A New Approach for the Treatment of Arthritis in Mice with a Novel Conjugate of an Anti-C5aR1 Antibody and C5 Small Interfering RNA

It is well accepted that the complement system plays an important role in the development of rheumatoid arthritis (RA) (1), and evidence suggests that the C5 component of complement may play a central role in disease progression (2). C5 is cleaved into C5a and C5b. C5a promotes inflammation via engagement of its receptors C5aR1 and C5aR2, whereas C5b nucleates the assembly of the membrane attack complex (MAC, C5b-9). It appears that C5aR1 signaling is crucial for the progression of RA, because C5aR1^−/− mice do not develop appreciable disease (3). C5aR, (CD88) is expressed by immune cells such as neutrophils, dendritic cells, and macrophages (4), and is also expressed by the liver, kidney, brain, lung, and skin (reviewed in Ref. 5). Engagement of C5aR results in numerous proinflammatory processes, including chemotaxis, vasodilation, enhanced secretion of inflammatory mediators and reactive substances, and enhanced phagocytosis (4). A second C5a receptor, C5L2 (now known as C5aR2), has been identified but its role is controversial (6, 7).

Several groups have targeted the C5–C5aR signal transduction pathway in RA. The anti-C5 mAb BB5.1 decreased disease in the collagen-induced arthritis (CIA) mouse model (8). Other C5 neutralizing Abs prevented both CIA and collagen Ab-induced arthritis (CAIA) in mice (9). C5-deficient mice are highly resistant to CIA in some studies, but not others (10–12). In a recent study by Macor et al. (13), an anti-C5 Ab was developed that bound to mouse, rat, and human RA tissues but not healthy tissues. Clinical effects mediated by this Ab were modest. Using the CAIA model, we have shown that C3 and C5 components of the complement cascade play an important role in disease development (3, 14, 15). Of note, we found that >80% of C5a in mice is derived from the alternative pathway (15).

Human trials with C5 and C5aR targeted therapeutics have been largely unsuccessful despite the abundance of C5 and C5aR1 within human RA joint tissues (16–18). Eculizumab, a humanized anti-C5 Ab, has shown excellent efficacy when used to treat paroxysmal nocturnal hemoglobinuria (19); however, its use in a phase IIb trial for the treatment of RA was unsuccessful (20). PMX53 was also unsuccessful in a small clinical trial testing its efficacy on patients with RA (20).

Small interfering RNAs (siRNAs) are a new and evolving class of biotherapeutics that is likely to find applications alongside traditionally used Abs, fusion proteins, and recombinant proteins. These dsRNAs (20–25 bp long) interfere with the expression of specific genes via the engagement of the RNA-inducing silencing complex, and they have been applied to the treatment of various diseases, including cancer, infection, and arthritis (21–23). Targeting of the siRNA along with minimization of off-target effects is a major challenge. Functionalized nanoparticles have been used successfully to deliver siRNAs in CIA by targeting integrins upregulated during angiogenesis (24). Abs have also been useful targeting agents for siRNAs. The conjugates of an Ab–siRNA (F105 Ab–protamine–siRNA HIV-gag) have been tested successfully in vitro and in vivo (25). Polo-like kinase siRNA conjugated to a single-chain fragmented Ab–protamine complex has been shown to suppress HER2+ breast cancer growth (26). Recently it has been demonstrated that an Ab–siRNA (Shamporter–siRNA nephrin or TRPC-6) conjugate could inhibit gene expression successfully in podocytes after i.v. administration in mice (27).

In this study, we examined the efficacy of siRNAs targeting the C5-C5aR signaling pathway. In particular, we explored the effect of conjugating C5 siRNAs to an anti-C5aR1 blocking Ab. We show that an anti-C5aR1 Ab–protamine–C5 siRNA conjugate is significantly more efficacious than the combination of identical siRNAs and unconjugated anti-C5aR1 Ab in the CAIA. These data provide a proof of concept that it is possible to block complement sufficiently with a bispecific therapeutic molecule to block disease progression effectively. Furthermore, these data demonstrate the utility of anti-C5aR1 Ab–protamine–C5 siRNA conjugates as potential therapeutic entities for the treatment of arthritis.

