Targeting of Liver Mannan-Binding Lectin–Associated Serine Protease-3 with RNA Interference Ameliorates Disease in a Mouse Model of Rheumatoid Arthritis

Rheumatoid arthritis (RA), a crippling inflammatory autoimmune disease of the joints, affects ∼0.24% of the world population (1). The severe morbidity of this condition indicates that this disease will impact the public health care system dramatically (2). Although the advent of biologics for the treatment of RA has improved outcomes, ∼40% of patients respond partially or not at all. Despite clinical improvement afforded by these biologic and small molecule therapeutics, the radiological damage in RA can continue over time because of the underlying inflammatory processes that are not fully controlled by the existing drugs on the market (3).

We and others have extensively probed the complement network to identify targets suitable for therapeutic development (4–7). Because of the complexity of human RA, current drug targets include cytokines (TNF-α, IL-1, and IL-6) as well as cells (B cells and T cells). We suggest that complement may serve as an additional important therapeutic target to add to the RA armamentarium, which also may confer benefit because complement activation exerts its function upstream in the inflammatory cascade.

The etiology of RA is complex and involves a combination of environmental, hormonal, genetic, and epigenetic elements leading to an increased likelihood of developing an immune response to posttranslationally altered peptides. Citrullinated peptides are considered a prime example, and the presence of anticitrullinated peptide Abs (ACPAs) is both disease-specific and significantly correlated with radiological damage in patients with RA (3). However, the development of autoantibodies to altered self alone is not sufficient to trigger RA. Rather, a second hit has been postulated to be required for localization of this immune response to the joint (8). Ultimately, it appears that the deposition of Ab within the synovial space leads to the initiation of a joint-specific immune response, which becomes self-sustaining. Our hypothesis is that the liver plays a critical role in this second hit process through the secretion of potent proinflammatory complement proteins.

The complement system is composed of many soluble proteins and receptors, a substantial proportion of which are produced by the liver (9). Other tissues, however, can secrete complement components, including epithelial, endothelial, and immune cells (10, 11). The complement system is activated through three different pathways: the classical pathway, the lectin pathway (LP), and the alternative pathway (AP). All of these pathways converge with the cleavage of C3 and C5 via C3 and C5 convertases, respectively, generating C3a, C3b, C5a, and C5b. C3a and C5a function as general activators of inflammatory cells via C3a and C5a receptors, whereas C5b promotes the assembly of the membrane attack complex C5b-9.

It is well accepted based on studies of complement component-deficient strains in mouse models of arthritis that this system plays an essential role in the development of experimental models of RA (12–16). Moreover, activated complement fragments have been found in the synovium of RA patients (15, 17, 18). Additionally, two mouse models of arthritis (i.e., collagen-induced arthritis and collagen Ab–induced arthritis [CAIA]) have extensively been used to successfully test many drugs currently being used for the treatment of RA (19, 20).

The LP of the complement system is activated when pattern recognition molecules (PRMs) binds to targets. These PRMs include mannan-binding lectin (MBL), collectin-L/K (collectin-10/11), H-ficolin (ficolin-3), L-ficolin (ficolin-2), or M-ficolin (ficolin-1) (21). Associated with the PRMs are a set of proteases that link PRM binding to activation of the complement system. To this end, the liver synthesizes three different proteases of the LP that play this role: MBL-associated serine protease (MASP)-1, MASP-2, and MASP-3 and two MBL-associated proteins, sMAp (MAp19 or MAP-2) and MAp44 (MAP-1) (22–25). The MASP1 gene generates via alternative splicing the two proteases MASP-1 and MASP-3 as well as MAp44 (26, 27), whereas the MASP2 gene generates MASP-2 and sMAp (25).

When PRMs bind to relevant patterns, MASP-1 is activated and then promotes the activation of MASP-2, which subsequently cleaves C4 and C2, thereby generating the C3 convertase, C4b2a (28). In vitro (not in vivo) studies have shown that MASP-1 is essential for bacterial LPS but not zymosan-induced AP (29), indicating that MASP-1 can regulate a specific AP subset and not the entire AP. MASP-3 is required for the cleavage of profactor D (proFD) into mature factor D (FD), an essential factor of the AP and amplification loop, which is engaged and boosts the activity of all pathways of the complement system (22, 28, 30–33). Mice lacking MASP-1/3 are strongly resistant to CAIA (14), whereas mice lacking MASP-2 are partially resistant (34). The individual roles of MASP-1 and MASP-3 could not be resolved in these studies, as the knockout of the MASP1 gene eliminates both splice variants. In addition, MASP-1/3^−/− exhibit developmental delays and small size. Both MASP-1/3^−/− mice lacking the enzymes as well as wild type (WT) mice treated with a mixture of commercial MASP-1/3 small interfering RNAs (siRNAs) were protected from the CAIA (22). Because of the interrelationships of the systems, the individual role of MASP-1 or MASP-2 or MASP-3, however, remained to be determined.

RNA interference (RNAi) is a naturally occurring silencing mechanism that has been used to develop a highly efficient and targeted technology to silence pathogenic genes using synthetic siRNAs or intracellularly generated siRNAs using short hairpin (sh)RNAs. In this study, we used two different RNAi delivery approaches to silence the MASP-3 transcript specifically in cells and in mouse liver. We used lentiviral constructs encoding shRNAs targeting MASP-3 for efficient and long-term silencing of MASP-3 in human cell lines (35).

The asialoglycoprotein receptor (ASGPR) is highly expressed on hepatocytes and binds triantennary N-acetylgalactosamine (GalNAc) ligands with nanomolar affinity (36, 37). The ASGPR is conserved across species, is highly expressed on hepatocytes (∼500 K/cell) (38–40), and has a 15-min recycling time following internalization (41). These properties of the ASGPR facilitate rapid and specific uptake of GalNAc-siRNAs conjugates into hepatocytes (reviewed in Ref. 42). The GalNAc-siRNA conjugate technology has been used to target multiple liver-expressed genes and has demonstrated robust silencing in preclinical animal models and clinical studies. As an example, 90% knockdown was obtained of the liver-expressed transthyretin gene in a phase I study of the GalNAc-siRNA ALN-TTRsc (reviewed in Ref. 42). More recently, it has been shown that a single dose of Inclisiran (ALN-PCSsc), a GalNAc-conjugated siRNA targeting proprotein convertase subtilisin-kexin type 9 (PCSK9) robustly reduced serum PCSK9 levels, resulting in the lowering of a target for the lowering of low-density lipoprotein cholesterol for up to 6 mo after a single dose (43). These studies provide direct clinical proof of concept that GalNAc-conjugated siRNAi are a good choice for hepatocyte-targeted delivery of GalNAc–MASP-3–siRNA duplexes to silence liver-derived MASP-3 expression and thus to specifically investigate the role of MASP-3 in inflammatory arthritis.

