; likewise, CFHR4/CFH ratios were not different, suggesting that the deleterious expression effect is in the synovium. We have not analyzed using IHC the expression of CFHR4 protein in the synovium of RA patients 6 mo after treatment due to nonavailability of biopsies. Nonetheless, MIHC confirmed that CFHR4 was present in early RA biopsies in the vicinity of CFH (data not shown). We do not know what cell types expressed CFH and CFHR4 protein in the early RA synovium, for this will require triple IHC staining. Nonetheless, we can speculate based on the individual IHC staining patterns of CFHR4, CFH proteins, and immune cells that macrophages present in the synovial lining or in the subsynovial lining area might be expressing these proteins (Fig. 7H). The endothelial cells forming blood vessels might also be highly expressing CFH (Fig. 7D). The exact role of CFHR4 in regulation of the CS is not clear, although it may be an antagonist of CFH and compete for binding to C3b on the surface of human cells (62). If this is the case, CFHR4 binding to the C3b on the cell surface promotes complement activation.

One of the limitations of our study is that we have not analyzed the complement protein levels based on the pathotype classification or specific cell types in the AMP RA samples. Future planned studies using spatial transcriptomic– and neighborhood analysis–related complement genes and proteins locally in the synovial cells in their native, early RA, and RA state will address these important questions, inclusive of synovium heterogeneity. Another limitation of our study is that we have not examined, due to technical limitations, the presence of ACPAs or RF in the early RA synovial biopsies, which could explain the imbalance or compartmentalization of complement proteins in the early RA biopsies. Nonetheless, we have found plasma cells in the early RA biopsies (data not shown) and, therefore, ACPAs can be present and, if so, then this will be consistent with the previous observations (63).

Complement receptors C3AR1 and C5AR1 are the receptors for split complement components, C3a and C5a (i.e., anaphylatoxins), and these can mediate proinflammatory signals in macrophages (64, 65), a dominant cell type present in the synovial lining of early RA patients (data not shown). Not only have increased levels of C5a has been reported in the synovial fluid of RA patients, but also disease activity in RA patients correlates with complement activation (65); however, there are no reports regarding the status of C3AR and C5AR1 gene expression in early RA synovium. The significant gene expression and positive correlations among C3AR1, C5AR1, and DAS28-ESR in the synovium, but not in the blood, indicate that synovium and/or synovial lining is the prime target in the pathogenesis of arthritis. The inverse correlation of C5 transcripts with DAS28-ESR and positive correlation of C5AR1 with DAS28-ESR were striking. Our unpublished data also show that the C5AR1 receptor is highly and significantly expressed compared with C5AR1 and CD21 receptors in the synovium from early RA patients, and therefore remains a valid therapeutic target. Regardless, pharmacological manipulation of C5 complement or C5AR1 has been less successful for the treatment of RA (52, 53). Although a number of small-molecule complement inhibitors and inhibitory Abs have been developed and work well in vitro, they have failed in RA clinical trials. The reasons for this are unknown. It may be that the absence of sufficient C5 to generate C5a restricts the ability of C5AR1 to mediate injury, or that the cell-specific expression of C5 and thus local activation in regional subdomains of the synovium does not match sufficiently with the presence of its receptor, or that the prior antagonist used was not pharmacologically active. Our data do suggest, however, that anti-C5AR1 might be effective as a part of a precision medicine approach if applied selectively at the very early stage of RA.

There were significant differences and an increase in the expression of C3AR1 in all three pathotypes (lymphoid > myeloid > fibroid) in the synovium but not in the blood (Supplement Fig. 2E, 2F). In contrast, the fibroid pathotype, compared with the lymphoid and myeloid pathotypes both in the synovium as well as in the blood, had a lower level of C5AR1 expression (Supplement Fig. 2G, 2H), which is consistent with our hypothesis that local complement activation in the synovium is more important than the systemic activation, for these two receptors had positive correlations with the DAS28-ESR (Fig. 2A, 2C). Our previous studies in mouse models of arthritis have shown that mice lacking C3AR1 and C5AR1 genes were resistant to arthritis development (21). Both C5AR1 and C5AR1 are widely distributed on neutrophils, basophils, eosinophils, mast cells, and monocytes/macrophages (66), and engagement of C3a with C3AR in human monocytes/macrophages along with TLR signaling induced proinflammatory cytokines such as IL-1β, IL-6, and TNF-α (67, 68). These cytokines are well known to play an important role in the pathogenesis of RA. In contrast, engagement of C5a with C5AR1 serves as a chemoattractant for neutrophils, monocytes, eosinophils, and basophils (69). In contrast, we have not seen any correlation between C5AR2 gene expression, a decoy receptor for C5a, and DAS28-ESR (Fig. 2E, 2F). Again, these early RA synovial gene expression data are consistent with mouse models of arthritis in which C5AR2 plays no role in arthritis development in the CAIA model (22). The C5a/C5AR1 axis is well known to drive the infiltration or influx of immune and inflammatory cells into the synovial fluid and synovium of RA and psoriatic arthritis patients, and C5AR1 blocking inhibits leukocyte migration into the synovial fluid (65).