Anti-C5aR1 mAb (clone 20/70) was purchased from LifeSpan Biosciences (Seattle, WA). This 20/70 clone has been well described as an anti-C5aR1 blocking Ab by several investigators (28–32). It functions by binding to C5aR1 and sterically inhibiting its interaction with C5a. Protamine was conjugated by BIOO Scientific to divalent anti-C5aR1 mAb (3 mg) using the BIOO T3-Max Conjugation kit (Austin, TX) with 35% efficiency according to the manufacturers’ instructions. This conjugation efficiency can vary greatly from one experiment to another. Protamine binds nucleic acids and is positively charged. Protamine plays no role in binding of the C5 siRNA to cells. Instead, protamine is used as a linker for the binding of the C5 siRNA to the anti-C5aR1 Ab. The chemistry makes use of amines such as that found on lysine side chains. Any available lysine side chain amine group may be conjugated to protamine through this process. In brief, the Ab is dialyzed and combined with kit components at a specific temperature as suggested by the manufacture of conjugation kit. The conjugation reaction runs for 14–16 h followed by the addition of buffer, which stops the reaction and places the complex in an environment suitable for siRNA loading and in vivo administration. After removal of unconjugated protamine by gel filtration chromatography, complexes are assessed with SDS-PAGE. Conjugation efficiency was determined by the amount of material that increased in m.w.

The conjugate of anti-C5aR1 mAb (20/70)–protamine (150μg) was incubated with Accell C5 siRNA (8 μg) or Accell Non-Targeting siRNA (Scrambled siRNA) (8 μg; Dharmacon, GE Healthcare) at 4°C for 30 min. These Accell siRNAs are nuclease resistant and are stable in vivo (Dharmacon, GE Healthcare, personal communication). BIOO Scientific estimates that on average, three protamine molecules are bound to an IgG after conjugation and that one protamine fragment can bind 20–30 siRNAs (L. Ford, BIOO Scientific, personal communication). In these studies, we used a mixture of four siRNAs (2 μg of each) to combine with anti-C5aR1–protamine. The unconjugated siRNAs were not removed because of unavailability of this methodology and for scientific reasons explained later. The advantage of this method of construction was that the fewer steps were involved, leading to decreased loss of material and that binding properties of the Ab generally remain unaltered.

C5aR1 expression was measured on a macrophage cell line (RAW 264.7) using the clone 20/70–protamine conjugate at a concentration of 0.4 μg per 1 × 10⁶ cells. Ab conjugate binding was visualized using a FITC-conjugated secondary goat–anti-rat IgG (Life Technologies, Grand Island, NY) diluted 1:400. A dose-dependent curve using FACS analysis was also generated using various concentrations (1, 0.5, 0.250, 0.125, 0.0625 and 0.0312 μg/ml) of the conjugated and unconjugated anti-C5aR1 Ab. A matched Rat IgG2b-FITC–conjugated isotype control (BD Biosciences, San Jose, CA) was also used. C57 BL/6 wild type (WT) mice (n = 3) were injected i.v. with PBS (50 μl) or anti–type II collagen (CII) mAbs (4 mg) and LPS (50 μg), and mice were sacrificed at 24 h. Liver and spleen were dissected, a single-cell suspension was made, and RBCs were removed. C5aR1 expression was assessed using a PE-conjugated clone 20/70 mAb at a concentration of 0.2 μg/1 × 10⁶ cells (BioLegend, San Diego CA) along with matched Rat IgG2b-PE–conjugated isotype control (BD Biosciences; dilution 1:2000). Liver single-cell suspension might contain not only hepatocytes but also Kupffer cells (macrophages). All samples were analyzed using an FC500 flow cytometry machine.

RAW cells were cultured in 24-well plates at a density of 1 × 10⁵ cells and transfected with either 1 μM (13.2 ng) or 2 μM (26.4 ng) Accell C5 siRNA using siRNA delivery media as directed (Dharmacon; GE Healthcare, Waltham, MA). RNA was extracted using the RNAeasy Kit (Qiagen, Valencia, CA). Separately, RNA was extracted from liver samples derived from CAIA experiments and was similarly analyzed. PCR determination of mRNAs for C5 and C5aR1 from RAW cells and C57BL/6 liver as well as the mRNA for TNF-α and matrix metalloproteinase-3 (MMP-3) from the left knee joints were performed by RT-PCR using 40 cycles according to published methods as described (33, 34). All quantitative real-time PCR (QRT-PCR) data were analyzed using a cDNA-based standard curve made by using liver and RAW cell RNA. Primer sequences are available upon request. The standard curves for mRNA encoding C5 and C5aR were constructed using mRNA from mouse liver.