Our first objective in this study was to demonstrate that human shMASP-3 (hu shMASP-3) RNA can be used to silence MASP-3 in human cells expressing high levels of MASP-3. Our second objective in this study was to generate a GalNAc–MASP-3–siRNA duplex targeting the liver and to test its efficacy in vivo in mice with and without arthritis. Our hypothesis is that liver-derived MASP-3 is essential for the development of joint damage and that its targeted hepatic inhibition can lead to the attenuation of arthritis.

The human glioblastoma multiforme brain tumor cell line T98G and the human liver carcinoma cell line HepG2 were grown in Eagle’s MEM (Invitrogen by Thermo Fisher Scientific, Carlsbad, CA) containing 5% FBS (HyClone, Logan, UT) and maintained at 37°C in 5% CO₂ (44). Liver from WT C57BL/6 mouse was dissected, washed in 1× PBS, and used to make the RNA. Total RNA from the normal human liver was obtained from outside sources as a residual material. Total RNA was extracted from the human cell lines or from the mouse liver using the Qiagen RNeasy Kit. RNA quality was validated prior to use by agarose gel electrophoresis. Human or mouse MASP-1, MASP-2, and MASP-3 expression in cells and in liver were determined by quantitative RT-PCR (qRT-PCR) as described previously (5, 45). Primer sequences used for amplification of human or mouse MASP-1–, MASP-2–, and MASP-3–encoding RNA are available on request from the authors. Standard curves to calculate the expression of MASP-1, MASP-2, and MASP-3 were made using RNA from the HepG2, T98G cell lines, WT mouse liver, and from the human liver.

T98G or HepG2 cells or mouse liver tissues were lysed, and total RNA was prepared using RNeasy Mini Kit (Qiagen, Germantown, MD). Absence of RNA degradation was assessed by using a 1% agarose gel, and cDNA was prepared (5, 45). MASP-1, MASP-2, and MASP-3 mRNA were measured by RT-PCR. Quantitation of 18S rRNA (18S rRNA) was used as an internal control in each experiment, and all data were expressed in picograms per nanogram mRNA/18S rRNA. All mRNA samples were analyzed in duplicate. Samples were analyzed by amplifying at 40 cycles according to the methods published previously in Nature Protocols (5, 45). All qRT-PCR data were analyzed by using cDNA based on the standard curve made by using liver RNA from a healthy human subject or normal WT mice.

hu shMASP-3 RNA targeting human MASP-3 (NM_139125.3) (mission TRCN0000429394; Sigma-Aldrich, St. Louis, MO) and shScramble RNA (mission pLKO.1-CMV-puro, SHC002; Sigma-Aldrich) were obtained from the Functional Genomic Core of the University of Colorado at Denver.

hu shScramble RNA was used as a negative control in each experiment. The human embryonic kidney 293 cells transformed with the SV40 large T Ag (HEK293T) packaging cell line was used to package hu shMASP-3 RNA and to generate lentivirus viral particles. Briefly, 2 μg of shMASP-3 plasmid was combined with 2 μg of packaging vector (3:1 M ratio of pΔ8.9 and pCMV-VSV-G at 100 ng/μl), diluted in 400 μl of Opti-MEM (31985070; Life Technologies), and incubated with 12 μl of 1 mg/ml polyethylenimine (Polysciences, Warrington, PA) for 15–20 min at room temperature. The entire contents of the tube were added to HEK293FT cells and incubated overnight. After transfection for 16 h, the media on the cells were replaced with complete DMEM. Forty-eight hours later, the media containing viral particles were collected from transfected HEK293FT cells, transferred to a 5-ml syringe, and filtered through a 0.45-μM cellulose acetate filter (VWR, Radnor, PA), and used immediately.

Transduction efficiency was established at various time points using GFP-expressing viral particles, TurboGFP (SHC003; Sigma-Aldrich). A total of 3 × 10⁶ T98G or HepG2 cells per well (12-well plate) were infected with 1.6 × 10³ infectious units (16-μl volume) of lentivirus particles expressing either shMASP-3 RNA or shScramble RNA. After 24 h, viral media were replaced with media containing puromycin (10 μg/ml).

To explore the short-term (at day 15) or the long-term (at day 60) effect of hu shMASP-3 RNA on MASP-3 silencing, the following in vitro experiments were conducted. In the first experiment to examine the expression of MASP-3 in T98G or HepG2 cells, after transfection with shMASP-3 RNA or shScramble RNA, cells were harvested and lysed to make the total RNA at day 15. To further explore the extended long-term effect of hu shMASP-3 RNA or of hu shScramble RNA on MASP-3 protein at day 60, MASP-3 expression in T98G cell lysates was assessed by qRT-PCR. We have not used HepG2 cells to show the longer-term effect at day 60 because of the longer-term toxicity of hu shMASP-3 in this particular cell line.

Normal human serum and FD-depleted sera were obtained from Complement Technology, Tyler, TX. Human and mouse serum samples were stored at −70°C and were thawed only one time prior to use. Human serum deficient in MASP-2 and MASP-1/3 for these studies were kindly provided by Dr. P. Garred (Copenhagen, Demark) and Dr. S. Thiel (Aarhus, Denmark), respectively. We have established a colony of FD^−/−, MASP-1/3^−/−, and MASP-2^−/− mice on C57BL/6 background. Sera from WT, FD^−/−, MASP-1/3^−/−, and MASP-2^−/− mice were obtained through retro-orbital bleeding according to the approved protocol by the Institutional Animal Care and Use Committee. All mice used in this study were on C57BL/6 background. WT C57BL/6 mice were also purchased from The Jackson Laboratory to obtain serum for various studies. All mice were kept in our barrier facility and fed breeder chow provided by the Center for Laboratory Animal Care, University of Colorado Anschutz Medical Campus.

The presence or absence of ProFD and FD was detected by Western blotting of mouse and human serum deficient in either MASP-1/3 or MASP-2. Sera from FD^−/− mice and also human sera depleted in FD were also examined, which served as negative controls. Normal sera from WT C57BL/6 mice and normal human sera were used as positive controls in all experiments. Before Western blotting, human serum (25 μl) was preincubated with affinity-purified goat anti-human FD (2 μl) (R&D Systems, Minneapolis, MN), allowing ProFD and FD to be immunoprecipitated with Protein A/G beads (12 μl) (Santa Cruz Biotechnology, Dallas, TX). The immunoprecipitates were then treated with peptide-N-glycosidase F (PNGase F) overnight at 37°C according to the manufacturer’s suggested protocol (New England BioLabs, Ipswich, MA). The samples were analyzed by SDS-PAGE and Western blot. After transfer, the blots were incubated for 1 h with biotinylated anti-human anti-FD Ab as an alternative Ab that binds to a different epitope than the primary anti-human FD Ab used for precipitation (dilution 1:500) (R&D Systems). Finally, the proFD and FD bands at ∼26 kDa were detected using streptavidin HRP conjugate (dilution 1:1000) (R&D Systems) and luminescent substrate. The band of proFD was visually slightly bigger in kDa than the FD band, and these differences were only visible by using 12% Bis-Tris Gel (14, 22, 28).