CR1 was highly positively correlated with DAS28-ESR in the early RA synovium but not in the blood (Table III). CR1 binds to C3b and C4b to regulate the CS and also binds to complement initiation cascade molecules such as MBL2 and C1q (70). CR1 on erythrocytes promotes the clearance of complement-fixed immune complexes, and in the highly vascularized synovium its high expression is consistent with this role. Interestingly, CR1 competes with the MASPs for binding to the MBL2 and FCNs to inhibit the LP of the complement (70). Consistent with our hypothesis, the high levels of MBL2 protein present in the early RA synovium compared with FCN3 indicate that LP locally might be initiating the disease. In addition, regulatory complement mechanisms such as high gene expression of CR1, which inhibit CP and AP C3 and C5 convertases, might be operating locally to protect synovium from complement attack. CR1 is expressed on many immune cells such as T cells, B cells, neutrophils, and monocytes in the synovium. CR1 ligation inhibits B cell activation and differentiation into plasma cells (71, 72). Therefore, high gene expression of CR1 in the synovium might be playing an important role in complement regulation to keep B cells in check and at low levels in early RA. Notably, we have previously shown that TT32, a CR2/CR1 chimeric inhibitor, prevented the development of arthritis in mice (73).

We also found that there were significant positive correlations (p-adj values) between gene expression of four out of six FcγRs and DAS28-ESR in the synovium, but none in the blood (Table IV). This striking observation reconfirms the new roles of FcγRs in cytokine production, which are important in controlling the local versus systemic inflammation, as these receptors are required for shaping the innate and adaptive immune responses (74). This cytokine production is controlled by the balance between activating and inhibitory FcγRs. More than half of the patients were ACPA positive, and new evidence showed that neural sensory FcγRs can mediate joint pain in arthritis even in the absence of visible inflammation (75). Surprisingly, we have not seen any correlation between the gene expression of FcγRIIb and DAS28-ESR in the early RA synovium as well as in the blood (Table IV). FcγRIIb is the only known inhibitory receptor in the FcγRs family, as the remaining five are activating receptors. FcγR mRNA expression data show that more activating FcγRs are dominant locally rather than systemically compared with inhibitory receptors in early RA patients. This can tilt the balance of cytokine production at an early stage of the disease. We speculate that the control of cytokine production locally in the synovium might be lost at a very early stage of the disease due to the dominance of activating FcγR gene expression. The high FcγR gene expression might be due to the infiltrating inflammatory immune cells in the early RA synovium such as macrophages and monocytes, which were significantly higher as compared with T cells, B cells, neutrophils, and mast cells (data not shown). Future spatial transcriptomic analysis will show which specific cell types express specific FcγR transcripts. In addition, it has been shown that C5a–C5AR1 interactions regulate FcγR expression and function and, therefore, can modulate IC-mediated inflammation (76) and also contribute to the synovial inflammation in the presence of autoantibodies. The i.v. Ig anti-inflammatory mechanism is dependent on the induction of FcγRIIb by sialylated Fc. C5a induces FcγRIII expression on macrophages while inhibiting FcγRIIb, and i.v. Ig inhibits the effects of C5a on FcγRs (77, 78). Thus, the strong correlations among gene expression of FcγRs, C3AR1, C5AR1, and DAS28-ESR locally in the synovium of early RA patients indicate that FcγRs and C5AR1 might be more valuable and superior therapeutic targets combined together instead of singly.

In conclusion, in early RA, we have shown that there is differential clinical disease-related complement factor mRNA expression of a subset of genes locally in the synovium compared with the blood. Nonetheless, we do not rule out the possibility that complement blood gene transcripts in other systemic autoimmune diseases associated with complement protein consumption such as lupus will be informative. In addition to autoantibodies, CFP-dependent AP or amplification loop activation is likely to be operational. The high levels of CFH protein in early RA are likely present in a failed attempt to regulate the AP, as suggested by a positive correlation between CFB expression and DAS28-ESR, and they also may be involved in the clearance of ICs, which are known to be present in the synovium but not in the blood of early RA (79, 80). Finally, our data can explain the failure of complement C5-based therapeutics in RA, which have been based on systemic clinical data rather than information gathered from the synovium. Based on our data, it is more likely that therapeutics targeted at both C3 as well as C5 activation will be necessary to see a beneficial effect of complement inhibition.

We are grateful to the Radiology Department nursing staff and sonographer nurses at the University of Colorado Anschutz Medical Campus, Aurora, for identifying and generating ultrasonic images of the synovitis and also helping to obtain synovial biopsies from early RA and RA patients. We thank William Apruzzese from the AMP/RA/SLE network for identifying suitable early RA biopsies for shipping. We are also grateful to Elizabeth Smith, AMC Histopathology Clinical Core, for helping to standardize various complement protein–related Abs along with their antigen retrieval procedures. We also thank the University of Colorado Human Immune Monitoring Shared Resource group and, specifically, Troy Schedin for generating composite images.

N.K.B. has a patent for the treatment of inflammatory diseases using an anti–C5aR1ab-C5siRNA conjugate. V.M.H. receives royalties, consulting income, stock, and stock options in complement therapeutics companies. The other authors have no financial conflicts of interest.