To show the effect of anti-C5 Ab or C5/C5aR1 siRNAs or conjugated anti-C5aR1 mAb with or without protamine and C5 siRNAs, a mouse model of RA known as CAIA was used. CAIA, which represents the effector phase of inflammatory arthritis, is complement dependent and has been used to test the role of deleterious cytokines and the therapeutic efficacy of various drugs, which are now in clinical use for the treatment of RA (35). Many biologicals which have revolutionized the treatment for RA have been tested previously in mouse models of arthritis. For example, drugs such as infliximab, adalimumab, and etanercept have been tested using the CAIA mouse model (35). Eight-week-old C57 BL/6 WT mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Arthritis was induced in these mice using a mixture of anti-collagen Abs and LPS according to our previously published studies (33). Four separate CAIA experiments were performed using a total of 55 mice. For the first study, CAIA in WT and C5aR2^−/− was induced according to our published studies (33), and the clinical disease activity (CDA) was examined by a blinded observer according to published studies (33, 36). These mice were originally obtained and genotyped again as C5L2^−/− from Drs. Craig Gerard and Bao Lu (Boston Children’s Hospital, Boston MA). Recently, because of the change in nomenclature by the International Complement Society (7), for this experiment and for consistency these mice have been designated as C5aR2^−/−. In the second experiment, to examine the effect of an anti-C5 inhibitory Ab on arthritis, CAIA in both the cohorts of WT mice was induced according to our published studies (33, 36). Mice were injected i.p. two times (i.e., on days 3 and 7) with an inhibitory anti-C5 Ab (BB5.1; 750 μg/mouse) or IgG1. For the third experiment, each mouse was injected i.v. three times (i.e., on days −5, 0, and 3) with a dose of the respective commercially available siRNAs. In this experiment we used a mixture of four siRNAs targeting either C5a or C5aR1. The total siRNA used for each target was 8 μg per mouse (i.e., 2 μg of each). Some mice were also injected simultaneously with a combined dose of C5 and C5aR1 siRNAs (8 μg + 8 μg = 16 μg per mouse). For the fourth CAIA experiment, to examine the effect of anti-C5a Ab–protamine–C5 siRNA conjugate on arthritis, each mouse was injected i.p. three times (i.e., on days −5, 0, and 3) with PBS or a conjugate of anti-C5aR1 mAb–protamine–C5 siRNA (150 μg mAb + 8 μg C5 siRNA per mouse; conjugated complex) or of anti-C5aR1 mAb–no protamine–C5 siRNA (150 μg mAb + 8 μg C5 siRNA per mouse; unconjugated mixture). Mice injected with an unconjugated mixture served as negative controls. All mice were weighed before, during, and after the induction of CAIA. All mice were sacrificed at day 10.

Knee joints from fore limbs, the right hind limb with knee joint, ankle, and paw were fixed in 10% neutral buffered formalin. Toluidine-blue stain was used to assess histopathology scores for inflammation, pannus formation, and cartilage and bone damage as described (37). C3 immunohistochemistry was performed as described (37). Monocyte–macrophage and neutrophil infiltration was counted as described (3).

The absolute levels of TNF-α, IL-1β, and MMP-3 mRNAs were measured from knee joints of mice at day 10 treated with PBS or anti-C5aR1 Ab–protamine–C5 siRNA or anti-C5aR1 Ab–no protamine–C5 siRNA using QRT-PCR according to previously described method (33). The standard curves for TNF-α, IL-1β, and MMP-3 mRNAs were made using RAW cells stimulated with LPS (5 μg/ml) for 24 h. All data were expressed in picograms per nanogram of 18S rRNA.

Normality of all data was determined using a null hypothesis for W statistics. The p values (p < 0.05 indicated by asterisks in the figures) were calculated using Student t test (unpaired two-tailed) within GraphPad Prism 4. Data in the graphs and histograms are shown as the mean ± SEM.

CAIA was induced as mentioned in the 2Materials and Methods in C5aR2^−/− and WT mice (both on a C57BL/6 background). The CDA, at day 10, in C5aR2^−/− and WT was 8.6 ± 0.82 and 9.0 ± 0.71, respectively (Fig. 1A), and these differences were not statistically significant (p < 0.82). The prevalence, at day 10, in C5aR2^−/− and WT mice was 100% (Fig. 1B). These results show that C5aR2 plays no role in CAIA, in contrast to C5aR1, which plays an important role in CAIA (3).