To examine the cleavage of ProFD by rMASP-3 in vivo in MASP-1/3^−/− mice, (male and female) were injected only once i.v. in the tail with active or inactive rMASP-3 (serine changed to alanine) (50 μg/mouse). After 24 h of injection with rMASP-3 protein, sera from all MASP-1/3^−/− (n = 4) male and female mice were examined ex vivo for AP activation. The AP activation in serum at various dilutions (1:6 and 1:12 in EGTA/Mg buffer) was determined by rabbit erythrocyte lysis analysis as described earlier (32). Serum from a WT control mice was used as a positive control at various dilutions (1:6, 1:12, 1:24, and 1:48 in EGTA/Mg buffer) in duplicate to show the specificity of rabbit erythrocyte lysis for AP activation. The absorbance was read at 405 nm.

The serum samples before and after treatments with rMASP-3 were also analyzed for the cleavage of proFD into mature FD using Western blot analysis.

Fifty mouse siRNA sequences were designed based on the available bioinformatics information to target the MASP3 splice variant of the MASP1 gene. GalNAc-siRNA conjugates were synthesized using solid-phase synthesis methods as previously described (46). Cos-7 cell line was used to examine the effect of MASP-3 siRNA duplexes on the expression of MASP-3 in vitro. A customized fusion psiCHECK vector (Blue Heron Biotech, Bothell, WA) of luciferase reporter gene and mouse MASP3 was diluted with Opti-MEM for a final concentration of 1.2 ng/ml. MASP-3–siRNA and control siRNAs were diluted eight times at a 6-fold serial dilution beginning at 10 nM. The siRNA dilutions were cotransfected with the psiCHECK vector (1:1 ratio) and preincubated for 15 min with Lipofectamine (1:10 ratio) (Invitrogen by Thermo Fisher Scientific, Waltham, MA). The cell suspension was added to each well with the lipid/siRNA mix and incubated for 48 h at 37°C in an atmosphere of 5% CO2. Dual-Glo Reagent (Promega, Madison, WI) was used as the luminescence according to the manufacturer’s protocol and measured using Biotek Synergy HT plate reader. Each data point was tested with at least four biological replicates. Renilla luciferase reporter was used as the positive control. The magnitude of siRNA activity was then normalized relative to the negative control (psiCHECK2-MASP3 transfected but either nontargeting siRNA or PBS). All transfections were done at n = 2 or greater.

The psiCHECK2 Dual-Glo system enables detection of siRNA-mediated silencing of target sequences fused to a Renilla luciferase reporter. The sequence of interest (MASP-3) was cloned into the multiple cloning region (XhoI-NotI sites) located downstream of the Renilla STOP codon in the 3′UTR. RNAi-mediated cleavage and degradation of the fusion mRNA can be measured by a loss in Renilla signal following siRNA treatment. The psiCHECK2 vector also contains a second reporter, firefly luciferase, which is driven by a different promoter and allows for normalization of Renilla expression. When MASP-3 siRNAs are transfected in Cos-7 cells followed by silencing of MASP3 gene, the levels of luciferase signal also go down. IC₅₀ of each MASP-3 duplex was calculated from the expression curve. The IC₅₀ value has been listed for each MASP-3 duplex on the plots. For example, duplex 4 (IC₅₀ = 1.2 pM) is the more effective, as lower concentration of the duplex is enough to silence MASP-3 expression by 50%, whereas duplex 1 (IC₅₀ = 41.2 pM) is less effective in comparison.

The effect of MASP-3 siRNA by duplex 1 (duplex 3) and duplex 2 (duplex 4) was initially characterized in 8-wk-old naive C57BL/6J mice (Charles River Laboratories), and later on, these duplexes were further chemically modified. Single 1 or 10 mg/kg doses of duplex 1 and duplex 2 were administered s.c. Livers were collected on day 7 postinjection. The mRNA was isolated using RNeasy Plus Mini Kit (Qiagen). High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) was used to convert the mRNA samples. Levels of MASP3 expression were analyzed using qRT-PCR reagents compatible with Roche Lightcycler 480. The MASP3 mRNA was measured using a customized TaqMan assay (Thermo Fisher Scientific) and normalized to the endogenous GAPDH expression (Thermo Fisher Scientific).

In another experiment, C57BL/6J mice were injected with duplex 1 or with duplex 2 on day −5, day 0, and day 3 followed by liver collection at day 10, as this would be a relevant time point in a CAIA model (see below). Serum was obtained prior to injections to examine the levels of circulating MASP-3 by Western blot and ELISA. Liver was collected on day 10 and examined for MASP-3 expression using qRT-PCR.

Lastly, WT mice (n = 10) were injected with GalNAc–luciferase–siRNA or with duplex 2 of GalNAc–MASP-3–siRNA on day −10, day 0, and day 3. Sera from mice were obtained prior to injections to examine the circulating levels of MASP-3 protein by Western blot and ELISA. Liver was collected on day 10 and tested for MASP-3 mRNA expression using qRT-PCR. The impact of GalNAc–MASP-3–siRNA duplexes on MASP-1 levels was tested by examining the serum for MASP-1 levels by ELISA.

Infected and control T98G cell culture supernatants at day 60 were collected and concentrated by centrifuging in Centricon Plus-70 Filter Units at 3000 rpm for 30 min (Millipore, Bedford, MA). MBL/MASP-1 or MBL/MASP-3 complexes were pulled out by incubating 20 μl of concentrated supernatant with 20 μl of mannan-agarose (Sigma-Aldrich) at 4°C overnight. The beads were washed and then subjected to 4–12% Tris-Glycine Gel (Invitrogen by Thermo Fisher Scientific) SDS-PAGE. Similarly, to analyze the effect of GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA on MASP-3 in vivo, 20 μl of serum from each mouse was incubated with mannan-agarose as described above. All samples were transferred to 0.2 μm of microporous Immobilon P^SQ Transfer Membrane, polyvinylidene difluoride (PVDF) membrane (7 cm × 8.4 cm size) (Merck Millipore, Tullagreen, Carrigtwohill, County Cork, Ireland) using Tris-Glycine Transfer Buffer (Invitrogen by Thermo Fisher Scientific) for 1 h at 100 V. The PVDF membrane was blocked in 5% dry milk in 1× PBS containing 0.05% Tween 20 for 1.5 h and incubated overnight at 4°C with either 1) human anti–MASP-3 primary Ab (1:200 dilution) (Santa Cruz Biotechnology) 5% dry milk for analysis of human MASP-3 derived from T98G cells or 2) rabbit anti–MASP-3 primary Ab at a 1:200 dilution (produced in-house at Dr. Thiel’s laboratory) in 5% dry milk for analysis of MASP-3 in mouse serum. All blots were washed 3 times in 1× PBS containing 0.5% Tween 20 then incubated with either 1) HRP-conjugated goat anti-mouse (1:5000 dilution) (Santa Cruz Biotechnology) or 2) HRP-conjugated goat anti-rabbit (1:5000) (Jackson ImmunoResearch, West Grove, PA) diluted in 5% milk containing 0.5% Tween 20.

Finally, the membranes were developed from 1 to 5 min using a 1:1 mixture of SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific). Different concentrations of a truncated form of the recombinant human MASP-3 protein were also used as a positive control along with normal human serum. The blots were scanned using GeneGnome XRQ Chemiluminescence Imaging System and GeneTools analysis software from Syngene (Frederick, MD), and the density of each band was quantified using the Quantity One software (Bio-Rad, Hercules, CA).

The procedure for measuring MASP-1 was identical to that described above. The PVDF membrane was blocked and then probed with rabbit anti–MASP-1 primary Ab (1:200) in 5% dry milk. Sera from WT and MASP-1/3^−/− mice were used as positive and negative controls, respectively.

Serum samples from various cohorts from WT C57BL/6 mice with no disease or with CAIA injected s.c. with GalNAc–luciferase–siRNA or with duplex 1 or with duplex 2 of GalNAc–MASP-3–siRNA were analyzed by time-resolved immunofluorometric immunoassays as previously described (26, 32, 47) by using in-house Abs. Sera were diluted 1:50 to measure the levels of MASP-3 in mouse serum. For MASP-3, sera from WT and MASP-1/3^−/− mice were used as positive and negative controls, respectively.

WT mice were injected s.c. with 10 mg/kg on days −10, -5, and on day 0 either with GalNAc–luciferase–siRNA (n = 22) or GalNAc–MASP-3–siRNA (duplex 2) (n = 22). Then, CAIA was induced in WT mice by using a mixture of five mAb to bovine collagen type II (Arthrogen-CIA, Chondrex) resuspended in sterile Dulbecco’s PBS according to our previously published studies (13, 15, 34, 48). All WT mice received i.p. injections of 8 mg of Arthrogen on day 0 and 50 μg of LPS from Escherichia coli strain 0111B4 on day 3. All mice were sacrificed at day 10 post–LPS injection. The total duration of the CAIA experiment was day 20 after the first siRNA injection. The severity of clinical disease activity (CDA) in all groups of WT mice was determined every day by two trained laboratory personnel acting independently and blindly according to our previously published study (4). The in vivo CAIA experiments were repeated three times (total n = 44) using an identical dose of duplexes and an identical injection schedule.

At day 10, both forepaws and the entire right hind limb, including the paw, ankle, and knee were surgically removed and fixed immediately in 10% buffered formalin (Biochemical Sciences). The preparation of knee joint samples and histological analysis using toluidine blue (T blue) dye were performed as previously described (14, 15). All sections were read by a trained observer who was also blinded to the treatment and to the disease activity score of each mouse. The joint sections from mice were scored for the changes in inflammation, pannus, cartilage damage, and bone damage on a scale of 0–5. C3 deposition in the joints (synovium and cartilage) was localized with polyclonal goat anti-mouse C3 antisera (ICN Pharmaceuticals) and scored as previously described (13). Macrophages infiltration in the knee joints were localized with rat anti-mouse F4/80 Ab (Bio-Rad) and scored using a scale from 0 to 4.

All joints from one CAIA experiment (n = 14) used for histopathology studies after treatments with GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA described above were collected for further immunohistochemical (IHC) localization of MASP-3 protein. All five joints (i.e., both forelimbs and the right hind limb) (knee joint, ankle, and hind paw) were processed for IHC studies and for the MASP-3 deposition in the joints. Liver and brain from WT were used as positive controls, whereas liver from MASP-1/3^−/− mice was used as negative control. Five micron–thick paraffin sections were prepared for immunodetection and stained with rabbit anti-mouse MASP-3 biotinylated Ab at a dilution of 1:200. Sections required modest retrieval in 10 mM sodium citrate pH 6 for 10 min at 110°C in the NxGen Decloaker (Biocare Medical, Concord, CA) with a 10 min cool down. MASP-3 immunodetection was performed on the Benchmark XT autostainer (Ventana Medical Systems, Tucson, AZ) at an operating temperature of 37°C with primary incubation for 32 min using a modified iVIEW DAB (Ventana Medical Systems) Detection Kit. The iVIEW secondary Ab was replaced with full-strength Rabbit ImmPRESS polymer (Vector Laboratories, Carpinteria, CA), and the enzyme was replaced with diluted Rabbit ImmPRESS polymer (Vector; 50% v/v in PBS pH 7.6). All sections were counterstained in hematoxylin for 2 min, blued in 1% ammonium hydroxide, dehydrated in graded alcohols, cleared in xylene, and cover glass mounted using synthetic resin. Negative controls to confirm the specificity of the immunostaining included omission of the primary Ab incubation step in the IHC protocol, substitution of the primary Ab diluent. Because of the overall low deposition of MASP-3 in the joints, an ordinal scale was used to assess levels.

All data were analyzed using GraphPad Prism 4.0 program. The CDA data from all individual WT mice were included in the final analysis of histolopathology and immunohistochemistry. The density of bands from Western blots was determined by using Bio-Rad Quantity One program after scanning and subtracting the background from blots. To analyze mRNA data from T98G or HepG2 cells and in mouse and human liver, t test was used. The test of ANOVA was used to find the statistical differences comparing more than three independent treatment groups. All data were expressed as the mean ± SEM with p < 0.05 considered significant, and original p values have been shown.

We examined two human cell lines (i.e., T98G and HepG2 for MASP mRNA expression). Overall, human HepG2 liver cells expressed MASP-1, MASP-2, and MASP-3, and T98G cells expressed MASP-1 and MASP-3 but no detectable MASP-2 (Fig. 1A). Human and mouse liver expressed all three mRNAs (Fig. 1B). Human liver HepG2 and brain T98G cells expressed more MASP-3 than MASP-1 mRNA (Fig. 1A). Mouse liver tissue showed significant (p = 0.05) higher expression levels of all MASP mRNAs (i.e., MASP-1, MASP-2, and MASP-3 expression) compared with the human liver (Fig. 1B). We used both HepG2 and T98G cells to test our ability to decrease human MASP-3 mRNA expression in a specific fashion.