We then examined the effect of an anti-C5 mAb on CAIA. CAIA was induced as described above. WT mice were injected with 750 μg of either mouse IgG1 or anti-C5 Ab on days 3 and 7 (Fig. 2A). The CDA, at day 10, in WT mice treated with IgG1 and anti-C5 inhibitory Ab was 9.5 ± 0.5 and 3.22 ± 0.577, respectively (Fig. 2A), and these differences were statistically significant (p < 0.0017). This decrease in the CDA of WT mice treated with anti-C5 inhibitory Ab was 66% compared with mice treated with IgG1 alone. The prevalence of disease both in IgG1 and anti-C5 mAb-treated WT mice was 100% from day 5 through day 10 (data not shown). These data recapitulate in the CAIA model the concept that the C5 system plays an important role in disease progression.

To examine the effect of C5 or C5aR1 siRNA on CAIA, another CAIA experiment was performed. In this study, commercially available C5, C5aR1, or scrambled siRNAs were used (Fig. 2B–D). Mice were injected i.v. on days −5, 0, and 3 with 2 μg of either C5 siRNA, C5aR1 siRNA, or combined (4 μg total) C5/C5aR1 siRNAs. We found that the combined use of C5/C5aR1 siRNAs versus single C5 and C5aR1 siRNAs significantly (p < 0.05) protected mice from CAIA (Fig. 2B). The CDA in WT mice treated with scrambled siRNA, C5 siRNA, C5aR1 siRNA, and combined C5/C5aR1 siRNAs was 8.50 ± 1.5, 7.4 ± 1.9, 6.6 ± 1.69, and 3.60 ± 0.40, respectively (Fig. 2B). The prevalence, at day 10, in all treatment groups was 100% in all groups (Fig. 2C). These CDA data validate the concept that targeting both C5 and its receptor, C5aR1, is more efficacious that targeting either singly.

Efficacy of siRNAs was tested using liver tissue from these mice upon completion of the experiment on day 10. RNA was prepared as described in 2Materials and Methods. C5 and C5aR1 mRNA levels were then determined with QRT-PCR using Taqman probes and absolute amounts calculated from standard curves as described above. Repeated dosing of 2 μg C5 siRNA resulted in a final reduction of C5 mRNA by 45% (Fig. 3A) with no effect on C5aR1 mRNA (Fig. 3B). Conversely, repeated dosing of 2 μg C5aR1 siRNA had no effect on C5 mRNA while reducing C5aR1 mRNA by 46.8%. Combinations of siRNAs showed similar results. These data suggest that even relatively large doses of siRNAs are capable of reducing endogenous mRNA levels by only 50% in this model.

The efficiency of the protamine conjugation to anti-C5aR1 Ab was determined by SDS-PAGE as described in the 2Materials and Methods. Both heavy and light chains were conjugated as evidenced by the upward shift in the size of these two bands (Fig. 4, lane 3). Overall the conjugation efficiency of protamine to the anti-C5aR1 Ab was 35%. Unconjugated anti-C5aR1 was removed using gel filtration chromatography as described in 2Materials and Methods. C5 siRNAs were then incubated with the conjugated anti-C5aR1 Ab for 30 min at 4°C as described in 2Materials and Methods, and this mixture was used directly for all in vitro and in vivo studies described below. Unconjugated C5 siRNAs were not separated from the conjugated C5 siRNA to the anti-C5aR1 Ab conjugate, because C5 siRNA not bound to the conjugated Ab can also inhibit the C5mRNA in cells not expressing C5aR1on their surface. These data show a reasonable binding efficiency of anti-C5aR ab (20/70) to the protamine without altering the binding affinity of the Ab to its receptors.

Lysine conjugation can markedly affect the binding affinity of an Ab. We thus performed flow cytometry to compare the binding of our anti-C5aR1 Ab–protamine–C5 siRNA conjugate to unmodified Ab. The RAW macrophage line robustly expresses C5aR1 and produces a strong fluorescence signal as compared with isotype control when unmodified Ab is used (Fig. 5A). The siRNA-loaded conjugated complex generates a virtually identical signal (Fig. 5B). To characterize further the binding properties of conjugated anti-C5aR1 Ab, we generated full binding curves for both conjugated and unconjugated Abs. RAW cells were incubated with varying amounts of either unconjugated or conjugated anti-C5aR1 Ab, and binding was determined with FACS. Binding curves were virtually identical, with a slight (not statistically significant; p > 0.05) increase for the binding of conjugated Ab (Supplemental Fig. 1). These data confirm that conjugation has no effect on binding affinity for this Ab.

For a better understanding of the distribution of C5aR1 expression, we created a state of systemic inflammation such as that seen in our arthritis model by injection of anti-collagen Ab and LPS as described in 2Materials and Methods. C5aR1 expression is low in liver and spleen cells derived from untreated mice (Fig. 5C, 5E) but is induced within 24 h upon anti-collagen Ab and LPS treatment (Fig. 5D, 5F). These FACS data show that there is an increase in the expression of C5aR1 in vivo during inflammatory conditions versus steady state.

We next validated the ability of our complex to deliver siRNA to a C5aR1-expressing cell. For this experiment, RAW cells were treated with C5 siRNA bound to the anti-C5aR1–protamine conjugate. Negative controls included cells treated with unconjugated anti-C5aR1 Ab plus scrambled siRNA and untreated cells. The expression levels of C5 mRNA and C5aR1 mRNA were examined at 72 h. C5 mRNA was downregulated to 66% in RAW macrophages (Supplemental Fig. 2A) without affecting C5aR1 mRNA levels (Supplemental Fig. 2D) as measured with QRT-PCR. The effect of this conjugate on the C5 mRNA levels was also visually confirmed by running a 2% agarose gel in a semiquantitative assay (Supplemental Fig. 2B) with no off-target effect on the GAPDH mRNA (Supplemental Fig. 2C). These data confirm that our conjugate delivers functional siRNA to C5aR1 expressing cells and does not affect the expression of the targeting gene (C5aR1).

We then assessed the efficacy of this conjugate in vivo, again using the CAIA model. Each mouse was injected i.p. three times (i.e., on days −5, 0, and 3) with scrambled siRNA, a mixture of anti-C5aR1 Ab plus C5 siRNA, or the conjugate of anti-C5aR1 Ab–protamine–C5 siRNA (conjugate). As before, we measured CDA daily until day 10. At day 10, the CDAs were 2.0 ± 0.966, 9.4 ± 1.60, and 11.6 ± 0.244 for conjugate, unconjugated mix, and scrambled siRNA, respectively (Fig. 6A). Disease prevalence was also measured daily and is shown in Fig. 6B. Interestingly, 25% of the conjugate group did not show clinical evidence of disease. Thus, the conjugate of anti-C5aR1 Ab–protamine–C5 siRNA significantly (p < 0.05) reduced the CDA by 83% when compared with scrambled siRNA and by 79% when compared with the unconjugated mix consisting of exactly the same dose of Ab plus siRNA as in the conjugate. All mice weighed before, during, and after the induction of CAIA, and there was no change in weight in mice treated with the conjugate of anti-C5aR1 Ab–no protamine–C5 siRNA (data not shown). These data show that delivering C5 siRNA systemically through specific targeting inflammatory cells using C5aR1 receptors is superior to existing anti-C5 Abs or using C5 siRNAs or C5aR1 siRNAs individually.

All mice were sacrificed at day 10, and joints from this study were processed for histopathologic and immunohistochemical analysis. Both forelimbs and the right hind limb (five joints) were processed for histopathologic analysis and for the measurement of local C3 deposition (Fig. 6C, 6D). Five joints from scrambled siRNA or anti-C5aR1 Ab–protamine–C5 siRNA or unconjugated anti-C5aR1 Ab plus C5 siRNA treated mice were examined for inflammation, pannus formation, cartilage damage, and bone damage (Fig. 6C). The all-joint mean (AJM) histopathology scores for all three groups were PBS (4.02 ± 0.02), anti-C5aR1 Ab–no protamine–C5 siRNA (3.18 ± 0.58), and anti-C5aR1 mAb–protamine–C5 siRNA (0.8 ± 0.3). Individual scores for inflammation, pannus formation, cartilage, and bone damage were significantly (p < 0.05) decreased in mice treated with the conjugate of anti-C5aR1 Ab–protamine–C5 siRNA compared with mice treated with scrambled siRNA or unconjugated anti-C5aR1 Ab plus C5 siRNA. Overall, AJM scores for histopathologic analysis were significantly (p < 0.001) reduced by 85% in mice treated with conjugated anti-C5aR1 Ab–protamine–C5 siRNA as compared with scrambled siRNA or unconjugated anti-C5aR1 Ab plus C5 siRNA. Representative histopathology pictures of the knee joints from these mice are shown in Supplemental Fig. 3A–C).