Lentiviruses expressing either GFP, hu shMASP-3 RNA, or a control hu shScramble RNA were prepared as described in 2Materials and Methods. Efficiency of infection was first qualitatively assessed using GFP expression. Lentiviral particles consisting of a mixture of GFP and either hu shMASP-3 RNA or control shRNA were incubated with HepG2 or T98G cells for 24 h followed by washing and feeding with fresh media. Puromycin was added to the media to select for infected cells. GFP expression was visualized after either 24 h or 15 d in T98G cells (Supplemental Fig. 1). The percentage of T98G cells expressing GFP increased with lentiviral doses of 0.2, 0.4, and 1.6 × 10³ particles (data not shown). For this reason, all subsequent experiments in this series were performed with 1.6 × 10³ particles using HepG2 or T98G cells. After lentiviral infection, day 15 cultures from HepG2 and T98G were then examined for expression of MASP-1, MASP-2, and MASP-3 mRNA. Cultures were lysed, total RNA purified, and MASP mRNA levels measured as described in 2Materials and Methods. MASP-1, an alternative splice variant of the MASP-1/3 gene, was unaffected by infection with either control or shMASP-3 RNA lentiviruses (Fig. 2A, 2D). Similarly, MASP-2 gene expression was unaffected in HepG2 cells (Fig. 2E), but the direct effect of shMASP-3 RNA could not be determined on T98G cells, for these cells expressed very low or undetectable MASP-2 (Fig. 2B). In HepG2 cells, there was a significant (p < 0.001) 70% decrease at day 15 in the expression of MASP-3 by hu shMASP-3 RNA compared with cells alone or control Scramble shRNA (Fig. 2F). In striking contrast, in T98G cells, shMASP-3 RNA–expressing lentiviruses (but not the control lentivirus) caused the level of MASP-3 mRNA to drop by 95% (Fig. 2C). No further decrease in the expression of MASP-3 was seen by using higher doses of lentivirus (data not shown). These data show that human MASP-3 can be silenced specifically by using hu shMASP-3 RNA in human cell lines without significantly affecting the expression of MASP-1 or MASP-2, if these are present.

To establish the durability of shMASP-3 RNA infection, T98G cells were incubated with lentiviral particles, washed, and cultured for 60 d. Supernatants were then collected, concentrated and examined for MASP-3 protein using Western blot analysis (Supplemental Fig. 2). There was an almost complete lack of MASP-3 protein in the supernatants of T98G cells at day 60 (Supplemental Fig. 2A, lane 5) compared with the cells transfected with shScramble RNA (Supplemental Fig. 2A, lane 4). Purified recombinant human MASP-3 catalytic fragment in varying amounts was used as a positive control (Supplemental Fig. 2B, lanes 2–5). These Western blot data show that hu shMASP-3 RNA specifically targeted human MASP-3 in T98G cells and reduced its expression substantially after 2 mo in culture.

To explore the role of MASP-3 in the cleavage of proDF to mature FD, we examined sera from both mice and humans deficient in the alternatively spliced MASP-1/3 gene by immunoprecipitation followed by Western blot analysis (Fig. 3). We found that both mouse and human serum contained only ProFD in the complete absence of MASP-1/3 (Fig. 3A, lane 2, and Fig. 3B, lane 2). In contrast, no proFD but only FD was present in human sera with MASP-2 deficiency (Fig. 3, lane 3). Serum from MASP-2^−/−-deficient mice also contained FD, but did not contain proFD (Fig. 3, lane 2). Sera from FD^−/− were used as a negative control, and no band of either FD or proFD is present (Fig. 3A, lane 1, and Fig. 3B, lane 1).

To determine the requirement of MASP-3 in vivo to reconstitute the AP, MASP-1/3^−/− mice were injected with an active recombinant MASP-3 (rM3) or with an inactive or mutated rM3 (rM3 S/A), and AP activity was examined after 24 h by well-known rabbit erythrocyte lysis assay specific for measuring the AP activity (32). AP is normally compromised in MASP-1/3^−/− mice versus WT mice because of proFD presence in their circulation (Fig. 4A, 4B). Exogenous active rMASP-3 compared with the mutated rMASP-3 clearly reconstituted the AP activity in the sera from MASP-1/3^−/− mice at 24 h (Fig. 4C). The ex vivo AP activity in sera from MASP-1/3^−/− mice after reconstitution was equivalent to the WT mouse serum (Fig. 4A, 4C). There were no differences in the ability of active rM3 to reconstitute the AP activity both in male and female mice (Fig. 4C). Furthermore, at 24 h, exogenous rM3 not only activated the AP but also cleaved the proDF in vivo in MASP-1/3^−/− mice (Fig. 4D, lane 7 versus lane 5). These data show that MASP-3 is specifically required for the cleavage of proDF into DF in vivo and results in the subsequent capacity for AP activation.

Mouse-specific siRNA sequences were identified using Alnylam bioinformatic tools. Chemically stabilized duplexes conjugated to a triantennary GalNAc on the 3′ end of the sense strand were synthesized using solid-phase synthesis methods (46). MASP-3–siRNAs were tested in vitro using a psiCHECK vector expressing MASP3 gene in Cos-7 cells. The silencing of MASP-3 was examined based on the inhibition of luciferase activity following transduction with various GalNAc–MASP-3–siRNA duplexes. Four duplexes show IC₅₀ silencing of 41.2, 4.6, 34.4, and 36 pM, respectively (Fig. 5).

Three independent in vivo experiments were done to determine the efficiency of GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA duplexes on the expression of MASP-3. In the first experiment, to examine the short-term effect of GalNAc–MASP-3, WT mice (n = 15) were injected s.c. with a single dose of either PBS or GalNAc–MASP-3 duplexes 1 or 2 at day 0 and sacrificed at day 7 (Fig. 6). Duplex 2 was most effective at blocking MASP-3 at both 1 and 10 mg/kg doses. Both duplexes have no effect on the expression of MAp44 (Supplemental Fig. 3A) and MASP-1 (Supplemental Fig. 3B).

In the second experiment, to examine the long-term effect of GalNAc–MASP-3–siRNA, WT mice (n = 6) were injected s.c. three times either with GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA duplexes 1 or 2 at day –5, at day 0, and at day 3. The mice were then sacrificed at day 10 (Fig. 7). ELISA results confirmed that at day 10, there was a decrease in the levels of MASP-3 protein, which ranged from 54 to 64% of control in mice injected with GalNAc–MASP-3–siRNA duplex 2 (Fig. 7A). A standard curve was generated after reconstituting the sera from MASP-1/3^−/− mice with known quantities of rM3 to show the specificity of MASP-3 ELISA (Fig. 7B). Sera from these mice were also analyzed by Western blot, confirming decreases in the range of 60% as measured by densitometric analysis (Fig. 7C, lane 7). Additionally, duplexes 1 and 2 caused a marked decrease in liver MASP-3 mRNA abundance by day 10 as compared with control (Fig. 7D).