C3 scores for the three groups were 5.98 ± 0.31, 3.5 ± 0.7, and 1.05 ± 0.68, respectively (Fig. 6D). This C3 deposition in the synovium and on the cartilage surface was also significantly (p < 0.05) reduced in mice treated with conjugated anti-C5aR1 Ab–protamine–C5 siRNA as compared with mice treated with unconjugated anti-C5aR1 Ab plus C5 siRNA or scrambled siRNA (Fig. 5D). Overall, AJM scores (synovium and cartilage) for C3 deposition were reduced by 82% and 70% mice in mice treated with conjugated anti-C5aR1 Ab–protamine–C5 siRNA as compared with scrambled siRNA or unconjugated anti-C5aR1 Ab plus C5 siRNA, respectively. Individually, C3 deposition in the synovium of mice treated with the conjugate of anti-C5aR1 Ab–protamine–C5 siRNA as compared with mice treated with unconjugated anti-C5aR Ab plus C5 siRNA or scrambled siRNA was decreased by 70% and 79%, respectively (Fig. 6D). Representative C3 deposition from the knee joints from mice treated with PBS or anti-C5aR–no protamine–no C5 siRNA or anti-C5aR1 Ab–protamine–C5 siRNA are shown in Supplemental Fig. 3D–F.

The infiltration of neutrophils and macrophages in knee joint synovium from mice treated with scrambled siRNA, with unconjugated anti-C5aR1 Ab plus C5 siRNA, or with conjugated anti-C5aR1 Ab–protamine–C5 siRNA was determined immunohistochemically using specific cell surface markers according to our previously published studies (3) (Fig. 6E, 6F). The percent of neutrophils and macrophages was decreased significantly (p < 0.05) in the synovium of CAIA mice treated with conjugated anti-C5aR1 Ab– protamine–C5 siRNA in comparison with the mice treated with scrambled siRNA or with unconjugated anti-C5aR Ab plus C5 siRNA (Fig. 6E, 6F). Neutrophil counts for the three groups were 3.4 ± 0.24, 3.0 ± 0.7, and 0.8 ± 0.58, respectively. The decrease in the percentages of synovial neutrophils was 88% (p < 0.005) and 85% (p < 0.001), respectively, in CAIA mice treated with conjugated anti-C5aR1 Ab–protamine–C5 siRNA treated mice as compared with scrambled siRNA or with unconjugated anti-C5aR Ab plus C5 siRNA (Fig. 6E). Macrophages showed a similar trend with counts of 2.20 ± 0.37, 1.40 ± 0.5, and 0.25 ± 0.22, respectively, for the three groups (Fig. 6F). The decrease in synovial macrophages in conjugated anti-C5aR1 Ab–protamine–C5 siRNA treated mice was 82% (p < 0.005) and 66% (p < 0.001), respectively, as compared with scrambled siRNA or with unconjugated anti-C5aR Ab plus C5 siRNA treated mice with CAIA (Fig. 6F). Representative pictures of macrophage and neutrophil IHS from the knee joints of mice treated with PBS or anti-C5aR1 Ab–no protamine–no C5 siRNA or anti-C5aR1 Ab–protamine–C5 siRNA are shown in Supplemental Fig. 3D–F, Supplemental Fig. 3G–I, and Supplemental Fig. 3J–L, respectively. A putative model of the effect of the conjugate on various cells and arthritis has been shown in Fig. 7.

Cytokine mRNA levels for TNF-α, IL-1β,and MMP-3 were determined with QRT-PCR using cDNA prepared from the mRNA from the knee joint of the left hind limb of mice treated with PBS or anti-C5aR1 Ab–protamine–C5 siRNA. No significant differences were seen in the levels of mRNA for TNF-α in the knee joint of CAIA mice treated with either scrambled siRNA or conjugated anti-C5aR1 Ab–protamine–C5 siRNA mice with disease (data not shown). However, a significant decrease of 76% (p < 0.033) and 78% (p < 0.04) in the mRNA levels for IL-1β and MMP3, respectively, were seen in knee joints from mice treated with conjugated anti-C5aR1 Ab–protamine–C5 siRNA compared with mice treated with scrambled siRNA. The absolute levels of IL-1β in the knee joints of mice treated with scrambled siRNA or conjugated anti-C5aR1 Ab–protamine–C5 siRNA were 168.5 ± 40.0 and 36.7 ± 23.7, respectively. The absolute levels of MMP3 in the knee joints of mice treated with scrambled siRNA or conjugated anti-C5aR1 Ab–protamine–C5 siRNA were 322.6 ± 88.8 and 78.9 ± 6.7, respectively. A baseline level of mRNA encoding IL-1β and MMP-3 was seen in the knees of heathy, untreated mice (data not shown). These cytokine mRNA data show that the conjugate of anti-C5aR1 Ab–protamine–C5 siRNA affected the proinflammatory cytokines locally in the knee joints of mice with disease.