In the third experiment, WT mice (n = 15) were injected three times with GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNAs duplex 1 or duplex 2 at day –10, day 0, and at day 3 followed by sacrificing all mice at day 10 (Fig. 8). In this fashion, the time between first and second doses was extended from 5 to 10 d. Serum MASP-3 dropped by ∼15% by day 10 (p = 0.015) as compared to the 60% drop seen in the previous experiment (compare Figs. 7A and 8A). MASP-1 protein levels were completely unaffected (Fig. 8B). Sera from mice injected with duplex 2 were also analyzed by Western blot (Fig. 8C). Densitometry measured the decrease in MASP-3 to be 32.5% at day 10. Liver from mice injected with duplex 1 or duplex 2 of GalNAc–MASP-3–siRNA also showed a 33.0% (p = 0.044) decrease in the expression of MASP-3 after treatment (Fig. 8D). These short-term and long-term experiments in healthy WT mice show that duplex 2 of GalNAc–MASP-3–siRNA performed better compared with duplex 1; therefore, we decided to use duplex 2 for CAIA study.

To examine the efficacy of our GalNAc–MASP-3–siRNA duplex 2 for the treatment of RA, we employed the CAIA model in which mice are injected with pathogenic anti-collagen Abs on day 0 and then injected with LPS on day 3. Synovitis develops rapidly and peaks by day 10. Mice are then sacrificed on day 10 for analysis. In this experiment, WT mice (n = 22) were injected s.c. on day –10 before the induction of disease and again on day –5 and on day 0 with either GalNAc–luciferase–siRNA or with GalNAc–MASP-3–siRNA (duplex 2) (n = 22) as described in 2Materials and Methods (Fig. 9). CDA was assessed daily. Strikingly, initial disease progression in the MASP-3 siRNA group was markedly slowed as compared with the luciferase siRNA–injected group until day 6 where the CDA was decreased by 69% (p < 0.001). Between days 6 and 7, there was a reproducible acceleration in disease in the MASP-3 siRNA injection group followed by a general disease plateau for both groups. By day 10, the average CDA for the luciferase siRNA–injected group was 10.40 ± 0.545, whereas the average CDA for the MASP-3 siRNA–injected group was 5.13 ± 0.798; thus, a decrease of 50% was observed (p < 0.001) (Fig. 9A). In all three separate experiments, there was a consistent decrease in the CDA in mice treated with MASP-3 siRNA. Individual CDA in each mouse was also consistent with the overall decrease in the CDA in mice treated with GalNAc–MASP-3–siRNA (data not shown). The prevalence of disease at day 10 in WT mice injected with GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA duplex 2 was 100 and 90%, respectively (Fig. 9B). There was a significant (p = 0.05) effect on the weight of mice treated with GalNAc–luciferase–siRNA or GalNAc–MASP-3–siRNA duplex 2 at day 0 and at day 10 (Fig. 9C). All mice survived to the experimental endpoint, and these in vivo CAIA were repeated three times (n = 44) with almost similar results; data shown in this paper are from all experiments. Overall, these data indicate that GalNAc–MASP-3–siRNA duplex 2 partially inhibited the development of CAIA in mice.

Joints derived from the CAIA experiments no. 1 (n = 14) out of three experiments described above were collected for further analysis. All five joints (i.e., both forelimbs and the right hind limb) (knee joint, ankle and hind paw) were processed for histopathology studies and for the measurement of C3 deposition in the joints (Fig. 10A). All data are shown as an all joint mean (AJM) score (0–5). Histopathology scores for inflammation, pannus formation, cartilage damage, bone damage show a 53% significant (p = 0.006) decrease (Fig. 10A). These data are consistent with the improvement in CDA (Fig. 9A). There was a significant decrease in inflammation 45% (p = 0.007), pannus formation 53% (p = 0.006), cartilage damage 55% (p = 0.0049) and bone damage 61% (p = 0.0019) in CAIA mice treated with GalNAc–MASP-3–siRNA duplex 2 compared with CAIA mice treated with GalNAc–luciferase–siRNA (Fig. 10A). Representative pictures shown of T blue staining related to histopathology from the knee joints (Fig. 11A, 11B) and ankle (Fig. 11C, 11D) of mice treated with GalNAc–MASP-3–siRNA duplex 2 or GalNAc–luciferase–siRNA.

Both the knee joints were examined for C3 deposition in the synovium as well as on the surface of cartilage (Fig. 10B). All data are shown as an AJM score (0–4) and also the total score (0–8). There was a 35% (p = 0.031) and 37% (p = 0.017) decrease in the C3 deposition in the synovium and also on the surface of cartilage, respectively (Fig. 10B), in mice treated with GalNAc–MASP-3–siRNA duplex 2 (n = 7) compared with mice treated with GalNAc-luciferase. These data are also consistent with the decrease in CDA in mice treated with GalNAc–MASP-3 versus GalNAc-luciferase (n = 7). Representative pictures shown of C3 deposition from the knee joints (Fig. 12A, 12B) of mice treated with GalNAc–MASP-3–siRNA duplex 2 or GalNAc–luciferase–siRNA.

There was a nonsignificant 28% decrease (p = 0.39) in macrophage infiltration in the synovium of knee joints of mice with CAIA treated with GalNAc–MASP-3–siRNA duplex 2 compared with mice treated with GalNAc-luciferase. Representative pictures shown of macrophage staining from the knee joints (Fig. 13A, 13B) of mice with CAIA treated with GalNAc–MASP-3–siRNA duplex 2 or GalNAc–luciferase–siRNA.

IHC analysis and images of MASP-3 protein in the joints of CAIA mice treated with GalNAc–MASP-3–siRNA duplex 2 compared with GalNAc–luciferase–siRNA showed lower levels of MASP-3 deposition in the knee joints synovium and forepaws synovium (Fig. 14A–I). No deposition of MASP-3 was found in the cartilage surface. Overall, the MASP-3 deposition was lower in the knee joints compared with the forepaws of mice with disease (Fig. 14E–H). Interestingly, MASP-3 deposition was also not seen in the synovial fibroblast-like cells but in different cells, and the identity of these cells is unknown at present. Notably, these MASP-3–positive cells were also present in the vicinity of fat cells known to secrete proFD (Fig. 14H). For MASP-3 IHC localization, liver and brain from WT were used as positive controls (Fig. 14A–D), whereas liver from MASP-1/3^−/− mice was used as negative control (Fig. 14I). These IHC data show that MASP-3 is present in the joints, but its deposition is low, and also it decreases with GalNAc–MASP-3–siRNA duplex 2 treatment in mice with CAIA (Fig. 14E–H).