Previously, we found that genetic disruption of C5aR1 resulted in a virtual block of disease progression in the CAIA model, with 50% of mice showing no evidence of disease and the remaining mice developing disease that was barely detectable (3). In this study, we focused on the possibility of translating this observation into a useful therapeutic. Before focusing on C5aR1, however, we addressed the new observation that C5L2 is a second C5a receptor (C5aR2) (38). We found no differences in the CDA between C5aR2^−/− mice and WT mice in the CAIA model, further demonstrating that the effects of C5a are mediated through C5aR1, which is consistent with our previous study (3).

These studies led us to hypothesize that a reason for the incomplete inhibition of disease progression when using Abs to target either C5 or C5aR1 is that a small amount of signaling is sufficient to drive disease. We tested our hypothesis by considering that C5aR1 signal transduction is a function of both the concentration of ligand (C5a) and the availability of receptor (C5aR1). Indeed, using siRNAs targeted to these two mRNAs, we found that dual targeting was synergistic as compared with targeting ligand or receptor individually. The efficacy of combined siRNAs was somewhat similar to that seen with BB5.1. We cannot discount the possibility that modified siRNAs might perform with a higher efficacy. Furthermore, we cannot ensure that both Ab and siRNA dosages were at similar points in their respective dose-response curves. Nevertheless, these data support the concept that the simultaneous inhibition of both C5 and inhibition of C5a binding to C5aR1 improves efficacy, presumably through a better inhibition of C5aR1 signaling.

By blocking both C5 mRNA and binding of C5a to C5aR1, we expected to synergistically decrease the possibility of activation C5-C5aR1 axis. C5aR1(20/70) has been shown to block the binding of rC5a to C5aR1, making this a reasonable choice for our Ab (31). Concerning the use of other anti-C5aR1 Abs, we hypothesize that C5 siRNA conjugated to any other C5aR1 Ab, which can bind and block the binding of rC5a to the C5aR1, will show a similar effect. Theoretically, C5 siRNA bound to asialoglycoprotein receptor Ab should have the same effect on hepatocytes. Similarly, a newly developed anti-C5aR1 inhibitory Ab by Novo Nordisk Park can be conjugated (39). Most biological activities of C5a are mediated through C5aR1 (40), and it is widely expressed in inflammatory cells, and the agents that act on C5aR1 hold great potential as therapeutics (41).

We next considered the possibility of combining Ab-mediated targeting with siRNAs. This mode of inhibition would be especially useful if C5a were being used in an autocrine fashion to drive activation. Several groups have described the use of Abs as delivery vehicles to target siRNAs, sharing the concept of using positively charged protamine attached to the Ab to bind negatively charged siRNAs (25, 27). To our knowledge, this technology has not yet been applied as a treatment for a disease state in vivo. We chose to replicate aspects of this technology using a bivalent C5aR1 mAb (clone 20/70) shown by several groups to block C5aR1 signal transduction (28–32). We used amine conjugation technology to attach protamine to the Ab. Free amine groups on lysine side chains that are accessible become conjugated. If the CDR contains lysine residues, this can dramatically alter the binding properties of the Ab, necessitating the control experiments performed in Fig. 5 and Supplemental Fig. 1.