Sera from CAIA mice treated with GalNAc–MASP-3–siRNA or with GalNAc–luciferase–siRNA were analyzed at day 10 for MASP-3 protein using Western blot analysis (Fig. 15A). Overall, there was a 47% decrease in the levels of MASP-3 protein, which, interestingly, was consistent with the 50% decrease in CDA as well as histopathology scores (Figs. 9A, 10A, 15A). To determine the levels of MASP-1 and MASP-3 proteins in the circulation time–resolved immunofluorometric assay was used as described in 2Materials and Methods (Fig. 15B, C). There was a significant (p = 0.033) decrease of 14.4% in the levels of MASP-3 at day 0 in mice injected with GalNAc–MASP-3–siRNA duplex 2 compared with mice injected with GalNAc–luciferase–siRNA (Fig. 15B). At day 10, there was a significant (p = 0.011) decrease of 42% in the levels of MASP-3 in mice injected with GalNAc–MASP-3–siRNA duplex 2 compared with mice injected with luciferase siRNAs (Fig. 15B). No significant differences were seen in the levels of MASP-1 at any time point before and after injections and also before and after the induction disease (Fig. 15C). The levels of MASP-2 and MAp44 were not examined because of the limited amount of serum from each mouse available for these studies. These data show that there was an effect on the levels of MASP-3 without substantial off-target effects on MASP-1 in mice injected with GalNAc–MASP-3–siRNA duplex 2 or GalNAc–luciferase–siRNA (Fig. 15A–C).

Upon sacrifice, livers from the above CAIA mice were snap frozen. Total RNA was extracted to examine the expression of MASP-1 and MASP-3 from mice treated with GalNAc–MASP-3–siRNA duplex 2 or GalNAc–luciferase–siRNA (Fig. 15D, 15E). There was a significant (p = 0.025) decrease in the expression of MASP-3 in the liver of mice with CAIA treated with GalNAc–MASP-3–siRNA duplex 2 compared with the mice with CAIA treated with GalNAc–luciferase–siRNA (Fig. 15D). Overall, there was a 31% decrease in the MASP-3 expression in the liver of CAIA mice treated with GalNAc–MASP-3–siRNA duplex 2 compared with the mice with CAIA treated with GalNAc–luciferase–siRNA. No significant change (p = 0.35) was seen for MASP-1 mRNA abundance in the liver from CAIA mice treated with GalNAc–MASP-3–siRNA compared with the mice with CAIA treated with GalNAc–luciferase–siRNA (Fig. 15E). These data show that GalNAc–MASP-3–siRNA duplex 2 decreased the expression of MASP-3 and not of MASP-1 in the liver.

MASP proteins (MASP-1 and MASP-2) were initially identified through their interaction with MBL (49–51). MASP-3 was identified shortly thereafter, also through interactions with MBL (23, 25). MASP-1 and MASP-3 are alternatively splice forms derived from the MASP1/3 gene along with the additional splice variant MAp44, whereas MASP-2 is encoded by the separate MASP2 gene. Although MASP-1 and MASP-3 are derived from the same gene, their functions are quite different. MASP-1 has recently been shown to activate MASP-2 via a cleavage event, thus enabling the MASP-2–mediated cleavage of C2 and C4 and the construction of the C3 convertase C4b2a for the LP (52, 53). MASP-1 is also suggested to be responsible for the activation of MASP-3, which, in contrast to MASP-2, is responsible for the cleavage of proFD to active FD for the AP (54). Thus MASP-3 is an important regulator of the AP, whereas MASP-1 is a more general MASP activator. Gene deletion experiments that generated MASP1/3^−/− mice resulted in both inactivated MASP-1 and MASP-3 functions, rendering our understanding of the individual roles of these two enzymes uncertain. Specific deficiency of MASP-3 is rare but has been observed in humans. Mutations in MASP1 exon 12 render MASP-3 inactivated and are responsible for the development of Malpuech–Michels–Mingarelli–Carnevale syndrome (55–57). This indicates that MASP-3 plays a developmental role as well as its role in the activation of the AP of complement.

In our studies of the role of MASPs in an experimental RA, we have previously shown that disruption of the MASP1 gene, resulting in the loss of both MASP-1 and MASP-3, almost completely blocked disease in the CAIA model (14). Subsequently, we found that overexpression of MAp44, a competitive inhibitor of MASP-1, MASP-2, and MASP-3, nearly completely blocked the development of disease (5). Dissociation of the effects of MASP-1 and MASP-3 genetically was technically challenging, however, requiring the selective disruption of exons unique to MASP-3.

Progress in the design of siRNAs, however, afforded us a second option, namely, to design siRNAs that could distinguish between MASP-1 and MASP-3 mRNAs. We approached this in two ways. Working with the functional genomics core of the University of Colorado, we designed shRNA sequences that could be expressed by lentiviral infection. The lentiviral construct was successful at targeting human MASP-3 mRNA but not MASP-1 or MASP-2. HepG2 and T98G cells chosen for both of these are high MASP-3–expressing cell lines (44). We saw no significant change in MASP-1 transcripts, whereas MASP-3 transcripts were almost entirely absent after lentiviral infection in T98G cells (Fig. 2C). Furthermore, the effect was durable and lasting for months in T98G cells (Supplemental Fig. 2A). In HepG2 cells, no effect was seen on MASP-1 and MASP-2, whereas there was a significant decrease in MASP-3 (Fig. 2F). High levels of MASPs in mouse liver compared with human liver indicate species-specific differences.

To evaluate specific silencing of MASP3 in vivo and mouse CAIA model, we used Alnylam Pharmaceuticals' liver-specific GalNAc-siRNA delivery technology. The triantennary GalNAc ligand engages the hepatocyte-specific ASGPR receptor enabling siRNA uptake. siRNAs conjugated to a GalNAc have been used extensively to silence liver-expressed genes in both preclinical species and in the clinic. This platform has demonstrated robust clinical translation across multiple liver target genes (42). For example, Inclisiran (ALN-PCSsc) significantly reduced the levels of PCSK9, a target for low-density lipoprotein (LDL) cholesterol, for at least 6 mo (43). In contrast, lentiviral constructs are rapidly cleared when administered systemically (58), making efficient knockdown of a target by this means difficult. Multiple GalNAc-siRNA conjugates specific for mouse MASP3 were designed and evaluated in in vitro studies and healthy mice to select a duplex for testing in the CAIA model.

The MASP-3 duplex chosen for our in vivo studies decreased liver MASP-3 mRNA by ∼50% 7 d after a single dose (Fig. 6). In our first CAIA study, however, providing three siRNA doses on days –5, 0, and +3 (relative to induction of disease) resulted in an almost complete elimination of MASP-3 mRNA (Fig. 7D). MASP-3 protein in the serum had decreased by 50% (Fig. 6A). This may be due, in part, to other tissue sources for MASP-3 and to the half-life of MASP-3 protein in the serum. CAIA disease progression was attenuated by the GalNAc–MASP-3–siRNA treatment in mice but not completely blocked. Histopathology scores for inflammation, pannus, cartilage damage, and bone damage were also consistent with the CDA. C3 deposition in the knee joints were also decreased and consistent with the decrease in MASP-3 deposition in the synovium of knee joints and forepaws. The attenuation of disease is consistent with the diminished presence of serum MASP-3 (Fig. 15B).

We have shown that CAIA is dependent on the AP as FD^−/−, MBL A/C/FD^−/−, C1q^−/−/FD^−/−, and MASP-1/3^−/− mice are resistant to CAIA (14, 59). An essential step in the AP is the cleavage of proFD to FD. Several groups have shown, using in vitro methods, that MASP-3 appears to be the essential enzyme that activates FD. One group has generated inhibitors specific for MASP-1, 2, and 3, and using these inhibitors has shown that MASP-3 alone appears to be responsible for proDF cleavage (31). Using various sera depleted for MASP-3 to varying extents compromises the conversion of proFD to DF (32). We show, in vivo, cleavage of proFD to DF and reconstitution of the AP activation in MASP-1/3^−/− mice by active rMASP-3 (Fig. 4). This demonstrates that MASP-3 is sufficient to reconstitute the AP in vivo but does not yet demonstrate that it is necessary. As MASP-1 and MASP-3 are generated by alternative splicing of a single gene, mice in which one or the other splice forms are genetically disrupted will be needed to formally prove this hypothesis in vivo.

Examining the time course of disease development in GalNAc–MASP-3 siRNA-treated mice, we note that disease is markedly inhibited through day 6 (69%) (Fig. 9). Then, a rapid acceleration of disease occurs on day 7 leading to the final inhibition of disease to 50% of control (Fig. 9A). We postulate that a mini breakthrough of MASP-3 expression may have occurred at this point. Alternatively, one can speculate that there is a possibility that an extrahepatic MASP-3 contribution by brain or inflammatory cells could have played a role in the breakthrough. Although a blood–brain barrier might exist, during acute inflammatory conditions, there could be a mini breakthrough in this blood–brain barrier. Although technically infeasible in this study, a future study will be designed to examine serum daily for MASP-3 levels to test this hypothesis.

Importantly, GalNAc–MASP-1–siRNA had no significant effect on CAIA even after inhibiting more than 95% of MASP-1 expression in the liver as well as MASP-1 protein in the circulation (data not shown), suggesting that the contribution of the MASP1/3 gene to CAIA is primarily through MASP-3 but not through MASP-1, although both are predominately generated by the liver. Given the suggested role of MASP-1 as an activator of MASP-3, this result demonstrates the existence of other activators of MASP-3. We also examined the effect of GalNAc–MASP-2 siRNA (data not shown) and obtained results similar to those we obtained with MASP-2^−/− mice (34). Given that the liver is also the major source of MASP-2, this result was expected. Interestingly, it has been shown using paired samples that high levels of MASP-3 versus MASP-2 are present in the synovial fluid as well as in the plasma of RA patients (17). Studies combining siRNAs targeting combinations of MASPs are planned for the future.

A limitation of our study is that liver-targeted MASP-3–siRNA duplexes did not completely inhibit MASP-3 expression as we observed MASP-3 protein in the circulation of mice after siRNA treatment. This may be due to the contribution of other tissues during acute inflammatory conditions or alternately may be due to the inherent stability of MASP-3 protein. Indeed, although we know that transcripts from the MASP1 gene derive from other tissues besides liver (44), we do not yet know the tissue specificity of the splicing event. Furthermore, we have no information on the rate of turnover of MASP-3 during steady state or during inflammation. All of these parameters will affect our interpretation of the data presented in this study. Thus, the partial inhibition of disease may simply be due to the partial removal of serum MASP-3. This question can only be answered by the construction of a MASP-3 exon tissue-specific knockout mouse strategy, allowing removal of all MASP-3 or tissue-specific expression while leaving MASP-1 and MAp44 intact.

The liver is seldom thought of as an organ associated with joint disease in RA. Given its role as a major source of soluble complement components (10, 11, 34), we suggest that this role should be reconsidered. Recently, it has been proposed that the etiology of RA involves a combination of genetics and epigenetics leading to an increased likelihood of developing an immune response to posttranslationally altered self (8). Deposition of Ab within the synovial space leads to the initiation of a joint specific immune response. This, however, is not sufficient. For the response to become self-sustaining, we propose that a second hit of liver-specific secretion of complement components necessary for the AP is necessary. Our treatment schedule (injections on days –5, 0, and +3 or on day –10, day –5, or day 0) has focused on the role of complement in the early phase of RA disease development. It will be important to employ our GalNAc–MASP-3–siRNA therapeutics in established disease to test whether complement remains a driver at all times or just at early times.

Given the importance of complement for innate immunity, it is likely that a therapeutic that strongly inhibits this system may increase susceptibility for infections. In this regard, it is, perhaps, beneficial that the GalNAc–MASP-3–siRNA only partially blocks MASP-3 protein levels. Interestingly, a recent study found that relatively small amounts of FD are sufficient to activate the AP in mice (60). Because MASP-3 deposition was low in the joints, we speculate that a small amount of MASP-3 can activate the AP locally or systemically. This idea is supported by the fact that MASP-1/3^+/− mice, which have suboptimal levels of MASP-3, were equally susceptible to CAIA when comparing with the WT mice (data not shown).

In conclusion, this represents the first study, to our knowledge, in vivo demonstrating an important role of MASP-3 (separate from MASP-1) in the development of RA. It supports the contention that the liver plays an important role in RA and supplies a candidate enabling the second hit to drive disease. Finally, it provides a further demonstration of the therapeutic potential for novel GalNAc–MASP-3–siRNA duplexes.

We are thankful to Umarani Pugazhenthi, the qRT-PCR facility the University of Colorado at Denver Cancer Center Shared Resource for designing the primers for MASPs for qRT-PCR analysis. We are also grateful to Dr. Molishree U. Joshi, the Functional Genomics Facility of the University of Colorado at Denver Cancer Center Shared Resource for preparing hu shMASP-3 RNA and human shScramble RNA lentiviral particles. We are thankful to Gaurav Mehta, Joseline Ramos Ramirez, Nhu Ho, and Numair Khan of University of Colorado at Denver for assisting in some studies related to this work.

The authors have no financial conflicts of interest.