We next considered the sources of C5a. Whereas macrophages and neutrophils within the arthritic joint are highly likely to be local sources of C5a, the liver and spleen are well known systemic sources. Given that we are targeting cells expressing C5aR1, we wanted to determine the level of C5aR1 expression in these tissues. Liver and spleen cells from healthy animals expressed low levels of C5aR1 on their surface. However, under inflammatory conditions such as arthritis, there was a marked upregulation of C5aR1 in crude liver and spleen preparations. This would suggest that a state of systemic inflammation is sufficient to induce C5aR1 expression in liver, in turn, indicating that our anti-C5aR1 Ab–protamine–C5 siRNA conjugate would be targeted to liver and arthritic joints at a very early stage of the disease initiation. A model of how the conjugate affects cells systemically and within the arthritic joint is shown in Fig. 7. We have not examined the surface expression of C5aR1 in the knee joints of arthritic mice because of the unavailability of suitable anti-C5aR1 Ab for immunohistochemical staining.

Arthritis is a heterogeneous disease and its origin is considered to be systemic, but the outcome is local inflammation in a subset joints. We think the therapeutic effect of anti-C5aR1 Ab–protamine–C5 siRNA is systemic because it might have inhibited C5mRNA systemically on all C5aR1-expressing cells present in the liver, spleen, macrophages, and neutrophils present locally in the knee joints. However, low C3 deposition and the presence of low levels of C5mRNA in the knee joints of mice treated with this conjugate indicate that the effects of the conjugate were local, because infiltration of macrophage and neutrophil was also decreased significantly. This is possible only with low generation of C5a systemically and locally.

We interpret our results as consistent with the hypothesis that there exists an autocrine or paracrine loop involving C5aR1 signaling and C5 production within the affected joint. Other interpretations are possible, however, and should be considered. For example, it is possible that Ab-mediated delivery of the C5 siRNA is the critical therapeutic process, and the actual choice of Ab is less important. It may be the case that any Ab capable of interacting with Fc receptors (expressed on follicular dendritic cells, macrophages, neutrophils, and mast cells) will suffice. Alternatively, the targeting of some macrophage receptor other than C5aR1 may suffice. In FACS experiments, we find that the inclusion of Fc block has no effect on conjugated anti C5aR1 binding. Given this finding, it is likely that the majority of interactions of the conjugated Ab in vivo will be with C5aR1 receptors. Because the unconjugated mixture of anti-C5aR1 Ab and siRNA is far less effective than the conjugated complex, as shown in Fig. 6A, it is reasonable to assume that the siRNA is being targeted primarily to C5aR1-expressing cells, with only a minor component being targeted to Fc receptors. We have not addressed the degree to which our conjugate is interacting with Fc receptors in this study. Future experiments will test these questions directly to establish the degree of flexibility that this therapeutic strategy may entertain.

In summary, we have shown that C5aR2 does not seem to play a role in arthritis disease progression, whereas C5aR1 and C5 appear to be central to disease development. Targeting both C5 ligand and C5aR1 simultaneously is clearly an improvement over targeting either component separately, which may have led to inhibition of therapeutic use for the treatment of RA. This finding suggests that only small amounts of C5aR1 signaling were required to drive disease. As C5aR1 activation would be proportional to the product of the concentration of ligand times the concentration of receptor, it is perhaps not surprising that a combination approach would be more effective. What is surprising is that conjugation would play such a critical role in efficacy for the treatment of arthritis. Future work will determine whether this vastly improved efficacy is due to increased siRNA delivery via Ab uptake, the targeting of siRNAs to locally C5aR1 expressing cells (as opposed to nonspecific uptake via Fc receptors), or the combined block of C5aR1 signaling and C5a production in the same cell populations. C5 siRNA might get into C5aR1 positive cells through C5aR1 receptor–mediated endocytosis, but the exact mechanism is unknown. Given that the use of Ab-based therapeutics in the clinic has been refined over several decades, we envision that this strategy shows promise as a potential new therapeutic entity for the treatment of RA.

We thank Umarani Pugazhenthi (University of Colorado Anschutz Medical Campus PCR Core) for performing QRT-PCR from the cell lines and knee joints of mice with CAIA and Karen Helm (University of Colorado Anschutz Medical Campus Flow Cytometry Core) for performing flow cytometry to examine the expression of C5aR on various cells. Protamine was conjugated to anti-C5aR (20/70) mAb by BIOO Scientific and we thank Dr. Lance P. Ford and Dr. Suresh Subramanya of BIOO Scientific for providing scientific and technical expertise related to the conjugation of siRNA to anti-C5aR mAb–protamine conjugated molecules. We also thank Dr. Craig Gerard and Dr. Bao Lu for providing C5aR2^−/− (formerly named and obtained as C5L2^−/−) mice on a C57BL/6 background for these studies.

The authors have no financial conflicts of interest.